JP2008295193A - Device, system and method for controlling power demand - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device, a system and a method for controlling power demand. <P>SOLUTION: A demand modelling calculation section 14 constitutes a customer model for anticipating a determination action of cooperation by each customer device 12 relative to power demand control from result data showing a result relating to the determination action of cooperation by each customer device 12 relative to a request of power demand control in the past. An optimization calculation section 15 optimizes the evaluation value of an objective function to evaluate the goal achievement status as a system, based on a prediction by each customer model constituted of the demand modelling calculation section 14. From the optimized result, a request for the power demand control to the customer device 12 and a reward to present to the request are determined. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to control of power demand in a power supply system that supplies power to a consumer from a power supply device.

  Electric power demand control for controlling the consumption of electric power by consumers has attracted attention, and some have already been implemented. The mainstream of power demand control implemented so far is introduced to individual customers in order to prevent excess contract power during the daytime in summer.

  On the other hand, with the spread of information home appliances in recent years, home appliances such as air conditioners can be controlled by communication. Therefore, a wide-area system using communication is also proposed in power demand control. Patent Document 1 discloses a wide area power demand control system using broadcast communication. In the system, when the amount of power consumption in a specific area exceeds a predetermined threshold, a power saving request is sent to the consumers in that area in a simultaneous broadcast. Customers who have received a power saving request will make efforts to save power. According to this system, it is expected that power demand in a wide area is efficiently suppressed.

In addition, with the movement of electricity liberalization, small businesses have entered the power business. Small-scale businesses have a greater influence on individual customers than large-scale businesses, and there is an increasing demand for detailed power demand control through mutual cooperation among multiple customers.
JP 2002-345177 A

  As disclosed in Patent Document 1, a method for transmitting a power-saving request by broadcasting the power-saving request to the extent that the entire specific area is roughly grasped and a desired power-saving can be performed based on the prediction is widely used. In a large-scale power supply system that supplies power, the prediction error can be kept small. However, a small-scale power supply system has a large prediction error, which may prevent efficient demand suppression with high control accuracy.

  If the prediction is for a continuous quantity such as power demand, a highly accurate prediction model can be realized relatively easily by applying multiple regression analysis or neural network. However, when predicting how a consumer responds to power demand control, it is difficult to apply these methods because the prediction target is discrete. Moreover, in the prediction of a consumer's response to power demand control, since there is little actual data compared with power demand prediction, it is very difficult to predict with high accuracy.

  An object of the present invention is to provide a system and method capable of improving the control accuracy of power demand.

In order to achieve the above object, a power demand control device of the present invention is a power demand control device that controls power consumption in a plurality of consumer devices including devices that consume power from a power supply device,
Demand that constitutes a consumer model for predicting the cooperative behavior of each consumer device for power demand control from performance data indicating the performance of the cooperative behavior of each consumer device in response to requests for past power demand control A modeling operation unit;
Based on the prediction by each customer model configured by the demand modeling operation unit, optimize the evaluation value of the objective function for evaluating the degree of achievement of the objective as a system, and from the optimization result, An optimization calculation unit that determines a request for power demand control to the consumer device and a reward to be presented in response to the request.

  According to the present invention, the customer model of each customer device is configured from the actual performance data of the cooperation of each customer device, and the evaluation value of the objective function as the system is optimized based on the prediction of each customer model. Therefore, it becomes possible to perform power demand control that captures individual consumers, and the control accuracy of the power demand as a system can be improved.

  The demand modeling operation unit may use a machine learning technique using a support vector machine when configuring the consumer model.

  According to this, it is possible to construct a customer model with excellent generalization ability by applying a support vector machine, which is a machine learning method with excellent identification performance.

  Further, the input to the consumer model includes the power consumption in the consumer device before the power demand control is performed and the reward to be given to the cooperation to the power demand control. Also good.

  The input to the consumer model may include weather conditions or date / time information when power demand control is performed.

  According to this, the customer's cooperation behavior for power demand control may depend on the weather situation and date / time, and in that case, the weather situation and date / time will be used for judgment to improve the prediction accuracy of the cooperation decision behavior. Can do.

  In addition, the optimization calculation unit, as an initial solution, a combination of a consumer device that makes a request for power demand control and a reward that is given to the consumer device given to the customer device in cooperation with the request, An optimum solution may be determined by repeatedly calculating an evaluation value of the objective function and adjusting the solution, and requesting power demand control from the consumer device based on the optimum solution.

  In addition, each customer model may output a predicted result of a cooperative decision action by each customer device as a probability or an expected value, and an evaluation value of the objective function may be obtained as a probability or an expected value.

  According to this, by using the probability or the expected value, it is possible to reduce the risk due to irregularity of the actual data and improve the control accuracy of the power demand.

  In addition, the optimization calculation unit may perform optimization using the power supply cost or the operator profit as an objective function.

  According to the present invention, the accuracy of power demand control of the power supply system can be improved.

  Embodiments for carrying out the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram illustrating a configuration example of the power supply system according to the present embodiment. Referring to FIG. 1, a power demand control device 11, a customer device 12, and a power supply device 13 are included. The power demand control device 11 includes a demand modeling calculation unit 14 and an optimization calculation unit 15. The consumer device 12 is a device group provided by the consumer, and includes a communication device 16, a demand adjustment device 17, and a power demand device 18.

  The power demand control device 11 and the customer device 12 are connected by a communication line 19 such as a communication network, and can communicate with each other. A power line 20 is connected from the power supply device 13 to the customer device 12, and power is supplied through the power line 20.

  The communication device 16 receives the control signal from the power demand control device 11 and presents the cooperation request information and the cooperation reward information included in the control signal to the customer or notifies the demand adjustment device 17. The communication device 16 is composed of, for example, a personal computer, a server, and a mobile phone. The cooperation request information is information for requesting the customer to cooperate in power demand control such as power saving. The cooperation reward information is information for notifying the consumer of the reward given for cooperation in response to the request based on the cooperation request information.

  The power demand device 18 is air conditioning, lighting, and other electric devices that operate by consuming electric power.

  The demand adjustment device 17 is a device that adjusts the power consumption of each power demand device 18. For example, the demand adjusting device 17 automatically adjusts the power consumption of the power demand device 18 based on a control signal from the power demand control device 11 from a device operated manually by a switch, a remote controller, or the like. There are various things to do.

  If the demand adjusting device 17 is manually operated, it is determined whether or not a customer who has received a request for power demand control presenting cooperation request information and cooperation reward information cooperates with the request. When cooperating with the request, the consumer operates the demand adjusting device 17 to adjust the power consumption of the power demand device 18.

  If the demand adjustment device 17 does not require human intervention, the demand adjustment device 17 that has received a request for power demand control including cooperation request information and cooperation reward information from the power demand control device 11 via the communication device 16 is set in advance. Judge whether or not to cooperate with the request in accordance with the judgment criteria that have been established. Judgment criteria are set by a consumer, for example. If it is determined to cooperate, the demand adjusting device 17 adjusts the power consumption of the power demand device 18.

  Whether or not to cooperate in power demand control may be appropriately determined by the customer or the demand adjusting device 17. If a consumer cooperates with power demand control, he / she can receive the reward presented in the cooperation reward information. The reward is, for example, payment of a reward or discount of a basic fee. When a customer receives a request for power demand control, he or she can cooperate with the request as much as possible. Electric power consumption can be controlled without imposing an unreasonable burden on consumers. In addition, it is expected that consumers can raise their awareness of cooperation in improving the efficiency of power use, and cost reduction and efficient use of power will be promoted.

  In the power demand control device 11, the demand modeling operation unit 14 generates a customer model that models each customer's cooperative behavior from past data of past cooperation of each customer in response to a request for power demand control. The consumer model of each consumer is held in the power demand control device 11 as a consumer model group 21. The actual data may be notified from the consumer device 12 to the power demand control device 11 by a control signal in response to the control signal for power demand control, for example, and the demand data measured for billing etc. You may decide from the power consumption in a house. The demand modeling operation unit 14 updates the customer model of the customer model group 21 by machine learning. The response from the customer device 12 to the power demand control by the power demand control device 11 is fed back and reflected in the customer model.

  The optimization calculation unit 15 performs an optimization process using the customer model generated by the demand modeling calculation unit 14, and generates a control signal including cooperation request information and cooperation reward information based on the obtained result. To the consumer device 12. The optimization process is, for example, a process of calculating what reward should be presented to which consumer to request cooperation. An objective function for evaluating the degree of achievement of the objective adapted to the objective of controlling the power demand by the system may be used for the optimization process. For example, the power supply cost may be an objective function, and the objective function value may be minimized. Alternatively, the difference between the sum of the charges collected from each consumer and the power supply cost may be used as an objective function, and the objective function may be maximized. Since the power supply cost depends on the characteristics of the power supply device 13, the optimization calculator 15 stores information on the characteristics of the power supply device 13 in advance, and the information may be used for calculating the power supply cost.

  The power supply device 13 is various devices that supply power to the power demand device 18 of the consumer, and is, for example, a power generation device or a power storage device. There are base load generators and peak load generators as generators for supplying power generated by the generator. Various power supply devices 13 differ in characteristics such as output variability and power generation efficiency. These characteristics affect power supply costs.

  FIG. 2 is a flowchart showing the operation of the power supply system according to the present embodiment. Here, an example is shown in which the difference between the sum of the charges collected from each consumer and the power supply cost (hereinafter referred to as the operator profit) is set as an objective function, and the objective function is maximized.

  Referring to FIG. 2, first, the optimization calculation unit 15 searches for an optimal solution that maximizes the objective function using the customer model of each customer device 12 and each power supply device 13 (step). 101).

  As an example, by inputting various pieces of condition information to the customer model, the predicted change in power consumption is output. Further, the fee collected from the consumer can be calculated from the information on the reward given to the consumer who cooperated with the power demand control and the predicted amount of change in power consumption. The total power supply cost can be calculated from the amount of change in the power consumption of each consumer. The operator profit can be obtained by subtracting the total power supply cost from the sum of the charges collected from each consumer. What is necessary is just to perform an optimization process so that this operator profit may be maximized. In the optimization, for example, which consumer is requested to cooperate in power demand control may be used as a parameter, and the content of remuneration for cooperation in power demand control may also be used as a parameter.

  Returning to FIG. 2, next, the optimization calculation unit 15 determines whether or not the optimal solution obtained in Step 101 can achieve an effect equal to or higher than the reference value (Step 102). The effect can be judged using the value of the operator profit as an index. The reference value may be set in advance. If it is not determined that an effect greater than the reference value can be obtained, the process returns to step 101 and the optimization process is repeated again.

  If it is determined that an effect equal to or greater than the reference value is obtained, the optimization calculation unit 15 generates cooperation request information and cooperation reward information based on the optimal solution, and sends a control signal including such information to the cooperation. It transmits to the consumer apparatus 12 of the consumer of the request | requirement object (step 103).

  Thereafter, the customer modeling operation unit 14 monitors the achievement status of power demand control in each customer who has sent the control signal (step 104). The achievement status of the power demand control can be determined from the amount of change in power consumption corresponding to the power demand control. The amount of change in power consumption may be calculated from the response to the control signal or the change in power consumption.

  Further, in step 104, the customer modeling calculation unit 14 reflects the change in the power consumption of each customer obtained as the achievement status of power demand control in the customer model as actual data, and updates the customer model. (Step 105). Once the customer model is updated, the search for the optimal solution is repeated next time using the updated customer model in step 101.

  The process shown in FIG. 2 may be performed regularly at all times, or may be performed when it is determined that power demand control is necessary based on predetermined conditions. Usually, power demand control is often effective when implemented so as to suppress power consumption at the peak of power consumption during the daytime in summer.

  As described above, various types of power supply devices 13 may be mixed. For example, many power generators that are operated at a base load are relatively highly efficient. On the other hand, for adjustment at peak load, a power generator with good output variability but low power generation efficiency, that is, high power supply cost is often used. Therefore, if the power generator for adjustment is operated with the highest possible power generation efficiency by power demand control or the power generation amount of the power generator for adjustment is suppressed as much as possible, the profit of the operator and the power supply cost in the entire system can be reduced. It can also be optimized.

  FIG. 3 is a diagram illustrating an example of a customer model. Referring to FIG. 3, input information and output information to the customer model 31 are shown.

  In the example of FIG. 3, input information includes weather information 32, date / time information 33, demand information 34, and cooperation reward information 35.

  The weather information 32 is information indicating weather conditions such as weather and temperature when the power demand control is performed. The behavior of the customer's cooperation with respect to the power demand control may depend on the weather situation, and in that case, the prediction accuracy of the behavior of determining the cooperation can be improved by using the weather information for the judgment.

  The date / time information 33 is information such as a day of the week and a time zone when performing power demand control, a holiday flag indicating whether it is a holiday or a weekday. The behavior of customer cooperation for power demand control may depend on the date and time. In this case, the prediction accuracy of the behavior for determining the cooperation can be improved by using the date and time information for judgment.

  The demand information 34 is information indicating the power consumption of the target consumer immediately before the power demand control is performed.

  The cooperation reward information 35 is information indicating a reward given for cooperation to power demand control. For example, the reward amount according to the amount of restraint of power consumption, the discount rate or discount amount of the basic fee, or the metered fee Unit price etc. can be considered. By temporarily reducing the pay-as-you-go fee, fine adjustment is possible when power is surplus.

  In the example of FIG. 3, there is response information 36 as output information.

  The response information 36 is information indicating the response content predicted by the target consumer when receiving power demand control. The response information 36 includes, for example, a prediction result as to whether or not the target customer cooperates in response to the request when receiving power demand control, a probability of cooperation in response, or a change in power consumption due to cooperation. Consists of expected values.

  FIG. 4 is a diagram for explaining an example of a method of classifying unknown data by machine learning of a customer model. Here, an example of machine learning using a support vector machine will be described. In this example, for simplification of explanation, it is further simplified than the customer model shown in FIG. 3, and the input information to the customer model is only the temperature and the cooperation reward amount. The cooperation reward amount indicates a reward amount presented as a reward for cooperation for power demand control. The output from the consumer model is a response determination indicating whether or not the consumer responds to cooperation when an arbitrary temperature and cooperation reward amount are input.

  There are two types of actual data. In FIG. 4, ○ (white circle) indicates actual data from which cooperation response was obtained from the customer, and × (X) indicates no response from customer. The actual data is shown.

  The demand modeling operation unit 14 calculates a boundary (hyperplane: curve in FIG. 4) for classifying two types of performance data by machine learning using a support vector machine. At that time, the demand modeling operation unit 14 uses, as a boundary, a hyperplane that maximizes the distance (margin) from the closest sample, as shown in FIG. 4, from the result data given as a learning sample. To do. In FIG. 4, a solid line or broken line arrow indicates a margin.

  As shown in FIG. 4, a two-dimensional feature space focusing only on the temperature does not form a simple boundary in relation to the air conditioner operation status, but can be similarly considered in a generalized multi-dimension. In the case of multi-dimensions, the two-dimensional example shown in FIG. The parameters may be determined.

  In this embodiment, the number of sample data is small, and the purpose of machine learning is not to classify sample data well, but to generalize so that unknown data can be classified based on known sample data. In this embodiment, a support vector machine is applied so that unknown data can be classified based on known sample data. The support vector machine is a machine learning technique with excellent discrimination performance and can be applied to regression analysis. Regression analysis using various prediction models is widely used. In the present embodiment, by applying a support vector machine to the regression analysis, it is possible to construct a customer model with excellent generalization ability.

  Consumers can sometimes make unusual responses that go against past responses. Therefore, it can be considered that irregular data is included in the sample of the performance data. For example, comfort is given priority over power saving during important visitors or special events, and air conditioning is operated.

  In order to deal with such a case with less risk, a customer model that outputs a probability or an expected value may be constructed. When a support vector machine is used, it can be output with probability by applying the output to a sigmoid function. Moreover, the expected value of the amount of change in power consumption for power demand control can be calculated by multiplying the predetermined power consumption value calculated from the contract power and the previous power consumption by this probability.

  In this embodiment, an example in which regression analysis and a support vector machine are used has been described, but the present invention is not limited to this. As another example, a customer model may be configured using a neural network, and unknown data may be classified by machine learning using a steepest descent method or a back propagation law. Neural networks have nonlinear characteristics and are effective for predicting models of complex data.

  Moreover, in this embodiment, it does not distinguish between the consumer who adjusts the power consumption of the power demand device 18 manually and the consumer whose demand adjustment device 17 adjusts the power consumption of the power demand device 18 without manual intervention. However, the present invention is not limited to this. As another example, we distinguish between consumers who manually adjust power consumption and those who automatically adjust power consumption, and apply an appropriate method for building a customer model to each. A simple machine learning method may be applied.

  FIG. 5 is a flowchart showing optimization processing in the present embodiment. Here, it is determined to which consumer what kind of reward should be presented to request cooperation.

  First, the optimization calculation unit 15 determines the cooperation reward amount that optimizes the evaluation value, assuming that the control signal for requesting the same power demand control is sent to all customers (step 201). The evaluation value is, for example, power supply cost or business profit. Since this calculation has few decision variables, it can be calculated relatively easily using the learned customer model group 21.

  Next, the optimization calculation unit 15 determines a combination of a customer expected to respond to the cooperation request information and the cooperation reward information as an initial solution (step 202). At this time, one initial solution may be determined or a plurality of initial solutions may be determined.

  Next, the optimization calculation unit 15 applies the solution to calculate the evaluation value of the objective function (step 203). Furthermore, the optimization calculation unit 15 determines whether or not an end condition has been reached (step 204).

  If the end condition has not been reached, the optimization calculation unit 15 adjusts the solution using an optimization metaheuristic technique (step 205), and returns to step 203. On the other hand, if the end condition has been reached, the optimization calculation unit 15 determines an optimal solution from the results of the processing so far (step 206).

  In the present embodiment, a solution to be evaluated next is determined using an optimization method. As a method for solving the combination optimization problem as in the present embodiment, the metaheuristic technique exemplified in the present embodiment is suitable. The metaheuristic method is a method that makes it possible to obtain a highly accurate approximate solution of a global optimal solution within a limited time, and representative examples include a tabu search and a genetic algorithm.

  The combination is changed little by little by adjusting the cooperation reward amount for each customer using the optimization method or changing the customer as the target of power demand control. The generation of a new solution and the evaluation of the solution are repeated until the termination condition is satisfied. The termination condition may be defined by, for example, the number of repetitions and the repetition time. When the termination condition is satisfied, the solution with the best evaluation value may be selected as the optimum solution from the solutions evaluated so far. In the iterative process, past search history and information on a plurality of search solutions may be used in an appropriate manner according to the nature of the optimization method used. By appropriately using these pieces of information, a more accurate approximate solution can be obtained within a limited time.

  In the present embodiment, the example in which the initial solution is determined as requesting power demand control to a consumer who is expected to cooperate in step 202 is shown, but the present invention is not limited to this. As another example, the initial solution may be determined by requesting power demand control from a consumer whose possibility of cooperating with the request for power demand control is a predetermined value or more.

  When calculating the evaluation value in step 203, the power supply cost may be used as the objective function, or the operator profit may be used as the objective function. The expected value of power supply cost or business profit may be used as the evaluation value.

  As described above, according to the present embodiment, the customer model of each customer is configured from the actual performance data of each customer's cooperation, and the objective function as a system is based on the prediction of each customer model. Since the evaluation value is optimized, it is possible to perform power demand control that captures individual consumers, and to improve the control accuracy of the power demand as a system.

  Further, in this embodiment, by setting the output of the customer model and the evaluation value of the objective function as the probability or expected value, it is possible to reduce the risk due to irregularity of the actual data and improve the control accuracy of the power demand. it can.

  In addition, in the present embodiment, detailed power demand control is performed by capturing individual consumers, so that power demand control can be appropriately performed not only at the time of power peak but also at other times of power surplus. . For example, it is possible to enable efficient operation of the power generator for adjustment.

  In the description of the present embodiment, when determining the initial optimization solution as shown in FIG. 5, it is assumed that the same power demand control is requested from all customers, and the cooperation reward amount that optimizes the evaluation value is determined. In addition, a combination of the cooperation fee amount and a customer who is expected to respond to the cooperation request information is determined as an initial solution. However, the present invention is not limited to this example, and the initial solution can be determined by various other methods. As another example, the cooperation reward amount of the initial solution may be determined on the assumption that the same power demand control is requested to some predetermined customers. Further, the cooperation reward amount of the initial solution may be determined on the assumption that different power demand control is requested for each consumer. Further, the initial solution may be one or plural. Moreover, you may decide to use the past optimal solution as an initial solution.

It is a block diagram which shows the structural example of the electric power supply system by this embodiment. It is a flowchart which shows operation | movement of the electric power supply system by this embodiment. It is a figure which shows an example of a customer model. It is a figure for demonstrating an example of the method of classifying unknown data by machine learning of a customer model. It is a flowchart which shows the optimization process in this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Power demand control apparatus 12 Consumer apparatus 13 Power supply apparatus 14 Demand modeling calculation part 15 Optimization calculation part 16 Communication apparatus 17 Demand adjustment apparatus 18 Power demand apparatus 19 Communication line 20 Power line 21 Customer model group 31 Customer model 32 Weather information 33 Date and time information 34 Demand information 35 Cooperation reward information 36 Response information

Claims (9)

A power demand control device that controls power consumption in a plurality of consumer devices including devices that consume power from a power supply device,
Demand that constitutes a consumer model for predicting the cooperative behavior of each consumer device for power demand control from performance data indicating the performance of the cooperative behavior of each consumer device in response to requests for past power demand control A modeling operation unit;
Based on the prediction by each customer model configured by the demand modeling operation unit, optimize the evaluation value of the objective function for evaluating the degree of achievement of the objective as a system, and from the optimization result, A power demand control device comprising: a power demand control request to a consumer device; and an optimization calculation unit that determines a reward presented for the request.   The power demand control apparatus according to claim 1, wherein the demand modeling calculation unit uses a machine learning method using a support vector machine when configuring the consumer model.   The input to the consumer model includes a power consumption in the consumer device before executing power demand control and a reward to be given to cooperation for the power demand control. The power demand control apparatus according to 1 or 2.   The power demand control apparatus according to any one of claims 1 to 3, wherein the input to the consumer model includes weather conditions or date information when power demand control is performed.   The optimization calculation unit uses a combination of a consumer device that makes a request for power demand control and a reward given to the customer device in cooperation with the customer device for the request as an initial solution, and the objective function 5. The method according to claim 1, wherein an optimal solution is determined by repeating calculation of an evaluation value and adjustment of the solution, and electric power demand control is requested from the consumer device based on the optimal solution. Power demand control device.   6. Each of the customer models outputs a predicted result of a cooperative decision action by each of the customer devices as a probability or an expected value, and the evaluation value of the objective function is obtained as a probability or an expected value. The power demand control apparatus according to claim 1.   The power demand control apparatus according to any one of claims 1 to 6, wherein the optimization calculation unit performs optimization using a power supply cost or a business profit as an objective function. The power demand control device according to any one of claims 1 to 7,
Upon receiving a request for power demand control, the cooperation for the request for power demand control is determined based on a determination in consideration of the reward presented for the cooperation to the request, and the power consumption in the device is determined according to the determination result. Adjusting a plurality of consumer devices, and having a power demand control system. A power demand control method for controlling power consumption in a plurality of consumer devices including equipment that consumes power from a power supply device,
Constructing a consumer model for predicting the cooperative behavior of each consumer device for power demand control from the performance data showing the performance of the cooperative behavior by each consumer device for the past demand demand control,
Based on the prediction by each customer model, the evaluation value of the objective function for evaluating the degree of achievement of the objective as the system is optimized,
A power demand control method for determining a request for power demand control to the consumer device and a reward to be presented for the request from a result of optimization.

Images