JP7601859B2 - Method for stabilizing the electrical energy supply grid - Google Patents

Description

The entire contents of the priority application German Patent Application No. 102019212773.8 are incorporated herein by reference.

The present invention relates to a method for stabilizing an electrical energy supply network having at least one control area, the method being conceived with respect to at least one control area.

An overview of the structure and function of the electric energy supply network is given in non-patent document 1.

From a system-technical point of view, the electrical energy system includes the totality of all technical facilities necessary for the realization of the generation, transmission and distribution of electrical energy within the system boundary for specific control-technical reasons. With regard to frequency and active power control, the area delimited by the system boundary is called the control area.

The power output per time unit fed into the grid, i.e. the conversion of energy, e.g. rotational or thermal energy, into electrical energy (corresponding to energy generation), is defined with a negative sign. The power output per time unit taken out of the grid, i.e. the conversion of electrical energy from the grid into other forms of energy, e.g. thermal or kinetic energy (e.g. rotational energy), (corresponding to energy consumption), is defined conversely with a positive sign.

In a deregulated electricity market, the power trading business, or more precisely the energy trading business, realized for the time being without knowledge of grid coordination, requires the establishment of so-called balancing groups and balancing areas. In each balancing group and each balancing area, the predicted discharge power and the contractually secured supply power are adjusted or balanced according to a plan so that they are equal. The integration of all balancing groups into the system balance allows the grid coordination to maintain and account for the balance between the generated and consumed electrical energy according to a plan.

A balancing group therefore contrasts supply power with discharge power, which must remain largely balanced at all times, either for physical reasons or from a frequency regulation perspective. The grid operator compensates for any remaining differences with supply reserves or balancing power, as we will refer to more precisely below.

Both the supply and discharge power transitions are planned based on specific "schedules". When implementing all the schedules of a balancing group, and therefore the entire control area, control area variances always arise as a result of unexpected deviations from the predicted load progression or unplanned losses of generating capacity.

The control area variance (also called the control area balance or system balance) is therefore the sum of all supply outputs and discharge outputs in the control area. For each balancing group, the sum of all supply outputs and discharge outputs in the balancing group is referred to as the variance balancing group aggregation or balancing group variance.

The grid operator must eliminate the control area difference by using positive or negative reserves, so-called balancing reserves, within the scope of its frequency maintenance duties. The grid operator generally obtains the right to purchase the balancing reserves in advance after a prior open tender for power plants operating at partial load, so-called spinning reserves, located inside and outside its control area. In the meantime, a reserve is provided by these plants. When used, a corresponding amount of balancing power, positive and negative, is further provided. For the provision of balancing reserves, capacity costs are charged at a capacity price, and for the supply of balancing power, power costs are charged at a metered price. The capacity costs are added to the grid usage costs, and the power costs are charged to the balancing group responsible for the supply of balancing power on a causal basis.

Furthermore, in the case of large renewable energy installations, so-called supply management output adjustments can be used if meteorological conditions would result in a higher output from the renewable energy installation, in particular from a wind power installation, than can be distributed or transmitted within the grid. Supply management consists in reducing the output of the renewable energy installation in the relevant grid section by the amount necessary to locally prevent a grid bottleneck.

Supply reserves are mainly generated by flexible production, i.e. by the controllable supply output of controllable renewable energies such as biogas plants, hydroelectricity or hydrogen and fossil or nuclear energies, and by variable production, i.e. by the variable supply output of uncontrollable renewable energies that are dependent on the weather, such as wind or solar energy.

Conventional systems for calculating the required reserve capacity are designed to be purely reactive or to make only very short-term, forward-looking adjustments. Allowing schedule deviations that are beneficial for the system at the balancing group level is in this context referred to as passive balancing. The starting point here is information about the control area balance that exists in real time. There is no time in advance for trading balancing power in advance.

So far, the time for intraday trading has simply been shortened to reduce imbalances and the need for balancing power due to fewer forecast errors on production and consumption. In this case, forecasting algorithms similar to those disclosed in the present invention are used to minimize the balancing power, especially for renewable energy consumption and production.

The patent document WO 2005/023333 shows a method for operating an entire energy grid in the case of distributed energy production. The grid is made up of three supply levels where energy is produced and distributed. The method provides a mechanism for stabilizing the grid through the use of balancing power and describes the cooperation of distributed producers.

In the prior art, the imbalance between production and demand is resolved primarily by supply reserves from flexible power plants. In this case, priority is given to energy from conventional power plants based on fossil fuels, which leads to increased _CO2 production. The procurement problems linked to the supply of balancing power as power plant reserves and linked to the procurement of supply reserves are assigned to each balancing group manager in the form of adjustment power in response to the balancing group difference. In this procedure, the demand for balancing power cannot be planned, and the size of the supply of balancing power is not dynamically determined depending on the situation. The metered price is only generated when the balancing power is requested and is determined after the conclusion of the balancing power transaction.

International Publication No. 2014/166524

Schwab, A.J., “Elektroenergiesysteme-Erzeugung, Uebertragung und Verteilung elektrischer Energie“, 5. Auflage, Springer Verlag, Berlin, 2017

The objective of the present invention is to improve the planning capability of balancing power required to eliminate control area differences in an electric energy supply network and the stability of the electric energy supply network.

The problem is solved by a method according to claim 1, a computer program according to claim 8 and a computer program product according to claim 9. Advantageous configurations of the invention are contained in the dependent claims.

The invention is based on the idea of improving the plannability of pre-supply reserves, i.e. the predicted control area difference, through the prediction of beneficial energy storage and flexibility control for the system, using both flexible and variable production. This results in increased plannability, increased stability of the electrical energy supply network and an increased share of variable renewable energies.

を、人工知能の領域の手法、好ましくはマルコフ連鎖を用いた関数を通じて決定するステップ、
－ 時間的間隔［ｔ_Ｎ，ｔ_Ｎ＋Δｔ］において必要なバランシング電力に関する予測ΔＥ_{ＲＬｖｏｒａｕｓ}を、時間的間隔［ｔ_Ｎ，ｔ_Ｎ＋Δｔ］における制御域差異に関する予測

から決定するステップ、
－ 予測されたバランシング電力ΔＥ_{ＲＬｖｏｒａｕｓ}を調達するステップ、及び、
－ 電気エネルギー供給網に予測されたバランシング電力ΔＥ_{ＲＬｖｏｒａｕｓ}を時間的間隔［ｔ_Ｎ，ｔ_Ｎ＋Δｔ］において供給するステップ。 The method according to the invention for stabilizing an electrical energy supply network having at least one control area comprises, for the at least one control area, the following steps:
inputting a time t _N in the future and the length of the interval Δt,
determining the time evolution ΔP _RZ (t) of the control range difference in the time interval [t _N , t _N +Δt] through methods in the field of artificial intelligence, preferably functions using Markov chains;
determining the balancing power ΔE _RL required in the time interval [t _N , t _N +Δt] from the time progression ΔP _RZ (t) of the control range difference in the time interval [t _N , t _N +Δt];
- A prediction regarding the control range difference in the time interval [ _tN , _tN +Δt].

through a technique in the field of artificial intelligence, preferably a function using a Markov chain;
the prediction ΔE _RLvoraus for the balancing power required in the time interval [t _N , t _N +Δt], the prediction ΔE RLvoraus for the control range variance in the time interval [t _N , t _N +Δt]

determining from
Procuring a predicted balancing power ΔE _RLvoraus ; and
- Supplying a predicted balancing power ΔE _RLvoraus to the electric energy supply grid in the time interval [t _N , t _N +Δt].

The method according to the invention allows for good quantitative planning of the balancing power requirements, which can then be quantitatively traded as secondary balancing power, for example also through the intraday power market. The share of variable renewable energies in the balancing power can also be increased.

The prediction for the control zone difference is preferably determined by the calculation of a Markov chain, in particular by the calculation of a higher order Markov chain. For this purpose, the algorithm CondEx is preferably used, which is derived from a basic physical model with assumptions based on direct probability. The algorithm CondEx can be trained by itself and is able to recognize and predict elaborated patterns.

と間隔の長さΔｔとの積として決定される。 In a preferred configuration of the invention, the balancing power ΔE _RL required in the time interval [t _N , t _N +Δt] is determined as the integral of the time evolution ΔP _RZ (t) of the control range difference over the time interval [t _N , t _N +Δt]. In addition, a prediction ΔE _RLvoraus for the balancing power required in the time interval [t _N , t _N +Δt] is determined as the integral of the time evolution ΔP RZ (t) of the control range difference over the time interval [t _N , t _N +Δt].

and the length of the interval Δt.

In this way, the control area variance and its prediction are determined in a mathematically straightforward manner.

In various configurations of the method according to the invention, which are modifications of the last-mentioned embodiment and are described below, conditions for system profitability are determined. In this case, system profitability is understood to be the degree of consistency of the advance reserve (in the context of the forecast of the state of the supply network), i.e. the agreement between the forecast reserve and the actually required reserve.

As a predicted, uncertainly determined magnitude, the pre-balancing power is in some of these embodiments determined to a value different from zero only under certain conditions, depending on the criteria selected with respect to beneficialness to the system. The conditions may be that the predicted control area variance has the same sign as the balance between the predicted positive and negative supply reserves and/or has a different sign from the predicted grid frequency variance, and/or that there is no supply management or the predicted pre-balancing reserve is negative.

と間隔の長さΔｔとの積として決定されるが、その他の場合はゼロである。 In a preferred configuration of the invention, the balancing power ΔE _RL required in the time interval [t _N , t _N +Δt] is determined as the integral of the time progression ΔP RZ (t) of the control range difference over a sub-interval of the time interval [t _N , t _N +Δt] consisting of the time t at which the first condition for the usefulness of the system is fulfilled or at which the first conditions for the usefulness of the system are cumulatively fulfilled. Furthermore, if the second condition for the usefulness of the system is fulfilled or at which the second conditions for the usefulness of the system are cumulatively fulfilled, the prediction ΔE RLvoraus of the balancing power required in the time interval [t _N , t N +Δt] is determined as the integral of the time progression ΔP _{RZ (} t) of the control range difference over the sub-interval of the time interval [t _N , t _N +Δt] consisting of the time _t at which the first condition for the usefulness of the system is fulfilled or at which the first conditions for the usefulness of the system _are cumulatively fulfilled.

and the interval length Δt, and is zero otherwise.

の符号が、時間的間隔［ｔ_Ｎ，ｔ_Ｎ＋Δｔ］におけるプラスの供給予備力とマイナスの供給予備力との予測された収支Ｐ_{ＲＬｐｒｏｇｎ} ^ｔ _Ｎの符号と同じであることにある。 In a further preferred configuration of the invention, the first condition or one of the first conditions for the profitability of the system consists in the sign of the control range difference ΔP _RZ (t) at a particular time instant t being the same as the sign of the balance P _RL (t) between positive and negative reserve power at time instant t. The second condition or one of the second conditions for the profitability of the system consists in the predicted control range difference in the time interval [t _N , t _N +Δt].

has the same sign as the forecasted balance P _RLprogn ^t _N of positive and negative reserve power in the time interval [t _N , t _N +Δt].

の符号が、時間的間隔［ｔ_Ｎ，ｔ_Ｎ＋Δｔ］における予測された供給網周波数差異Δω_{ｐｒｏｇｎ} ^ｔ _Ｎの符号とは異なることにある。 In a further preferred configuration of the invention, the first condition or one of the first conditions for the profitability of the system consists in the sign of the control range difference ΔP _RZ (t) at a particular time instant t being different from the sign of the predicted grid frequency difference Δω _progn (t) at time instant t. The second condition or one of the second conditions for the profitability of the system consists in the sign of the predicted control range difference in the time interval [t _N , t _N +Δt] being different from the sign of the predicted grid frequency difference Δω progn (t) at time instant t.

The difference is that the sign of the predicted grid frequency difference Δω progn t N in the time interval [t _N , t _N +Δt] is different from the sign of the predicted grid frequency difference Δω _progn ^t _N in the time interval [t N , t N +Δt].

がマイナスであることにある。 In a further preferred configuration of the invention, the first condition or one of the first conditions for profitability for the system consists in that the supply management signal Einman(t) at a particular time t has a "not supply management" value or the advance supply reserve ΔP _RZ (t) at time t is negative. The second condition or one of the second conditions for profitability for the system consists in that the predicted supply management signal Einman _progn (t _N ) for the relevant region at time t _N has a "not supply management" value or the predicted control zone difference in the time interval [t _N , t _N +Δt] is negative.

is negative.

The first condition regarding the usefulness for the system mentioned above may be applied individually or also cumulatively when calculating the required balancing power ΔE _RL or when predicting the required balancing power ΔE _RLvoraus .

The various conditions of system usefulness according to various embodiments of the present invention increase the potentially possible share of secondary balancing power that is plannable within the framework of the smart balancing mechanism and therefore quantitatively tradable through the intraday electricity market, to a relative share of up to a potential 77% in the control area balance to be adjusted, depending on the criteria used for system usefulness. This reduces the share of "must run capacity" based on traditional fossil energy sources based on increased plannability.

The plannability of the management instruments for procuring reserves also increases the possibility of integrating the weather-influenced forecasts of variable renewable energies into a unified process of forecasting, trading and procurement of balancing power from renewable energies. In this case, it will be important to reduce _CO2 with respect to balancing power from conventional plants. The share of balancing power from renewable energies in the control area balance can thereby potentially rise to 77%, depending on the criteria regarding the usefulness for the system and the availability of renewable energies.

In this way, by using storage and flexibility to the system's advantage, a jointly planned pre-balancing of power from flexible and variable production through forecasting of grid-defining variables: grid bottlenecks, grid frequency, reserve capacity and control area variances, is traded within the framework of a smart balancing process in the sense of a pre-determined power and associated actual metered price, allowing an optimized combined use of flexibility and renewable energies.

In a further preferred configuration of the invention, the time interval between the current time and the time _tN at which the time interval begins, during which the required balancing power ΔE _RLvoraus is predicted, is at least one hour.

Conversely, therefore, the implementation of the mechanism for determining balancing power, aimed at the plannability of the imbalance between energy production and energy consumption in the electric energy supply network, is brought forward by at least one hour before the time of requesting balancing power. Not only the price for the procurement of capacity (capacity price), but also the power to be procured is quantitatively determined and traded, including the unit price, so that procurement can be carried out quantitatively after determining in advance the unit price, which can be minimized due to plannability.

The present invention develops balancing power procured by renewable energies while simultaneously minimizing balancing power on a wide scale, for example in the double-digit percentage range. The application-specific use of the preferably used forecasting algorithm CondEx places the decisive application scenario of artificial intelligence methods in the context of renewable energies, integrated plannability, tradability and stabilization of the energy system.

The potential for reducing _CO2 emissions that this offers must be considered as broad.

The method according to the present invention can further be automated through the technique of intelligent communication using typology-based messages between actors (demonstrators) connected to each other via communication lines in the energy supply system.

In summary, the invention and its preferred embodiments are based on a combination of forecasts regarding the various variables that determine the balance between energy production and energy consumption and the benefits to the system of the flexibility to be used. A large part of the balancing power can be traded in terms of pre-determining reserves at least one hour ahead ("hour ahead") within the framework of a smart balancing mechanism and planned for the market size in terms of price as well as energy quantity. In this case, a larger share of renewable energy can be used as such, with adjustments through the intraday market being pre-empted.

In the following, examples of the present invention will be described in detail with reference to the accompanying drawings. The examples should be understood as illustrative of the present invention and not as limiting. Shown are the following figures:

FIG. 13 shows a one-hour forecast of the control area variance for the control area of the supply network operator TransnetBW. FIG. 1b shows the predicted control area variance for the control area according to FIG. 1a after connecting the prediction of the usefulness of the control area variance for the system; FIG. 2 shows a one-hour forecast (in signs) of the balance of reserve supply capacity in the control area of the supply network operator TransnetBW. FIG. 2b shows a one-day-ahead forecast of grid frequency variance for the control area according to FIG. 2a. FIG. 13 shows the hour-ahead forecast of pre-provision reserves in the control area of the supply network operator TransnetBW after connecting the forecast of the beneficial effect of pre-provision reserves on the system through the same sign as the predicted control area variance. 3b shows the pre-fetch reserve for the control area according to FIG. 3a after connecting the prediction of the benefit to the system of the pre-fetch reserve through the opposite sign to the predicted grid frequency difference ("day ahead"). FIG. 13 is an interaction diagram of a demonstrator in a field test of “hour ahead” smart balancing supported by prediction using CondEx. FIG. 1 illustrates a circuit for predicting and adjusting demand for pre-supply reserves. FIG. 1 illustrates a circuit for predicting and adjusting demand for pre-supply reserves beneficial to the system. FIG. 1 illustrates an agent-based domain model of the business logic layer implemented in CondEx for use case modeling with roles and intelligent communication for type-based messages following a publish-subscribe model. FIG. 13 is a demonstrator interaction diagram for a grid operator's field test of hour-ahead planning supported by forecasting and trading of pre-balancing power using CondEx.

Examples of the present invention are described in detail below.

のイントラデイ予測は、Ｎ次のマルコフ連鎖の条件付き期待値として、ここでは、

として表され、時間的間隔

によって積分される。 A computer implemented control range difference over N time steps of length Δt.

The intraday forecast of is given by the conditional expectation of the Nth order Markov chain, where

and the time interval

It is integrated by

を用いて、

現時点として、

を、制御域差異が知られる又は公開される時間的ずれとして、事前バランシング電力に関して以下の式が有効である。 At this time,

Using

As of now,

Let be the time lag at which the control range difference is known or published, then the following equation is valid for pre-balancing power:

The method according to the present invention uses known mathematical techniques, also called artificial intelligence techniques, to predict events relating to future scenarios.

The field of application is smart balancing based on the prediction of control area variances. By being able to act in a beneficial way for the system in the form of anticipatory follow-up adjustments, the stabilization of the system can be regulated, for example in sub-networks of energy suppliers.

が評価され得る。 Informed forecasts regarding control area variances, instead of information that may possibly be provided by the supply network operator

can be evaluated.

Differences within balancing groups will be left alone if they are beneficial to the system for national economic minimization of regulated and balanced power, and will otherwise be resolved by linking with the intraday market, as is customary in the past. In general, flexibility in this regard may be used to the benefit of the system.

With a smart balancing process configured in this way, a time window of at least one hour is available in the method according to the present invention for achieving system-beneficial stabilization of the electrical energy supply grid.

Therefore, for schedule deviations that are beneficial to the system, frequency differential countermeasures and reserve capacity requirements are planned "hour ahead."

ＭＷにおける離散化ステップサイズが、特に優れた結果をもたらす。 In the application of one-hour-ahead prediction of pre-balancing power, the steps of the above equations can be implemented, for example, in an embodiment with N=4 and Δt=15 minutes. In this case, experiments have shown that the accuracy of the one-hour-ahead prediction of the control range difference is remarkably high, with a correlation coefficient of 0.42 to the measured control range difference. In this case, the power range

The discretization step size in MW gives particularly good results.

では、制御域に障害が発生し、時点

では、事前供給予備力の要求によって制御域が調整される。この際、

は、制御域差異が知られる又は公開される時間的なずれである。当該調整は、市場を通じた、スマートバランシングプロセスの枠内における、１時間の取引時間内での事前供給予備力の取引に対応する。 The steps in the above equation are performed according to the regulation circuit of FIG.

In this example, a failure occurs in the control area, and

In the , the control area is adjusted according to the advance supply reserve requirement.

is the time lag when the control area difference is known or disclosed. The adjustment corresponds to the trading of advance reserves within a trading hour through the market, within the framework of the smart balancing process.

が生じる。 The resulting control variables are

This occurs.

In Fig. 1a, a qualitatively good agreement between measurement and prediction is observed for the rising and falling edges, in the center, shifted in time by one time step and therefore synchronized for the prediction period of four time steps.

による削減可能性は、プラスのバランシング電力については６３％であり、マイナスのバランシング電力については５５％である。 For the period from mid-August to mid-September 2018, the balance of control area variances

The reduction potential due to is 63% for positive balancing power and 55% for negative balancing power.

For the realization of the embodiment, a known algorithm, here CondEx, is correspondingly extended to conditional expectation, where N is the number of time steps to be predicted, based on an Nth order Markov chain. The real-time and robustness of the method is adaptable to computational power and databases/information.

In the case of application of hour-ahead planning and trading of pre-balancing power by the grid operator, in order to avoid bottlenecks in reserves with respect to reserve capacity, the grid operator uses the hour-ahead information obtained on the pre-balancing power to review and, if necessary, adjust its balancing power allocation through the intraday market. The procurement of balancing power traded in the intraday market can result from a smart balancing step, as described above, but may also be done through already established mechanisms for procuring reserve capacity (see Figure 8).

A further area of application is the qualification of the usefulness of advance reserves to the system through forecasting the sign of the balance between positive and negative reserves.

の符号が、予測されたプラスの供給予備力とマイナスの供給予備力との収支

の符号と同じである時間のみが、積分の際に考慮されることによって、システムにとっての有益性に関する上述の適用事例のフィルタリングが行われる。 In this case, the predicted control area difference

The sign of indicates the balance between the predicted positive and negative reserve capacity.

Only times where the sign is the same as , are considered in the integration, thereby filtering the above mentioned applications for usefulness to the system.

は、長さΔｔのＮの時間ステップにわたる供給予備力のイントラデイ予測

を用いて、Ｎ次のマルコフ連鎖の条件付きの期待値

として決定される。 In this case, the balance between positive and negative reserve capacity

is the intraday forecast of reserve capacity over N time steps of length Δt

Using the conditional expectation of the Nth order Markov chain,

is determined as:

の符号の１時間前予測が、図２ａ（測定と予測との相関は０．２８）において、例えば供給網事業者ＴｒａｎｓｎｅｔＢＷがリアルタイムに公開しているように、行われる。この場合、出力範囲

ＭＷにおける離散化ステップサイズが、特に優れた結果をもたらす。 In the case of applying the one-hour-ahead forecast of the sign of the balance between positive and negative reserve capacity with N=4, the balance between positive and negative reserve capacity is

A one-hour-ahead forecast of the sign of is made in FIG. 2a (correlation between measurement and forecast is 0.28), as published in real time, for example, by the grid operator TransnetBW. In this case, the output range

The discretization step size in MW gives particularly good results.

の符号が、予測された供給網周波数差異

の符号と異なる時間のみが、積分の際に考慮されることによって、システムにとっての有益性に関する上述の適用事例のフィルタリングが行われる。 A further field of application is the determination of the usefulness of pre-supply reserves to a system through the prediction of the sign of grid frequency discrepancies.

The sign of is the predicted grid frequency difference.

Only times that differ in sign from are considered in the integration, thereby filtering the above mentioned application cases for usefulness to the system.

供給網周波数差異

が、長さΔｔのＰの時間ステップ又は２４時間にわたる供給網周波数差異の予測

を用いて、Ｐ次のマルコフ連鎖の条件付き期待値として決定される。

Power grid frequency difference

is a prediction of the grid frequency difference over P time steps or 24 hours of length Δt

It is determined as the conditional expectation of the P-th order Markov chain using

Therefore, qualitative criteria for advance determination of reserve capacity that allows statements to be made about whether frequency will meet forecasts are determined "day ahead" of commencement.

が行われる。この場合、供給網周波数領域

Ｈｚにおける離散化ステップサイズが、特に優れた結果をもたらす。 In the application example of 1-day-ahead grid frequency forecasting, according to Fig. 2b (correlation is 0.7), the 1-day-ahead grid frequency forecasting with P = 1

In this case, the grid frequency domain

A discretization step size in Hz gives particularly good results.

A further area of application is the definition of the benefits to the system of advance supply reserves through supply management forecasting.

の算定の際に考慮されることによって、システムにとっての有益性に関する上述の適用事例のフィルタリングが行われる。 In this case, the pre-balancing power is calculated only when the supply management signal in the relevant region is predicted to be "not supply management" or when the pre-balancing power reserve is negative.

The above mentioned application cases are filtered in terms of their usefulness to the system by being taken into account in the calculation of

The supply management signal is determined as the conditional expectation of an M-th order Markov chain using intraday forecasts over M time steps of length 1 hour. This is particularly interesting for highly heterogeneous systems, e.g. wind oversupply in areas with many wind farms connected to other systems.

であり、計算上可能な状態

が数学的に、１に切り下げ又は切り上げされる。 In the application example of 5-hour-ahead forecasting of supply management, an M=5 (5 hours ahead) intraday bottleneck forecast is made as possible (frequent supply management in municipalities) based solely on the real-time published supply management signals of the "supply network traffic lights" of the supply network operator SchleswigHolsteinNetz AG, with an average correlation coefficient between the actual and predicted signals of 0.35, and is used as a preliminary step to determine the advance supply reserve in order to exclude the simultaneous occurrence of bottlenecks. In this case, the discretization in the state space between both states, "1" for "supply management" and "0" for "not supply management", is

is a computationally feasible state.

is mathematically rounded up or down to 1.

A further application area is the hour-ahead forecasting beneficial to the system of pre-balancing power, where the above mentioned application cases are filtered for their beneficialness to the system through a combination according to the regulation circuit of FIG. 6.

In addition to the adjustment circuit of FIG. 5, the signs of the balance between positive and negative reserve power, the sign of the grid frequency difference, and the cascade type of supply management queries are made through forecasting.

において、制御域差異

によって妨げられ、時点

において、事前供給予備力

の要求によって調整される。この際、

は、制御域差異が知られる、又は、公開される時間的なずれである。当該調整は、市場を通じた、スマートバランシングプロセスの枠内での１時間の取引時間内での事前供給予備力の取引に相当する。Ｎ＝４、Δｔ＝１５分、Ｐ＝１及びＭ＝５で、制御域差異の１時間前予測と、プラスのバランシング電力に関して４％、マイナスのバランシング電力に関して２％の制御域収支に対する、

による削減可能性と、が得られる。この削減可能性は選択的に、スマートバランシングプロセスの枠内での市場における予測されたシステムにとって有益な事前バランシング電力全体の取引を通じて、上述したように、図４に対応して、バランシンググループ責任者によって利用し尽くされる、又は、同じく上述したように、送電網事業者による１時間前プランニング及び事前バランシング電力の取引を通じて利用し尽くされる。この際、バランシンググループ責任者は、インバランスを解消せずに残すことによって、そのバランシンググループ内の場合によって存在し得るインバランスから成る事前バランシング電力を用いることができる。 The control area is

In the control area difference

At the time

In the case of

The amount of the tax will be adjusted according to the request of the

is the time lag when the control zone variance is known or published. The adjustment corresponds to trading of advance reserves through the market within a one-hour trading window within the framework of the smart balancing process. For N=4, Δt=15 min, P=1 and M=5, with a one-hour-ahead forecast of the control zone variance and a control zone balance of 4% for positive balancing power and 2% for negative balancing power,

4 , which can be used up by the balancing group manager through trading of the entire pre-balancing power that is predicted to be beneficial for the system in the market within the framework of the smart balancing process, or through hour-ahead planning and trading of the pre-balancing power by the grid operator, also as described above. In this case, the balancing group manager can use the pre-balancing power that consists of possible imbalances in the balancing group by leaving the imbalances unresolved.

の時間的間隔を可能にする。 The steps from the first equation, either alone (i.e. without applying the system-beneficiality condition) or in combination with one or more steps for verifying the system-beneficiality of pre-dispatch reserves (i.e. applying the system-beneficiality condition), may typically be used for trading of pre-balanced electricity to be procured by the grid operator or for the appropriate determination of the balancing group manager in the smart balancing process:

This allows for a time interval of

The predicted control area variance, taking into account the prediction of the benefit to the system, is shown by way of example in FIG. 1b. In FIG. 3a, the advance reserve is shown in comparison with the control area variance. The predicted balance of positive and negative reserve must have the same sign as the predicted control area variance in FIG. 1a. Additionally, the benefit to the system is confirmed through the opposite sign of the predicted grid frequency variance, as is clear from FIG. 3b.

If two control areas have opposing signs for their control area balances, the pre-provision reserve is determined to a value different from zero and activated only for those control areas where the additional orientation to the sign of the grid frequency difference makes the pre-provision reserve beneficial for the system (as opposed to the grid frequency difference). For the other of both control areas, the imbalance in the corresponding balancing group is not eliminated in that time step. Therefore, only the simultaneous orientation to the grid frequency and the control area difference ensures the beneficial effect on the system for each control area.

As a simple measure of system benefit, the prediction without the benefits based on the first equation is first adjusted with the previous measurements of the control area difference. This gives a corrected reduction potential in the control area balance of 58% positive balancing power and 38% negative balancing power with pre-balancing power. The incorrectly adjusted proportions not beneficial to the system associated with the prediction without the benefits of pre-balancing power in the control area balance are -8% for positive balancing power and -17% for negative balancing power.

When the prediction excluding the system benefits is combined with the prediction including the system benefits for the sign of the predicted balance of positive secondary balancing power and negative secondary balancing power based on the second equation, a pre-balancing power reduction potential in the control area balance of 56% for positive balancing power and 28% for negative balancing power is obtained. The misaligned non-system-beneficial percentages of the pre-balancing power in the control area balance with this combination of predictions are -4% for positive balancing power and -20% for negative balancing power.

When the forecast excluding system benefits is combined with the forecast including system benefits for the system beneficial sign of the predicted grid frequency difference relative to the sign of the predicted control area balance, a pre-balancing power reduction potential in the control area balance of 71% for positive balancing power and 55% for negative balancing power is obtained. The misaligned non-system beneficial fraction of the control area balance pre-balancing power with this combination of forecasts is 13% for positive balancing power and 17% for negative balancing power.

Of particular importance in this benchmark is the experimental quantification of the method's error in the form of a "fraction of pre-balancing power that is not beneficial for a misaligned system," which is a number with respect to magnitude, about 20% of the pre-balancing power.

The method operates on pre-balancing power according to a forecast based on the method including and excluding the benefits to the system described above, with a remaining efficiency of 80%. The efficiency of the possibility of reducing the control area balance (pre-determined) due to pre-balancing power according to the method described above including one of the first conditions on benefits to the system is 32% when divided by the sign of the balance between positive and negative reserve power and 6% when divided by the grid frequency difference. Excluding the conditions on benefits to the system, the efficiency is 59%.

The success of the technology was quantified using simulation scenarios, where hour-ahead determination of pre-supply reserves and beneficial control of the system were technically implemented in field tests (on-site). For the field tests of interacting demonstrators, the use case modeling system CondEx, realized by agents communicating between different computers ("remotely"), was used.

The agent interaction diagram for hour-ahead smart balancing supported by prediction is shown in Figure 4.

The planning and procurement of pre-balancing power by the grid operator is shown in Figure 8.

The domain model of the underlying software architecture of CondEx in Figure 7 shows the division of the business logic and use case modeling demonstrators with agents playing a role in controlling the energy supply system and communicating with each other through typology-based messages.

Type-based messaging is an interface that makes it easier to identify "Intents" and allows the possibility of their execution. Within this interface, a paradigm of intelligent communication is realized, which combines aspects of artificial intelligence in the form of (predictive) algorithms from the domain of data science with software technical concepts of distributed, selectively synchronous or parallel (asynchronous) processes and corresponding middleware. Type-based messaging allows the integration of object-oriented programming principles such as encapsulation (in the sense of reusability and extensibility) with modern actors, microservices and cloud-based technologies through a publish-subscribe model.

An agent can "see and understand" as an autonomous and intelligently acting subject. It is networked with its surroundings, for example through sensors for weather data and for load measurements, and from the measurements it derives predictions in the sense of understanding the system in the future.

A further application example is performing predictions using artificial neural networks.

Technologies exist that allow multi-layer neural networks to use multi-dimensional forecast data, such as weather forecasts, as input for predicting time-series-based variables that determine the system, by combining recurrent neural networks with traditional perceptrons or tree-based long short-term memories (TreeLSTMs).

Furthermore, predictions can be made using any predictive method in the field of data science or machine learning.

In a further configuration of the invention, the results of the determination of the pre-supply reserve and the predicted system benefit according to the above-mentioned method are used as input variables for the forecast of the regulated power price, which estimates public market data on trading of balancing power and regulated power prices. This can be done by a method using artificial neural networks. In a special case, it is possible to limit the input variables to the regulated power prices traded up to the time of creating the forecast. In this case, it is possible to use the first-mentioned forecasting method for the control area difference (without the condition on the system benefit). The predicted regulated power price can be evaluated as a criterion for deciding whether or not the imbalance in the balancing group is eliminated within the framework of the smart balancing process, instead of the pre-supply reserve based on the first-mentioned forecasting method or instead of the system benefit based on the further forecasting method described above.

The currently valid mixed pricing method (MPV) is believed to be the cause of the serious system instability that occurred in the German grid over three days in June 2019, as reported, for example, in the Frankfurter Allgemeine Zeitung on July 2, 2019. Extreme grid conditions have been accepted to a large extent so far because the balancing power costs, here the metered price (i.e. the cost of requesting balancing power) and thus the cost of regulating power by MPV, could be cheaper than intraday procurement and intraday management to actively resolve the imbalance (following the timetable). What intraday management enables and promotes benefits for the system can be resisted by the hour-ahead smart balancing according to the present invention.

MPV penalizes renewable energies, with their high metered prices and low capacity prices (i.e. procurement prices), and demand management against conventional power plants, with their low metered prices and high capacity prices. This can be avoided by taking into account the possibility of demand when determining the weighting factors for the metered prices, instead of the uniform weighting factors for the metered prices in MPV. This leads to the demand for the introduction of a balancing electricity market, as is evident from the European Commission's guidelines on electricity balancing.

Based on the above-mentioned forecasting method (including or excluding the condition on the benefit to the system), the pre-balancing power is determined as the integral of the conditional (demand) possibility and the possibility on the benefit to the system and the imbalance (as a cause of the demand). Thus, the pre-balancing power implements the above-mentioned approach of isolating the unified weighting factor for the metered price in MPV through the use of the possibility on the electricity or the adjusted electricity price by the algorithm. Furthermore, the pre-balancing power is the possibility of trading not only the price but also the quantity of electricity, which is especially significant in the context of the balancing electricity market.

In structuring the mechanism based on the demandability of various energy types and flexibility, the various energy types and flexibility are classified according to traditional and renewable energy source classes, as well as according to technologies for demand management and sector linkages.

Claims (6)

A method for stabilization of an electric energy supply network having at least one control area, comprising, for at least one said control area, the following steps:
inputting a time t _N in the future and the length of the interval Δt ,
- Prediction of the control zone difference in the time interval [tN, tN+Δt] through _{a function} using an Nth order Markov chain, where N is the number of time steps to be predicted.
determining a prediction
is the step determined as the conditional expectation of the Markov chain,
a prediction regarding the control range difference in said time interval [t _N , t _N +Δt];
determining a forecast ΔE _RLvoraus for the balancing power required in said time interval [t _N , t _N +Δt] from
Procuring a predicted balancing power ΔE _RLvoraus ; and
supplying said electrical energy supply network with a predicted balancing power ΔE _RLvoraus in said time interval [t _N , t _N +Δt];
The method comprises: The prediction ΔE _RLvoraus for the balancing power required in the time interval [t _N , t _N +Δt] is a prediction ΔE RLvoraus for the control range difference in the time interval [t _N , t _N +Δt].
and the length of the interval Δt. When one condition for system benefit is met or when multiple conditions for system benefit are cumulatively met, the prediction ΔE _RLvoraus for the balancing power required in the time interval [t _N , t _N +Δt] is equal to the prediction ΔE RLvoraus for the control range difference in the time interval [t _N , t _N +Δt].
and the interval length Δt, and is zero otherwise.
The condition or one of the conditions for the system's usefulness is a predicted control range difference in the time interval [t _N , t _N +Δt].
has the same sign as the sign of the forecasted balance P _RLprogn ^t _N of positive and negative reserve power in said time interval [t _N , t _N +Δt],
The condition or one of the conditions for the system's usefulness is a predicted control range difference in the time interval [t _N , t _N +Δt].
the sign of the predicted grid frequency difference Δω progn t N in said time interval [t _N , t _N +Δt] is different from the sign of the predicted grid frequency difference Δω _progn ^t _N in said time interval [t N , t N +Δt]; and/or
The condition or one of the conditions for the usefulness of the system is that the predicted supply management signal Einman _progn (t _N ) in the corresponding area at the time t _N has a value of “not supply management” or that the predicted control area difference in the time interval [t _N , t _N +Δt] is
The method of claim 1 , wherein is negative . 4. The method according to claim 1 , wherein the time interval between the current time and the time _tN at which the ΔE _RLvorous time interval begins, during which the required balancing power is predicted, is at least one hour. A computer program for carrying out the method according to any one of claims 1 to 4 when said method is carried out on a computer. A computer program product having program code stored on a machine readable medium and comprising the computer program of claim 5 .

Images