JP7069732B2 - Estimator program, estimation method and estimation device - Google Patents

Description

The present invention relates to an estimation program, an estimation method and an estimation device.

Conventionally, in the fields of energy such as electricity and gas, and the field of sensing by an ad hoc network in which meter reading sensors are fixedly arranged, predictions are made from past histories and plans are made.

For example, in the field of electricity charges, a stationary storage battery system controlled by HEMS (Home Energy Management System), BEMS (Building Energy Management System), CEMS (Community Energy Management System), or the like is known. In such a stationary storage battery system, the parameters of the power demand forecast model are optimally fitted to the past actual power demand results, and the power demand forecast obtained from the obtained power demand forecast model is charged and discharged. The plan is being calculated.

Japanese Unexamined Patent Publication No. 2012-194935 Japanese Unexamined Patent Publication No. 2010-21347

However, in the above technique, the charge / discharge plan is created according to the power demand forecast fitted with the parameters of the forecast model without considering the degree of influence on the electricity rate, so that the accuracy of the achievable electricity rate reduction is limited. be.

Generally, the charge / discharge plan of the storage battery is planned so that the battery is charged at a time when the electricity charge is low and discharged at a time when the electricity charge is high. However, since the storage battery cannot reverse power flow to the grid, it cannot be discharged more than expected at the stage of making a plan. Therefore, if the accuracy of power demand prediction is lowered, the accuracy of the charge / discharge plan itself is also lowered, and the amount of discharge is limited, so that the amount of reduction in electricity charges is reduced.

One aspect is to provide an estimation program, estimation method and estimation device that can improve the accuracy of demand forecasting.

In the first proposal, the estimation program optimizes the parameters of the likelihood function of the forecast model selected by a predetermined information criterion for the past demand actual data to the computer, and the estimation is obtained by the optimization. Using the probability distribution of the demand forecast obtained from the forecast model, the process of estimating the first demand forecast is executed. The estimation program causes a computer to execute a process of calculating the degree of influence of a specific item that affects the expected value predicted by the first demand forecast. The estimation program estimates the second demand forecast by using the probability distribution of the demand forecast based on the forecast model obtained by optimizing the parameters of the likelihood function to which the weight by the influence degree is added to the computer. To execute.

According to one embodiment, the accuracy of demand forecasting can be improved.

FIG. 1 is a diagram for explaining the power demand forecast according to the first embodiment. FIG. 2 is a diagram illustrating a flow of optimization according to the first embodiment. FIG. 3 is a functional block diagram showing a functional configuration of the information processing apparatus according to the first embodiment. FIG. 4 is a diagram illustrating an approach for sensitivity analysis. FIG. 5 is a flowchart showing the flow of processing. FIG. 6 is a diagram showing an example of charge / discharge plan 1 when the prediction parameters are optimized by the maximum likelihood method. FIG. 7 is a diagram showing an example of charge / discharge plan 2 when the prediction parameter is optimized by the weighted likelihood. FIG. 8 is a diagram illustrating a predicted distribution of the amount of reduction in electricity charges according to the charge / discharge plan 1. FIG. 9 is a diagram illustrating a predicted distribution of the amount of reduction in electricity charges according to the charge / discharge plan 2. FIG. 10 is a diagram illustrating a mean square error of each charge / discharge plan. FIG. 11 is a diagram for explaining the evaluation of the predicted distribution of the electricity rate reduction amount. FIG. 12 is a diagram illustrating the overall processing of the second embodiment and the first embodiment. FIG. 13 is a functional block diagram showing a functional configuration of the information processing apparatus according to the second embodiment. FIG. 14 is a diagram illustrating a statistical basis for weighted maximum likelihood estimation. FIG. 15 is a diagram illustrating the effect of the second embodiment. FIG. 16 is a diagram illustrating a hardware configuration example of the information processing apparatus.

Hereinafter, examples of the estimation program, estimation method, and estimation device disclosed in the present application will be described in detail with reference to the drawings. The present invention is not limited to this embodiment.

[Power demand forecast]
FIG. 1 is a diagram for explaining the power demand forecast according to the first embodiment. The information processing device 10 shown in FIG. 1 is an example of an estimation device that executes power demand forecasting of a stationary storage battery system, and is, for example, a server or a personal computer.

The information processing device 10 inputs electricity charges for each time zone, past actual power demand data, and a demand forecast model (hereinafter, may be simply referred to as a forecast model) which is an example of the likelihood function to be used. As a result, the charge / discharge plan 1 is generated using the probability distribution of the power demand forecast obtained by performing the parameter fitting of the demand forecast model.

Subsequently, the information processing apparatus 10 generates a plurality of random numbers from the power demand forecast distribution in order to express the power demand forecast expressed as a probability distribution as a plurality of scenarios. Then, the information processing apparatus 10 slightly fluctuates each scenario for each time, and may be described as a minute fluctuation rate of the electricity charge reduction amount with respect to the charge / discharge plan when the scenario is achieved (hereinafter, referred to as a sensitivity coefficient). ) Is calculated.

After that, the information processing apparatus 10 averages the sensitivity coefficients of each scenario for each time zone to obtain the degree of influence of the achievable electricity charge reduction amount for each time zone. Then, the information processing apparatus 10 uses the probability distribution of the power demand forecast derived from the power demand forecast model obtained by performing the parameter fitting of the weighted demand forecast model again with the sensitivity coefficient for each time zone as a weight. The charge / discharge plan 2 is generated.

That is, the information processing apparatus 10 improves the power demand forecast accuracy in an important time zone by fitting the demand forecast model by weighting the likelihood function of the demand forecast model with a sensitivity coefficient. In other words, the information processing apparatus 10 reduces the variance of the predicted probability distribution. Then, the information processing apparatus 10 improves the accuracy of the amount of electricity charge reduction that can be achieved by calculating the charge / discharge plan for the power demand forecast.

[Optimization flow]
Subsequently, the creation of the charge / discharge plan described with reference to FIG. 1 will be described. FIG. 2 is a diagram illustrating a flow of optimization according to the first embodiment. As described with reference to FIG. 1, the information processing apparatus 10 according to the first embodiment generates a charging / discharging plan for a capacitor that increases the reduction amount of electricity charges based on the power demand forecast.

Specifically, as shown in FIG. 2, the generation of the charge / discharge plan consists of two phases, the parameters of the prediction model and the optimization evaluation. More specifically, the information processing apparatus 10 optimizes the parameters of the likelihood function of the prediction model by the maximum likelihood method, and solves the probability optimization problem for the probability distribution of the power demand forecast obtained from the prediction model. , Generate a charge / discharge plan 1.

After that, the information processing apparatus 10 uses the result of the charge / discharge plan 1 to perform a sensitivity analysis on the amount of electricity charge reduction that can be achieved in the demand forecast for each time zone, and determines the importance of the power demand forecast for each time zone. Quantitatively calculated as a coefficient (β _i ). Then, the information processing apparatus 10 weights the likelihood function of the prediction model with a sensitivity coefficient (β _i ), optimizes the power demand forecast model again, and predicts the power demand obtained from the prediction model. The charge / discharge plan 2 is generated by solving the probability optimization problem again for the probability distribution of.

In this way, the information processing apparatus 10 generates the charge / discharge plan after improving the power demand forecast accuracy in the important time zone.

[Functional configuration]
FIG. 3 is a functional block diagram showing a functional configuration of the information processing apparatus 10 according to the first embodiment. As shown in FIG. 3, the information processing apparatus 10 includes a communication unit 11, a storage unit 12, and a control unit 20. The communication unit 11 is a processing unit that controls communication of other devices, such as a network interface card or a wireless interface.

The electric power demand forecast data DB 13 is a database that stores past electric power usage records. Specifically, the power demand forecast data DB 13 stores the power usage record for each time zone stored by the administrator or the like. The storage battery data DB 14 is a database that stores information about the storage battery that is the target of the charge / discharge plan. Specifically, the storage battery data DB 14 stores the storage capacity of the storage battery, the maximum charge / discharge amount, and the like stored by the administrator or the like.

The electricity rate plan data DB 15 is a database that stores electricity rates in the area for which the charge / discharge plan is created. Specifically, the electricity rate plan data DB 15 stores the electricity rate unit price for each time stored by the administrator or the like. The charge / discharge plan DB 16 is a database that stores the generated charge / discharge plan, and is updated by the resolving unit 27 described later.

The control unit 20 is a processing unit that controls the entire information processing device 10, and is, for example, a processor. The control unit 20 includes an optimization unit 21, a sampling unit 22, a solution unit 23, a sensitivity analysis unit 24, a reoptimization unit 25, a resampling unit 26, and a resolution unit 27. The optimization unit 21, the sampling unit 22, the solution unit 23, the sensitivity analysis unit 24, the reoptimization unit 25, the resampling unit 26, and the resolution unit 27 are executed by, for example, an example of an electronic circuit of a processor or a processor. This is an example of the process. Further, the solution unit 23 is an example of the first estimation unit, the sensitivity analysis unit 24 is an example of the calculation unit, and the resolving unit 27 is an example of the second estimation unit.

The optimization unit 21 is a processing unit that optimizes the parameters of the prediction model by the maximum likelihood method. Specifically, the optimization unit 21 selects an appropriate prediction model based on an information criterion such as AIC (Akaike Information Criterion). Subsequently, the optimization unit 21 evenly performs parameter fitting of the prediction model for the power demand data collected within a predetermined period, and outputs the parameter fitting to the sampling unit 22.

For example, the optimization unit 21 performs parameter fitting by a linear Gaussian state space model and a setting method of each parameter. More specifically, the optimization unit 21 performs parameter fitting using the equations (1) and (2). Here, each parameter will be described. The coefficient matrices F, G, H are heuristically given from the prior information of the object. η is obtained by updating from the measured value at any time by the Kalman filter. Q and R are statistically obtained from past data by the maximum likelihood method. Then, the optimization unit 21 determines the likelihood function (formula) that maximizes the log-likelihood of the log-likelihood function for the linear Gaussian state-space model in which the parameters are set in this way. (3)) is generated. That is, the optimization unit 21 generates a prediction probability distribution by the maximum likelihood method.

The sampling unit 22 is a processing unit that executes sampling for the optimized prediction model. Specifically, the sampling unit 22 performs sampling on the prediction probability distribution generated by the optimization unit 21 by using the Monte Carlo method or the like. Then, the sampling unit 22 outputs the sampling result of the prediction probability distribution to the solution unit 23 and the sensitivity analysis unit 24.

The solution unit 23 is a processing unit that generates an optimized charge / discharge plan 1 by solving a probability optimization problem with respect to the prediction by the prediction model. That is, the solution unit 23 formulates an optimization problem for obtaining a charge / discharge plan that optimizes the amount of electricity charge reduction based on the prediction model. Specifically, the solution unit 23 acquires the storage battery data from the storage battery data DB 14, and acquires the electricity charge from the electricity rate plan data DB 15. Then, the solution unit 23 solves the probability optimization problem for the sampling result of the prediction probability distribution input from the sampling unit 22 using the storage battery data and the electricity charge, and generates the charge / discharge plan 1. , Is output to the sensitivity analysis unit 24.

For example, the solution unit 23 solves an optimization problem that maximizes E [objective function] shown in the equation (4). Here, the objective function in [] of the equation (4) is a function for calculating the electricity rate reduction amount, and indicates the expected value of the electricity rate reduction amount on that day. “X _i ” is a determining variable indicating the amount of discharge at time i, and when it is negative, it indicates the amount of charge. "P _i " indicates the unit price of the electricity charge at time i, and "H" is a long-term forecast period which is a look-ahead number. Further, the if statement of the equation (4) is a reverse power flow condition that the discharge cannot be performed more than the power demand, and "ξ _i " is the power demand at time i.

The constraints for solving the optimization problem of Eq. (4) are shown in Eqs. (5) and (6). Equation (5) shows the restriction that the limit amount that can be charged and discharged in one hour at each time is not exceeded, "U _c " is the limit amount that can be charged in one hour, and "U _dc " can be discharged in one hour. It is the limit amount. Equation (6) indicates the restriction that the stored amount at each time does not exceed the stored capacity, “si” is the stored amount at time _i , and “ _Us ” is the stored capacity. Further, the constraint conditions of the equation (5) and the equation (6) may be collectively described as “Ax ≦ b”.

Here, the optimization problem of the equation (4) cannot be easily solved because of the reverse power flow condition of the if statement. Therefore, we introduce a slack variable "w" called a wait decision that reflects the effect when each sample is realized. Equation (7) is an equation showing the definition of a slack variable. Here, "w" in the equation (7) indicates a sample at the j-th time of the demand forecast by the Monte Carlo method.

Then, the solution unit 23 introduces a slack variable into the equation (4) and executes the formulation by the linear programming method (LP). Equation (8) is an equation that specifically formulates the target optimization problem by introducing a slack variable. As shown in the equation (8), the introduction of the slack variable adds two constraints of the slack variable in addition to the constraint condition of the equation (4). By solving the optimization problem of the equation (8), the solution unit 23 generates a charge / discharge plan 1 indicating the charge amount or the discharge amount at each time i.

The sensitivity analysis unit 24 performs sensitivity analysis on the amount of electricity charge reduction that can be achieved in the demand forecast for each time zone, quantitatively calculates the importance of the power demand forecast for each time zone as a sensitivity coefficient, and reoptimizes it. This is a processing unit that outputs to unit 25. Here, a calculation example of the sensitivity analysis will be described with reference to FIG. FIG. 4 is a diagram illustrating an approach for sensitivity analysis.

As shown in FIG. 4A, the sensitivity coefficient is the rate of change of the optimum value for minute fluctuations of one thing. That is, it is the rate of change in each time zone, such as the fluctuation at time 1 and the fluctuation at time 2. However, the sensitivity coefficient is defined for the uncertainty of one thing, not for a wide probability distribution as shown in FIG. 4 (b). .. Further, as shown in FIG. 4 (c), even if the sensitivity coefficient to the expected value of the electricity charge reduction amount for the minute fluctuation of the average of the demand forecast distribution is simply calculated, it is one averaged one. It becomes a sensitivity coefficient for the demand forecast, and does not become a sensitivity coefficient for the demand forecast in which uncertainty is expressed as a probability distribution.

Therefore, as shown in FIG. 4D, the sensitivity analysis unit 24 uses a large number of random numbers from the power demand forecast distribution in order to express all the uncertainties of the power demand forecast expressed by the probability distribution as a plurality of scenarios. generate.

Then, as shown in FIG. 4 (e), the sensitivity analysis unit 24 slightly changes each scenario for each time, and if the scenario is achieved, the minute change in the amount of electricity charge reduction with respect to the charge / discharge plan. Check the rate (sensitivity coefficient). After that, the sensitivity analysis unit 24 averages these to obtain the degree of influence of the achievable electricity price reduction amount for each time zone.

For example, a case where there are three scenarios will be described as an example. At time 1, the sensitivity analysis unit 24 determines the rate of change in electricity charges when scenario 1 is tentatively realized, the rate of change in electricity charges when scenario 2 is tentatively realized, and the rate of change in electricity charges when scenario 3 is tentatively realized. The average value with the fluctuation rate of is calculated as the degree of influence at time 1 (β ₁ : sensitivity coefficient). Similarly, at time 2, the sensitivity analysis unit 24 determines the fluctuation rate of the electricity rate when scenario 1 is tentatively realized, the fluctuation rate of electricity price when scenario 2 is tentatively realized, and the case where scenario 3 is tentatively realized. The average value with the fluctuation rate of the electricity rate is calculated as the degree of influence at time 2 (β ₂ : sensitivity coefficient). In this way, the sensitivity analysis unit 24 calculates the degree of influence (β _i ) in each time zone.

More specifically, the probability optimization problem shown in Eq. (8) is expressed by Eq. (9). Here, "C" is a constraint condition. "F" is a function representing the amount of electricity charge reduction with "x", which is a vector of the charge / discharge plan to be obtained, and "ξ", which is a random variable vector (demand forecast) according to a multidimensional probability distribution, as variables. ..

Here, equation (10) is obtained by approximating the optimization problem of equation (9) using N demand forecast scenarios. Subsequently, when the charge / discharge plan is set to “x ^* ” and the “x” in the equation (10) is replaced, the equation (11) is obtained.

Subsequently, the sensitivity analysis unit 24 analyzes the expected value of the optimum electricity rate reduction amount. Specifically, the equation (13) is generated by Taylor-expanding the equation (12) in which the prediction is slightly changed from the optimum value with respect to the time of the prediction. When this equation (13) is summarized for each variable component, it can be expressed as equation (14).

As a result, the degree of influence (β _i ) in each time zone can be expressed by the equation (15) from the equation (14). That is, this equation (15) is the degree of influence that can be quantitatively obtained. Therefore, in order to suppress fluctuations in the optimal electricity rate reduction amount, it is sufficient to focus on improving the prediction accuracy at times with a large impact (β _i ). It should be noted that each of the functions of f in the equation (15) is a sensitivity coefficient for the amount of electricity charge reduction when the scenario N is realized.

Returning to FIG. 3, the reoptimization unit 25 is a processing unit that reoptimizes the parameters of the prediction model using the likelihood weighted by the degree of influence (β _i ). Specifically, the reoptimization unit 25 weights the predictive model likelihood function shown in Eq. (3) with a sensitivity coefficient (β _i ) and parameter-fits it (Equation (16)). To generate. In this way, the accuracy of power demand forecasting in important time zones is improved. By calculating the charge / discharge plan for this power demand forecast, the accuracy of the amount of electricity charge reduction that can be achieved is improved. Since the optimization method is the same as that of the optimization unit 21, detailed description thereof will be omitted.

The resampling unit 26 is a processing unit that executes sampling on the prediction model optimized by the reoptimization unit 25 and outputs the sampling result to the resolving unit 27. Since the processing of the re-sampling unit 26 is the same as that of the sampling unit 22, detailed description thereof will be omitted.

The resolving unit 27 is a processing unit that generates an optimized charge / discharge plan 2 by solving a probability optimization problem with respect to the prediction by the prediction model and stores it in the charge / discharge plan DB 16. Since the method for answering the optimization problem is the same as the method described in the solution unit 23, detailed description thereof will be omitted. Further, by solving the optimization problem here, it can be confirmed that the equation (7) is satisfied. That is, since the formulated equation (8) is LP, the optimum solution exists on the boundary of the feasible region. Furthermore, since P _i ≧ 0 is a maximization problem, the slack variable “ ^{wi j} _” tries to take a large value on the boundary. Therefore, the equation (7) is satisfied.

[Processing flow]
FIG. 5 is a flowchart showing the flow of processing. As shown in FIG. 5, when the process is started (S101: Yes), the optimization unit 21 selects an appropriate prediction model (S102) and optimizes the parameters of the prediction model by the maximum likelihood method (S103). ).

Subsequently, the solution unit 23 solves the optimization problem by using the storage battery data, the electricity charge, the sampling of the probability distribution of the prediction executed by the sampling unit 22, and the like, and generates the charge / discharge plan 1 (S104).

After that, the sensitivity analysis unit 24 generates random numbers according to the demand forecast distribution and approximates the demand forecast distribution in a plurality of scenarios (S105). Then, the sensitivity analysis unit 24 calculates a minute volatility (sensitivity coefficient) of the electricity rate reduction amount for each scenario (S106). Subsequently, the sensitivity analysis unit 24 calculates the degree of influence (β _i ) in each time zone (S107).

After that, the reoptimization unit 25 optimizes the parameters of the prediction model with the weighted likelihood (S108). Then, the re-solution unit 27 solves the optimization problem by using the storage battery data, the electricity charge, the sampling of the probability distribution of the prediction executed by the re-sampling unit 26, and the like, and generates the charge / discharge plan 2 (S109).

[effect]
As described above, the information processing apparatus 10 generates a large number of random numbers from the power demand forecast distribution, and represents the uncertainty of the demand forecast as a plurality of scenarios. The information processing apparatus 10 slightly fluctuates each scenario for each time, and examines the sensitivity coefficient of the electricity charge reduction amount with respect to the charge / discharge plan when the scenario is achieved. By averaging these, the degree of influence of the amount of electricity cost reduction that can be achieved for each time zone is calculated. Therefore, the information processing apparatus 10 can detect an important time zone in relation to the reduction of electricity charges.

Then, the information processing apparatus 10 improves the power demand forecast accuracy in an important time zone by fitting the prediction model by weighting it with a sensitivity coefficient. As a result, the information processing apparatus 10 can improve the accuracy of the demand forecast.

(Comparison)
Here, the charge / discharge plans of one day are compared. Specifically, the general charge / discharge plan 1 generated by the solution unit 23 and the charge / discharge plan 2 considering the degree of influence (β _i ) generated by the resolving unit 27 are compared. The horizontal axis of FIGS. 6 and 7, which will be described later, is time, the plus on the vertical axis is the discharge amount, and the minus on the vertical axis is the charge amount.

(Comparison of charge / discharge plans)
FIG. 6 is a diagram showing an example of charge / discharge plan 1 when the prediction parameters are optimized by the maximum likelihood method. As shown in FIG. 6, in the charge / discharge plan 1, the discharge plan that exceeds the predicted average value is planned from around 10:00 to around 25:00 when the electricity rate is high, so that the accuracy of the plan is low. Further, in this charge / discharge plan 1, the mean square error of the prediction of all time zones is 0.649, and the mean square error of the prediction of the time zone of β _i > 0 is 0.759.

On the other hand, FIG. 7 is a diagram showing an example of charge / discharge plan 2 when the prediction parameter is optimized by the weighted likelihood. As shown in FIG. 7, in the charge / discharge plan 2, the discharge plan is planned so that the discharge amount is the same as the predicted average value from around 10:00 to around 25:00 when the electricity charge is high, so that the accuracy of the plan is high. .. Further, in this charge / discharge plan 2, the mean square error of the prediction of all time zones is 0.495, and the mean square error of the prediction of the time zone of β _i > 0 is 0.519.

Therefore, when FIG. 6 and FIG. 7 are compared, the charge / discharge plan 2 has higher planning accuracy, and the mean square error of the prediction in all time zones and the time zone of β _i > 0 is smaller, so that the prediction accuracy is higher. ..

(Comparison of predicted distribution)
Next, the predicted distribution of the amount of electricity charge reduction by the charge / discharge plan of one day will be compared and explained. Specifically, the predicted distribution of the electricity charge reduction amount in the charge / discharge plan 1 and the predicted distribution of the electricity charge reduction amount in the charge / discharge plan 2 considering the degree of influence (β _i ) are compared.

FIG. 8 is a diagram for explaining the predicted distribution of the electricity charge reduction amount according to the charge / discharge plan 1, and FIG. 9 is a diagram for explaining the predicted distribution of the electricity charge reduction amount according to the charge / discharge plan 2. The horizontal axis of FIGS. 8 and 9 shows the amount of electricity charge reduction, and the vertical axis shows the frequency of the predicted distribution.

Comparing FIGS. 8 and 9, the charge / discharge plan 2 has a smaller frequency within the range of "-100 to -50", which cannot reduce the electricity charge, and is within the range of "0 to 100", which can reduce the electricity charge. The frequency is high. In the charge / discharge plan 1, the expected value is "21.27 yen" and the standard deviation is "42.50 yen", and in the charge / discharge plan 2, the expected value is "52.95 yen" and the standard deviation is "43. Since it is "32 yen", the charge / discharge plan 2 can further reduce the electricity charge.

(Comparison from a long-term perspective)
Next, the charge / discharge plans for 150 days are compared. FIG. 10 is a diagram illustrating a mean square error of each charge / discharge plan. As shown in FIG. 10, in the prediction 1 (charge / discharge plan 1), the mean square error of the prediction in all time zones is "0.3363", and the mean square error of the prediction in the time zone of the degree of influence (β _i )> 0. Was "0.3415". On the other hand, in Prediction 2 (Charge / Discharge Plan 2), the mean square error of the prediction in all time zones is "0.3426", and the mean square error of the prediction in the time zone with the degree of influence (β _i )> 0 is “0. It became "3370".

Further, FIG. 11 is a diagram for explaining the evaluation of the predicted distribution of the electricity rate reduction amount. As shown in FIG. 11, in Prediction 1 (Charging / Discharging Plan 1), the average expected value was “87.07 yen” and the average standard deviation was “32.57 yen”. On the other hand, in Prediction 2 (Charging / Discharging Plan 2), the average expected value was "108.35 yen" and the average standard deviation was "31.15 yen".

According to FIG. 10, in the evaluation for 150 days, the mean square error in all time zones is smaller in the prediction 1. Therefore, the charge / discharge plan 1 has higher prediction accuracy. However, in the time zone where the degree of influence is large, the charge / discharge plan 2 has a smaller mean square error, so that the total reduction amount can be expected to be larger. Further, as shown in FIG. 11, the average expected value is higher in the charge / discharge plan 2.

Therefore, in the charge / discharge plan 2, the prediction accuracy for the entire 150 days is slightly worse, but the prediction accuracy for the time zone having a large influence is higher. That is, in the charge / discharge plan 2, the prediction accuracy of the time zone having a low influence in the charge / discharge plan is only lowered, and a significant increase in the amount of electricity charge reduction can be expected.

In the first embodiment, in the optimization of the decision-making plan based on the prediction, the influence (importance) of the prediction probability distribution on the decision-making is obtained by the sensitivity analysis and weighted, and the parameters of the prediction model are fitted again. As a result, we explained an example of improving the accuracy of prediction of important parts and improving the accuracy of decision-making planning based on it.

By the way, the sensitivity analysis of Example 1 analyzes only the influence of the predicted probability distribution on the purpose (objective function value) of the decision-making plan. However, what is actually updated is the parameters of the predictive model. Therefore, in Example 2, an example will be described in which the accuracy of the sensitivity analysis and the prediction accuracy of the important part are improved by analyzing the degree of influence from the parameters of the prediction model on the prediction probability distribution, and the objective function value is improved.

That is, in Example 2, in addition to the degree of influence that each time has on the power demand forecast described in Example 1, the degree of influence that each parameter of the prediction model has on the power demand forecast is also calculated, and these influences are used. The weighted likelihood is used to reoptimize the parameters of the predictive model and then generate a charge / discharge plan.

[overall structure]
FIG. 12 is a diagram illustrating the overall processing of the second embodiment and the first embodiment. As shown in FIG. 12, in the first embodiment, the parameters of the prediction model are optimized by using the maximum likelihood method, Monte Carlo sampling is performed on the prediction model, and the electricity charge reduction amount (charge) is obtained from the sampling result of the prediction probability distribution. Determine the discharge plan 1). Then, by performing a sensitivity analysis on the electricity rate reduction amount, the degree of influence of each time on the electricity rate reduction amount is calculated, and then the parameters of the prediction model based on the weighted likelihood are refitted to perform charge / discharge planning. Generate 2.

On the other hand, in Example 2, in order to improve the prediction accuracy of important time zones, the degree of influence on the prediction of each time from the parameters of the prediction model is also taken into consideration when performing the sensitivity analysis for the amount of electricity charge reduction. By doing so, the parameters of the predictive model are improved. That is, by re-optimizing the parameters of the likelihood function by using the degree of influence from the parameters of the likelihood function of the prediction model on the first demand forecast, the accuracy of the achievable electricity price reduction can be improved. Plan.

[Functional configuration]
FIG. 13 is a functional block diagram showing a functional configuration of the information processing apparatus 10 according to the second embodiment. As shown in FIG. 13, the information processing apparatus 10 includes a communication unit 11 that executes communication, a storage unit 12 that is an example of a storage device that stores programs, data, and the like, a processor, and the like, as in FIG. 3 of the first embodiment. It has a certain control unit 20. Since the difference from the first embodiment is the processing executed by the sensitivity analysis unit 30, the sensitivity analysis unit 30 will be described in the second embodiment.

In the first embodiment, the sensitivity analysis unit 24 performs a sensitivity analysis on the amount of electricity charge reduction that can be achieved in the demand forecast for each time zone, and quantifies the importance of the power demand forecast for each time zone as a sensitivity coefficient (β _i ). Calculated. In the second embodiment, when the sensitivity analysis unit 30 performs the sensitivity analysis, the sensitivity coefficient ( β ⁱ ) is calculated. Since the processing after calculating the sensitivity coefficient is the same processing as in Example 1 using the sensitivity coefficient (β _i ), detailed description thereof will be omitted.

The sensitivity analysis unit 24 defines each of the equations (17). “V _f ” in the equation (17) is each time, for example, v ₁ = 1 o'clock. “Ξ _f ” is a power demand forecast (random variable) at each time, and “u ^* ” is a discharge amount at each time, which is a determinant determined by optimization. “P” is the power demand forecast distribution up to H o'clock, that is, the simultaneous probability density function. "Α" is the frequency of the test distribution at each time, "β" is the importance of the power demand forecast at each time, and "G" is the electricity rate reduction until the H period based on the power demand forecast. It is the sum of the foreheads. The subscript of each variable in the equation (17) is an index representing the time.

Then, the sensitivity analysis unit 24 defines the objective function of the plan optimization as the equation (18) based on the definition of each equation of the equation (17). Here, the sensitivity coefficient (β ⁱ ) to be obtained is given by Eq. (19). Equation (18) is an expected value of the sum of the amount of electricity price reduction up to the H period based on the power demand forecast.

Here, the learning method of the prediction model will be described. FIG. 14 is a diagram illustrating a statistical basis for weighted maximum likelihood estimation. Weighted maximum likelihood estimation is supervised learning under a covariate shift, for example, a type of transfer learning. As shown in the upper figure of FIG. 14, the training distribution which is the distribution of the training data showing the relationship between the temperature and the frequency prepared in advance is set as p _tr (v), and the test distribution p _te (v) which is the distribution of the test data. ) Is set arbitrarily.

The figure below shows the relationship between temperature and electricity demand. In the figure below, the data are plotted according to the training distribution. Of the data to be plotted, the appearance of the data belonging to the training distribution is defined as the equation (a), and the appearance of the data belonging to the test distribution is defined as the equation (b).

Based on this definition, "θ", which summarizes the parameters of power demand forecast, and " _ptr ( ^vi )", which is the frequency of appearance of the training distribution (data frequency distribution) at each time. The relationship can be defined as in Eq. 20). From the equation (20), the relationship of the equation (21) can be obtained, and the equation (22) can be obtained by partially differentiating both sides of the equation (21) with respect to "α ⁱ ". Then, the sensitivity coefficient (β ⁱ ) can be defined by the equations (19) to (22) as in the equation (23).

After that, using the sensitivity coefficient (β ⁱ ) of the formula (23) calculated by the sensitivity analysis unit 24, processing by the reoptimization unit 25, processing by the resampling unit 26, and resolving unit 27, as in the first embodiment. Is executed, and the optimized charge / discharge plan 2 is generated.

As described above, in Example 2, the sensitivity coefficient (β ^{i), which was a heuristic value in Example 1, can be rewritten to the sensitivity coefficient (β i} ₎ , which is an accurate value. FIG. 15 is a diagram illustrating the effect of the second embodiment. FIG. 15 (a) shows the final charge / discharge plan according to the first embodiment, and FIG. 15 (b) shows the final charge / discharge plan according to the second embodiment. Comparing the charge / discharge plan according to the first embodiment and the charge / discharge plan according to the second embodiment, the overall prediction error is larger in the second embodiment. However, in the important time zone from 9:00 to 23:00, that is, the time zone that can contribute to the reduction of electric energy, the error of Example 2 is smaller than that of Example 1. As a result, as for the overall electricity rate reduction amount, the reduction amount is larger in Example 2 than in Example 1.

Therefore, in the second embodiment, it is possible to generate a charge / discharge plan having a larger reduction amount than in the first embodiment, which can generate a charge / discharge plan having a larger reduction amount than the general method. It should be noted that the electricity rate reduction that can be achieved for each time zone by averaging the minute fluctuation rate (sensitivity coefficient) of the electricity rate reduction amount with respect to the charge / discharge plan when the scenario is achieved, which is the method described in Example 1. The method of calculating the degree of influence of the forehead can also be applied to the calculation of the degree of influence on the prediction probability distribution from the parameters of the prediction model in the second embodiment. For example, an average value reflecting the degree of influence on the charge / discharge plan 1 from the parameter of the likelihood function of the minute fluctuation of the random variable vector can be calculated as the degree of influence for each time zone.

By the way, although the examples of the present invention have been described so far, the present invention may be carried out in various different forms other than the above-mentioned examples. Therefore, different embodiments will be described below.

[Scope of application]
In the above embodiment, the forecast of electric power demand has been described as an example, but the present invention is not limited to this, and other demand forecasts having different charges depending on the time zone, such as gas demand forecast, can be applied. It should be noted that the same can be applied not only to the demand forecast in which the charge differs depending on the time zone, but also to the demand forecast in which the charge differs depending on the place, for example. In this case, by replacing the time zone of the first embodiment with a place. , Can be processed in the same way.

[Predictive model]
The methods such as the prediction model and sampling described above are examples, and other known methods can also be used. Further, the method of solving the optimization problem is also an example, and other known methods can also be used.

[system]
In addition, among the processes described in this embodiment, all or part of the processes described as being automatically performed can also be performed manually. Alternatively, all or part of the processing described as being performed manually can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above document and drawings can be arbitrarily changed unless otherwise specified.

Further, each component of each of the illustrated devices is a functional concept, and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of distribution / integration of each device is not limited to the one shown in the figure. That is, all or a part thereof can be functionally or physically distributed / integrated in any unit according to various loads, usage conditions, and the like. Further, each processing function performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.

[Hardware configuration]
FIG. 16 is a diagram illustrating a hardware configuration example of the information processing apparatus 10. As shown in FIG. 16, the information processing apparatus 10 includes a communication interface 10a, an HDD (Hard Disk Drive) 10b, a memory 10c, and a processor 10d.

The communication interface 10a is a network interface card or the like that controls communication of other devices. The HDD 10b is an example of a storage device for storing programs, data, and the like.

Examples of the memory 10c include RAM (Random Access Memory) such as SDRAM (Synchronous Dynamic Random Access Memory), ROM (Read Only Memory), flash memory and the like. Examples of the processor 10d include a CPU (Central Processing Unit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), a PLD (Programmable Logic Device), and the like.

Further, the information processing device 10 operates as an information processing device that executes an estimation method by reading and executing a program. That is, the information processing apparatus 10 has a program that executes the same functions as the optimization unit 21, the sampling unit 22, the solution unit 23, the sensitivity analysis unit 24, the reoptimization unit 25, the resampling unit 26, and the resolution unit 27. Run. As a result, the information processing apparatus 10 executes the same functions as the optimization unit 21, the sampling unit 22, the solution unit 23, the sensitivity analysis unit 24, the reoptimization unit 25, the resampling unit 26, and the resolution unit 27. Can be executed. The program referred to in the other embodiment is not limited to being executed by the information processing apparatus 10. For example, the present invention can be similarly applied when other computers or servers execute programs, or when they execute programs in cooperation with each other.

This program can be distributed over networks such as the Internet. In addition, this program is recorded on a computer-readable recording medium such as a hard disk, flexible disk (FD), CD (Compact Disc) -ROM, MO (Magneto-Optical disk), or DVD (Digital Versatile Disc). It can be executed by reading from a recording medium.

10 Information processing device 11 Communication unit 12 Storage unit 13 Power demand forecast data DB
14 Storage battery data DB
15 Electricity rate plan data DB
16 Charge / discharge plan DB
20 Control unit 21 Optimization unit 22 Sampling unit 23 Solution unit 24, 30 Sensitivity analysis unit 25 Reoptimization unit 26 Resampling unit 27 Resolution unit

Claims (7)

On the computer
The parameters of the likelihood function of the forecast model selected by a predetermined information criterion are optimized for the past demand actual data, and the probability distribution of the demand forecast obtained from the forecast model obtained by the optimization is used. , Estimate the first demand forecast,
Using the first demand forecast, a sensitivity analysis to the achievable electricity price reduction amount of the demand forecast for each time zone is performed to calculate the influence of the demand forecast for each time zone .
It is characterized in that the process of estimating the second demand forecast is executed by using the probability distribution of the demand forecast based on the forecast model obtained by re-optimizing the parameters of the likelihood function weighted by the influence degree. The estimation program. The process of estimating the first demand forecast solves an optimization problem with respect to the forecast by the optimized forecast model by using the past power demand data, the electricity charge for each time zone, and the storage capacity value. Then, estimate the first charge / discharge plan for the electricity demand forecast,
In the calculation process, the sensitivity analysis is executed using the first charge / discharge plan, and the degree of influence of the demand forecast in each time zone is calculated.
The estimation program according to claim 1, wherein the second charge / discharge plan for the electric power demand forecast is estimated by the process of estimating the second demand forecast. In the estimation program according to claim 1, the process of estimating the second demand forecast uses the degree of influence on the first demand forecast from the parameters of the likelihood function weighted by the degree of influence. , An estimation program that reoptimizes the parameters of the likelihood function. In the estimation program according to claim 2, the process of estimating the first demand forecast generates a plurality of random variable vectors that approximate the probability distribution from the probability distribution of the probability function, and for each time zone. Each of the plurality of random variable vectors is slightly varied, and the average value of the minute fluctuations of the plurality of random variable vectors is calculated as the degree of influence for each time zone.
The process of estimating the second demand forecast is an estimation program characterized in that the parameters of the likelihood function to which the average value is added for each time zone are optimized again. In the estimation program according to claim 4, the process of calculating the degree of influence for each time zone is the first step from the parameters of the likelihood function of the minute fluctuations of the plurality of random variable vectors for each time zone. An estimation program characterized in that an average value reflecting the degree of influence on demand forecast is calculated as the degree of influence for each time zone. The computer
The parameters of the likelihood function of the forecast model selected by a predetermined information criterion are optimized for the past demand actual data, and the probability distribution of the demand forecast obtained from the forecast model obtained by the optimization is used. , Estimate the first demand forecast,
Using the first demand forecast, a sensitivity analysis to the achievable electricity price reduction amount of the demand forecast for each time zone is performed to calculate the influence of the demand forecast for each time zone .
It is characterized by executing a process of estimating a second demand forecast by using a probability distribution of a demand forecast based on the forecast model obtained by re-optimizing the parameters of the likelihood function weighted by the degree of influence. The estimation method. The parameters of the likelihood function of the forecast model selected by a predetermined information criterion are optimized for the past demand actual data, and the probability distribution of the demand forecast obtained from the forecast model obtained by the optimization is used. , The first estimation unit that estimates the first demand forecast,
A calculation unit that calculates the degree of influence of the demand forecast for each time zone by performing a sensitivity analysis for the achievable electricity price reduction amount of the demand forecast for each time zone using the first demand forecast.
Using the probability distribution of the demand forecast based on the forecast model obtained by re-optimizing the parameters of the likelihood function weighted by the influence degree, the second estimation unit for estimating the second demand forecast An estimation device characterized by having.

Images