JP7844381B2 - Energy operation planning device and energy operation planning method - Google Patents

Description

This invention relates to an energy utilization planning device and an energy utilization planning method.

To ensure efficient power operation, a power operation plan is created that outlines guidelines for future equipment operation and control command values based on predicted future conditions. An example of such a plan is a generator operation plan. Based on predicted power demand at each point in the plan period, the generator operation plan determines the operation or shutdown state of each generator and the power grid, as well as the generator output, in order to minimize total power generation costs, while satisfying operational constraints for each generator and power system. Furthermore, if electricity trading is conducted in the power market, the operation plan may be determined based on predicted electricity market prices to maximize not only total power generation costs but also revenue from electricity trading.

Such a method for planning the operation of a generator is disclosed, for example, in Non-Patent Document 1. Non-Patent Document 1 discloses a technology for calculating an operation plan that minimizes the total power generation cost while satisfying operational constraints of each generator and power system, such as the supply-demand balance where the demand and supply of electricity match, and the minimum continuous start-up time and minimum continuous stop-down time where the generator maintains its state for a certain period of time after starting up and stopping. In calculating this operation plan, even if there is only one generator, if the two states of starting up and stopping the generator are considered for all n time cross-sections of the operation plan, a vast number of 2 ^{^n} combinations of operation plans can be considered. For this reason, an optimization method that can quickly determine an operation plan that minimizes the total power generation cost or maximizes profits from among the vast number of combinations of operation plans has been essential.

On the other hand, for renewable energy sources such as solar power, whose output depends on weather conditions, future power generation forecasts are subject to prediction errors, which represent a discrepancy between the forecast and the actual power generation during actual operating hours. Further increases in the amount of renewable energy introduced are planned for the future. However, as the amount of renewable energy introduced increases, the impact of prediction errors also expands. Therefore, the discrepancy between actual power generation and the forecast value becomes larger, making it difficult to meet operational constraints such as the balance of electricity supply and demand. Furthermore, the prediction errors for renewable energy also increase the prediction error, which represents the discrepancy between the predicted value and the actual contracted price of electricity in the market.

If operations are modified after implementation based on an operational plan due to prediction errors in renewable energy, it could lead to an increase in total power generation costs and a deterioration in profitability. In such uncertain situations where actual operational conditions cannot be accurately predicted, planning techniques that simulate uncertainties such as renewable energy generation and electricity market prices using a vast number of prediction scenarios have been proposed in Non-Patent Documents 2 and 3. According to these techniques, operational plans are created not based on a single assumed scenario, but on scenarios with different predicted values. Therefore, if the deviation of actual power generation is small in any of the scenarios, uncertainties such as prediction errors are assumed, making it possible to mitigate the impact of uncertainty on the operational plan.

M. Carrion and J. M. Arroyo, "A computationally efficient mixed-integer linear formulation for the thermal unit commitment problem," in IEEE Transactions on Power Systems, vol. 21, no. 3, pp. 1371-1378, Aug. 2006 S. Takriti, J. R. Birge and E. Long, "A stochastic model for the unit commitment problem," in IEEE Transactions on Power Systems, vol. 11, no. 3, pp. 1497-1508, Aug. 1996 Yao Zhang, et al., “Chance-Constrained Two-Stage Unit Commitment under Uncertain Load and Wind Power Output Using Bilinear Benders Decomposition”, IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 5(2017) Yoav Goldberg, “A Primer on Neural Network Models for Natural language Processing,” Journal of Artificial Intelligence Research 57, pp.354-357(1016)

As mentioned above, in creating power operation plans, uncertainty scenarios were anticipated in advance to minimize the impact of uncertainties such as prediction errors in future conditions. These uncertainty scenarios include not only fluctuations in renewable energy generation, but also scenarios caused by short-term weather changes, such as a brief period of extreme heat despite it being the rainy season. Furthermore, for longer-term plans, scenarios caused by long-term factors, such as a mild winter contrary to predictions, would also be included.

The more uncertain scenarios anticipated in advance, the larger the computational scale required to create the operational plan becomes. In this case, there was a concern that the process of creating the operational plan would require an enormous amount of computation time, making it impractical for actual operation.

Therefore, to reduce the computation time required to create the operational plan, it is necessary to reduce the number of uncertainty scenarios. However, simply reducing the number of uncertainty scenarios will also decrease the accuracy of uncertainty simulation, resulting in insufficient uncertainty assumptions at the time of operational plan creation. If an unexpected uncertainty scenario occurs after the operational plan has been created, the operational plan must be revised and the generators operated accordingly. As a result, there are concerns about an increase in total power generation costs and a deterioration in profitability.

This invention was made in view of these circumstances, and aims to create an energy operation plan by selecting scenarios that do not affect the results even when scenarios that anticipate uncertainty are removed.

The energy operation planning device according to the present invention comprises: an estimated plan creation unit that inputs information necessary for creating an energy operation plan as multiple scenarios into an estimation model and creates an estimated plan by estimating the energy operation plan; an evaluation unit that evaluates the similarity between scenarios based on the results of the estimated plan; a scenario selection unit that selects scenarios from among the scenarios input into the estimation model so that the scenarios are not similar based on the similarity evaluation by the evaluation unit; and a planning unit that creates an energy operation plan using the selected scenarios.

According to the present invention, by selecting scenarios in which the results are not affected even if scenarios that assume uncertainty are removed, it becomes possible to create an energy operation plan that can improve economic efficiency while suppressing the effects of uncertainty within a practical computation time.
Other issues, configurations, and effects not mentioned above will be clarified by the following description of the embodiments.

This is a block diagram showing an example of the functional configuration of a power operation planning device according to the first embodiment of the present invention. This is a block diagram showing an example of the internal configuration of the decision variable evaluation and scenario selection unit for the estimated plan according to the first embodiment of the present invention. This figure shows an example of the configuration of a power system according to the first embodiment of the present invention, and an example of the hardware configuration of a power operation planning device according to the first embodiment. This flowchart shows an example of the process by which a power operation planning device according to the first embodiment of the present invention selects an uncertainty scenario and creates a power operation plan. This figure shows an example of a screen display of the operating cost distribution for uncertainty according to the first embodiment of the present invention. This graph shows the relationship between the number of scenarios according to the first embodiment of the present invention and the accuracy of the planning results for multiple scenarios that simulate uncertainty. This figure shows an example of the functional configuration of a power operation planning device according to a second embodiment of the present invention. This is a block diagram showing an example of the internal configuration of the decision variable evaluation and scenario selection unit for the estimated plan according to a second embodiment of the present invention. This flowchart shows an example of processing by the calculation unit of a power operation planning device according to a second embodiment of the present invention. This is a conceptual diagram of loss risk according to a second embodiment of the present invention.

The following describes embodiments preferred for carrying out the present invention. It should be noted that the following are merely examples of embodiments, and the invention itself is not intended to be limited to the specific details described below.
Furthermore, in the following description, the same or similar elements and processes are denoted by the same reference numerals, and redundant explanations are omitted. Also, in later embodiments, only the differences from previously described embodiments are explained, and redundant explanations are omitted. In addition, the following descriptions of embodiments and the configurations and processes shown in each figure are intended to provide an overview of the embodiments to the extent necessary for understanding and implementing the present invention, and are not intended to limit the manner in which the present invention can be implemented. Furthermore, each embodiment and each modification can be combined in whole or in part within a range that is consistent with each other and does not depart from the spirit of the present invention.

This specification describes an energy operation planning device that assists in the development of energy operation plans by operators with multiple energy facilities or operators capable of coordinating the operation plans of energy facilities. Below, an example of a power operation planning device for a power business to which the present invention applies is described.

[First Embodiment]
<Functional Configuration of Power Operation Planning System 1>
First, an example of the functional configuration of the power operation planning device 1 will be explained with reference to Figures 1 and 2.
Figure 1 is a block diagram showing an example of the functional configuration of the power operation planning device 1 according to the first embodiment.
The power operation planning device 1 (an example of an energy operation planning device) comprises a power generation and grid information unit 2, a calculation unit 3, and a screen display and result storage unit 4, and performs processing for power operation planning.

The power generation and grid information unit 2 has the function of storing information necessary for creating a power operation plan, as shown in Figure 3 below, namely the grid information database DB1 and the power operation plan database DB2. This power generation and grid information unit 2 stores, for example, power generation equipment information, periodic inspection information, grid information, supply and demand information, adjustment capacity information, renewable energy information, assumed accident information, and scenario information.

Power generation equipment information includes information such as equipment constants that indicate the characteristics of each generator.
Periodic inspection information refers to information regarding the shutdown or restriction of the generator's output for maintenance and inspection purposes.
System information includes information such as the maximum transmission capacity of interconnection lines in the power grid.
Supply and demand information includes electricity demand, which is the required amount of power generation, and the probability distribution range of demand, including forecast errors.

Adjustment capacity information is information used to adjust for deviations from assumed conditions, such as power operation plans.
Renewable energy information includes information such as the characteristics of renewable energy sources, the amount of electricity generated, and the probability distribution range of the amount of electricity that can be generated, including prediction errors.
The anticipated accident information includes details about generators and equipment that will be shut down due to malfunctions preventing them from operating.
Scenario information refers to information about scenarios assumed by the operator, or scenarios statistically created from a range of possible probability distributions, including supply and demand information and prediction errors in renewable energy.

Next, we will describe an example of the functional configuration of the arithmetic unit 3.
The calculation unit 3 includes a scenario-based plan estimation unit 31, a decision variable evaluation and scenario selection unit 32 for the estimated plan, and an uncertainty plan calculation unit 33.

In the scenario-based planning estimation unit 31 (an example of an estimation plan creation unit), information necessary for creating a power operation plan, such as generator equipment information, supply and demand information, and scenario information, is input from the power generation and grid information unit 2 as multiple scenarios into an approximate estimation model. This estimation model is a model that shows the relationship between scenarios and planning results. For example, the estimation model is a model that simulates the relationship between the information necessary for creating a power operation plan and a power operation plan created by strictly solving the optimization problem of equation (1) described later, and is composed of a method different from the method used to create the power operation plan. Based on this simulated relationship, the estimation model outputs a planning result (also called an "estimated plan") that estimates the power operation plan from the input. The estimation model is, for example, a relaxed optimization problem that relaxes the optimization problem of equation (1) for creating a power operation plan. The estimation model may also be a machine learning model that has learned the relationship between the information necessary for creating a power operation plan and a power operation plan that assumes uncertainty. Details of the relaxed optimization problem and machine learning model will be described later.

The scenario-based planning estimation unit 31 outputs the output of the estimation model as an estimated plan that takes into account the uncertainty of the power operation plan. Here, the scenario-based planning estimation unit 31 takes information necessary for creating the power operation plan, such as generator equipment information, supply and demand information, and scenario information, as input as a scenario. As a scenario, for example, there are multiple scenarios where the predicted power demand (the amount of electricity people need) is inaccurate, to account for cases where the predicted power demand is incorrect. For these scenarios, the power operation plan is, for example, a plan that determines when and how much power to generate in order to decide the operation of the generators on a given day. The planning result of this plan is information (also called the "planning result") that shows when and how much power will be generated according to the power operation plan. In other words, the planning result is the output derived from the plan, and by showing the operator when to start or stop the generators, it serves as information for the operator to decide on the operation of the generators on a given day.

Furthermore, the plan results include elements that the operator does not currently consider important, as these can be adjusted by the operator according to future circumstances. While multiple scenarios may arise if the operator anticipates several different situations, if the differences between scenarios are small, deleting one of the scenarios with small differences may not change the elements the operator considers important in the power operation plan. Therefore, extracting all scenarios stored in the power generation and grid information unit 2 would result in redundant information. To reduce the calculation time required for creating the power operation plan described later, it is necessary to extract the necessary scenarios by excluding redundant ones, based on the estimated plan output by the scenario-based plan estimation unit 31.

Here, an example of the functional configuration of the decision variable evaluation and scenario selection unit 32 for the preliminary plan will be explained with reference to Figure 2.
Figure 2 is a block diagram showing an example of the internal configuration of the decision variable evaluation and scenario selection unit 32 for the preliminary plan.

The preliminary plan's decision variable evaluation and scenario selection unit 32 comprises a similarity evaluation unit 321 for plan result features and a scenario selection unit 322.

The similarity evaluation unit 321 of the planning result features evaluates the similarity between scenarios based on the results of the estimated plans created by the scenario-based planning estimation unit 31. Here, the similarity evaluation unit 321 evaluates the similarity based on the estimated plans, assuming that the effects on the energy operation plan are similar for each scenario. For example, if the same planning result is estimated for different scenarios, then the planning result that is not an estimate for the scenario on which that estimated plan result was based is also evaluated as similar. Here, when proceeding with operations according to the plan, the planning results include elements that, even if the plan content is modified during operation, do not result in violations of operational constraints or additional costs, and therefore do not affect operations. When evaluating the planning results, instead of evaluating all elements of the planning results, it is also possible to exclude elements that do not affect operations even if changed, and only evaluate the decision variables that would result in violations of operational constraints or additional costs if not decided during planning.

The scenario selection unit 322 selects scenarios from those input to the estimation model based on the similarity evaluation of the plan result features by the similarity evaluation unit 321, ensuring that the scenarios are not similar. For example, the scenario selection unit 322 deletes similar scenarios based on the similarity evaluation of the plan result features by the similarity evaluation unit 321 and selects the remaining scenarios. In this process, the scenarios that form the basis of the estimated plan results evaluated as similar are considered duplicate scenarios. The scenario selection unit 322 determines one of the duplicate scenarios to be unnecessary, removes the duplicate scenario, and selects the remaining scenarios. The scenarios selected by the scenario selection unit 322 are output to the uncertainty plan calculation unit 33.

Here, the scenario selection unit 322, based on the scenarios input by the scenario-based plan estimation unit 31, selects scenarios based on a probability similarity evaluation index, where the value decreases as the probability of the scenario occurring decreases and the similarity evaluation with other scenarios increases. It prioritizes deleting scenarios with low probability similarity evaluation index values, and then selects the remaining scenarios. Therefore, even if the scenario selection unit 322 deletes a scenario, its probability of occurrence is low, minimizing the impact on the results of the power operation plan created from the remaining scenarios.

Returning to Figure 1, we continue the explanation.
The uncertainty planning calculation unit 33 (an example of a planning unit) calculates a power operation plan that takes uncertainty into account using the scenarios selected by the preliminary plan determination variable evaluation and scenario selection unit 32. The uncertainty planning calculation unit 33 outputs the uncertainty scenarios and power operation plan to the screen display and result saving unit 4.

The screen display and result storage unit 4 displays the uncertainty scenario and power operation plan output from the uncertainty plan calculation unit 33 on the screen of the display unit 21, as shown in Figure 3, which will be described later. The screen display and result storage unit 4 also saves the results of the power operation plan to the power operation plan database DB2, as shown in Figure 3, which will be described later.

<Hardware configuration of power operation planning device 1>
Figure 3 shows an example of the configuration of the power system 100 and an example of the hardware configuration of the power operation planning device 1 according to the first embodiment.

The upper part of Figure 3 shows an example of the configuration of the power system 100. The power system 100 is a system in which multiple generators 130 and loads 150 are interconnected via busbars (nodes) 110, transformers 120, transmission lines 140, etc. The system information representing the configuration of the power system 100 is stored in the system information database DB1 shown in the lower part of Figure 3.

The power operation planning device 1 acquires various information, such as power generation equipment information stored in the power generation and system information unit 2 shown in Figure 1, and multiple measurement data, such as load 150, from the power system 100. The power operation meter information acquired by the power operation planning device 1 from the power system 100 via the communication network 300 is stored in the power operation planning database DB2 shown at the bottom of Figure 3.

In Figure 3, various measuring instruments are appropriately installed on the busbar 110 for the purpose of protecting, controlling, and monitoring the power system 100. Signals detected by the measuring instruments are transmitted to the communication unit 23 of the power operation planning device 1 via the communication network 300. The dashed lines in the figure represent how the generator 130 and load 150 transmit signals detected by the measuring instruments wirelessly to the communication network 300.

The lower part of Figure 3 shows an example of the hardware configuration of the power operation planning device 1 for the power system 100.
The power operation planning device 1 is composed of a computer system. This power operation planning device 1 includes a display unit 21, an input unit 22, a communication unit 23, a CPU 24, a memory 25, and various databases (system information database DB1, power operation planning database DB2), and each unit is connected to a bus line 26.

The display unit 21 is, for example, a display device. Alternatively, the display unit 21 may be configured to use, for example, a printer device or an audio output device, either in place of or in conjunction with a display device.

The input unit 22 is configured to include at least one of the following: a pointing device such as a keyboard or mouse, a touch panel, or a voice instruction device.

The communication unit 23 includes circuits and communication protocols for connecting to the communication network 300. The communication unit 23 also communicates with weather systems, power market systems, and aggregators that monitor and control multiple distributed power sources and consumers, such as VPPs (Virtual Power Plants).

The CPU 24 works in cooperation with the memory 25 to execute programs and implement each part of the power operation planning device 1, as well as to issue instructions for image data to be displayed and to search for data in various databases. The CPU 24 may be configured as one or more semiconductor chips, or as a computer device such as a computing server.

Memory 25 is configured, for example, as RAM (Random Access Memory) and stores computer programs, calculation result data, image data, etc., necessary for each process. Alternatively, memory 25 may be ROM (Read Only Memory), HDD (Hard Disk Drive), SSD (Solid State Drive), non-volatile memory, etc. In this case, memory 25 is used as an example of a non-transient storage medium readable by the computer storing the program executed by the power operation planning device 1. The data stored in memory 25 is saved in a database, sent to the display unit 21 for display, or transmitted as operation control commands to each piece of equipment, such as generators, via the communication network 300. Since each piece of equipment is operated based on the operation control command value, the control command value must be a value that satisfies the operation constraints.

<Selection of uncertainty scenarios and creation of power operation plans>
Next, the process of selecting uncertainty scenarios for the power operation planning device 1 and creating a power operation plan will be explained with reference to Figure 4.

Figure 4 is a flowchart illustrating an example of the process by which the power operation planning device 1, according to the first embodiment, selects an uncertainty scenario and creates a power operation plan. The power operation planning device 1 inputs the necessary input information for creating the power operation plan into an estimation model and has a function to select a scenario based on the estimated plan output from the estimation model. Subsequently, the power operation planning device 1 creates a power operation plan that takes uncertainty into account, based on the selected scenario. The process from step S11 shown in Figure 4 is described below.

In step S11 of Figure 4, the planning estimation unit 31, based on the scenario shown in Figure 1, extracts information necessary for creating a power operation plan, such as power generation equipment information, supply and demand information, and scenario information, from the power generation and grid information unit 2 as an initial scenario. The scenario information extracted as an initial scenario includes scenarios showing cases where the assumptions for power demand and renewable energy generation differ between the forecast and the actual operation due to prediction errors, or a range of possible probability distributions including prediction errors.

Furthermore, the process by which the scenario-based planning and estimation unit 31 extracts the initial scenario may use a scenario desired by the operator of the power operation planning device 1, such as past power demand data. Alternatively, the operator of the power operation planning device 1 may use past time-series data with similar conditions, such as the same season and period, as the initial scenario.

Furthermore, if the scenario information represents a range of possible probability distributions including prediction errors, the scenario-based planning estimation unit 31 may create an uncertainty scenario showing the prediction error based on the probability distribution. This uncertainty scenario can be created using statistical scenario creation methods such as Monte Carlo simulation, based on the probability distribution range for each time point. In this way, the scenario-based planning estimation unit 31 extracts the information necessary for creating a power operation plan that considers uncertainty from the power generation and grid information unit 2.

In step S12 of Figure 4, the scenario-based planning estimation unit 31 estimates the power operation plan based on the scenario using an estimation model that shows the relationship between the input information necessary for creating the power operation plan extracted in step S11 and the results of the power operation plan. Note that when estimating the power operation plan, the estimation model may be used for each scenario to estimate the plan results corresponding to that scenario.

Two methods are envisioned as examples of this estimation model: (1) a method utilizing a relaxation optimization problem, and (2) a method utilizing a pre-trained machine learning model. These two methods are described below. Note that the estimation of planning results below refers to the scenario-based planning calculation unit 31 estimating a power operation plan based on the scenario.

(1) Method utilizing a relaxed optimization problem The method of estimating the planning results by utilizing a relaxed optimization problem involves approximately solving the optimization problem in power operation planning that assumes the uncertainty shown in equation (1) below. The objective function of the optimization problem in equation (1) is to minimize the total power generation cost of all generators during the planning period. The definitions of the variables that make up equation (1) and the constraints of equation (1) are also described below.

Tend: End time of the plan
Ngen: Number of generators
ai, bi, ci: Power generation cost coefficients for generator i
Pits: Scenario s, generator i, power output at time t
uits: Scenario s, generator i, discrete variables Δuits indicating start/stop at time t; Scenario s, generator i, discrete variables Δuits indicating start/stop at time t;
SUCi: Startup cost of generator i
S: Number of expected scenarios
Prs: Probability of scenario s occurring
PNLts: Penalties for violations of operational constraints

<Restrictions>
- Maximum and minimum generator output (the output of each generator is within the range from the maximum output to the minimum output)
- Supply and demand balance (the demand for each share matches the total power generation output)
- Minimum continuous operating time and minimum continuous shutdown time (restarting or restarting will occur after the minimum continuous time has elapsed).
- Operating period and shutdown period (the generator will be shut down or operation will continue during the specified period)
- Operating reserve capacity and necessary adjustment capacity (the capacity to correct the discrepancy between actual operation and supply and demand plans)

Here, if the scenario-based planning and estimation unit 31 attempts to solve equation (1) exactly, it will take a very long time to compute. This is because it needs to solve equation (1) for all scenarios. Furthermore, it searches for the optimal values and combinations of discrete variables from a vast number of solution combinations for the numerous discrete variables included in equation (1). Therefore, the scenario-based planning and estimation unit 31 estimates the planning results by solving a simplified relaxed optimization problem of equation (1). By simplifying the complex characteristics of equation (1), the relaxed optimization problem can be solved in less time than solving the original optimization problem.

The method for solving this relaxed optimization problem involves solving a relaxed optimization problem by approximating all the complex discrete characteristics of equation (1) as continuous, rather than strictly solving the optimization problem related to power operation planning in equation (1).

The discrete variable uits, which represents a vast number of combinations depending on which generator i is 1 or 0 at which time t, is approximated by a continuous variable where 0 < uits_con < 1. Therefore, discrete combinations of uits do not arise. As a result, the scenario-based planning and estimation unit 31 can solve the relaxed optimization problem in a short time. The approximate power operation plan obtained by solving this relaxed optimization problem is treated as the estimated plan.

(2) Method using a pre-trained machine learning model In the method of estimating the planning results using a pre-trained machine learning model, the relationship between the input information for creating the power operation plan and the power operation plan that has been created is pre-trained in the machine learning model. By inputting the input information into this pre-trained machine learning model, the machine learning model estimates the planning results, that is, creates an approximate power operation plan.

One method for estimating planning results based on machine learning is, for example, the method utilizing the neural network shown in equation (2) of Non-Patent Document 4. In this case, the machine learning model is expressed as shown in equation (2) below.

g ⁽¹⁾ : Activation function
W ⁽ⁿ⁾ : Weight coefficients for n levels
b ⁽ⁿ⁾ : Bias coefficient for n levels
x: Input information
y: Output information

Equation (2) simulates the relationship between input information x and the target of estimation y. In equation (2), w and b are adjusted so that the output information y, obtained by applying a nonlinear function g ⁽ⁿ⁾ to (wx+b), which is the weighted sum wx plus bias b, corresponds to the power operation plan, which is the target of estimation, for input information x such as a scenario.

When the scenario-based planning and estimation unit 31 estimates the planning result from the input information x, it is assumed that examples of estimation targets (planning results) corresponding to the input information x are prepared in advance, and the learning model of equation (2) is trained with these examples to adjust w. The scenario-based planning and estimation unit 31 can estimate the planning result corresponding to the input x by inputting the input information x to the trained equation (2). This estimated planning result is treated as a rough plan.

As described above, the scenario-based planning unit 31 can estimate the plan results by approximating the power operation plan using methods such as methods based on relaxation optimization problems or machine learning. Here, the operator may be notified of the estimated power operation plan results, which take uncertainty into account, by displaying these estimated plans on the screen. Compared to strictly solving equation (1), which is the optimization problem for power operation planning that assumes uncertainty, obtaining the plan results in a shorter time can improve the operator's work efficiency.

When outputting the estimated plan to the screen, the scenario-based plan estimation unit 31 calculates the probability of each estimated plan occurring and the operating cost index for the estimated plan. An example of the screen display based on these probabilities and operating cost indexes is shown in Figure 5. Figure 5 is a graph showing the operating cost distribution against uncertainty in the power operation plan, with the horizontal axis representing operating costs and revenues, and the vertical axis representing frequency. Revenues are shown on the right side of the horizontal axis, and losses on the left side. The operating cost distribution shown in Figure 5 is displayed on the display unit 21 shown in Figure 3 by the screen display and result saving unit 4.

The scenario-based planning and estimation unit 31 can report to the operator whether operating costs will fluctuate due to uncertainty or whether there is a possibility of loss, by showing the operating cost distribution to the operator through the screen display and result saving unit 4. For example, if the frequency of a certain scenario is skewed towards losses, the operator can determine, when creating a power operation plan, that there is a high probability of loss if that scenario occurs in actual operation.

Returning to the explanation of the flowchart shown in Figure 4.
Here, the estimated plan result obtained in step S12 is an approximate plan result, and therefore deviates from the plan result created by strictly solving the power operation plan optimization problem in equation (1). Since these approximate plan results do not provide the accuracy of a plan that can actually operate power equipment, in the following steps (S13 to S15), the approximate plan results are used as criteria for selecting scenarios. In particular, based on the approximate plan result estimated in step S12, scenarios with similar and overlapping plan results are excluded. The scenario selection unit 322 selects the minimum number of scenarios necessary to accurately simulate uncertainty. The scenario selection based on these approximate plans is shown in the following steps S13 to S15.

In step S13 of Figure 4, the similarity evaluation unit 321 of the planning result features shown in Figure 2 performs a similarity evaluation to determine whether the planning results are similar and overlapping, based on the estimated planning results in step S12. The similarity evaluation unit 321 uses, for example, the index in equation (3), which extracts elements from each scenario s for the objective function in equation (1), as a criterion for determining whether the estimated planning results are similar and overlapping. Equation (3) assumes the total power generation cost for each scenario.

Equation (3) extracts elements corresponding to one scenario from the sum of the elements from 1 to s in the Σs Pr term of equation (1). Therefore, equation (3) does not contain Σs and is specific to a particular scenario s.

Furthermore, the similarity evaluation unit 321 of the planning result features may also use specific elements as evaluation indicators in addition to the elements described above for each scenario s. For example, in the planning results, there are elements that can be adjusted according to the situation, such as generator output, which can be adjusted in a short time, even if any of the scenarios actually occur. While these elements can be adjusted according to the situation, there are also elements that cannot be adjusted according to the situation after the plan is created due to constraints on generator operation, such as slow adjustments, or due to policies and systems in power generation operation. Since elements that cannot be adjusted according to the situation are more important to the operator than elements that can be adjusted according to the situation, these elements that cannot be adjusted according to the situation after the plan is created due to constraints on generator operation or systems in power generation operation may also be used as evaluation indicators.

Here, the similarity evaluation unit 321 of the planning result features treats elements that cannot be adjusted according to the situation after the creation of the power operation plan as evaluation indicators and as decision variables determined in the power operation plan. These decision variables are elements whose values determined in the power operation plan cannot be changed during the operation of the power operation plan. A specific example of a decision variable is the starting and stopping of generators, which cannot be adjusted according to the situation even if the prediction error changes immediately before operation on the day after the power operation plan is created the day before the day of operation. In addition, the trading volume of the electricity market, which is finalized the day before and cannot be adjusted on the day of operation, also becomes a decision variable. Furthermore, in the case of starting and stopping _generators , if the power operation planning device 1 creates the power operation plan the day before the actual operation day, the variables for the few hours immediately before operation become the decision variables in equation (1). The reason for this is that even if a power operation plan for generators has been created, if the generators are already operating according to the plan and the generators scheduled to operate in the near future are the ones that will be operated soon, it is not possible to change the plan, such as starting a different generator than planned. In the similarity evaluation unit 321 of the planned result features, "planned result features" refers to the decision variables.

Furthermore, if the amount of fuel purchased before operation begins, which is necessary for power generation, is used as the determining variable, then instead of the determining variable, variables that influence the determining variable, such as fuel consumption in each scenario, may be used as evaluation indicators.
Here, in equation (1), the decision variables were defined for each scenario in order to evaluate the planning results according to each scenario. However, after the creation of the power operation plan, the decision variables are elements that cannot be adjusted according to the occurrence of each scenario. For this reason, when creating the power operation plan that takes uncertainty into account in step S16 described later, it is assumed that the decision variables are the same across all scenarios.

The similarity evaluation unit 321 of the planning result features evaluates the similarity and overlap of the planning results based on the evaluation index for each scenario s in the objective function and the decision variables determined from the power operation plan results in the planning results estimated for each scenario. In this process, the evaluation method shown in equation (4) below is used.

s: Similarity and scenarios to be evaluated (scenarios to be deleted)
s': Comparison target with scenario s
Prs: Probability of scenario s occurring
ys: A vector of normalized evaluation metrics (In each scenario s, the evaluation metrics and decision variables in the objective function are normalized and combined into a single vector).

The similarity evaluation unit 321 of the planned result features evaluates the target scenario s using equation (4) based on its proximity to scenario s', which has a small probability of occurrence Prs or the closest distance (||・|| Euclidean norm) of the evaluation index y. Here, the evaluation value in equation (4) is used as the probability similarity evaluation index.

If the probability of scenario s occurring, Pr _s, is small, it is considered unlikely to actually occur. Removing such scenarios has little impact on the planning results. Also, if the distance between scenario s and the closest scenario s' is small, it means that there are scenarios that will result in similar planning outcomes. Therefore, the similarity evaluation unit 321 of the planning result features calculates a similarity evaluation index from equation (4) for each scenario as a similarity evaluation of the planning results according to the scenario. The similarity evaluation index calculated for each scenario is used as the similarity evaluation value.

In step S14 of Figure 4, the scenario selection unit 322 selects a scenario based on the similarity evaluation value calculated in step S13. The scenario selection unit 322 removes the scenario evaluated as overlapping due to having the smallest similarity evaluation value, and excludes it from scenario s or scenario s'. The probability of occurrence of the scenario removed by the scenario selection unit 322 is added to the probability of occurrence of the most similar scenario in equation (4).

Here, the relationship between the number of scenarios and the accuracy of the uncertainty simulated by multiple scenarios will be explained with reference to Figure 6. Figure 6 is a graph showing the relationship between the number of scenarios and the accuracy of the planning results for multiple scenarios simulating uncertainty. In the graph shown in Figure 6, the horizontal axis represents the number of scenarios, and the vertical axis represents the deterioration of the accuracy of the planning results according to the number of scenarios. Here, the deterioration of the planning results is evaluated, for example, by the expected value of the total power generation cost calculated using equation (1) based on the approximate planning results of all scenarios, and represents the change in the expected value that changes as the number of scenarios decreases from the total number of scenarios.

The initial state corresponds to scenario number A0, located at the rightmost end of the horizontal axis in the graph shown in Figure 6. Since scenario number A0 represents the initial state with the largest number of scenarios, the degradation of the planning results' accuracy is minimal. However, as duplicate scenarios are removed through the process in step S14 of Figure 4, the accuracy of the planning results gradually deteriorates, as indicated by the white arrows in the figure.

The scenario selection unit 322 evaluates the accuracy degradation of the planning results due to scenario exclusion, as shown in Figure 6, each time it excludes a scenario evaluated as highly similar based on the similarity evaluation index of equation (4). As long as the accuracy degradation remains within the threshold a%, scenarios are excluded, and the remaining scenarios are selected. Note that, above, the accuracy degradation of the planning results was evaluated using the expected cost calculated based on the scenario and the estimated planning results. However, instead of the expected cost, the expected value of the probabilistic similarity evaluation index shown by the similarity evaluation unit 321 of the planning result features may be used as the evaluation value. Alternatively, multiple elements may be treated as a single vector, as in ys in equation (4), and the evaluation may be based on the expected value of the distance between each scenario (||・|| Euclidean norm). By doing so, the accuracy of the power operation plan, which was created assuming uncertainty, can be guaranteed by excluding duplicate scenarios while keeping the degradation of the planning results within the threshold.

In step S15 of Figure 4, the scenario selection unit 322 evaluates whether the accuracy of the planning results shown in step S14 is maintained within the threshold of a%. The scenario selection unit 322 evaluates that the accuracy is within the threshold if the error representing the deterioration of the planning result accuracy is within a% of the specified error shown in Figure 6 (YES in S15). In this case, since there is no problem with the accuracy of the planning result, the scenario selection unit 322 returns to step S13 and repeats the deletion of scenarios with overlapping planning results. As the process from step S13 to S15 is repeated, the accuracy deteriorates from the initial state each time the number of scenarios decreases, as shown in Figure 6.

The scenario selection unit 322 evaluates that the error representing the deterioration of the accuracy of the plan result is not within the threshold if it exceeds a% of the specified error (NO in S15). In this case, since there is a problem with the accuracy of the plan result, the scenario selection unit 322 stops the scenario deletion process. Subsequently, the scenario selection unit 322 outputs the remaining scenarios that were not deleted to the uncertainty plan calculation unit 33.

In step S16 of Figure 4, the uncertainty planning unit 33 creates a power operation plan that takes uncertainty into account. Here, the uncertainty planning unit 33 creates the power operation plan by solving an optimization problem that considers uncertainty using equation (1) for scenarios where the plan results selected in step S15 do not overlap.

Unlike the approximate planning result in step S12, the uncertainty planning calculation unit 33 solves equation (1) without approximating its characteristics. In this case, the uncertainty planning calculation unit 33 obtains an exact solution by applying a commercial optimization solver to solve the optimization problem in equation (1). Subsequently, the uncertainty planning calculation unit 33 outputs the created power operation plan to the screen display and the result storage unit 4.

In step S17 of Figure 4, the screen display and result saving unit 4 displays the scenarios deleted in steps S13 to S15 due to duplicate planning results, the selected scenarios with no duplicate planning results, and the power operation plan results created in step S16 on the screen of the display unit 21 shown in Figure 3, and saves them to the power operation plan database DB2 shown in Figure 3.

<Effects of the First Embodiment>
In the power operation planning device 1 according to the first embodiment described above, when creating a power operation plan that takes into account uncertainties such as prediction errors, the relationship between a vast number of scenarios that simulate uncertainty and the plan results created based on those scenarios is evaluated in advance. Based on this evaluation, scenarios whose effects on the plan results overlap are reduced, and the remaining scenarios are selected. The power operation planning device 1 then creates a power operation plan that takes uncertainty into account for the selected scenarios. In this way, the power operation planning device 1 deletes only the scenarios whose effects on the plan overlap, that is, scenarios whose effects on the plan overlap, and selects scenarios that affect the plan results. Furthermore, even when deleting overlapping scenarios, the accuracy of the plan results is guaranteed by the evaluation, so even if the number of scenarios is reduced, the accuracy of the uncertainty simulation does not deteriorate, and the computation time required to create the power operation plan is reduced.

Here, in evaluating the effect of scenarios on the planning results, the power operation planning device 1 constructed an estimation model that shows the relationship between scenarios and planning results. This estimation model uses methods such as solving an approximate relaxation problem of the optimization problem used in power generation planning, or utilizing a machine learning model that has been pre-trained on the relationship between scenarios and planning results, to estimate planning results according to the scenario. Therefore, compared to the power operation planning device 1 creating a power generation plan by strictly solving the optimization problem related to power generation planning, it can estimate planning results faster and evaluate the similarity of scenarios based on the estimated planning results.

Furthermore, the power operation planning device 1 uses elements that are restricted from being modified during planned operation after plan creation due to equipment and operational regulations as decision variables in the approximate planning results output from the estimation model. These decision variables and the objective function value of the plan are evaluated, and scenarios with a low probability of occurrence and a high degree of overlap are deleted while evaluating the impact before and after deletion. In this way, because the power operation planning device 1 deletes only scenarios with overlapping planning results, it can ensure optimal accuracy of the elements to be determined in the plan while also speeding up the computation time required to create a power operation plan that considers uncertainty.

In conventional technology, when creating plans that anticipate uncertainty, the impact of scenarios on the plan results was not considered. Therefore, reducing the number of uncertainty scenarios assumed in the plan resulted in lower optimal accuracy of the plan results and the decision variables used in the plan. On the other hand, increasing the number of scenarios resulted in an enormous increase in the computation time required for plan creation. In contrast, the power operation planning device 1 according to the first embodiment carefully selects scenarios while evaluating their impact on the plan results, based on the approximate plan results and decision variables estimated by the estimation model, and creates a power operation plan. As a result, it becomes possible to achieve both improved optimality and computation speed, which was not possible with conventional technology.

In the above-described embodiment, electricity was used as an example of energy, and the present invention was applied to the power operation planning device 1. However, gas, wind power, hydropower, biomass, etc., may be used as examples of energy sources instead of electricity, and the present invention may be applied to the operation planning of these energies.

[Second Embodiment]
Next, an example of the configuration and processing of the power operation planning device 1A according to the second embodiment of the present invention will be described with reference to Figures 7 to 10. Note that explanations that overlap with those described in the first embodiment will be omitted.

<Functional Configuration of Power Operation Planning System 1A>
Figure 7 shows an example of the functional configuration of the power operation planning device 1A according to the second embodiment.

The calculation unit 3A of the power operation planning device 1A includes, in addition to the scenario-based plan estimation unit 31, the estimation plan determination variable evaluation and scenario selection unit 32A, and the uncertainty plan calculation unit 33 provided by the power operation planning device 1 according to the first embodiment, a unit 34 for evaluating the impact of unselected scenarios.

The unselected scenario impact evaluation unit 34 takes the planning results (power operation plan considering uncertainty) calculated by the uncertainty planning calculation unit 33 as input and evaluates the situation in which the scenario selection unit 322 did not select a scenario. The unselected scenario impact evaluation unit 34 outputs an evaluation value that assesses the planning results for scenarios not anticipated during planning by the uncertainty planning calculation unit 33. Scenarios not anticipated during planning are those not selected by the scenario selection unit 322.

The preliminary plan determination variable evaluation and scenario selection unit 32A re-selects scenarios based on evaluation values input from the unselected scenario impact evaluation unit 34. For example, after creating a power operation plan by fixing the determination variables for scenarios not selected by the scenario selection unit 322, if the unselected scenario impact evaluation unit 34 evaluates that the unselected scenarios will have an impact on operating costs, it adds the unselected scenarios to the scenarios selected by the scenario selection unit 322. The uncertainty plan calculation unit 33 then calculates the uncertainty power operation plan by adding the unselected scenarios.

Figure 8 is a block diagram showing an example of the functional configuration of the decision variable evaluation and scenario selection unit 32A for the preliminary plan.

The preliminary plan determination variable evaluation and scenario selection unit 32A includes a risk scenario selection unit 323, in addition to the similarity evaluation unit 321 and scenario selection unit 322 of the first embodiment. Furthermore, evaluation values are input to the risk scenario selection unit 323 from the unselected scenario impact evaluation unit 34.

The Risk Scenario Selection Unit 323 receives the evaluation values of the planning results for the scenarios selected by the Scenario Selection Unit 322 and the scenarios not anticipated during planning, as evaluated by the Unselected Scenario Impact Evaluation Unit 34. Based on the scenarios selected by the Scenario Selection Unit 322 and the estimated plans for those scenarios, the Risk Scenario Selection Unit 323 selects only those scenarios that affect the statistical risk evaluation indicators based on the operating cost distribution, and selects these scenarios as risk scenarios. In this case, the Risk Scenario Selection Unit 323 selects scenarios that fall within a specific distribution area of the operating cost distribution calculated based on the estimated plan for each scenario as risk scenarios. A specific distribution area is, for example, the area corresponding to the loss risk 50 shown in Figure 10, which will be described later. Alternatively, the Risk Scenario Selection Unit 323 may extract scenarios corresponding to statistical evaluation indicators from the operating cost distribution and select these scenarios as risk scenarios. These risk scenarios are then combined with the unselected scenario impact assessment unit 34, which evaluates scenarios not anticipated during plan creation and determines that they will impact operating costs based on the evaluation values of the plan results. These scenarios are then output to the uncertainty plan calculation unit 33.

As shown in Figure 7, the uncertainty planning calculation unit 33 creates a power operation plan that takes uncertainty into account, based on the scenarios selected by the risk scenario selection unit 323 that affect the statistical risk assessment indicators. The created power operation plan is stored in the power operation plan database DB2 shown in Figure 3 and output to the screen display and result storage unit 4.

The screen display and result saving unit 4 displays the uncertainty scenarios and power operation plans output from the uncertainty plan calculation unit 33 and the risk scenario selection unit 323, and saves the plan results to the power operation plan database DB2.

<Hardware configuration of power operation planning device 1A>
Since this is the same as the first embodiment, a detailed explanation will be omitted.

<Processing of the calculation unit of the power operation planning device 1A>
Figure 9 is a flowchart showing an example of processing by the calculation unit 3A of the power operation planning device 1A according to the second embodiment. The functions of power operation planning and impact evaluation of unselected scenarios performed by the calculation unit 3A will be explained using the flowchart in Figure 9.

The processes from steps S11 to S15 are the same as those in the first embodiment shown in Figure 4.

In step S15, the scenario selection unit 322 evaluates whether the accuracy of the planning results is maintained within the threshold range. If the accuracy of the planning results is not within the threshold range (NO in S15), the risk scenario selection unit 323 selects a risk scenario (S15B-1). Here, the risk scenario selection unit 323 evaluates the scenario selected by the scenario selection unit 322 and its approximate planning results (rough plan) using CVaR (Conditional Value at Risk), a statistical risk assessment index indicating loss risk. The risk scenario selection unit 323 selects only scenarios that correspond to loss risk as risk scenarios. Now, referring to Figure 10, the method for selecting scenarios according to risk will be explained.

Figure 10 is a conceptual diagram of loss risk. Figure 10 is similar to the example screen display of the operating cost distribution for uncertainty shown in Figure 5. Here, we will explain how to select scenarios according to risk based on this conceptual diagram of loss risk.

To simulate the uncertainty of power operation planning under multiple scenarios, the operating costs and revenues are expressed by equation (3) above, based on the approximate plans for each scenario shown in Example 1. Each scenario also has a frequency (probability of occurrence Prs). The horizontal axis of Figure 10 shows the operating costs or revenues when a particular scenario occurs, and the vertical axis shows the frequency with which these operating costs or revenues occur. Based on the frequency and operating costs for each scenario, these can be represented by histograms or probability density functions.

Fund managers may have requirements such as wanting to avoid situations where operating costs result in losses rather than generating profits. In this case, fund managers may want to evaluate only risk scenarios that correspond to loss risk 50 in Figure 10. The aforementioned CVaR is used as an indicator to evaluate such loss risk 50.

The risk scenario selection unit 323 evaluates the loss risk using CVaR and selects scenarios corresponding to the loss risk 50 in the operating cost distribution shown in Figure 10. By evaluating scenarios that correspond to loss risk, the risk scenario selection unit 323 can eliminate scenarios that do not correspond to loss risk, thereby reducing the number of scenarios required for evaluation and reducing the computation time required for the uncertainty plan calculation unit 33 to create the plan.

Furthermore, the uncertainty planning calculation unit 33 can create a power operation plan that considers uncertainty while minimizing loss risk by focusing only on scenarios that correspond to loss risk. The uncertainty planning calculation unit 33 can minimize costs by targeting scenarios s that correspond to loss risk in the objective function of equation (1).

In the processing of the risk scenario selection unit 322 described above, the operator pre-selects scenarios corresponding to the risks they wish to evaluate. As an example, the selection of scenarios corresponding to loss risks was explained. However, the risk scenario selection unit 323 may also select scenarios within the area specified by the operator in the operating cost distribution as risk scenarios. For example, the power operation planning device 1A creates a histogram based on the scenarios and the estimated approximate planning results. The operator specifies the distribution area they wish to evaluate for this histogram. Scenarios within this specified distribution area are selected as risk scenarios.

In step S16 of Figure 9, the uncertainty planning unit 33 creates a power operation plan that takes uncertainty into account, based on the risk scenario selected in step S15B-1.

In step S16B-1 of Figure 9, the unselected scenario impact evaluation unit 34 evaluates the impact of operating the equipment according to the power operation plan created in step S16. In particular, since scenarios other than the selected scenario may occur during operation, the unit evaluates the impact of such scenarios occurring. That is, the unselected scenario impact evaluation unit 34 evaluates the impact of scenarios other than those selected by the risk scenario selection unit 323 in step S15B-1, or the scenario selection unit 322 in step S14, occurring during operation.

As described above, the scenario selection unit 322 and the risk scenario selection unit 323 select scenarios based on approximate planning results estimated by the scenario-based planning estimation unit 31. Therefore, if the estimation accuracy of the approximate plan is low, the risk scenario selection unit 323 may not extract scenarios that affect the planning results. To avoid insufficient extraction of important scenarios that affect the planning results, the unselected scenario impact evaluation unit 34 evaluates the impact of applying scenarios other than those extracted by the scenario selection unit 322 and the risk scenario selection unit 323 to the planning results. Furthermore, the unselected scenario impact evaluation unit 34 evaluates whether the operating costs shown in equation (3) become excessive due to constraint violations or other reasons.

In particular, it is possible that the scenarios not selected above may only become apparent immediately before operation begins after the plan has been created. Therefore, when evaluating the unselected scenarios, the uncertainty plan calculation unit 33 fixes the variables that cannot be changed during operation after the plan is created (the decision variables defined in the first embodiment) to their planned values. Then, using the remaining variables, the uncertainty plan calculation unit 33 recreates the power operation plan targeting the scenarios not selected by the scenario selection unit 322. Based on the results of this replanning, the impact is evaluated assuming that the unselected scenarios occur during operation.

Compared to the power operation plan created according to the first embodiment, the power operation plan recreated with fixed decision variables to anticipate post-planning operation has fewer adjustment elements because the decision variables cannot be changed. Therefore, depending on the scenario that occurs, the operational constraints may be violated, potentially leading to an excessively high operational cost in equation (3). The unselected scenario impact evaluation unit 34 evaluates whether scenarios not selected by the scenario selection unit 322 will result in an excessively high operational cost in equation (3) due to violations of the operational constraints, and outputs an evaluation value.

In step S16B-2 of Figure 9, the unselected scenario impact evaluation unit 34 determines whether the impact of the unselected scenario is within the threshold. Here, it evaluates whether the evaluation value in step S16B-1, which assumes a scenario other than the selected scenario occurs during operation, becomes excessive due to a constraint violation.

If the impact of the unselected scenarios exceeds the threshold (NO in S16B-2), the unselected scenario impact evaluation unit 34 extracts scenarios where the operating cost exceeds a pre-set threshold. In this case, the scenarios that exceed the threshold are those other than the risk scenarios selected in S15B-1. The process then returns to step S15B-1 and continues.

In other words, the risk scenario selection unit 323 adds the scenarios determined to cause constraint violations in step S16B-2 based on evaluations conducted by the unselected scenario impact evaluation unit 34, which assumes operational conditions, to the scenario selected in step S15B-1, and outputs this to the uncertainty plan calculation unit 33. The uncertainty plan calculation unit 33 then creates a power operation plan considering uncertainty again in step S16, and the unselected scenario impact evaluation unit 34 evaluates the impact of the unselected scenarios. This process is repeated until there are no more constraint violations in step S16B-2.

If the impact of the unselected scenarios is within the threshold (YES in S16B-2), the uncertainty plan calculation unit 33 outputs the power operation plan considering the uncertainties created in step S16 to the screen display and result storage unit 4. Similarly, the scenario selection results in step S14, the risk scenario selection results in step S15B-1, and the scenarios added during the impact evaluation of the unselected scenarios in step S16B-1 are also output.

In step S17 of Figure 9, the screen display and result saving unit 4 saves the power operation plan, which takes uncertainty into account, to the power operation plan database DB2 and displays it on the screen. At this time, the power operation plan database DB2 stores the scenario selection results from step S14, the risk scenario selection results from step S15B-1, and the scenarios added during the impact assessment of the unselected scenarios in step S16B-1, and the selection results are displayed on the screen.

Furthermore, the power generation and grid information unit 2 may be configured as a storage device or cloud server located outside the power operation planning device 1. In this case, the calculation unit 3 of the power operation planning device 1 will acquire necessary information from the storage device or cloud server as appropriate when executing processing. The uncertainty plan calculated by the calculation unit 3 may be stored on the storage device or cloud server.

<Effects of the second embodiment>
In the power operation planning device 1A according to the second embodiment described above, in addition to scenario selection according to the first embodiment, an operational cost distribution (revenue) is created based on the estimated planning results according to the scenario, showing the frequency of occurrence of the scenario and the operational cost assumed for that scenario. From this operational cost distribution, the power operation planning device 1A selects scenarios that fall within a specific area of the distribution specified by the operator, such as the risk of revenue and loss, and creates a power operation plan that takes uncertainty into account. As a result, the power operation planning device 1A has the effect of reducing the number of scenarios outside the specific area, prioritizing the improvement of risks that the operator considers important to evaluate when creating the power operation plan, and speeding up the calculation time for plan creation.

Here, scenario selection is performed based on the estimation of planning results according to the scenario, as shown in the first embodiment. If the estimation accuracy of the estimated planning results is low, the overlap of planning results may not be properly evaluated, and scenarios that affect the planning results may not be sufficiently extracted during scenario selection. Therefore, to avoid insufficient extraction of important scenarios, the power operation planning device 1A evaluates the occurrence of scenarios that were not selected during the actual operation of the power operation plan, which takes uncertainty into account and is created based on the selected scenarios.

During operation after this plan is created, there are decision variables that cannot be adjusted immediately before operation. Therefore, in evaluations simulating operation, these decision variables are set to their planned values, limiting the adjustment elements. By adding scenarios where costs become excessive due to constraint violations, etc., in the evaluation simulating operation, to newly selected scenarios, the aforementioned lack of scenario extraction can be corrected, and the accuracy of uncertainty simulation can be improved with high precision.

It should be noted that the present invention is not limited to the embodiments described above, and various other applications and modifications can be taken as long as they do not deviate from the gist of the present invention as described in the claims.
For example, the embodiments described above are detailed and specific explanations of the configuration of the apparatus and system in order to clearly illustrate the present invention, and are not necessarily limited to having all the configurations described. Furthermore, it is possible to replace some of the configurations of the embodiments described here with the configurations of other embodiments, and it is also possible to add the configurations of other embodiments to the configuration of one embodiment. In addition, it is possible to add, delete, or replace some of the configurations of each embodiment with other configurations.
Furthermore, the control lines and information lines shown are those deemed necessary for explanatory purposes, and not all control lines and information lines are necessarily shown in the actual product. In reality, it is safe to assume that almost all components are interconnected.

1, 1A…Power Operation Planning Device, 2…Power Generation and Grid Information Unit, 3…Calculation Unit, 4…Scenario Display and Result Storage Unit, 31…Scenario-Based Plan Estimation Unit, 32…Decision Variable Evaluation and Scenario Selection Unit for Estimation Plan, 33…Uncertainty Plan Calculation Unit, 34…Excluded Scenario Impact Evaluation Unit, 321…Similarity Evaluation Unit for Plan Result Features, 322…Scenario Selection Unit, 323…Risk Scenario Selection Unit, DB1…Grid Information Database, DB2…Power Operation Planning Database

Claims (14)

A preliminary plan creation unit inputs information necessary for creating an energy operation plan as multiple scenarios into an estimation model and creates a preliminary plan that approximates the energy operation plan,
An evaluation unit that evaluates the similarity between the scenarios based on the results of the aforementioned budget plan,
A scenario selection unit selects scenarios from among the scenarios input to the estimation model such that the scenarios are not similar, based on the evaluation of similarity by the evaluation unit.
An energy operation planning device comprising: a planning unit that creates the energy operation plan using the selected scenario; The evaluation unit evaluates the similarity of the effects on the energy operation plan for each scenario based on the results of the preliminary plan,
The energy operation planning device according to claim 1, wherein the scenario selection unit deletes similar scenarios based on the similarity evaluation of the evaluation unit and selects the remaining scenarios. The energy operation planning device according to claim 2, wherein the estimated plan creation unit creates an operation cost distribution consisting of the probability of occurrence of the scenario and the operation cost of the estimated plan based on the estimated plan. The energy operation planning device according to claim 3, further comprising a risk scenario selection unit that selects scenarios that affect the statistical risk assessment index based on the operating cost distribution as risk scenarios, based on the scenario selected by the scenario selection unit and the estimated plan. The estimation model is a model that simulates the relationship between the information necessary for creating the energy operation plan and the estimated plan which is estimated assuming uncertainty, and is composed of a method different from the method used in creating the energy operation plan, as described in any one of claims 1 to 4. The energy operation planning device according to claim 2, wherein the evaluation unit evaluates the similarity of overlapping scenarios using decision variables, which are elements whose values determined in the energy operation plan cannot be changed, when the energy operation plan created by the planning unit is put into operation. The system includes an unselected scenario impact evaluation unit that evaluates situations in which the scenario selection unit did not select a scenario,
The energy operation planning apparatus according to claim 6, wherein the planning unit creates the energy operation plan by fixing the decision variables for the scenarios not selected by the scenario selection unit, and then, if the unselected scenario impact evaluation unit evaluates that the scenarios not selected by the scenario selection unit have an impact on the operation costs, the unselected scenarios are added to the scenarios selected by the scenario selection unit. The scenario selection unit, with respect to the scenarios input by the estimated plan creation unit, removes the scenarios with the lowest probability of occurrence and the highest similarity in the similarity evaluation with other scenarios, based on a probability similarity evaluation index, prioritizing the removal of scenarios with the lowest probability similarity evaluation index, and extracting the remaining scenarios, as described in claim 2. The energy operation planning device according to claim 8, wherein, when the scenario selection unit deletes the scenario and extracts the remaining scenarios, it selects a scenario such that, before and after deleting the scenario, the expected value of the operating cost of the energy operation plan based on the scenario and its estimated plan, or the change in the probability similarity evaluation index shown by the evaluation unit in the similarity evaluation with other scenarios, is within a threshold. The energy operation planning device according to claim 3, further comprising a display unit for displaying the aforementioned operating cost distribution. The energy operation planning device according to claim 5, wherein the estimation model is a relaxed optimization problem obtained by relaxing the optimization problem for creating the energy operation plan. The energy operation planning device according to claim 5, wherein the estimation model is a machine learning model that has learned the relationship between the information necessary for creating the energy operation plan and the estimated plan. The system includes a display unit that shows at least one of the deleted scenario and the selected scenario.
The energy utilization planning device according to claim 2. An energy operation planning method executed by an energy operation planning device comprising a preliminary plan creation unit, an evaluation unit, a scenario selection unit, and a plan creation unit,
The aforementioned preliminary plan creation unit inputs information necessary for creating the energy operation plan as multiple scenarios into an estimation model and creates a preliminary plan that approximates the energy operation plan,
The evaluation unit performs the step of evaluating the similarity between the scenarios based on the results of the estimated plan,
The scenario selection unit performs the step of selecting scenarios from among the scenarios input to the estimation model such that the scenarios are not similar, based on the evaluation of similarity.
An energy operation planning method comprising the step of a planning unit creating the energy operation plan using the selected scenario.