Energy demand prediction device, learning device, energy demand amount prediction method, and learning method

The energy demand forecasting system enhances prediction accuracy by correlating people flow with energy demand using machine learning, effectively addressing inaccuracies in facilities with irregular events.

JP2024168069APending Publication Date: 2024-12-05SHIMIZU CORP

Patent Information

Application Number
JP2023084464
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-23
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Existing energy demand forecasting methods struggle to maintain accuracy, particularly in facilities with irregular events or complex user dynamics, leading to decreased prediction precision.

Method used

An energy demand forecasting system that utilizes machine learning to predict energy demand by correlating future people flow information with historical energy demand data, using models like neural networks to enhance prediction accuracy.

Benefits of technology

Improves energy demand prediction accuracy by integrating people flow data, enabling precise forecasting even in facilities with unpredictable user patterns, facilitating better energy management and grid stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024168069000001_ABST
    Figure 2024168069000001_ABST
Patent Text Reader

Abstract

To improve prediction accuracy of energy demand.SOLUTION: An energy demand prediction device comprises a prediction unit which predicts an energy demand amount in an object area in future by inputting human flow information obtained by predicting human flow in future to the object area in a learned model for predicting the energy demand amount of the object area by learning relation between the human flow information representing the human flow to the object area and the energy demand amount of the object area.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an energy demand prediction device, a learning device, an energy demand prediction method, and a learning method. [Background technology]

[0002] In the energy management of buildings or groups of buildings, it is important to predict energy demand. This prediction uses information such as the weather and the day of the week. In this case, electricity demand can be predicted with an accuracy of about ±5%, and heat demand with an error range of about ±10 to 20%. When the energy is electric power, there is a device that predicts electric power demand (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2023-010235 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the method of predicting facility usage based on the day of the week, which is a major factor in energy demand forecasting, can result in a decrease in the accuracy of energy demand forecasts in cases where the number of users is difficult to predict based on the day of the week alone (such as facilities that host many irregular events) or in cases where facility uses are complex and people move between facilities.

[0005] The present invention has been made in consideration of the above circumstances, and its purpose is to provide an energy demand prediction device, a learning device, an energy demand prediction method, and a learning method that can improve the accuracy of energy demand prediction. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, one aspect of the present invention is an energy demand prediction device having a prediction unit that predicts future energy demand in a target area by inputting people flow prediction information that predicts future people flow through the target area into a trained model that predicts energy demand in the target area by learning the relationship between people flow information that represents the flow of people through the target area and the energy demand in the target area.

[0007] Another aspect of the present invention is a learning device having a people flow information acquisition unit that acquires people flow information representing people flow in a target area, an energy demand acquisition unit that acquires energy demand in the target area, and a learning unit that generates a learned model that predicts energy demand by performing machine learning using the people flow information and the energy demand.

[0008] Another aspect of the present invention is an energy demand prediction method that causes a computer to execute a prediction step of predicting energy demand in a target area in the future by inputting people flow prediction information that predicts future people flow in the target area into a trained model that predicts energy demand in the target area by learning the relationship between people flow information that represents the flow of people in the target area and the energy demand in the target area.

[0009] Another aspect of the present invention is a learning method that causes a computer to execute a pedestrian flow information acquisition step of acquiring pedestrian flow information representing the flow of people in a target area, an energy demand acquisition step of acquiring energy demand in the target area, and a learning step of generating a trained model that predicts energy demand by performing machine learning using the pedestrian flow information and the energy demand. Effect of the Invention

[0010] As described above, according to the present invention, it is possible to improve the prediction accuracy of energy demand. [Brief description of the drawings]

[0011] [Figure 1] 1 is a schematic diagram showing the configuration of an energy demand prediction system S. [Diagram 2] 4 is a flowchart illustrating an operation of the energy demand prediction device 10 in a learning phase. [Diagram 3] 13 is a flowchart illustrating an operation of the energy demand prediction device 10 in the execution phase. [Figure 4] FIG. 1 is an explanatory diagram illustrating a trained model. [Diagram 5] FIG. 11 is a diagram illustrating an example of people flow prediction information in a target area. [Figure 6] FIG. 11 is a diagram showing an example of people flow prediction information for each building in a target area. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] An energy demand prediction system S according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic diagram showing the configuration of an energy demand prediction system S. In the energy demand prediction system S, an energy demand prediction device 10, a people flow prediction device 20, and a terminal device 30 are communicatively connected via a network NW.

[0013] The energy demand prediction device 10 includes a communication unit 101 , a memory unit 102 , a people flow information acquisition unit 103 , an energy demand acquisition unit 104 , a learning unit 105 , a prediction unit 106 , and a control unit 107 . The communication unit 101 communicates with the people flow prediction device 20 and the terminal device 30 via the network NW, and transmits and receives various data.

[0014] The storage unit 102 stores various types of data. For example, the storage unit 102 stores people flow information and energy demand. The people flow information is information that indicates the number of people in a target area at each time. The people flow information may include people passing through the target area and people staying in the target area. This people flow information is an actual value obtained by measuring the flow of people in the target area. This actual value may be a measurement result measured by a people flow sensor or the like, or may be a value obtained by manually measuring the flow of people. The target area is provided with at least one entrance / exit through which at least one of the people can enter or exit the target area. If the target area has two or more entrances / exits, the number of people present in the target area may be further classified according to which entrance / exit the people entered through. Furthermore, if the target area has two or more entrances / exits, the people flow information may indicate how many people present in the target area entered through which entrance / exit and exited through which exit.

[0015] The target area may be a region or a building within the region. If the target area is a region, the entrance / exit may be a road, passageway, etc. adjacent to the target area. If the target area is a building, the entrance / exit may be a front door, a back door, a passageway, etc. Furthermore, if the target area is a train station, the entrance / exit may be the north exit, south exit, west exit, east exit, etc. of the station.

[0016] The energy demand represents the amount of energy used in the target area. In this embodiment, the energy is described as district heating and cooling, and includes heat and electricity (power) supplied to the area, but may be either heat or electricity (power). The energy demand stored in the storage unit 102 is a past actual value.

[0017] The storage unit 102 is configured with a storage medium, for example, a hard disk drive (HDD), a flash memory, an electrically erasable programmable read only memory (EEPROM), a random access read / write memory (RAM), a read only memory (ROM), or any combination of these storage media. The storage unit 102 may be, for example, a non-volatile memory.

[0018] The people flow information acquisition unit 103 acquires the people flow information by reading out the people flow information stored in the storage unit 102 .

[0019] The energy demand acquisition unit 104 acquires the energy demand of the target area from the storage unit 102. When the energy demand is stored in the storage unit 102, the energy demand acquisition unit 104 acquires the energy demand from the storage unit 102.

[0020] The learning unit 105 executes machine learning using the people flow information and the energy demand, thereby generating a trained model for obtaining an energy demand forecast amount that predicts the energy demand amount in the future. The pedestrian flow information and the pedestrian flow prediction information may each include a position where a person enters the target area (an entrance / exit position) and a position where a person leaves the target area (an entrance / exit position). In this case, the learning unit 105 may generate a trained model that has learned the relationship between the pedestrian flow information including the entry and exit positions for the target area and the future energy demand for the target area. Furthermore, when there are multiple target areas, the learning unit 105 may generate a different trained model for each of the different target areas.

[0021] The machine learning technique may be, for example, multiple regression analysis, a Kalman filter, a neural network (deep learning), a random forest, a support vector machine, or the like, but for example, a neural network (deep learning) is used.

[0022] The prediction unit 106 predicts future energy demand in the target area by inputting pedestrian flow prediction information that predicts future pedestrian flow through the target area into a trained model that predicts energy demand in the target area by learning the relationship between pedestrian flow information representing pedestrian flow through the target area and energy demand in the target area.

[0023] The prediction unit 106 may predict energy demand by inputting people flow prediction information, including the positions where people enter the target area and the positions where people leave the target area, into the learned model.

[0024] The control unit 107 controls each unit of the energy demand prediction device 10 .

[0025] The communication unit 101, people flow information acquisition unit 103, energy demand acquisition unit 104, learning unit 105, prediction unit 106, and control unit 107 of the energy demand prediction device 10 may be composed of a processing device such as a CPU (central processing unit) or a dedicated electronic circuit.

[0026] The pedestrian flow prediction device 20 predicts the future number of visitors and passers-by in a target area by simulating pedestrian flow using past pedestrian flow information. Since past pedestrian flow information for pedestrian flow prediction is provided by mobile / communication companies and related companies, such pedestrian flow information may be utilized. When the people flow prediction device 20 performs a simulation, an AI prediction model may be created and the people flow may be predicted using the AI ​​prediction model. The function of pedestrian flow prediction (pedestrian flow simulation) may be implemented using a system that is already provided.

[0027] The terminal device 30 transmits instruction information for predicting people flow to the energy demand prediction device 10, and thereby obtains the result of predicting people flow according to the instruction information. The instruction information includes, for example, information specifying a target area and a time period for predicting people flow.

[0028] Next, we will explain the operation of the above-mentioned energy demand prediction device 10. The operation of the energy demand prediction device 10 is roughly divided into a learning phase and an execution phase. Learning Phase FIG. 2 is a flowchart illustrating the operation of the energy demand prediction device 10 in the learning phase. The people flow information acquisition unit 103 of the energy demand prediction device 10 reads out people flow information of a target area from the storage unit 102 (step S101). In addition, the energy demand acquisition unit 104 reads out energy demand of the target area according to the instruction information from the storage unit 102 (step S102). The learning unit 105 executes machine learning using the people flow information and the energy demand to generate a trained model that obtains an energy demand forecast amount that predicts the energy demand amount in the future (step S103). The learning unit 105 stores the generated trained model in the storage unit 102 in association with target area identification information that indicates for which target area the trained model can be used (step S104). Such learning may be performed for each target area.

[0029] Execution Phase FIG. 3 is a flowchart illustrating the operation of the energy demand prediction device 10 in the execution phase. The control unit 107 of the energy demand prediction device 10 acquires the instruction information transmitted from the terminal device 30 (step S201), extracts the target area included in the instruction information, and identifies the target area by reading out the trained model corresponding to the extracted target area from the memory unit 102 (step S202).

[0030] Next, the prediction unit 106 extracts the time period included in the instruction information, and acquires pedestrian flow prediction information predicting the pedestrian flow for each time period included in the extracted time period from the pedestrian flow prediction device 20 (step S203). Here, the prediction unit 106 acquires information indicating the target area and the time period from the pedestrian flow prediction device 20, thereby obtaining pedestrian flow prediction information for each time period included in the time period in the target area.

[0031] When the prediction unit 106 acquires the people flow prediction information in the target area, the prediction unit 106 inputs the information into the identified trained model to obtain an energy demand prediction amount that predicts the energy demand in the target area (step S204). The control unit 107 transmits the obtained predicted energy demand amount to the terminal device 30 that is the source of the instruction information (step S205).

[0032] According to the above-described embodiment, it is possible to obtain an energy demand forecast by inputting people flow prediction information into a trained model. For example, by using a digital twin of a city or the like to set an area around a building or a group of buildings as a target area, and using people flow prediction information for the target area as an input parameter for an energy demand forecast model (trained model), it is possible to predict energy demand within the target area with high accuracy.

[0033] A machine learning model such as a deep neural network (DNN) can be used for the energy demand prediction model. In addition, by acquiring long-term (e.g., 1 to 2 weeks) people flow prediction information from the people flow prediction device 20, it is possible to improve the accuracy of long-term energy demand prediction.

[0034] FIG. 4 is an explanatory diagram for explaining a trained model. The trained model includes an input layer, an intermediate layer, and an output layer, and learns the relationship between input data input as input items (input parameters) of the input layer and output data output from the output layer by updating weight coefficients in the intermediate layer. The input parameters input to the input layer include at least people flow information. In addition to the input parameters, traffic access, events, weather, time zones, days of the week, seasons, solar radiation, temperature, etc. may be used as parameters.

[0035] Furthermore, as one of the utilization methods, the energy demand prediction by the energy demand prediction device 10 in the above-described embodiment can be more effectively implemented by incorporating it into an energy management system for district heating and cooling. For example, in recent years, variable renewable energy (hereinafter, VRE) such as solar power generation has been rapidly increasing in Japan, and there are cases where the time periods during which the supply and demand balance of local electricity is disrupted are increasing. As one countermeasure, it is effective to provide a district heating and cooling system that supplies heat and electricity to the region with the above-mentioned energy storage equipment with a capacity of several days or a large-scale cogeneration system (hereinafter, CGS). For example, the energy demand prediction device 10 uses the above-mentioned energy demand prediction device 10 to obtain an energy demand forecast amount according to people flow prediction information, calculates the difference between the obtained energy demand forecast amount and the VRE forecast amount that predicts the VRE in the region, identifies a future time period when VRE is likely to be surplus, stores energy during the time period when VRE is surplus, and consumes it during the time period when electricity demand is tight or operates a CGS, thereby contributing to the leveling of grid power.

[0036] In addition, the energy demand forecast amount predicted using such people flow forecast information may be used for operation control not only on the day of forecast or the day after the forecast, but also on a long-term basis such as on a weekly basis. This makes it possible to make use of highly accurate energy demand forecasts. In particular, in arenas, large exhibition halls, commercial facilities, hotels, etc., even in cases where events are held irregularly regardless of the day of the week, there are extreme population fluctuations in the area during events, and complex people flows occur such as people who visit one facility using another facility, because the increase / decrease in people and their movements on an hourly basis greatly affect the energy consumption tendency of the facility, it is effective to include people flow forecasts in the parameters and perform energy demand forecasts.

[0037] Here, as the number of users (people flow) in an area or building within the target area increases, the power consumption of the computers, electronic devices, electrical equipment, lighting, etc. used by the people increases, and as the number of people increases, the heat load that the air conditioning needs to handle (heat generated by people, and heat from the computers, electronic devices, electrical equipment, and lighting used by the people) increases, so the air conditioning energy also increases, and as the number of people decreases, the heat load also decreases. Thus, it can be said that there is a correlation between the number of people in the target area and the energy demand, in terms of the relationship between the number of people and the energy demand.

[0038] Here, there are two main forms when the target area is a region or a building. If the target area is a region If the flow of people entering and leaving the area from moment to moment during the time period included in the instruction information can be predicted, the predicted flow of people entering and leaving the target area can be used for energy demand prediction. Here, the total number of people within the target area is calculated from people flow predictions of the number of people entering and leaving the area, and this total number is used to predict energy demand for the entire target area, making it possible to predict total energy consumption within the area (within a premises, etc.). Fig. 5 is a diagram showing an example of people flow prediction information in a target area. In Fig. 5, the horizontal axis is time, and the vertical axis is number of people. In this way, in the case where the people flow prediction information can predict the people flow (number of people) entering and leaving an area from moment to moment, it is possible to obtain the energy demand prediction amount for the target area at that time by inputting the people flow prediction information corresponding to that time into the trained model. In other words, the energy demand prediction amount for the target area can be obtained for each hour.

[0039] If the target area is a building If the flow of people entering and exiting buildings from moment to moment during the time period included in the instruction information can be predicted, the predicted flow of people for each building in the target area can be used to predict energy demand. Here, by utilizing the number of building occupants calculated from the people flow prediction for each building, it becomes possible to predict the energy consumption of each building / facility. This makes it possible to improve the prediction accuracy of the total energy consumption within a building / facility by integrating the energy demand prediction for each building / facility. FIG. 6 is a diagram showing an example of people flow prediction information for each building in a target area. In FIG. 6, the horizontal axis is time, and the vertical axis is number of people. In this example, the people flow prediction information includes the flow of people (number of people) entering and exiting each building in the target area at each time, and by inputting this people flow prediction information into a trained model corresponding to each building, it is possible to obtain the predicted energy demand amount for each building at that time. In other words, the predicted energy demand amount for each building can be obtained for each time. In this figure, during time period A indicated by symbol A, a large event is being held at the stadium, and among the multiple buildings, the stadium has the highest flow of people. Also, during time period B shortly before the time period indicated by symbol A, among the multiple buildings, the commercial facility has a high flow of people, and this can be presumed to be because people are staying at the commercial facility during the day and then moving to the stadium to participate in the event. Therefore, it is predicted that the energy demand forecast amount at the commercial facility is high during the day, and then the energy demand forecast amount at the commercial facility will decrease and the energy demand forecast amount at the stadium will increase.

[0040] In addition, the energy demand forecast amount predicted by the energy demand forecasting device 10 in the above-mentioned embodiment may be used for integrated management of district heating and cooling facilities. For example, in recent years, there has been a concept of constructing a VPP to integrate and manage multiple district heating and cooling facilities, and to perform power exchange and power saving adjustment and control. Integrated management can be greatly beneficial in cases where large-scale energy storage facilities or CGS are installed, but energy demand forecasts for such facilities have not necessarily been fully considered. Therefore, by using the above-mentioned energy demand forecasting device 10, it is possible to optimally control the integrated management of multiple district heating and cooling facilities by applying it to each district heating and cooling facility in a management system that integrates and manages each district heating and cooling facility and performing long-term energy demand forecasting. In other words, by using the total number of people in a target area to predict energy demand for the entire area, it can also be used to control the operation of equipment in the main district heating and cooling plant (such as energy storage equipment and CGS). In addition, when the target area is a building, by using the number of building occupants calculated from the people flow prediction for each building for not just one building but several buildings in the target area in the energy demand prediction for each building, the energy demand prediction for each building can be integrated to improve the accuracy of the energy demand prediction for the entire district heating and cooling system (main plant). It can also be used for the control of sub-plants and secondary equipment.

[0041] In the above embodiment, the case where the energy demand forecast amount is obtained based on the flow of people into the target area has been described, but when generating the trained model, the positions of people entering and leaving the target area may also be used as input parameters together with the flow of people to learn the trend of energy demand and generate the trained model. This makes it possible to obtain the energy demand forecast amount taking into account the positions of people entering and leaving the target area.

[0042] Furthermore, according to the embodiment described above, the accuracy of energy demand forecasts for a building or group of buildings, or for district heating and cooling can be improved, which makes it possible to level out energy (electricity) on the supply side, such as electric power companies, and reduce greenhouse gas emissions through energy-saving operation. In particular, when the target area is an arena, large exhibition hall, commercial facility, hotel, or other area where irregular events are held, there are extreme population fluctuations in the area during events, and complex human flows occur such as people who visit one facility then using other facilities, the prediction accuracy of energy demand forecasting is improved to a greater extent than when the energy demand prediction device 10 of this embodiment is not used. In addition, when combined with energy storage facilities or energy generation such as CGS, it not only enables BCP measures but also peak cutting and shift operation over long periods (weekly).

[0043] In the above embodiment, the learning unit 105 is provided in the energy demand prediction device 10. However, the learning unit 105 may be provided as a learning device outside the energy demand prediction device 10. In this case, the learning device may be provided with a people flow information acquisition unit 103 and an energy demand acquisition unit 104, and may obtain various information from the memory unit 102 of the energy demand prediction device 10. In addition, the case where the people flow information and the energy demand are stored in the storage unit 102 has been described, but they may be stored in a storage device external to the energy demand prediction device 10. The functions of the energy demand prediction device 10 may be provided in different devices, which are communicatively connected via a network or the like, to form a so-called cloud system.

[0044] Furthermore, according to the embodiment described above, instruction information is transmitted from the terminal device 30 and the energy demand forecast amount is obtained from the energy demand forecasting device 10, but the function of the terminal device 30 may be incorporated in another system. For example, the function of the terminal device 30 may be incorporated in a management system that performs integrated management of each district heating and cooling facility. This allows the energy demand forecasting device 10 to be linked to a system that performs various information processing using the energy demand forecast amount, such as a management system.

[0045] The energy demand forecasting device 10 in the above-described embodiment may be realized by a computer. In that case, a program for realizing this function may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into a computer system and executed to realize the function. The term "computer system" as used herein includes hardware such as an OS and peripheral devices. The term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into a computer system. The term "computer-readable recording medium" may also include a medium that dynamically holds a program for a short period of time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line, and a medium that holds a program for a certain period of time, such as a volatile memory inside a computer system that is a server or client in that case. The above-described program may be a program for realizing a part of the above-described function, or may be a program that can realize the above-described function in combination with a program already recorded in the computer system, or may be a program that is realized using a programmable logic device such as an FPGA (Field Programmable Gate Array).

[0046] Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and designs that do not deviate from the gist of the present invention are also included.

[0047] The "Sustainable Development Goals (SDGs)" are 17 international goals that were adopted at the United Nations Summit in September 2015. The energy demand prediction device 10 according to this embodiment can contribute to the achievement of, for example, goal 7 "Affordable and clean energy" among the 17 SDGs. [Explanation of symbols]

[0048] 10...energy demand prediction device, 20...people flow prediction device, 30...terminal device, 101...communication unit, 102...storage unit, 103...people flow information acquisition unit, 104...energy demand acquisition unit, 105...learning unit, 106...prediction unit, 107...control unit

Claims

1. A prediction unit that predicts the energy demand in a target area in the future by inputting people flow prediction information that predicts the future people flow in the target area into a trained model that predicts the energy demand in the target area by learning the relationship between people flow information that indicates the flow of people in the target area and the energy demand in the target area. An energy demand forecasting device having the above configuration.

2. the target area is a building; The prediction unit predicts the energy demand in the building in the future by inputting people flow prediction information that predicts people flow in the building into the trained model. The energy demand forecasting device according to claim 1 .

3. The pedestrian flow information and the pedestrian flow prediction information each include a position where a person enters the target area and a position where a person exits the target area, The trained model learns a relationship between people flow information including an entry position and an exit position for the target area and an energy demand for the target area, The prediction unit predicts the energy demand by inputting people flow prediction information including a position where a person enters the target area and a position where a person exits the target area to the trained model. The energy demand forecasting device according to claim 1 .

4. A people flow information acquisition unit that acquires people flow information representing people flow in a target area; an energy demand acquisition unit that acquires an energy demand in the target area; a learning unit that generates a trained model for predicting the energy demand by performing machine learning using the people flow information and the energy demand; A learning device having the above configuration.

5. On the computer, A prediction step of predicting the energy demand in the target area in the future by inputting people flow prediction information that predicts the future people flow in the target area into a trained model that predicts the energy demand in the target area by learning the relationship between people flow information that indicates the flow of people in the target area and the energy demand in the target area. The energy demand forecasting method includes:

6. On the computer, A people flow information acquisition step of acquiring people flow information representing people flow in a target area; an energy demand acquisition step of acquiring an energy demand for the target area; A learning step of generating a trained model for predicting the energy demand by performing machine learning using the people flow information and the energy demand. A learning method that allows you to execute the following:

Citation Information

Patent Citations

  • Power demand prediction device and power demand prediction method

    JP2023010235A

Cited By

  • system

    JP2026155559A

  • system

    JP7943438B2

  • Device and method

    WO2026185959A1