Power consumption amount prediction device and method

Abstract

This power consumption amount prediction device comprises: an acquisition unit that acquires an actual value of a power consumption amount of each of a plurality of consumers and weather information in a range including a location of each of the plurality of consumers, the weather information including at least information on an amount of solar radiation; an analysis unit that analyzes, for each of the consumers and as a capacity of a solar power generation facility, a coefficient of a term of the amount of solar radiation in a regression equation using, among the information acquired by the acquisition unit, at least the amount of solar radiation as an explanatory variable and the actual value of the power consumption amount as an objective variable; and a prediction unit that inputs the weather information including at least the amount of solar radiation and the capacity of the solar power generation facility to a prediction model trained in advance so as to predict the power consumption amount, predicts a future power consumption amount for the plurality of consumers, and outputs a result of the prediction.

Description

Power Consumption Prediction Device and Method

[0001] The disclosed technologies relate to a power consumption prediction device and a power consumption prediction method.

[0002] Retail electricity providers create demand plans by forecasting the amount of electricity consumed by their customers, and then procure and adjust power sources based on these plans. Traditionally, electricity consumption forecasts have used customer attributes such as contracted power and pricing plans, calendar information such as time and day of the week, and weather information such as temperature and humidity. However, the situation has changed dramatically with the expansion of solar power generation.

[0003] When solar power generation equipment is installed on a customer's building or property, the customer consumes the electricity generated by the solar power equipment themselves. Therefore, solar radiation, which has a very strong correlation with the amount of electricity generated by the solar power equipment, has become an indispensable explanatory variable for predicting electricity consumption. Smart meters observe that for customers who own solar power generation equipment, electricity consumption decreases when the intensity of solar radiation is high and increases when the intensity of solar radiation is low. On the other hand, even for customers who do not own solar power generation equipment, slight changes in electricity consumption due to the intensity of solar radiation are observed. During severe summer weather conditions, electricity consumption tends to increase more when the intensity of solar radiation is high, while during severe winter weather conditions, there is a tendency for electricity consumption to decrease slightly when the intensity of solar radiation is high.

[0004] As described above, the intensity of solar radiation and the status of solar power generation equipment ownership affect electricity consumption, so these factors need to be considered to reduce errors in demand planning. Solar radiation data can be obtained from weather information provided by weather companies, but information on the status of solar power generation equipment ownership is only available to some customers, or may not be available at all. Therefore, there are technologies that analyze the output of solar power generation equipment at individual customers based on past electricity consumption and weather information.

[0005] The technology described in Non-Patent Document 1 analyzes the supply capacity of solar power generation using three types of data for individual consumers: ownership status of solar power generation equipment, power consumption, and solar radiation. In this technology, the first step is to divide consumers into those who own solar power generation equipment and those who do not. In the second step, the power consumption patterns of consumers who do not own solar power generation equipment are clustered, and based on these patterns, the power consumption patterns (consumption components) of consumers who own solar power generation equipment are analyzed. Subsequently, in the third step, the power generation component is extracted from the power consumption and consumption components of consumers who own solar power generation equipment, and the capacity of the solar power generation equipment is analyzed by analyzing its relationship with solar radiation.

[0006] Furthermore, the technology described in Non-Patent Document 2 analyzes the amount of solar power generated using two types of data: the ownership status of solar power generation equipment by individual consumers and the amount of electricity consumed. In this technology, as with the technology in Non-Patent Document 1, the first step is to divide consumers into those who own solar power generation equipment and those who do not. In the second step, assuming that the unique electricity consumption patterns are similar regardless of whether or not solar power generation equipment is owned, the power generation component is extracted from the electricity consumption.

[0007] Chung Ming Cheung, Wen Zhong, Chuanxiu Xiong, Ajitesh Srivastava, Rajgopal Kannan, Viktor K. Prasanna, "Behind-the-Meter Solar Generation Disaggregation using Consumer Mixture Models", 2018 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, October 2018, DOI:10.1109 / SmartGridComm.2018.8587539.Fankun Bu, Rui Cheng, Zhaoyu Wang, "A Two-Layer Approach for Estimating Behind-the-Meter PV Generation Using Smart Meter Data", IEEE Transactions on Power Systems, Volume: 38, Issue: 1, January 2023, DOI:10.1109 / TPWRS.2022. 3164010.

[0008] As described above, there are technologies that analyze the unique power consumption and solar power generation of individual customers who own solar power generation equipment, based on their ownership status and power consumption. In addition, there are technologies that analyze the equipment capacity of customers who own solar power generation equipment using solar radiation data.

[0009] However, conventional technologies are effective only when some or all of the ownership status of solar power generation facilities is known, and cannot be applied when the entire ownership status is unknown. Furthermore, regardless of the ownership status of solar power generation facilities, the inherent amount of electricity consumed is essentially determined by demand behavior that changes depending on factors such as temperature, humidity, time of day, and calendar information such as weekdays or holidays, resulting in different load curves for individual consumers. Therefore, it is difficult to accurately analyze the capacity of solar power generation facilities using methods that match specific patterns without considering these factors. If the capacity of solar power generation facilities cannot be accurately analyzed, the error in predicting electricity consumption will also increase.

[0010] The disclosed technology was developed in light of the above points, and aims to predict future electricity consumption with low error, even when the ownership status and capacity of solar power generation facilities are completely unknown.

[0011] A first aspect of this disclosure is a power consumption forecasting device, comprising: an acquisition unit that acquires actual power consumption values ​​for each of a plurality of consumers and weather information including at least solar radiation information for a range including the location of each of the plurality of consumers; an analysis unit that analyzes the information acquired by the acquisition unit for each of the consumers by using the coefficient of the solar radiation term in a regression equation in which at least the solar radiation is the explanatory variable and the actual power consumption value is the dependent variable as the capacity of the photovoltaic power generation equipment; and a prediction unit that inputs the weather information including at least solar radiation and the capacity of the photovoltaic power generation equipment into a prediction model that has been pre-trained to predict power consumption, predicts future power consumption for the plurality of consumers, and outputs a prediction result.

[0012] A second aspect of this disclosure is a method for predicting electricity consumption, wherein a computer performs the following steps: acquires actual electricity consumption values ​​for each of a plurality of consumers and weather information for a range including the location of each of the plurality of consumers, including at least solar radiation information; analyzes the acquired information for each of the consumers using at least the solar radiation as an explanatory variable and the actual electricity consumption value as the dependent variable in a regression equation, using the coefficient of the solar radiation term as the capacity of the solar power generation equipment; inputs the weather information including at least solar radiation and the capacity of the solar power generation equipment into a prediction model that has been pre-trained to predict electricity consumption, predicts future electricity consumption for the plurality of consumers, and outputs the prediction results.

[0013] According to the disclosed technology, future electricity consumption can be predicted with low error even when the ownership status and capacity of solar power generation facilities are completely unknown.

[0014] This is a block diagram showing the hardware configuration of the power consumption forecasting device. This is a functional block diagram of the power consumption forecasting device. This is a diagram showing an example of the customer attribute database. This is a diagram showing an example of the power consumption actuals database. This is a diagram showing an example of the weather information database. This is a diagram to explain the processing of the analysis unit. This is a diagram showing an example of the PV capacity database. This is a diagram to explain the processing of the construction unit. This is a diagram to explain the processing of the forecasting unit. This is a flowchart showing the flow of the construction process. This is a flowchart showing the flow of the analysis process. This is a flowchart showing the flow of the forecasting process.

[0015] An example of an embodiment of the disclosed technology will be described below with reference to the drawings. In each drawing, identical or equivalent components and parts are given the same reference numerals. Furthermore, the dimensional ratios in the drawings are exaggerated for illustrative purposes and may differ from actual ratios.

[0016] Figure 1 is a block diagram showing the hardware configuration of the power consumption prediction device 10 according to this embodiment. As shown in Figure 1, the power consumption prediction device 10 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, storage 14, an input unit 15, a display unit 16, and a communication I / F (Interface) 17. Each component is connected to the others via a bus 19 so as to be able to communicate with each other.

[0017] The CPU 11 is a central processing unit that executes various programs and controls various components. Specifically, the CPU 11 reads a program from the ROM 12 or storage 14 and executes the program using the RAM 13 as a working area. The CPU 11 controls each of the above components and performs various calculations according to the program stored in the ROM 12 or storage 14. In this embodiment, the ROM 12 or storage 14 stores a power consumption prediction program, which will be described later.

[0018] ROM 12 stores various programs and data. RAM 13 temporarily stores programs or data as a working area. Storage 14 consists of storage devices such as HDD (Hard Disk Drive) and SSD (Solid State Drive) and stores various programs and data, including the operating system.

[0019] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used for various types of input. The display unit 16 is, for example, a liquid crystal display and displays various types of information. The display unit 16 may also function as the input unit 15 by employing a touch panel system.

[0020] Communication I / F17 is an interface for communicating with other devices. For this communication, wired communication standards such as Ethernet® and FDDI, or wireless communication standards such as 4G, 5G, and Wi-Fi® may be used.

[0021] Next, the functional configuration of the power consumption prediction device 10 will be described. Figure 2 is a functional block diagram of the power consumption prediction device 10. As shown in Figure 2, the power consumption prediction device 10 includes, as a functional configuration, an acquisition unit 22, an analysis unit 24, a construction unit 26, and a prediction unit 28. In addition, a predetermined storage area of ​​the power consumption prediction device 10 stores a photovoltaic (PV) capacity DB (database) 36 and a prediction model 38. Each functional configuration is realized by the CPU 11 reading the power consumption prediction program stored in the ROM 12 or storage 14, expanding it in the RAM 13, and executing it.

[0022] The acquisition unit 22 acquires attribute information for each of the multiple customers (hereinafter referred to as "customer attributes") from the customer attribute DB 30. Figure 3 schematically shows an example of the customer attribute DB 30. In the example in Figure 3, the customer attribute DB 30 contains information about customer attributes such as "customer number," "contracted power," "location," and "regional mesh code." The "customer number" is an identification number that identifies the customer. The "contracted power" is a value that indicates the customer's contracted power and capacity. The "location" is a value that identifies the customer's location, for example, latitude and longitude. The "regional mesh code" is information that identifies the mesh containing the location indicated by the "location" when the geographical area is divided into meshes.

[0023] In addition to the information mentioned above, the Customer Attribute DB30 may also include information such as "Rate Menu," "Demand Cluster," "Industry," and "Self-Generated Supplemental Power." "Rate Menu" is a value that indicates the rate menu to which the customer has contracted, from among rate menus that differ depending on whether it is for lighting or power, standard or holiday high-operating hours, etc. "Demand Cluster" is a value that identifies the cluster to which the customer belongs when classified based on load curves analyzed from past power consumption. "Industry" is a value that indicates the industry of the corporation if the customer is a corporation. "Self-Generated Supplemental Power" is a value that indicates the status of ownership of self-generation equipment and the details of the supply contract.

[0024] Furthermore, the acquisition unit 22 acquires the actual power consumption values ​​for each of the multiple consumers from the power consumption data database 32. Figure 4 schematically shows an example of the power consumption data database 32. In the example in Figure 4, the power consumption data database 32 contains information such as "consumer number," "date and time," and "actual power consumption value." "Date and time" is the date and time when the power was consumed, and is the start date and time of a 30-minute time interval. "Actual power consumption value" is the actual amount of electricity consumed [kWh] in the 30-minute time interval specified by "date and time."

[0025] Furthermore, the acquisition unit 22 acquires weather information from the weather information DB 34 that covers a range including the locations of each of the multiple consumers and includes at least solar radiation information. Figure 5 schematically shows an example of the weather information DB 34. In the example in Figure 5, the weather information DB 34 includes information such as "location," "date and time," "temperature," "humidity," and "solar radiation." "Location" is a value that identifies the observation point of the weather information, for example, latitude and longitude. "Date and time" is the date and time of the observation target of the weather information, and is the start date and time of a 30-minute time interval. "Temperature" is the average value [°C] of the temperature in the 30-minute time interval specified by "date and time." "Humidity" is the average value [%] of the relative humidity in the 30-minute time interval specified by "date and time." "Solar radiation" is the total value [kWh / m²] of solar radiation in the 30-minute time interval specified by "date and time." ２ ] (or "W / m ２ "J / m ２ )

[0026] In addition to the above-mentioned weather information, the weather database DB34 may also include information such as "precipitation," "snowfall," and "weather." "Precipitation" is the total amount of precipitation [mm] in a 30-minute time unit specified by "Date and Time." "Snowfall" is the total amount of snowfall [cm] in a 30-minute time unit specified by "Date and Time." "Weather" is a label indicating the weather, such as clear skies, partly cloudy, or rainy, in a 30-minute time unit specified by "Date and Time."

[0027] The analysis unit 24 analyzes each customer using the solar radiation term in a regression equation, which uses the solar radiation amount (at least included in the weather information) from the information acquired by the acquisition unit 22 as the explanatory variable and the actual power consumption value as the dependent variable, as the PV capacity.

[0028] Referring to Figure 6, the processing of the analysis unit 24 will be explained in more detail. As shown in A of Figure 6, the analysis unit 24 adds calendar information to the actual power consumption values ​​for each customer acquired by the acquisition unit 22. The calendar information associates the year, month, and day information with the day of the week for that date, and holiday information if that date falls on a public holiday. The calendar information also includes hour and minute information and includes a frame number that indicates which frame the time period indicated by the year, month, day, hour, and minute corresponds to when a day is represented by a predetermined number of frames divided into unit time units. For example, if 30 minutes is considered one frame, the frame numbers are assigned as follows: 30 minutes from the start time 0:00 is 1, 30 minutes from the start time 0:30 is 2, 30 minutes from the start time 1:00 is 3, ... 30 minutes from the start time 23:30 is 48. The analysis unit 24 associates the date and time in the power consumption data database 32 with the date and time in the calendar information, thereby adding information such as the day of the week, holidays, and time slot number to the power consumption data.

[0029] As shown in Figure 6B, the analysis unit 24 extracts the actual power consumption values ​​for a target date and time based on calendar information. The conditions for the target date and time are specified in advance by the user. For example, the analysis unit 24 extracts the actual power consumption values ​​for daytime hours on weekdays, excluding Saturdays, Sundays, and public holidays.

[0030] As shown in Figure 6C, the analysis unit 24 extracts weather information from the acquisition unit 22 for the same date and time as the date and time from which the actual power consumption value was extracted. The analysis unit 24 then combines the extracted actual power consumption value and weather information for each customer, associating them based on the date and time. The analysis unit 24 combines the data into a format such as (customer number, date and time, frame number, actual power consumption value, temperature, humidity, solar radiation). In this example, temperature and humidity are also used as weather information in addition to solar radiation.

[0031] As shown in D of FIG. 6, the analysis unit 24 searches for a parameter group of an arbitrary polynomial by regression analysis, and acquires a parameter corresponding to the amount of solar radiation as the PV capacity. For example, the analysis unit 24 uses the frame number (F of the combined data described above ｔ ), the actual power consumption value (D ｔ ), the temperature (A ｔ ), the humidity (H ｔ ), and the amount of solar radiation (R ｔ ), to define the polynomial shown in the following equation (1).

[0032]

[0033] The analysis unit 24 uses the combined data for each consumer to search for the optimal solution of the parameter (coefficient) a for which the polynomial of equation (1) holds, by regression analysis. １ to a ７ Among the searched parameter groups, the analysis unit 24 acquires the value of the parameter a corresponding to the term of the solar radiation amount as the PV capacity. The analysis unit 24 associates the PV capacity acquired for each consumer with the consumer number, and stores the PV capacity in the PV capacity DB 36. FIG. 7 schematically shows an example of the PV capacity DB 36. ４

[0034] Note that, as described above, it is not essential to extract the actual power consumption value and the weather information at the target date and time based on the calendar information. However, by selecting a daytime time zone (for example, 10:00 to 14:00) in which the intensity of the solar radiation amount increases and the power generation amount of the photovoltaic power generation increases, a good parameter a ４ value can be acquired.

[0035] The construction unit 26 constructs a prediction model 38 for predicting future power consumption when weather information including at least the amount of solar radiation and the PV capacity are input.

[0036] ​The processing of the construction unit 26 will be described more specifically with reference to FIG. 8. As shown by E in FIG. 8, the construction unit 26 reads information on the PV capacity for each consumer from the PV capacity DB 36, and associates and combines the consumer attributes and the actual power consumption values acquired by the acquisition unit 22 with the consumer numbers as a reference. The construction unit 26 combines the data into a data structure such as (consumer number, regional mesh code, rate menu, PV capacity, date and time, actual power consumption value).

[0037] Further, as shown by F in FIG. 8, the construction unit 26 aggregates the combined data to create demand statistical information in which values such as the actual power consumption value are smoothed, in order to improve the learning efficiency when training the prediction model 38 using the created training data. At this time, the aggregation method is configured so as not to impair the characteristics of the load curve of each individual consumer. Specifically, the construction unit 26 divides consumers into groups based on the consumer attributes and the date and time, and aggregates numerical items such as the actual power consumption value, contracted power, and PV capacity of the consumers included in each group by taking an average or the like.

[0038] The grouping of consumers may be performed, for example, by three elements of a regional mesh code, a rate menu, and date and time, or by three elements of a regional mesh code, a demand cluster, and date and time. In addition, when performing prediction by ownership status of photovoltaic power generation, the classification may be performed by four elements of a regional mesh code, a demand cluster, ownership status, and date and time, or a combination thereof. The aggregated values are the contracted power included in the consumer attributes, the power consumption included in the actual power consumption, the photovoltaic power generation facility capacity output by the photovoltaic power generation facility capacity analysis unit, and the like.

[0039] As shown by G in FIG. 8, the construction unit 26 associates the demand statistical information with the weather information acquired by the acquisition unit 22 based on the date and time, combines the data, and creates training data. The construction unit 26 creates training data having a structure such as (consumer group number, regional mesh code, rate menu, aggregated value of PV capacity, date and time, aggregated value of actual power consumption value). The consumer group number is an identification number for identifying each group of consumers divided into groups when creating the demand statistical information.

[0040] As shown in Figure 8H, the construction unit 26 trains the prediction model 38 using the created training data. Specifically, the construction unit 26 uses at least the aggregate values ​​of solar radiation and PV capacity from the training data as explanatory variables, and the aggregate values ​​of actual power consumption as the target variable. The construction unit 26 then constructs the prediction model 38 by updating the parameters of the prediction model 38 so that the power consumption prediction result output when the explanatory variables are input to the prediction model 38 approaches the target variable. The construction unit 26 stores the parameters of the constructed prediction model 38 in a predetermined memory area.

[0041] The load factor derived from equation (2) below may be used as the dependent variable. Load factor (L ｔ ) represents the contracted power converted to kW (P) and the actual value of electricity consumption (D). ｔ It is calculated by the following. Contracted power is a fixed value allocated to each individual customer, and if there is actual electricity consumption data for one year or more, it may be converted from the actual electricity consumption data.

[0042]

[0043] Machine learning algorithms such as LightGBM and RandomForest, or deep learning neural networks, may be used to build the model. Furthermore, lag variables created from demand statistics and weather data may be incorporated into the training data. For example, since actual power consumption figures may exhibit seasonal fluctuations, incorporating lag variables into the training data can improve the stability of predictions. Lag variables may include historical load rates per time frame, daily statistics of load rates, temperature sensitivity, etc.

[0044] The prediction unit 28 inputs weather information, including at least solar radiation, and PV capacity into the prediction model 38 to predict future power consumption for multiple consumers and outputs the prediction results.

[0045] Referring to Figure 9, the processing of the forecasting unit 28 will be explained in more detail. As shown in I to K of Figure 9, the forecasting unit 28 creates forecast data using the same processing as the construction unit 26 shown in E to G of Figure 8. However, since the forecasting unit 28 forecasts future power consumption, it uses customer attributes that reflect contract forecasts. For example, if the business plan includes new contracts or cancellations of customers, the forecasting unit 28 selects the target customer attributes based on that business plan. The forecasting unit 28 also selects past power consumption data to match the selected customer attributes. Furthermore, the forecasting unit 28 uses weather forecasts for meteorological information.

[0046] As shown in L in Figure 9, the prediction unit 28 inputs the created prediction data into the prediction model 38 and obtains the prediction results of future power consumption output from the prediction model 38. The prediction unit 28 stores the prediction results obtained for each customer group in the power consumption prediction result DB 40. The data structure of the power consumption prediction result DB 40 is the same as that of the power consumption actual DB 32. However, the customer number is replaced with the customer group number or the customer attribute corresponding to the customer group indicated by that customer group number, and the power consumption actual value is replaced with the prediction result. When all predictions are completed, the prediction unit 28 outputs the power consumption prediction result DB 40.

[0047] Next, the operation of the power consumption prediction device 10 will be explained. Figure 10 is a flowchart showing the flow of the construction process by the power consumption prediction device 10. Figure 12 is a flowchart showing the flow of the prediction process by the power consumption prediction device 10. The CPU 11 reads the power consumption prediction program from the ROM 12 or storage 14, expands it into the RAM 13, and executes it, thereby performing the construction process and the prediction process. Note that the prediction process is an example of the power consumption prediction method of this disclosure.

[0048] First, let's explain the construction process shown in Figure 10.

[0049] In step S10, the CPU 11, acting as an acquisition unit 22, acquires the customer attributes of each of the multiple customers from the customer attribute DB 30. The CPU 11, acting as an acquisition unit 22, also acquires the actual power consumption values ​​of each of the multiple customers from the power consumption actual DB 32. The CPU 11, acting as an acquisition unit 22, also acquires weather information for the area including the location of each of the multiple customers from the weather information DB 34.

[0050] Next, in step S20, the analysis process is executed. The analysis process will now be explained with reference to Figure 11.

[0051] In step S22, the CPU 11, acting as an analysis unit 24, adds calendar information to the actual power consumption values ​​for each customer acquired by the acquisition unit 22. Then, the CPU 11, acting as an analysis unit 24, extracts the actual power consumption values ​​for the target date and time from the actual power consumption values ​​acquired in step S10, based on the calendar information.

[0052] Next, in step S24, the CPU 11, acting as an analysis unit 24, extracts weather information from the weather information acquired in step S10 for the same target date and time as the target date and time from which the actual power consumption values ​​were extracted. Then, the CPU 11, acting as an analysis unit 24, combines the extracted actual power consumption values ​​and weather information for each customer, associating them with the date and time.

[0053] Next, in step S26, the CPU 11, acting as the analysis unit 24, uses the combined data for each customer to perform regression analysis to determine the parameters (coefficients) such that a polynomial like the one shown in equation (1) holds true. １ ~a ７ We search for the optimal solution.

[0054] Next, in step S28, the CPU 11, acting as the analysis unit 24, acquires the value of the parameter corresponding to the solar radiation term from the searched parameter group as the PV capacity. The CPU 11, acting as the analysis unit 24, stores the acquired PV capacity for each customer in the PV capacity DB 36, associating it with the customer number. Then the analysis process is completed, and the system returns to the construction process (Figure 10).

[0055] Next, in step S30, the CPU 11, acting as a construction unit 26, reads PV capacity information for each customer from the PV capacity DB 36 and combines it with the customer attributes and actual power consumption values ​​obtained in step S10, using the customer number as a reference.

[0056] Next, in step S32, the CPU 11, acting as a construction unit 26, groups consumers based on their attributes and the date and time, and then aggregates the actual power consumption, contracted power, PV capacity, etc., of the consumers included in each group by taking the average, etc., to create demand statistics information.

[0057] Next, in step S34, the CPU 11, acting as a construction unit 26, combines the demand statistics information and the weather information acquired in step S10, associating them based on date and time, to create training data.

[0058] Next, in step S36, the CPU 11, acting as the construction unit 26, uses at least the aggregated values ​​of solar radiation and PV capacity from the created training data as explanatory variables, and the aggregated values ​​of actual power consumption as the target variable. Then, the CPU 11, acting as the construction unit 26, constructs the prediction model 38 by updating the parameters of the prediction model 38 so that the power consumption prediction result output when the explanatory variables are input to the prediction model 38 approaches the target variable.

[0059] Next, in step S38, the CPU 11, acting as the construction unit 26, stores the parameters of the constructed prediction model 38 in a predetermined memory area. Then, the construction process ends.

[0060] Next, we will explain the prediction process shown in Figure 12.

[0061] In steps S10 to S32, the CPU 11, acting as the prediction unit 28, performs the same processing as the construction process described above (Figure 10).

[0062] Next, in step S44, the CPU 11, acting as a forecasting unit 28, combines demand statistics information and weather forecasts, which are weather information for a future date and time, by associating them with the date and time, and creates forecast data. The CPU 11, acting as a forecasting unit 28, also selects target customer attributes from the demand statistics information, reflecting future contract forecasts such as new contracts and cancellations, and obtains past power consumption actual values ​​corresponding to the selected customer attributes, and includes them in the forecast data.

[0063] Next, in step S46, the CPU 11, acting as the prediction unit 28, inputs the created prediction data into the prediction model 38 and obtains the prediction results of future power consumption output from the prediction model 38. Then, the CPU 11, acting as the prediction unit 28, stores the prediction results obtained for each group of consumers in the power consumption prediction result DB 40. Next, in step S48, the CPU 11, acting as the prediction unit 28, outputs the power consumption prediction result DB 40 once all predictions are complete, and the prediction process ends.

[0064] As described above, the power consumption prediction device according to this embodiment acquires the actual power consumption values ​​of each of the multiple consumers, and weather information including at least solar radiation information for the area including the location of each of the multiple consumers. The power consumption prediction device then analyzes the acquired information for each consumer, using the coefficient of the solar radiation term in a regression equation with at least solar radiation as the explanatory variable and the actual power consumption value as the dependent variable, as the PV capacity. The power consumption prediction device then inputs the weather information including at least solar radiation and the PV capacity into a prediction model that has been pre-trained to predict power consumption, predicts the future power consumption for the multiple consumers, and outputs the prediction results. This makes it possible to predict future power consumption with low error even when the ownership status and capacity of solar power generation facilities are completely unknown.

[0065] Furthermore, the power consumption forecasting device according to this embodiment constructs a forecasting model and performs forecasting using data aggregated based on customer attributes. Therefore, it is possible to construct a forecasting model that adapts to changes in the load curve and to predict future power consumption with high robustness.

[0066] In addition, the power consumption prediction process, which the CPU reads and executes in the above embodiment, may be executed by various processors other than the CPU. Examples of such processors include dedicated electrical circuits, which are processors with circuit configurations specifically designed to execute particular processes, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices) whose circuit configurations can be changed after manufacturing, and ASICs (Application Specific Integrated Circuits). Furthermore, the power consumption prediction process may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (for example, multiple FPGAs, and a combination of a CPU and an FPGA). More specifically, the hardware structure of these various processors is an electrical circuit that combines circuit elements such as semiconductor elements.

[0067] Furthermore, although the above embodiment describes a configuration in which the power consumption prediction program is pre-stored (installed) in the storage 14, the system is not limited to this configuration. The program may be provided in a form stored on a non-transitor storage medium such as a CD-ROM (Compact Disk Read Only Memory), DVD-ROM (Digital Versatile Disk Read Only Memory), or USB (Universal Serial Bus) memory. Alternatively, the program may be provided in a form that can be downloaded from an external device via a network.

[0068] The following additional information is disclosed regarding the embodiments described above.

[0069] (Note 1) A power consumption prediction device comprising: a memory; and at least one processor connected to the memory, wherein the processor acquires actual values ​​of the power consumption of each of a plurality of consumers and weather information including at least solar radiation information for a range including the location of each of the plurality of consumers; analyzes the acquired information for each of the consumers using the coefficient of the solar radiation term in a regression equation in which at least the solar radiation is the explanatory variable and the actual value of the power consumption is the dependent variable, as the capacity of the solar power generation equipment; and inputs the weather information including at least solar radiation and the capacity of the solar power generation equipment into a prediction model that has been pre-trained to predict power consumption, predicts future power consumption for the plurality of consumers, and outputs the prediction result.

[0070] (Appendix 2) A non-temporary storage medium storing a program executable by a computer to perform power consumption prediction processing, wherein the power consumption prediction processing acquires actual power consumption values ​​for each of a plurality of consumers and weather information including at least solar radiation information for a range including the location of each of the plurality of consumers, analyzes the acquired information for each of the consumers using the coefficient of the solar radiation term in a regression equation in which at least the solar radiation is the explanatory variable and the actual power consumption value is the dependent variable as the capacity of the solar power generation equipment, inputs the weather information including at least solar radiation and the capacity of the solar power generation equipment into a prediction model pre-trained to predict power consumption, predicts future power consumption for the plurality of consumers, and outputs the prediction results.

[0071] 10 Power Consumption Prediction Device 11 CPU 12 ROM 13 RAM 14 Storage 15 Input Unit 16 Display Unit 17 Communication Interface 19 Bus 22 Acquisition Unit 24 Analysis Unit 26 Construction Unit 28 Prediction Unit 30 Customer Attribute DB 32 Power Consumption Actual DB 34 Weather Information DB 36 PV Capacity DB 38 Prediction Model 40 Power Consumption Prediction Result DB

Claims

1. An energy consumption prediction device comprising: an acquisition unit that acquires actual values ​​of the energy consumption of each of several consumers and weather information including at least solar radiation information for a range including the location of each of the several consumers; an analysis unit that analyzes the information acquired by the acquisition unit for each of the consumers by using the coefficient of the solar radiation term in a regression equation in which at least the solar radiation is the explanatory variable and the actual value of the energy consumption is the dependent variable as the capacity of the solar power generation equipment; and a prediction unit that inputs the weather information including at least solar radiation and the capacity of the solar power generation equipment into a prediction model that has been pre-trained to predict energy consumption, predicts future energy consumption for the several consumers, and outputs the prediction result.

2. The power consumption prediction device according to claim 1, wherein the acquisition unit acquires the attributes of each of the plurality of consumers, and the prediction unit inputs the capacity of the solar power generation equipment aggregated according to the attributes of each of the plurality of consumers into the prediction model.

3. The power consumption prediction device according to claim 1 or 2, wherein the analysis unit analyzes the capacity of the solar power generation equipment for each consumer using information from the information acquired by the acquisition unit, specifically information from the time period when the intensity of solar radiation is high.

4. A method for predicting electricity consumption, performed by a computer, which involves acquiring actual electricity consumption data for each of several consumers and weather information for an area including the location of each of the several consumers, including at least solar radiation information; analyzing the acquired information for each consumer using a regression equation in which at least the solar radiation is the explanatory variable and the actual electricity consumption data is the dependent variable, with the coefficient of the solar radiation term as the capacity of the solar power generation equipment; inputting the weather information including at least solar radiation and the capacity of the solar power generation equipment into a prediction model pre-trained to predict electricity consumption, predicts future electricity consumption for the several consumers, and outputs the prediction results.

Images