Heat load prediction system, heat load prediction device, heat load prediction method, and program

Abstract

To provide a heat load prediction system capable of highly accurately predicting a heat load according to a facility and air-conditioning use timing, using prior air-conditioning reservation information, relating to central air-conditioning.SOLUTION: The heat load prediction system comprises: an acquisition unit for acquiring air-conditioning reservation information relating to a facility using central air-conditioning; and a heat load prediction unit for inputting input data including the air-conditioning reservation information acquired by the acquisition unit and supplemental information to a learning model of machine learning in which weighting can be changed based on the facility and air-conditioning use timing to acquire output data including information relating to a prediction result of the heat load output from the learning model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a heat load prediction system, a heat load prediction device, a heat load prediction method, and a program. [Background technology]

[0002] Heat load (demand) predictions for buildings are being made. In conventional technology for predicting heat load (demand) in buildings, the predicted value is obtained by extracting several pieces of actual data for days similar to the target prediction date from the history of past heat load actual data based on facility operation information (weekday operation / holiday operation) and weather forecast information, and then calculating the average value of these data through statistical processing.

[0003] Recently, there have also been cases where actual heat load information, facility operation information (weekday operation / holiday operation), and weather information have been machine-learned using AI (Artificial Intelligence), and the characteristics are then stored in a prediction system, which then inputs the facility operation information and weather forecast information for the next day into the prediction system, which then outputs a predicted heat load value for that next day.

[0004] The operation management device described in Patent Document 1 aims to improve the accuracy of power load prediction. Specifically, the operation management device includes an air conditioning heat load prediction unit that calculates an air conditioning heat load prediction value indicating the amount of heat expected to adjust the temperature to a preset temperature on a prediction date, a power generation output prediction processing unit that calculates power generation output prediction data indicating the power generated by a generator on the prediction date, and an operation plan creation unit that creates an air conditioning heat source operation plan indicating the allocation for each time of operation of each air conditioning heat source equipment device that generates the amount of heat indicated by the air conditioning heat load prediction value, calculates a power load prediction value indicating the power to be supplied to a load device to generate the predicted amount of heat indicated by the air conditioning heat load prediction value based on the air conditioning heat load prediction value, and determines the purchased power and the generated power corresponding to the power of the power load prediction value using the power generation output prediction data so that the power of the purchased power supplied from the purchased power source of the commercial power system for each predetermined fixed time period becomes an arbitrary target value, and creates a power equipment operation plan indicating a schedule of power to be output by the generator and the purchased power source (see Patent Document 1). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2012-80679 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, in the above-mentioned conventional technology, when the usage of the building is regular every day, such as during certain business hours (for example, 9:00 to 18:00, etc.), prediction is relatively easy, but in the case of facilities such as rental conference rooms or halls, where the usage is irregular every day based on the reservation status of the facility, it may be difficult to predict the heat load with only the information related to the above-mentioned conventional technology, and it is particularly difficult to predict the heat load over time. Also, the above-mentioned conventional technology has some shortcomings in terms of its application to central air conditioning.

[0007] In consideration of the above circumstances, an embodiment of the present disclosure provides a heat load prediction system, a heat load prediction device, a heat load prediction method, and a program for central air conditioning that can use advance air conditioning reservation information to perform heat load prediction with high accuracy according to the facility and the timing of air conditioning use. [Means for solving the problem]

[0008] One aspect of the present disclosure is a thermal load prediction system that includes an acquisition unit that acquires air conditioning reservation information for a facility that uses central air conditioning, and a thermal load prediction unit that inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use, and acquires output data including information regarding the thermal load prediction results output from the learning model.

[0009] One aspect of the present disclosure is a thermal load prediction device that includes an acquisition unit that acquires air conditioning reservation information for a facility that uses central air conditioning, and a thermal load prediction unit that inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use, and acquires output data including information regarding the thermal load prediction results output from the learning model.

[0010] One aspect of the present disclosure is a thermal load prediction method in which an acquisition unit acquires air conditioning reservation information for a facility that uses central air conditioning, and a thermal load prediction unit inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use, and acquires output data including information regarding the predicted results of the thermal load output from the learning model.

[0011] One aspect of the present disclosure is a program for enabling a computer constituting a thermal load prediction device to realize a function of acquiring air conditioning reservation information for a facility that uses central air conditioning, and a function of inputting input data including the acquired air conditioning reservation information and auxiliary information into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use, and acquiring output data including information regarding the thermal load prediction results output from the learning model. Effect of the Invention

[0012] According to the above-mentioned heat load prediction system, heat load prediction device, heat load prediction method and program, with regard to central air conditioning, heat load prediction can be performed with high accuracy according to the facility and the timing of air conditioning use using advance air conditioning reservation information. [Brief description of the drawings]

[0013] [Figure 1] 1 is a diagram illustrating a schematic configuration example of a heat load prediction system according to an embodiment. [Diagram 2] 1 is a diagram showing an outline of a central air conditioning system to which a heat load prediction system according to an embodiment is applied; [Figure 3A] FIG. 13 is a diagram illustrating an example of an irregular relationship between the thermal load record and the operation time. [Figure 3B] FIG. 11 is a diagram illustrating an example of a case where the relationship between the thermal load record and the operation time is regular. [Figure 4A] FIG. 4 is a diagram illustrating a table illustrating an example of reservation information according to the embodiment. [Figure 4B] FIG. 11 is a diagram illustrating a graph showing an example of reservation information according to the embodiment. [Diagram 5] FIG. 13 is a diagram illustrating an example of input and output of a trained model according to an embodiment. [Figure 6] FIG. 13 is a diagram illustrating an example of a heat load prediction using a trained model according to an embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present disclosure will be described in detail with reference to the drawings.

[0015] [Heat load prediction system] FIG. 1 is a diagram showing a schematic configuration example of a heat load prediction system 1 according to an embodiment. The thermal load prediction system 1 includes a thermal load prediction device 11 and an air conditioning management system 411. The air conditioning management system 411 includes an air conditioning reservation system 431 .

[0016] Here, in this embodiment, a configuration example is shown in which the heat load prediction system 1 includes an air conditioning management system 411 having the function of an air conditioning reservation system 431, but other configurations may be used. As another configuration example, when the thermal load prediction system 1 includes only the thermal load prediction device 11, the thermal load prediction system 1 corresponds to the thermal load prediction device 11. As another configuration example, the function of the air conditioning reservation system 431 may be provided as a system independent of the air conditioning management system 411. In this case, for example, the function of the air conditioning reservation system 431 may be provided in the heat load prediction system 1, and the air conditioning management system 411 may be provided as a system external to the heat load prediction system 1. In this embodiment, for convenience of explanation, the air conditioning management system 411 and the air conditioning reservation system 431 are described separately, but the air conditioning management system 411 and the air conditioning reservation system 431 may be integrated.

[0017] <Heat load forecasting device> The thermal load prediction device 11 includes an operation unit 111 , a display unit 112 , an input / output unit 113 , a storage unit 114 , and a control unit 115 . The control unit 115 includes an input processing unit 151 , a model learning unit 152 , a heat load prediction unit 153 , and an output processing unit 154 .

[0018] The operation unit 111 receives an operation from a user, and generates an operation signal in response to the received operation. The operation unit 111 outputs the generated operation signal to the control unit 115. Here, the operation unit 111 may have general-purpose members such as a mouse, a touch sensor, or a joystick, or may have dedicated members such as a button or a knob.

[0019] The display unit 112 displays various types of information on the screen in accordance with display data input from the control unit 115 . Here, the display unit 112 may have, for example, one or more of a liquid crystal display, an organic electroluminescence display, and the like.

[0020] The input / output unit 113 is connected wirelessly or wired to a device separate from the heat load prediction device 11 so as to input and output various data. The input / output unit 113 may be connected to the separate device via a communication network. The input / output unit 113 may have, for example, one or both of an input / output interface conforming to a predetermined input / output standard and a communication interface conforming to a predetermined communication method.

[0021] The storage unit 114 stores various types of data. The stored data may include information such as parameters used in processing by the control unit 115, data used in the processing, and data obtained by the processing. In this embodiment, the storage unit 114 stores a machine learning learning model M1. The learning model M1 can be in a state before learning, a state during learning, or a state after learning. Note that even if the learning model M1 has already been learned, it may be further subjected to reinforced learning. Here, the storage unit 114 includes a storage medium such as a RAM (Random Access Memory) and a ROM (Read Only Memory).

[0022] The control unit 115 performs various processes and controls for enabling the heat load prediction device 11 to fulfill its functions. Here, the control unit 115 may be realized as a computer system, for example, by a general-purpose processor executing processing indicated by instructions written in a specified program, or may be realized by a dedicated arithmetic circuit. Note that executing the processes indicated by the instructions written in a program may simply be called "executing a program" or "running a program."

[0023] The input processing unit 151 performs processing for inputting (obtaining) input data. Here, the input data may be, for example, one or more of data corresponding to an operation signal input from the operation unit 111, data input from the input / output unit 113, or data stored in the memory unit 114.

[0024] The model learning unit 152 uses the training data to perform machine learning on the learning model M1, which shows the association between the input data to the learning model M1 and the output data from the learning model M1. The training data is acquired by the input processing unit 151. The training data may be stored in the storage unit 114, for example. In this embodiment, the learning model M1 is stored in the storage unit 114. Here, any type of machine learning may be used as the machine learning, and for example, machine learning using a neural network may be used.

[0025] The thermal load prediction unit 153 predicts the thermal load using a trained learning model M1 (for convenience of explanation, in this embodiment, also referred to as a trained model) trained by the model learning unit 152. In this embodiment, the thermal load prediction unit 153 inputs input data to be predicted into a trained model, and acquires data (output data) output from the trained model as data related to the prediction result of the thermal load. Here, the output data may be, for example, data representing the predicted result of the heat load (itself), or may be data representing other items related to the predicted result of the heat load (for example, items that correspond one-to-one to the heat load). Note that the content of the output data depends on the content of the machine learning performed by the model learning unit 152 (i.e., the data representing the content to be output from the learning model M1).

[0026] The output processing unit 154 performs processing to output output data. The output processing unit 154 outputs, for example, the data acquired by the thermal load predicting unit 153 (data related to the prediction result of the thermal load). Here, the output processing unit 154 may perform one or more of the following processes, for example: displaying and outputting content based on the output data on the display unit 112; outputting the output data via the input / output unit 113; or outputting the output data to the memory unit 114 (here, storing the data).

[0027] [Central air conditioning system] In this embodiment, a case will be described in which the heat load prediction system 1 according to this embodiment is applied to a central air conditioning system.

[0028] FIG. 2 is a diagram showing an outline of a central air conditioning system to which the heat load prediction system 1 according to the embodiment is applied. FIG. 2 shows a building 301 and an air conditioning control system 411 . Here, the air conditioning control system 411 may be provided, for example, inside the building 301 or outside the building 301.

[0029] The building 301 has a room 311-1, a room 311-2, and a room 311-3. In the building 301, an air conditioner 331-1 corresponding to the room 311-1, an air conditioner 331-2 corresponding to the room 311-2, and an air conditioner 331-3 corresponding to the room 311-3 are installed. Furthermore, in the building 301, a heat source device 351 and a cooling tower 352 are installed as facilities common to all of the air conditioners 331-1 to 331-3. The heat source device 351 is configured using, for example, a refrigerator or a boiler. It should be noted that the air conditioners 331-1 to 331-3 are, for example, air handling units (AHUs), but may be other devices. Furthermore, the air conditioners may be called air conditioners.

[0030] Each of the air conditioners 331-1 to 331-3 blows out supply air and draws in return air for each of the rooms 311-1 to 311-3 (indoors). Outside air and return air are taken into each of the air conditioners 331-1 to 331-3. Hot and cold water flows through heat source equipment 351 in each of the air conditioners 331-1 to 331-3. In addition, cooling water flows from the heat source equipment 351 to the cooling tower 352 .

[0031] Here, in the central air conditioning system, heat source equipment 351 is installed in one place, and hot and cold water is sent to air conditioners 331-1 to 331-3 for air conditioning. In the central air conditioning system, efficient and efficient operation of the heat source equipment 351 in accordance with the heat load is required to achieve energy conservation.

[0032] In this embodiment, the air conditioning equipment installed inside the building 301 is provided with an air conditioning management system 411 that centrally manages the entire system. The air conditioning management system 411 has a function of managing air conditioning equipment (in the example of FIG. 2, air conditioners 331-1 to 331-3, heat source equipment 351, and cooling tower 352), and in this embodiment, shows a case where it has an automatic operation function based on advance reservation. The heat load prediction system 1 according to this embodiment utilizes such advance reservation information to realize accurate prediction of the heat load.

[0033] [Relationship between thermal load performance and operating hours] 3A and 3B, an example of the relationship between the thermal load record and the operation time is shown. In the graphs shown in FIG. 3A and FIG. 3B, the horizontal axis represents time (hours), and the vertical axis represents the thermal load record.

[0034] FIG. 3A is a diagram showing an example of a case where the relationship between the thermal load record and the operation time is irregular. The graph in FIG. 3A shows a characteristic 2011 and a characteristic 2012 that represent the relationship between the thermal load record and the operation time. The characteristics 2011 and 2012 represent, for example, data on different dates. As shown in FIG. 3A, in the example of FIG. 3A, the characteristics 2011 and 2012 are irregular.

[0035] FIG. 3B is a diagram showing an example in which the relationship between the thermal load record and the operation time is regular. The graph in FIG. 3B shows a characteristic 2111 and a characteristic 2112 that represent the relationship between the thermal load record and the operation time. The characteristics 2111 and 2112 represent, for example, data on different dates. As shown in FIG. 3B, in the example of FIG. 3B, the characteristics 2111 and 2112 are regular. It should be noted that although there is some difference between the characteristics 2111 and 2112, in this example, this difference is within the margin of error.

[0036] The technology according to the present embodiment can perform accurate thermal load prediction even when the relationship between the thermal load record and the operation time is irregular, for example, as shown in FIG. 3A. The technology according to this embodiment may be applied to a case where the relationship between the thermal load record and the operation time is regular, as shown in FIG. 3B, for example.

[0037] [Reservation information for autonomous driving function] The reservation information in the automatic driving function will be described. In this embodiment, the air conditioning equipment in each room and common area in the building has an advance operation reservation function. In this embodiment, in normal operation, the air conditioning equipment in each room and common area installed in the building basically operates automatically based on advance operation reservations.

[0038] Here, in this embodiment, in the thermal load prediction system 1, the rated capacity of each air conditioner to be reserved is registered in the air conditioning reservation system 431 (or the air conditioning management system 411) or the like, and can be acquired by the thermal load prediction device 11.

[0039] In this embodiment, reservation information is set for each time period in 30-minute intervals. There are 48 time periods in a 24-hour day. On / off commands can be set for each time period and for each air conditioner. It is possible to calculate {On / Off commands for 48 frames over 24 hours x rated capacity of the air conditioner} for each air conditioner. The operating rate [%] can then be found by adding up the values ​​calculated for each air conditioner for each of the 48 frames over 24 hours for all air conditioners and calculating the ratio to the total value of the rated capacity of all air conditioners (the value at 100%).

[0040] Here, in this embodiment, a case is shown in which all air conditioners in the building to be predicted are used as the subjects of calculation for heat load prediction, but as another example, a configuration may be used in which only some of the air conditioners in the building to be predicted are used as the subjects of calculation for heat load prediction. For example, only a representative air conditioner that reflects the operation time of the facility (building) to be predicted may be used as the calculation target for heat load prediction. For example, an air conditioner in a common area may be used as such a representative air conditioner.

[0041] <Examples of reservation information> Specific examples of reservation information are shown with reference to FIG. 4A and FIG. 4B.

[0042] FIG. 4A is a diagram showing a table T1 illustrating an example of reservation information according to the embodiment. Table T1 shows an example of calculating an operating rate, which is an example of information based on reservation information for each time period, for air conditioning. In this example, a building to be predicted has four rooms (conference rooms 1 to 4) to be predicted. The rated capacities of conference rooms 1 to 4 are 10 [kW], 10 [kW], 15 [kW], and 15 [kW], respectively.

[0043] In this case, the operation rate is calculated by the formula (1). Here, the denominator corresponds to the total value of the rated capacities of the four conference rooms 1 to 4.

[0044] [Number 1] Occupancy rate [%] = (total rated capacity of reserved conference rooms) / 50 (1)

[0045] In this example, as shown in table T1, the time period of one day, 00:00 to 24:00, is divided into 30-minute time periods. Specifically, the time period of one day is divided into the time period 00:00 to 00:30, the time period 00:30 to 01:00, ..., the time period 23:00 to 23:30, and the time period 23:30 to 24:00. In this example, for convenience of explanation, the start time of a day is set to 00:00 and the end time of a day is set to 24:00, but 24:00 corresponds to 00:00 of the next day.

[0046] In the example of table T1, the air conditioning reservation status for each of conference rooms 1 to 4 is shown for each time period. Specifically, the areas (columns) where there are reservations are marked with a circle (○), and the areas (columns) where there are no reservations are left blank. Table T1 shows the calculation results of (the total value of the rated capacity of the reserved conference rooms) and the utilization rate for each time period. In addition, in table T1, in order to simplify the illustration, only some of the time periods are illustrated, and the illustration of the remaining time periods is omitted.

[0047] FIG. 4B is a diagram showing a graph illustrating an example of reservation information according to the embodiment. In the graph shown in FIG. 4B, the horizontal axis represents time period, and the vertical axis represents operation rate [%]. In the example of Figure 4B, the start time of each time period is used as the time period on the horizontal axis. As a specific example, the operating rate for the time period from 08:30 to 09:00 is represented at the time of 08:30, and similarly for the other time periods. The example of FIG. 4B corresponds to the example of FIG. 4A, and shows the operating rates for time periods omitted in table T1 shown in FIG. 4A.

[0048] [Prediction of heat load using a trained model] <Input and output of trained model for heat load prediction> FIG. 5 is a diagram illustrating an example of input and output of a trained model according to an embodiment. Figure 5 shows a trained learning model M1 (trained model) that performs thermal load prediction, data (input data) representing input values ​​to the learning model M1, and data (output data) representing output values ​​from the learning model M1.

[0049] In this embodiment, the input data to the learning model M1 includes reservation data representing air conditioning reservation information (which may be called facility reservation information, etc.), which is reservation information related to air conditioning of a facility. In this embodiment, data representing the operating rate of the air conditioning (operating rate data) is used as the reservation data, but other data may also be used. The input data may also include other data, such as auxiliary data and performance data.

[0050] The auxiliary data may be, for example, data relating to one or more of the following information: temperature, humidity, number of occupants, zone information, or a calendar. The auxiliary data may be, for example, data related to the room being reserved (conference rooms 1-4 in the example of FIG. 4A), i.e., data such as the room's temperature, humidity, number of occupants, zone information, or a calendar (e.g., usage schedule). In addition, the auxiliary data may include, for example, data related to the day to be predicted (prediction date) or the period to be predicted (e.g., time period), that is, data such as the temperature (predicted value), humidity (predicted value), number of occupants (predicted value) for the day or period to be predicted. The auxiliary data may also include, for example, data relating to a past day or period (eg, time period), that is, temperature, humidity, number of occupants, etc., for a past day or period. Furthermore, as auxiliary data, for example, weather-related data other than temperature and humidity may be used.

[0051] The performance data is data that represents past performance data regarding heat load. As the performance data, for example, instead of or in addition to the data on the thermal load performance, data representing performance regarding other items may be used. As the other items, for example, power or the like may be used. Note that in this embodiment, for convenience of explanation, the performance data is shown as data with a different name from the auxiliary data, but the performance data is a type of auxiliary data, and they may be collectively referred to by the same name (for example, a name such as auxiliary data). In other words, in this embodiment, the various information included in the auxiliary data and the various information included in the performance data can be used as information auxiliary to the air conditioning reservation information (auxiliary information).

[0052] Here, when predicting the heat load, for example, the performance data may not be used as input data to the learning model M1 (trained model). For example, the reservation data and auxiliary data may be used as input data to the learning model M1 (trained model).

[0053] In addition, some or all of the input data to the learning model M1 may be subjected to dimensional compression (dimension reduction) before being input. Any method may be used as the dimensional compression method, and for example, a method of principal componentization (principal component analysis) may be used. For example, dimensionality reduction may be performed on some or all of the auxiliary data, and the data after dimensionality reduction may be used as input data to the learning model M1.

[0054] In this embodiment, the heat load prediction unit 153 of the heat load prediction device 11 in the heat load prediction system 1 inputs predetermined input data to a learning model M1 for heat load prediction and obtains output data from the learning model M1. The output data includes information about the result of the heat load prediction. The learning model M1 may be stored in the storage unit 114 of the heat load prediction device 11, for example.

[0055] Furthermore, in the thermal load prediction device 11, for example, among the input data to the learning model M1, data stored inside the thermal load prediction device 11 (for example, in the memory unit 114) (or data obtained by processing the data) is acquired internally, and data present outside is acquired from outside. As an example, the thermal load prediction device 11 may be communicatively connected to an external system (e.g., an air conditioning reservation system 431 or an air conditioning management system 411), and the thermal load prediction device 11 may acquire necessary data from the external system. As another example, the thermal load prediction device 11 may include functions of other systems (e.g., an air conditioning reservation system 431 or an air conditioning management system 411), and the thermal load prediction device 11 may obtain necessary data from those functions.

[0056] <Learning model> In this embodiment, the model learning unit 152 of the heat load prediction device 11 in the heat load prediction system 1 performs machine learning of the learning model M1 in advance (i.e., before heat load prediction is performed) so that desired output data is output from the learning model M1 when predetermined input data is input to the learning model M1. In the machine learning, for example, parameters (e.g., calculation coefficients) of the calculation performed in the learning model M1 are set to adaptive values.

[0057] In this embodiment, the model learning unit 152 performs machine learning of the learning model M1 using predetermined training data. The training data may include data (e.g., teacher data) that represents a correct value of a heat load prediction result. The training data may be, for example, data of actual measured values ​​of heat loads in the past.

[0058] Furthermore, when learning the learning model M1, just as when predicting thermal load using the learning model M1 (here, a trained model), the thermal load prediction device 11 acquires, for example, among the input data to the learning model M1, data stored inside the thermal load prediction device 11 (for example, in the memory unit 114) (or data obtained by processing the data) internally, and acquires data that is external from outside.

[0059] <Specific examples of learning models> FIG. 6 is a diagram illustrating an example of a heat load prediction using a trained model according to an embodiment. FIG. 6 shows a schematic overview of the calculations in the learning model M1. The calculation (model configuration) in the learning model M1 is not limited to this example, and models of various configurations may be used.

[0060] The learning model M1 includes calculation units 611-612, 621-622, and a calculation unit 631. The calculation unit 611 has a function of learning time-series fluctuations, and learns the relationship between past data and future data. The calculation unit 611 has a recurrent neural network 711 that learns monthly fluctuations, a recurrent neural network 712 that learns weekly fluctuations, and a recurrent neural network 713 that learns daily fluctuations.

[0061] Recurrent neural network 711 receives data including information indicating monthly fluctuations, performs a predetermined calculation based on the data, and outputs the result of the calculation to calculation unit 621. Recurrent neural network 712 receives data including information indicating fluctuations on a weekly basis, performs a predetermined calculation based on the data, and outputs the result of the calculation to calculation unit 621. The recurrent neural network 713 receives data including information indicating daily fluctuations, performs a predetermined calculation based on the data, and outputs the result of the calculation to the calculation unit 621.

[0062] The calculation unit 612 has a function of learning the waveform, and learns the waveform as an image. The calculation unit 612 has a convolutional neural network 721 that learns monthly fluctuations (waveform fluctuations), a convolutional neural network 722 that learns weekly fluctuations (waveform fluctuations), and a convolutional neural network 723 that learns daily fluctuations (waveform fluctuations).

[0063] The convolutional neural network 721 receives data including information representing monthly fluctuations, performs a predetermined calculation based on the data, and outputs the result of the calculation to the calculation unit 622. The convolutional neural network 722 receives data including information representing fluctuations on a weekly basis, performs a predetermined calculation based on the data, and outputs the result of the calculation to the calculation unit 622. The convolutional neural network 723 receives data including information representing daily fluctuations, performs a predetermined calculation based on the data, and outputs the result of the calculation to the calculation unit 622.

[0064] The calculation unit 621 calculates a weighted average of the calculation results output from the three recurrent neural networks 711 to 713 with a predetermined weight, and outputs the weighted average result to the calculation unit 631 as a time-series prediction result. The calculation unit 622 weight-averages the calculation results output from the three convolutional neural networks 721 to 723 with a predetermined weighting, and outputs the weighted average result to the calculation unit 631 as a prediction result based on the waveform.

[0065] The calculation unit 631 calculates a weighted average of the prediction result based on the time series output from the calculation unit 631 and the prediction result based on the waveform output from the calculation unit 622 with a predetermined weight, and outputs the weighted average result as the final prediction result (prediction result regarding thermal load) to the outside of the learning model M1.

[0066] <Input data> Here, the data input to the recurrent neural network 711 and the convolutional neural network 721 may be, for example, actual data for one or more months prior to the month in which the prediction target date is included, or, as a specific example, actual data for each of multiple consecutive months in the past may be used. For example, time-series numerical data is input to the recurrent neural network 711, and image data representing a waveform is input to the convolutional neural network 721.

[0067] In addition, the data input to the recurrent neural network 712 and the convolutional neural network 722 may be, for example, actual data for one or more weeks prior to the week in which the prediction target date is included. As a specific example, actual data for the same day of the week for multiple consecutive weeks in the past (for example, weeks in the same month) may be used. For example, time-series numerical data is input to the recurrent neural network 712, and image data representing a waveform is input to the convolutional neural network 722.

[0068] In addition, the data input to the recurrent neural network 713 and the convolutional neural network 723 may be, for example, actual data for one or more days prior to the prediction target date, and as a specific example, actual data for each of a number of consecutive days in the past (for example, days of the same week) may be used. For example, time-series numerical data is input to the recurrent neural network 713, and image data representing a waveform is input to the convolutional neural network 723.

[0069] In addition, for example, reservation data is also input to each of the three recurrent neural networks 711 to 713 and the three convolutional neural networks 721 to 723. The reservation data may be, for example, reservation data for a prediction target date. Furthermore, for example, auxiliary data may also be input to each of the three recurrent neural networks 711 to 713 and the three convolutional neural networks 721 to 723. In this embodiment, availability data is used as the form of reservation data used as input data to the learning model M1.

[0070] <Weighting example> In the example of Figure 6, in learning model M1, weighting is performed at the stage where the output from calculation unit 611 is input to calculation unit 621, weighting is performed at the stage where the output from calculation unit 612 is input to calculation unit 622, and weighting is performed at the stage where the outputs from calculation unit 621 and calculation unit 622 are input to calculation unit 631. In this embodiment, these weights can change depending on, for example, the facility to which the heat load prediction system 1 according to this embodiment is applied (the facility corresponding to the building 301 in this embodiment). Furthermore, in this embodiment, these weights can change depending on, for example, the timing of use of air conditioning in the facility (air conditioning use timing). In the learning model M1, when machine learning training is performed, the weightings are updated, and the values ​​of the weightings may change depending on, for example, one or both of the facility and the timing of air conditioning use.

[0071] As described above, the heat load prediction system 1 according to this embodiment is capable of performing highly accurate heat load prediction for central air conditioning in accordance with the facility and the timing of air conditioning use using advance air conditioning reservation information. In the heat load prediction system 1 according to this embodiment, even when the heat load record and operation time are irregular, for example, it is possible to perform heat load prediction with high accuracy by using advance reservation information.

[0072] In the thermal load prediction system 1 of this embodiment, advance reservation information (e.g., information on operating rates) is utilized, and actual thermal load data, as well as auxiliary data such as weather information and calendar information are used to perform machine learning using AI to extract and record characteristics related to thermal load usage, thereby improving the prediction accuracy for the target day (e.g., the following day). As a result, the heat load prediction system 1 of this embodiment can create an operation schedule based on the capacity required for the heat load demand predicted the previous day, etc., thereby realizing efficient energy-saving operation.

[0073] In the heat load prediction system 1 of this embodiment, for example, for a facility using central air conditioning, machine learning of the learning model M1 is performed using information such as the most recent temperature and humidity as well as reservation information regarding air conditioning (facility reservation information), and the accuracy of the heat load prediction value can be improved by changing the weighting based on each facility and the timing of use.

[0074] In the heat load prediction system 1 of this embodiment, by performing calculations based on the reservation status of facilities such as rental conference rooms or halls, it is expected that the accuracy of heat load predictions for facilities that are used irregularly from day to day can be improved. In the thermal load prediction system 1 according to the present embodiment, for example, by handling reservation information that takes into account the rated capacity of an air conditioner, it is possible to predict and grasp the thermal load demand more accurately. The heat load prediction system 1 according to this embodiment is expected to be particularly effective when introduced into a complex air conditioning management system (in this embodiment, the air conditioning management system 411) that manages the operation of multiple buildings and multiple heat sources, for example.

[0075] Here, the technology according to this embodiment may be applied to the management of any facility, such as a conference room, a hall, or a hotel guest room, and may also be applied to any energy management, such as energy saving.

[0076] [Configuration example] As an example configuration, the thermal load prediction system 1 includes an acquisition unit (in the examples of Figures 1 and 5, the input processing unit 151) that acquires air conditioning reservation information for a facility that uses central air conditioning (in the example of Figure 2, the facility corresponds to building 301), and a thermal load prediction unit (in the example of Figure 1, the thermal load prediction unit 153) that inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model (in the example of Figure 1, the learning model M1) whose weighting can change based on the facility and the timing of air conditioning use, and acquires output data including information regarding the predicted results of the thermal load output from the learning model. Therefore, with regard to central air conditioning, the heat load prediction system 1 can perform highly accurate heat load prediction according to the facility and the timing of air conditioning use using advance air conditioning reservation information.

[0077] As one configuration example, in the thermal load prediction system 1, the learning model includes a first calculation unit (in the example of Figure 6, calculation unit 611) that learns time series fluctuations, and a second calculation unit (in the example of Figure 6, calculation unit 612) that learns waveform images. Therefore, in the thermal load prediction system 1, by using both time-series fluctuation learning and waveform image learning in machine learning, it is possible to perform thermal load prediction with higher accuracy.

[0078] As an example of the configuration, in the heat load prediction system 1, the first calculation unit has a recurrent neural network (in the example of FIG. 6, recurrent neural networks 711 to 713). The second calculation unit has a convolutional neural network (in the example of FIG. 6, convolutional neural networks 721 to 723). Therefore, the thermal load prediction system 1 can perform highly accurate thermal load prediction by using neural networks suited for learning time-series fluctuations and learning waveform images, respectively.

[0079] As an example of the configuration, in the heat load prediction system 1, the recurrent neural network includes a neural network corresponding to monthly fluctuations (in the example of FIG. 6, recurrent neural network 711), a neural network corresponding to weekly fluctuations (in the example of FIG. 6, recurrent neural network 712), and a neural network corresponding to daily fluctuations (in the example of FIG. 6, recurrent neural network 713). The convolutional neural network includes a neural network corresponding to monthly fluctuations (in the example of FIG. 6, convolutional neural network 721), a neural network corresponding to weekly fluctuations (in the example of FIG. 6, convolutional neural network 722), and a neural network corresponding to daily fluctuations (in the example of FIG. 6, convolutional neural network 723). Therefore, the heat load prediction system 1 can perform heat load prediction with high accuracy by using a learning model that takes into account monthly, weekly, and daily fluctuations.

[0080] As one configuration example, in the heat load prediction system 1, the auxiliary information includes one or more of information on weather, information on the number of people in the room, information on zone information, information on a calendar, and information on actual power or heat load. Therefore, the heat load prediction system 1 can perform heat load prediction with high accuracy based on various information other than air conditioning reservation information.

[0081] As one configuration example, in the heat load prediction system 1, the air conditioning reservation information includes information on the operation rate of the central air conditioning. Therefore, the heat load prediction system 1 can perform heat load prediction with high accuracy based on information on the operation rate as air conditioning reservation information.

[0082] In this embodiment, it is also possible to provide a heat load prediction device. As one configuration example, the thermal load prediction device 11 includes an acquisition unit that acquires air conditioning reservation information for a facility that uses central air conditioning, and a thermal load prediction unit that inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use, and acquires output data including information regarding the thermal load prediction result output from the learning model. Therefore, with regard to central air conditioning, the thermal load prediction device 11 can perform highly accurate thermal load prediction according to the facility and the timing of air conditioning use using advance air conditioning reservation information.

[0083] In this embodiment, it is also possible to provide a thermal load prediction method. As one configuration example, in a thermal load prediction method, an acquisition unit acquires air conditioning reservation information for a facility that uses central air conditioning, and a thermal load prediction unit inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use, and acquires output data including information regarding the thermal load prediction result output from the learning model. Therefore, with regard to central air conditioning, the heat load prediction method can perform highly accurate heat load prediction according to the facility and the timing of air conditioning use using advance air conditioning reservation information. The heat load prediction method is, for example, a processing method performed by the heat load prediction device 11.

[0084] In this embodiment, it is also possible to provide a program that is executed by the processor of a computer. As one configuration example, the program is a program for enabling a computer constituting a thermal load prediction device to realize the following functions: acquiring air conditioning reservation information for a facility that uses central air conditioning; inputting input data including the acquired air conditioning reservation information and auxiliary information into a machine learning learning model in which weighting can change based on the facility and the timing of air conditioning use; and acquiring output data including information regarding the thermal load prediction results output from the learning model. Therefore, with regard to central air conditioning, this program can perform highly accurate heat load predictions according to the facility and the timing of air conditioning use using advance air conditioning reservation information. The computer is, for example, a computer constituting the heat load prediction device 11.

[0085] A program for implementing the functions of any component in any device may be recorded in a computer-readable recording medium, and the program may be read and executed by a computer system. The term "computer system" as used herein includes an operating system (OS) or hardware such as peripheral devices. The term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD (Compact Disc)-ROMs, and storage devices such as hard disks built into a computer system. The term "computer-readable recording medium" also refers to storage devices that hold a program for a certain period of time, such as volatile memory (RAM) in a computer system that serves as a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. The recording medium may be, for example, a non-transitory recording medium.

[0086] The above program may be transmitted from a computer system in which the program is stored in a storage device or the like to another computer system via a transmission medium, or by transmission waves in the transmission medium. Here, the "transmission medium" that transmits the program refers to a medium that has a function of transmitting information, such as a network such as the Internet or a communication line such as a telephone line. The above program may be for implementing some of the above functions. Furthermore, the above program may be a so-called differential file that can implement the above functions in combination with a program already recorded in the computer system. The differential file may be called a differential program.

[0087] In addition, the function of any component in any device may be realized by a processor. For example, each process in the present embodiment may be realized by a processor that operates based on information such as a program and a computer-readable recording medium that stores information such as a program. Here, the processor may, for example, realize the functions of each part by individual hardware, or may realize the functions of each part by integrated hardware. For example, the processor may include hardware, and the hardware may include at least one of a circuit that processes a digital signal and a circuit that processes an analog signal. For example, the processor may be configured using one or more circuit devices mounted on a circuit board, or one or both of one or more circuit elements. An IC (Integrated Circuit) or the like may be used as the circuit device, and a resistor or a capacitor or the like may be used as the circuit element.

[0088] Here, the processor may be, for example, a CPU. However, the processor is not limited to a CPU, and various processors such as a GPU (Graphics Processing Unit) or a DSP (Digital Signal Processor) may be used. The processor may be, for example, a hardware circuit using an ASIC (Application Specific Integrated Circuit). The processor may be, for example, a plurality of CPUs, or a hardware circuit using a plurality of ASICs. The processor may be, for example, a combination of a plurality of CPUs and a hardware circuit using a plurality of ASICs. The processor may include, for example, one or more of an amplifier circuit or a filter circuit that processes an analog signal.

[0089] Although the present disclosure has been described using the embodiments, the technical scope of the present disclosure is not limited to the above-mentioned embodiments. It is obvious to those skilled in the art that various modifications and alternative aspects can be adopted without departing from the spirit and scope of the present disclosure. [Explanation of symbols]

[0090] 1...heat load prediction system, 11...heat load prediction device, 111...operation unit, 112...display unit, 113...input / output unit, 114...storage unit, 115...control unit, 151...input processing unit, 152...model learning unit, 153...heat load prediction unit, 154...output processing unit, 301...building, 311-1 to 311-3...room, 331-1 to 331-3...air conditioner, 351...heat source equipment, 352...cooling tower, 411...air conditioning management system, 431...air conditioning reservation system, 611 to 612, 621 to 622, 631...calculation unit, 711 to 713...recurrent neural network, 721 to 723...convolutional neural network, 2011 to 2012, 2111 to 2112...characteristics, M1...learning model, T1...table

Claims

1. an acquisition unit that acquires air conditioning reservation information related to a facility using central air conditioning; a heat load prediction unit that inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can be changed based on the facility and the timing of air conditioning use, and acquires output data including information regarding the prediction result of the heat load output from the learning model; A heat load prediction system comprising:

2. The learning model includes a first calculation unit that has learned time series fluctuations and a second calculation unit that has learned waveform images. The heat load prediction system according to claim 1 .

3. The first calculation unit has a recurrent neural network, The second calculation unit has a convolutional neural network. The heat load prediction system according to claim 2 .

4. the recurrent neural network includes a neural network corresponding to monthly fluctuations, a neural network corresponding to weekly fluctuations, and a neural network corresponding to daily fluctuations; The convolutional neural network includes a neural network corresponding to monthly fluctuations, a neural network corresponding to weekly fluctuations, and a neural network corresponding to daily fluctuations. The heat load prediction system according to claim 3 .

5. The auxiliary information includes at least one of information regarding weather, information regarding the number of people in the room, information regarding zone information, information regarding a calendar, or information regarding actual power or heat load. The heat load prediction system according to claim 1 .

6. The air conditioning reservation information includes information on the operation rate of the central air conditioning, The heat load prediction system according to any one of claims 1 to 5.

7. an acquisition unit that acquires air conditioning reservation information related to a facility using central air conditioning; a heat load prediction unit that inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can be changed based on the facility and the timing of air conditioning use, and acquires output data including information regarding the prediction result of the heat load output from the learning model; A heat load prediction device comprising:

8. The acquisition unit acquires air conditioning reservation information regarding a facility using central air conditioning, A heat load prediction unit inputs input data including the air conditioning reservation information and auxiliary information acquired by the acquisition unit into a machine learning learning model in which weighting can be changed based on the facility and the timing of air conditioning use, and acquires output data including information regarding the prediction result of the heat load output from the learning model. Heat load forecasting methods.

9. A computer constituting the heat load prediction device includes: A function to obtain air conditioning reservation information for facilities using central air conditioning; A function of inputting input data including the acquired air conditioning reservation information and auxiliary information into a machine learning learning model in which weighting can be changed based on the facility and the timing of air conditioning use, and acquiring output data including information regarding the heat load prediction result output from the learning model; A program to achieve this.

Images