JP7793486B2 - Water demand forecasting device and water demand forecasting method - Google Patents

Description

The present invention relates to a water demand prediction device and a water demand prediction method.

Municipalities and other water utilities create water operation plans for water intake, water supply, etc. to ensure an appropriate water supply to consumers. When creating these water operation plans, they predict water demand in the target distribution area based on past water usage data.

If the above-mentioned water demand forecasts are accurate, it will be possible to appropriately avoid the need to revise plans or carry out emergency operations during operation, enabling stable operation.

Prior art related to water demand forecasting has been proposed, such as a water distribution volume planning and forecasting system (see Patent Document 1) that is inexpensive and can reduce the deviation between the predicted water distribution volume plan pattern and the actual measured water distribution volume.

This technology relates to a water distribution volume plan and prediction system that creates plans for the volume of water distribution to be supplied from water supply facilities to consumers. The system includes a water distribution volume plan creation unit that creates a water distribution volume plan pattern that predicts the volume of water distribution for a specified future period based on information on actual water distribution volume performance values acquired over time. If the day on which a water distribution volume plan pattern, which represents the predicted volume of water distribution, is created is a public holiday, the actual water distribution volume for the previous Saturday during a set period is calculated. If the day is a weekday other than a public holiday, the actual water distribution volume for the same day during a set period is calculated. The average of the calculated actual water distribution volume values is used as the water distribution volume plan pattern for that day. The system also includes a water distribution volume plan prediction pattern correction unit that corrects the predicted water distribution volume plan pattern for a specified future period to reduce the deviation from the actual water distribution volume performance values if the deviation between the predicted water distribution volume plan pattern and the actually measured actual water distribution volume exceeds the allowable deviation range.

JP 2012-99049 A

In recent years, smart water meters are being installed in ordinary households. These households can acquire and store water usage data in a timely manner. By aggregating and processing the data from these households, it is possible to roughly estimate water usage and water volume at appropriate intervals, such as by day or time period (for example, technology disclosed in JP 2014-134985 A).

The above estimates show that, although there are variations depending on the region and household, the amount of water used for each purpose in a day is roughly fixed.

However, when events occur that are not factors that are predicted in conventional water demand forecasts, such as the holding of a local event or an inaccurate weather forecast, it becomes difficult to predict short-term increases or decreases in demand in units of a few hours. In such cases, the accuracy of water demand forecasts may decrease, making appropriate operation difficult.

The object of the present invention is to provide technology that enables accurate water demand forecasting taking into account short-term demand fluctuations.

The water demand forecasting device of the present invention, which solves the above problem, comprises a storage device that stores water consumption data for each purpose of each consumer, and a water demand forecasting device for each consumer that stores water consumption data for each purpose of each consumer.
a process for calculating a difference in the time series data for each of the clusters relating to the applications after the transition timing of the peak time period and generating a first model that indicates a distribution trend of the difference per unit time; a process for calculating a differential value for each of the clusters relating to the applications at regular time intervals in the water usage data corresponding to each of the applications and generating a second model that indicates a transition trend of water usage for each of the clusters; and a computing device that detects a transition event of the peak time period using the second model, allocates the amount of deviation between predicted water usage data and actual water usage data for the detected applications over time in accordance with the first model, and applies the difference to the predicted data to correct it.

Furthermore, the water demand forecasting method of the present invention is characterized in that an information processing device includes a storage device that stores water usage data for each application at each consumer, and executes the following processes: clustering time-series data of daily water usage at each consumer indicated by the water usage data for each application, and identifying applications for which there are differences in peak hours between clusters; calculating differences in the time-series data between the clusters for the applications after the transition timing of the peak hours, and generating a first model that shows the distribution trend of the differences per unit time; calculating differential values for each of the clusters for the applications at regular time intervals in the corresponding water usage data, and generating a second model that shows the water usage transition trend for each of the clusters; and detecting transition events at the peak hours using the second model, allocating the amount of deviation between the predicted water usage data and the actual water usage data for the detected application over time according to the first model, and applying the deviation to the predicted data to correct it.

This invention makes it possible to accurately forecast water demand taking into account short-term demand fluctuations.

1 is a diagram illustrating a network configuration including a water demand prediction device according to an embodiment of the present invention. 1 is a diagram illustrating an example of a hardware configuration of a water demand prediction device according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an example of the configuration of a weather condition DB according to the present embodiment. FIG. 2 is a diagram illustrating an example of the configuration of a water usage log DB according to the present embodiment. FIG. 2 is a diagram illustrating an example of the configuration of an other infrastructure log DB according to the present embodiment. 1 is a diagram showing a flow of a water demand prediction method and an example of the configuration of functional units in this embodiment. FIG. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 2 is a diagram illustrating an example of a shift model in the present embodiment. FIG. 10 is a diagram illustrating an example of a shift deviation model in the present embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 4 is a diagram illustrating an example of performance information data according to the present embodiment. FIG. 2 is a diagram illustrating an example of water demand forecast time series data in this embodiment. FIG. 2 is a diagram showing an example of the flow of a water demand prediction method according to the present embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 2 is a diagram showing an example of the flow of a water demand prediction method according to the present embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 1 is a diagram illustrating an example of a processing concept in this embodiment. FIG. 10 is a diagram illustrating an example of output in this embodiment.

<System Configuration>
An embodiment of the present invention will be described in detail below with reference to the accompanying drawings. Fig. 1 is a diagram showing an example of a system configuration including a water demand prediction device 100 according to the present embodiment, and Fig. 2 is a diagram showing the functional configuration of the water demand prediction device according to the present embodiment.

The water demand prediction device 100 shown in Figures 1 and 2 is a computer that enables accurate water demand predictions that take into account short-term demand fluctuations.

As shown in FIG. 1, the water demand prediction device 100 of this embodiment is communicably connected to a user terminal 200 via a network 1. Therefore, these may be collectively referred to as a water demand prediction system 10.

The water demand prediction device 100 of this embodiment is an information processing device operated, for example, by a water utility such as a local government, or by a vendor that provides services to the water utility. This water demand prediction device 100 can also be considered a water demand prediction service provider.

The water demand prediction device 100 of this embodiment is connected to a user terminal 200 as a user interface, which is equipped with a display that displays the screen 301, and devices such as a keyboard and mouse for the user to input information.

On the other hand, the user terminal 200 is a terminal operated by a person who manages and operates the water infrastructure by utilizing the results of the water demand prediction, such as an employee of the local government mentioned above. This user terminal 200 is a device that displays the information output from the water demand prediction device 100. Specifically, it can be a personal computer, tablet terminal, smartphone, etc.

Other devices that may be connected to and linked with this network 1 include weather forecast systems that provide information such as weather conditions for the locations of water consumers, and systems from other infrastructure providers, such as electricity providers, other than water.

The weather forecast system is a server device that distributes information on conditions such as weather, temperature, humidity, wind direction, wind speed, rainfall, and snowfall for each time period of each day to the water demand prediction device 100, for example, at regular intervals.

In addition, the systems of other infrastructure providers are systems of providers such as electric power providers, gas providers, or telecommunications providers, and are server devices that distribute usage history, i.e., log data, of the infrastructure services that the providers provide to the above-mentioned consumers to the water demand prediction device 100, for example, at regular intervals.
<Hardware configuration>
As shown in the configuration examples of Figures 1 and 2, the water demand prediction device 100 of this embodiment includes an auxiliary memory device 101, a main memory device 103, a CPU 104 as a computing device, a display interface 105, an input/output interface 106, and a network interface 107.

Of these, the auxiliary storage device 101 is composed of an appropriate non-volatile storage element such as an SSD (Solid State Drive) or a hard disk drive.

The main memory device 103 is composed of volatile memory elements such as RAM (Random Access Memory).

The CPU (Central Processing Unit) 104 is a computing device that executes the program 102 stored in the auxiliary storage device 101 by reading it into the main storage device 103, thereby providing overall control of the device itself and performing various judgments, calculations, and control processes.

The display interface 105 is an interface that displays processing data between the user terminal 200 and the system.

The input/output interface 106 is an interface that accepts key inputs and mouse operations from the user and responds with processing results from the user terminal 200.

The network interface 107 is assumed to be a network interface card or the like that connects to an appropriate network 1 such as the Internet and handles communication processing with the user terminal 200 and other systems.

In addition to the program 102 for implementing the functions required for the water demand prediction device of this embodiment, the auxiliary storage device 101 also stores at least various databases and data such as a weather conditions DB 125, a water usage log DB 126, an other infrastructure log DB 127, water usage time series data by use 128, a shift model 129, a shift deviation model 130, and water demand forecast time series data 131. However, these will be described in more detail below.

In addition, by executing the above-mentioned program 102, a water use analysis unit 110, a demand change analysis unit 111, a unit period demand forecasting unit 112, a water use deviation monitoring unit 113, a deviation amount calculation unit 114, a demand forecast correction unit 115, and a water demand forecast visualization unit 116 are implemented.
<Data structure example>
Next, a description will be given of various databases and information used by the water demand prediction device 100 of this embodiment. Fig. 3 shows an example of the configuration of the weather condition DB 125 in this embodiment.

In this embodiment, the weather condition DB125 is a database that stores information obtained from weather forecast service providers, etc., regarding weather conditions in the area where water consumers are located.

This weather condition DB125 is a collection of records that link data such as weather, temperature, precipitation, and amount of sunshine, using, for example, the year, month, date, and time as keys.

Figure 4 also shows an example of the data configuration of the water usage log DB 126 in this embodiment. The water usage log DB 126 in this embodiment is a database that stores actual water usage information for each of the above-mentioned consumers.

This water usage DB126 is a collection of records that use a household ID that uniquely identifies a consumer as a key, and link values such as the date and time, and the amount of water usage observed on that date and time (the amount of usage since the previous observation time).

5 shows an example of the data configuration of the other infrastructure log DB 127 in this embodiment. The other infrastructure log DB 1276 in this embodiment is a database that stores usage records of infrastructure other than water infrastructure, such as electricity.

The other infrastructure log DB 127 is a collection of records that link values such as the date and time, and the amount of power usage observed at that date and time, using the household ID as a key.
The actual procedure for the water demand forecasting method of this embodiment will be described below with reference to the accompanying drawings. The various operations corresponding to the water demand forecasting method described below are realized by a program that is read into a main memory or the like and executed by the water demand forecasting device 100. This program is composed of code for performing the various operations described below.

Figure 6 shows an example of the flow of the water demand prediction method and the configuration of functional units in this embodiment. In this case, the water use analysis unit 110 of the water demand prediction device 100 receives the weather conditions DB 125, water usage log DB 126, and other infrastructure log DB 127 as inputs and generates water usage time series data 128 (see Figure 7) by purpose for each day for each household.

As already mentioned, the method for generating this water usage time series data 128 by purpose employs an analytical algorithm based on existing technology, and by inputting the above-mentioned data into this, time series data on water usage by purpose for each household and each day can be output.

Unlike other infrastructure such as electricity, water usage tends to be limited to short periods of time when it is used, with zero usage at other times. When this is combined with time information and electricity usage information, it is possible to analyze, for example, that "if a small amount of water is used in the morning with almost no electricity usage, it was used for toilet purposes."

Next, the processing in the demand change analysis unit 111 will be explained with reference to Figures 8 to 13. Here, the demand change analysis unit 111 adds up the water usage time series data for a certain purpose for all households for each day based on the above-mentioned water usage time series data by purpose 128 (s10).

Next, the demand change analysis unit 111 performs geometric clustering of the time series data obtained in s10 when graphed, for example, using the time series K-means method, and calculates the average value of the observed values at each time among the time series data belonging to each cluster (s11).

The demand change analysis unit 111 then compares the averaged time series data for each cluster obtained in s11 and checks whether there are any time series data for which the demand peaks (peaks of water usage) are significantly shifted in time (i.e., peaks whose occurrence times are separated by more than a predetermined threshold) (s12). Furthermore, at this time, the demand change analysis unit 111 defines, as a result of this check, the applications for which a time shift is recognized as the "shift amount," and the applications for which a time shift is not recognized as the "base amount" (s12).

When the demand change analysis unit 111 has performed the above steps s10 to s12 for all applications (s13: yes), it ends this flow.

Next, as shown in the flow in Figure 9, the demand change analysis unit 111 generates a shift model and a shift deviation model.

In this case, the demand change analysis unit 111 sums the usage data by purpose for each household for each day (s20).

In addition, the demand change analysis unit 111 overlays the curves shown by the water usage time series data by purpose for each day, and determines the time during the daytime hours when there is statistically significantly less change as time period U (the transition timing to the peak time period) (s21).

Next, the demand change analysis unit 111 clusters the curves indicated by the time series data for each day described above, and obtains f(t) and g(t) for the cluster, for example, as shown in Figure 10 (s22).

The demand change analysis unit 111 also determines the time after the above-mentioned time U at which a significant difference in the slopes of clusters f(t) and g(t) occurs as the start time of T, and the time at which the difference in the slopes of f(t) and g(t) again becomes significantly equal as the end time of T (s23). This T is the time period after the above-mentioned time period U has elapsed at which a significant difference in trend (difference in the slopes of the graphs) occurs between the clusters.

Next, the demand change analysis unit 111 integrates the differences between the above-mentioned clusters f(t) and g(t) at each time t1 to tn during the post-shift time period T to obtain Sn (s24).

The demand change analysis unit 111 also divides the above-mentioned difference value at each time by the above-mentioned Sn value, standardizes it, and uses the result as a shift model 129 (see Figure 11) that shows the temporal distribution of water usage during the shift time period for that use (s25).

Next, the demand change analysis unit 111 calculates the average value for each time of the water usage data included in the above-mentioned cluster to obtain the averages f(t) and g(t), and then calculates the differential value of the water usage data at each time from these averages f(t) and g(t) to create a shift deviation model 130 (see Figure 12) (s26), and then ends the processing.

Next, the processing in the water usage deviation monitor unit 113 will be explained based on the flow in Figure 13. It should be noted that prior to this flow, the unit period demand forecast unit 112 is assumed to have generated water demand forecast time series data 131 (Figure 15) based on the most recently observed water usage data, i.e., performance information data 140 (see Figure 14). This water demand forecast time series data 131 is assumed to have been obtained by analyzing the performance information data 140 using the algorithm in the existing technology already mentioned.

Here, the water usage deviation monitor unit 113 monitors whether there is any deviation between the current water usage amount and the predicted value f(t) for use "laundry" for use A (e.g., laundry), which is defined as the shift amount (s30).

By continuing the above-mentioned monitoring, the water usage deviation monitor unit 113 determines whether the above-mentioned deviation is equal to or greater than a predetermined set value (s31).

If the result of the above judgment indicates a deviation greater than a predetermined set value (s31: yes), the water usage deviation monitor unit 113 verifies, using the shift deviation model for use A (laundry), whether the trend in water usage for the use "laundry" up to now is similar to the behavior (slope of the graphed curve, peak point) of the shift deviation model g'(t) (s32).

If the above verification reveals similarity (s34: yes), the water usage deviation monitor unit 113 determines that there is a shift amount deviation, i.e., that the data is moving out of the scope of the initial prediction data and is transitioning to a different usage trend (s35), and ends this flow.

On the other hand, if the above verification results in no similarity, i.e., no shift amount deviation, the water usage deviation monitor unit 113 determines whether the water usage amount for the application defined as the base amount deviates from the predicted data by more than a predetermined threshold (s36).

If the result of the above determination indicates a deviation greater than or equal to the threshold (s36: yes), the water usage deviation monitor unit 113 determines that the deviation is below the base amount (s37) and ends this flow.

On the other hand, if the result of the above determination is that there is no deviation greater than the threshold (s36: no), the water usage deviation monitor unit 113 determines that there is currently no deviation from the predicted data regarding the base amount (s38), and ends this flow.

Figure 16 shows the processing of the deviation amount calculation unit 114 when it is determined that the shift amount is deviating, as described above.

In this case, the deviation calculation unit 114 monitors the change in water usage (Figure 17A) until the deviation ends (s40). To determine the end of the deviation, the shift deviation model is used, just as when the start of the deviation is detected, to determine whether the predicted data and actual usage data match or the slope of the graph (the derivative of the observed value at each time) are close to each other.

In addition, the deviation calculation unit 114 calculates the difference between the accumulated water volume in the predicted water usage data (dashed line in the graph in Figure 17B) and the accumulated water volume of the actual usage (solid line in the graph in Figure 17B) for the time period from the start to the end of the deviation as the shift volume (s41).

If it is determined in step s37 of the flow in Figure 13 that the base amount is deviating, the deviation amount calculation unit 114 will observe the deviation until it converges or for a short period of time (e.g., approximately 1 to 3 hours) as shown in the flow in Figure 18 (s50).

If the observation shows that the deviation has converged (s51: no), the deviation calculation unit 114 does not perform any particular processing and ends this flow.

On the other hand, if the above observation shows that the deviation does not converge (s51: yes), the deviation calculation unit 114 determines that the prediction is off overall, defines Eave as the "base volume" (s52), and terminates processing. Eave can be calculated using the formula shown in Figure 19. As shown in the graph in Figure 19, if a certain amount of time has passed since the deviation began and the deviation has not converged, it is recognized that the water usage amount predicted as the base needs to be corrected, and the average value of the increasing deviation is taken as the base volume.

In addition, the shift volume and base volume values obtained in the flow up to this point are notified to the demand forecast correction function 115 as demand deviation data 143 (see Figure 20).

The demand forecast correction function 115 applies each value (shift volume, base volume) indicated by the demand deviation data 143 to the unit period water demand forecast time series data 144 for the relevant use that has been used up to this point, and corrects it.

In other words, the "shift volume" of the demand deviation amount is distributed over time in accordance with the "shift model," and is added or subtracted according to the time indicated by the shift model. Also, if a "base volume" is presented, the base volume [%] is added or subtracted from the forecast amount for the following time period. Such corrections are made in the same way for each use.

The water demand forecast visualization unit 116 also aggregates the unit period water demand forecast time series data 144 for each use after the above-mentioned corrections, and outputs it to the user terminal 200 as a graph of the total forecast demand amount (see Figure 21). In addition to the graph, the screen on which this output is displayed also displays, in table format, the predicted values of water usage for each use, such as total usage, cooking, bathing, and toilet, for each date and time.

The above provides a detailed explanation of the best mode for carrying out the present invention, but the present invention is not limited to this and various modifications are possible without departing from the spirit of the invention.

This embodiment makes it possible to accurately forecast water demand taking into account short-term demand fluctuations.

The description herein makes at least the following clear: In other words, in the water demand prediction device of this embodiment, the storage device stores predicted data of water usage for the application, and the computing device compares actual water usage data for the application at each consumer with the predicted data, and if the actual water usage data deviates from the predicted data by more than a standard amount, determines that a shift in the time period of peak water usage has occurred if one of the second models for each of the multiple clusters is identified that has a similar water usage transition trend indicated by the actual water usage data; and aggregates the amount of deviation between the predicted data and the actual water usage data until the end of the deviation, allocates the result of the aggregation over time according to the first model, and applies the result of the aggregation to the predicted data of water usage for the application to correct it.

This allows the data necessary to revise forecast data to be collected and applied with high accuracy. Ultimately, this enables more accurate water demand forecasts that take short-term demand fluctuations into account.

Furthermore, in the water demand prediction device of this embodiment, the calculation device may further perform a process of identifying applications for which there is no difference in peak time periods among the multiple clusters obtained by the clustering as base demand applications, and if the result of the determination shows that there has been no shift in the peak time periods and the water usage data for the base demand application deviates from the predicted data for that application by more than a standard amount, correcting the predicted data by increasing or decreasing it according to the degree of deviation.

This makes it possible to accurately correct forecast data even for applications that do not require a time shift. Ultimately, this will enable even more accurate water demand forecasts that take short-term demand fluctuations into account.

Furthermore, in the water demand prediction device of this embodiment, the calculation device may further perform a process of outputting information regarding the prediction data before and after the correction to a specified device.

This will enable water utility personnel, such as local governments, to visually confirm any corrections to water demand forecasts, making it easier to make accurate decisions regarding subsequent operations. Ultimately, this will enable more accurate water demand forecasts that take short-term demand fluctuations into account.

In addition, in the water demand prediction method of this embodiment, the information processing device may store predicted data of water usage for the application in the storage device, compare the actual water usage data for the application at each consumer with the predicted data, and if the actual water usage data deviates from the predicted data by more than a standard, determine that a shift in the peak time period has occurred if one of the second models for each cluster can be identified that has a similar water usage transition trend indicated by the actual water usage data; and aggregate the amount of deviation between the predicted data and the actual water usage data until the end of the deviation, allocate the result of the aggregation over time according to the first model, and apply the result of the aggregation to the predicted data of water usage for the application to correct it.

Furthermore, in the water demand prediction method of this embodiment, the information processing device may identify applications for which there is no difference in peak time periods among the multiple clusters obtained by the clustering as base demand applications, and if the determination shows that there has been no shift in the peak time periods and the water usage data for the base demand application deviates from the predicted data for that application by more than a standard amount, the information processing device may further perform a process of correcting the predicted data by increasing or decreasing it depending on the degree of deviation.

Furthermore, in the water demand prediction method of this embodiment, the information processing device may further perform a process of outputting information regarding the prediction data before and after the correction to a specified device.

1 Network 10 Water demand prediction system 100 Water demand prediction device 101 Auxiliary storage device 102 Program 103 Main storage device 104 CPU (arithmetic unit)
105 Display interface 106 Input/output interface 107 Network interface 110 Water use analysis unit 111 Demand change analysis unit 112 Unit period demand forecast unit 113 Water use deviation monitor unit 114 Deviation amount calculation unit 115 Demand forecast correction unit 116 Water demand forecast visualization unit 125 Weather condition DB
126 Water usage log DB
127 Other infrastructure log DB
128 Time series data of water use by use 129 Shift model (first model)
130 Shift Deviation Model (Second Model)
131 Water demand forecast time series data 200 User terminal

Claims (8)

a storage device that stores water usage data for each purpose at each consumer;
a processing for clustering time series data of daily water usage for each of the consumers indicated by the water usage data for each application, and identifying applications for which there are differences in peak time periods between clusters; a processing for calculating differences in the time series data between the clusters for the applications after a transition timing of the peak time period, and generating a first model that indicates a distribution trend of the differences per unit time; a processing for calculating, for each of the clusters for the applications, a differential value for each fixed time period in the corresponding water usage data, and generating a second model that indicates a transition trend of water usage for each of the clusters; and a processing for detecting a transition event of the peak time period using the second model, allocating a time period for a deviation between predicted water usage data and actual water usage data for the detected application, and applying the deviation to the predicted data to correct the data;
A water demand prediction device comprising: The storage device includes:
Prediction data of water usage for the above-mentioned uses is held,
The computing device
a process of comparing actual water usage data relating to the application at each of the consumers with the predicted data, and if the actual water usage data deviates from the predicted data by more than a standard, determining that a shift in the peak time period has occurred if one of the second models for each of the clusters can be identified that has a similar water usage transition trend indicated by the actual water usage data;
a process of aggregating the amount of deviation between the predicted data and the actual water usage data until the end of the deviation, allocating the result of the aggregation over time according to the first model, and applying the result of the aggregation to the predicted data of the water usage for the application to correct it;
The following is a list of the tasks that can be performed:
The water demand forecasting device according to claim 1 . The computing device
Identifying a use that has no difference in the peak time period among the plurality of clusters obtained by the clustering as a use of base demand;
If the result of the determination indicates that no shift in the peak time period has occurred and the water usage data relating to the base demand application deviates from the predicted data for that application by more than a standard, a process of correcting the predicted data by increasing or decreasing it according to the degree of the deviation;
The following further implements the following:
The water demand prediction device according to claim 2 . The computing device
and further executing a process of outputting information about the prediction data before and after the correction to a predetermined device.
The water demand prediction device according to claim 1 . The information processing device
A storage device is provided to store data on the amount of water used by each customer.
a process of clustering time series data of daily water usage for each consumer indicated by the water usage data for each application, and identifying applications for which there are differences in peak time periods between clusters; a process of calculating differences in the time series data between the clusters for the applications after the timing of a transition to the peak time period, and generating a first model that indicates a distribution trend of the differences per unit time; a process of calculating, for each of the clusters for the applications, differential values for each fixed time period in the corresponding water usage data, and generating a second model that indicates a transition trend in water usage for each of the clusters; a process of detecting a transition event in the peak time period using the second model, allocating the amount of deviation between predicted water usage data and actual water usage data for the detected application over time according to the first model, and applying the deviation to the predicted data to correct it;
A water demand forecasting method comprising: The information processing device,
The storage device stores predicted data of the amount of water usage related to the intended use,
a process of comparing actual water usage data relating to the application at each of the consumers with the predicted data, and if the actual water usage data deviates from the predicted data by more than a standard, determining that a shift in the peak time period has occurred if one of the second models for each of the clusters can be identified that has a similar water usage transition trend indicated by the actual water usage data;
a process of aggregating the amount of deviation between the predicted data and the actual water usage data until the end of the deviation, allocating the result of the aggregation over time according to the first model, and applying the result of the aggregation to the predicted data of the water usage for the application to correct it;
6. The water demand forecasting method according to claim 5, wherein the water demand forecasting method further comprises: The information processing device,
Identifying a use that has no difference in the peak time period among the plurality of clusters obtained by the clustering as a use of base demand;
If the result of the determination indicates that no shift in the peak time period has occurred and the water usage data relating to the base demand application deviates from the predicted data for that application by more than a standard, a process of correcting the predicted data by increasing or decreasing it according to the degree of the deviation;
The water demand forecasting method according to claim 6, further comprising: The information processing device,
The water demand forecasting method according to claim 5, further comprising the step of outputting information about the forecast data before and after the correction to a specified device.