Load prediction device, load prediction method, and program
By constructing a load prediction device and using a database and learning model to predict the peak load at the start-up of the energy supply device, the problem of low prediction accuracy in the existing technology is solved, and high-precision load prediction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- FUJI ELECTRIC CO LTD
- Filing Date
- 2022-03-24
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies struggle to accurately predict the peak heat load of energy supply devices, such as heat source devices, during startup, especially due to insufficient data types leading to low prediction accuracy.
By constructing a load prediction device, utilizing a database that stores past load performance and operating schedules of energy supply devices, and combining sensor data, a learning model is used to predict peak loads at startup, including predictions of startup date and time.
It enables high-precision load prediction when energy supply devices are started up, improving the accuracy and reliability of the prediction.
Smart Images
Figure CN115409232B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to load prediction devices, load prediction methods, and programs. Background Technology
[0002] Previously known technologies exist for predicting the demand for various forms of energy, such as heat and electricity (e.g., Patent Documents 1-3). These technologies construct models based on past performance data and use these models to predict future demand. These technologies enable the prediction of the load on energy supply devices used to supply energy.
[0003] <Prior art documents>
[0004] <Patent Documents>
[0005] Patent Document 1: Japanese Invention Patent No. 3360520
[0006] Patent Document 2: Japanese Invention Patent No. 6187003
[0007] Patent Document 3: Japanese Patent Application Publication No. 2001-216001 Summary of the Invention
[0008] <Problem to be solved by this invention>
[0009] However, for example, in heat source devices that supply heat as energy, it is known that a peak-shaped heat load is generated during startup. This is generated in order to heat (or cool) the heat medium to a set temperature.
[0010] However, for example, Patent Documents 1 and 2 mentioned above are used to predict the stable demand (i.e., the demand under normal conditions) after the heat source device is started up, but they cannot predict the heat load of the heat source device at startup. On the other hand, for example, Patent Document 3 mentioned above can predict the heat demand at startup / shutdown of the heat source device, but the types of data are insufficient for accurate prediction. Specifically, since the prediction is based only on a few past load data, the accuracy of the predicted heat load of the heat source device at startup is low.
[0011] One embodiment of the present invention is made in view of the above points, and its purpose is to predict the load, including the load at startup of the energy supply device.
[0012] <Methods for solving problems>
[0013] To achieve the above objective, one embodiment of the load prediction device predicts the load of an energy supply device. The load prediction device has a first prediction unit that refers to a database storing at least one of the past load performance values of the energy supply device and the operating schedule of the energy supply device, and predicts the predicted value of the peak load generated when the energy supply device is started, as well as the date and time of the peak load.
[0014] <The Effects of the Invention>
[0015] The present invention is capable of predicting the load, including the load during the startup of energy supply devices. Attached Figure Description
[0016] Figure 1 This is a diagram illustrating an example of the configuration of a heat source device and a load device.
[0017] Figure 2 This is a diagram illustrating an example of the heat load on a heat source device.
[0018] Figure 3 This is a diagram illustrating an example of the overall configuration of the load prediction system according to this embodiment.
[0019] Figure 4 This is a diagram illustrating an example of the load prediction processing in this embodiment.
[0020] Figure 5 This is a diagram illustrating an example of the hardware configuration of the load prediction device according to this embodiment.
[0021] Explanation of reference numerals in the attached figures
[0022] 100 Heat source device
[0023] 200 load device
[0024] 310, 320 catheters
[0025] 400 sensor group
[0026] 410 Inlet Temperature Sensor
[0027] 420 Outlet Temperature Sensor
[0028] 430 Flow Sensor
[0029] 500 Load Prediction Device
[0030] 510 Startup Prediction Department
[0031] 520 Prediction Department during Stable Operation
[0032] 521 Study Department
[0033] 522 Forecasting Department
[0034] 600 Database Server
[0035] 610 Database
[0036] 700 terminal
[0037] 800 communication network
[0038] 901 Input Device
[0039] 902 Display Device
[0040] 903 External I / F
[0041] 903a Recording Media
[0042] 904 Communication I / F
[0043] 905 processor
[0044] 906 memory device
[0045] 907 bus Detailed Implementation
[0046] Hereinafter, one embodiment of the present invention will be described. Hereinafter, as an example of an energy supply device, a heat source device will be assumed, and the description will focus on predicting the heat load (especially the heat load during startup) when the energy supply is heat (including both hot and cold). However, this is just an example, and can also be applied to situations where various forms of energy other than heat, such as electricity and steam, are used as the source of energy, and the load on the energy supply device supplying that energy is predicted.
[0047] <Composition of Heat Source Device and Load Device>
[0048] exist Figure 1 An example of the configuration of the heat source device and load device assumed in this embodiment is shown. Figure 1 As shown, the heat source device 100 and the load device 200 are connected via conduits 310 and 320, through which water, an example of a heat medium, is circulated. It should be noted that the water is circulated via a pump (not shown) through conduits 310 and 320.
[0049] The heat source device 100 heats the cold water introduced through the conduit 310 and discharges it as warm water into the conduit 320. On the other hand, the load device 200 utilizes the heat of the warm water introduced through the conduit 320 and discharges it as cold water into the conduit 320. Examples of load devices 200 include air conditioning equipment in buildings such as buildings, load devices in factories (e.g., sterilization equipment for beverages), or load devices in specific areas.
[0050] Additionally, an inlet temperature sensor 410 is provided within the heat source device 100 (or conduit 310) to measure the temperature of cold water (or cold water to be introduced into the heat source device 100 from the conduit 310). Similarly, an outlet temperature sensor 420 is provided within the heat source device 100 (or conduit 320) to measure the temperature of warm water (or warm water discharged from the heat source device 100 into the conduit 320). Furthermore, a flow sensor 430 is provided within the conduit 320 to measure the flow rate of the warm water within the conduit 320.
[0051] Here, the temperature measured by the inlet temperature sensor 410 (inlet side temperature) is set as T1, the temperature measured by the outlet temperature sensor 420 (outlet side temperature) is set as T2, the flow rate measured by the flow sensor 430 is set as P, and the specific heat of the heat medium is set as C. The heat load L of the heat source device 100 is then calculated as follows.
[0052] L=|T2-T1|×P×C
[0053] It should be noted that, Figure 1 The configuration shown is an example where, for instance, the heat source device 100 can introduce warm water through conduit 310 and cool that warm water (that is, it can supply both hot and cold water). In this case, the load device 200 utilizes the cold water introduced through conduit 320 and discharges it as warm water from conduit 310. Additionally, in Figure 1 In the example shown, although water is used as a heat medium, it is undeniable that heat media other than water can also be used.
[0054] Moreover, in Figure 1 In the example shown, for simplicity, only one heat source device 100 and one load device 200 are recorded, but multiple devices can exist.
[0055] <Heat load of heat source device>
[0056] exist Figure 2 An example of the heat load of the heat source device 100 is shown. Figure 2The example shown illustrates a scenario where the heat source device 100 is started at time t1, stopped at time t4, and the heat supply from the heat source device 100 ceases at time t5. In this case, as... Figure 2 As shown, after starting at time t1, a peak-shaped heat load (in other words, a peak-shaped heat demand) is generated in the heat source device 100 at a certain time t2, and a stable heat load is generated after a certain time t3.
[0057] Here, the stable heat load between time t3 and time t4 consists of the heat demanded by the load device 200 (the actual load) and the heat loss in the conduit 320. On the other hand, the peak heat load between time t1 and time t3 consists of the heat demanded by the load device 200 (the actual load), the heat loss in the conduit 320, and the heat used to heat (or cool) the heat medium in the conduit 320 to a set temperature. That is, the peak heat load is generated by heating (or cooling) the heat medium in the conduit 320 to a certain set temperature. It should be noted that the set temperature is a temperature preset in the heat source device 100, which is the target temperature of the heat medium in the conduit 320 during operation (in other words, the target temperature at the outlet side of the heat source device 100).
[0058] Hereinafter, the time from the start-up of the heat source device 100 until a stable heat load is generated (that is, the time from time t1 to time t3) will be referred to as the start-up time; the time during which a stable heat load is generated (that is, the time from time t3 to time t4) will be referred to as the stable operation time; and the time from the start of the shutdown operation of the heat source device 100 until it stops (that is, the time from time t4 to time t5) will be referred to as the shutdown time. It should be noted that the stable operation time can also be referred to as the normal operation time, etc.
[0059] At this point, the objective of this embodiment is to perform high-precision prediction of the heat load during startup and the heat load during stable operation of the heat source device 100 (in particular, to perform high-precision prediction of the heat load during startup). It should be noted that since the heat load during stable operation can be predicted with high precision using prior art as described in, for example, Patent Document 1, prior art is also used in this embodiment for predicting the heat load during stable operation.
[0060] It should be noted that, in Figure 2 In the example shown, for simplicity, the heat load during stable operation is set to constant. However, this is just an example, and a non-constant heat load may also occur. That is, it is sufficient to generate a normal heat load (or a heat load that can be considered normal) during stable operation; the load does not need to be constant.
[0061] <Overall Structure of a Load Forecasting System>
[0062] Figure 3 The overall configuration of the load prediction system of this embodiment is shown in the figure. Figure 3 As shown, the load prediction system in this embodiment includes a load prediction device 500, a database server 600, a terminal 700, and a sensor group 400, and each of them is connected in a manner that enables them to communicate via a communication network 800.
[0063] Sensor group 400 includes various sensor groups such as inlet temperature sensor 410, outlet temperature sensor 420, and flow sensor 430. It should be noted that sensor group 400 also includes sensors for measuring the load values of heat source device 100.
[0064] The load prediction device 500 is a computer that predicts the thermal load of the heat source device 100 during startup and during stable operation, referring to the database 610 provided by the database server 600. Here, the load prediction device 500 includes a startup prediction unit 510 and a stable operation prediction unit 520. These units are implemented, for example, by having a computing device such as a CPU (Central Processing Unit) execute one or more programs installed on the load prediction device 500.
[0065] The startup prediction unit 510 is used to predict the heat load of the heat source device 100 during startup. The stable operation prediction unit 520 predicts the heat load during stable operation, for example, using the prediction model described in Patent Document 1. Here, the stable operation prediction unit 520 includes a learning unit 521 for constructing the prediction model (making the prediction model learn), and a prediction unit 522 for predicting the heat load during stable operation using the prediction model.
[0066] Database server 600 has a database 610 for storing various types of data. Database 610 stores, for example, calendar information such as load performance of heat source device 100, various sensor values, and schedules, as well as meteorological information including temperature. Data representing load performance and various sensor values are collected from sensor group 400 via communication network 800. Additionally, data representing calendar information and meteorological information are input from terminal 700 via communication network 800. However, data representing meteorological information can be collected from an external server, such as via the Internet. Here, calendar information includes the operating schedule of heat source device 100 (that is, the start-up date and time, stop date and time, and maintenance date and time of heat source device 100, etc.).
[0067] Terminal 700 stores the aforementioned calendar information, weather information, etc., in database 610 via communication network 800. It should be noted that the calendar information, weather information, etc., can be input by the user, for example, through a screen displayed on terminal 700, or through a file containing calendar information, weather information, etc.
[0068] It should be noted that, Figure 1 The overall configuration of the load forecasting system shown is an example. For instance, at least one of the load forecasting device 500, database server 600, and terminal 700 can be implemented by a single device.
[0069] <Load Prediction Processing>
[0070] exist Figure 4 The flowchart of the load prediction process in this embodiment is shown. Figure 4 As shown, the load forecasting process consists of a learning phase and a forecasting phase. Step S101 is the learning phase, and steps S102 to S103 are the forecasting phases. It should be noted that the learning phase is executed before the forecasting phase.
[0071] During stable operation, the learning unit 521 of the prediction unit 520 uses various data stored in the database 610 (data representing the load performance of the heat source device 100, various sensor values, calendar information, meteorological information, etc.) to learn a prediction model for predicting the load during stable operation (step S101). Here, the learning unit 521 can learn, for example, the prediction model described in Patent Document 1. It should be noted that the prediction model described in Patent Document 1 is a model implemented by a neural network (especially a recurrent neural network), which is used to predict the load during stable operation. For details on the structure of this prediction model and the learning method, please refer to Patent Document 1. However, it is not limited to the prediction model described in Patent Document 1; any model capable of predicting the load during stable operation can be used, and any prediction model can be constructed through learning.
[0072] At startup, the prediction unit 510 uses various data stored in the database 610 (data representing the actual load of the heat source device 100, various sensor values, calendar information, etc.) to predict the heat load of the heat source device 100 at startup (step S102). Details of this prediction method will be described later.
[0073] Next, the prediction unit 522 of the stable operation prediction unit 520 uses various data stored in the database 610 (data representing the actual load of the heat source device 100, various sensor values, calendar information, meteorological information, etc.) and the prediction model learned during the learning phase to predict the heat load during stable operation (step S103). For details of the prediction method based on this prediction model, please refer to Patent Document 1. However, it is undeniable that predictions can be made using models other than the prediction model described in Patent Document 1.
[0074] Thus, the load prediction device 500 of this embodiment distinguishes between the start-up and stable operation of the heat source device 100 and predicts the load accordingly. As will be described later, since the load prediction device 500 of this embodiment can perform high-precision prediction of the load during the start-up of the heat source device 10, it can perform high-precision prediction of the load during both the start-up and stable operation.
[0075] It should be noted that the load prediction device 500 of this embodiment can, for example, display the prediction results obtained in steps S102 to S103 above on the terminal 700, and can also control various machines or devices based on the prediction results. For example, based on the prediction results, the operation of devices or machines that supply fuel or electricity to the heat source device 100 can be controlled, the operation of the heat source device 100 itself can be controlled, and the operation of the load device 200 can also be controlled.
[0076] <Methods for predicting thermal load during startup>
[0077] The following describes the method for predicting the heat load at startup in step S102 above. Here, we will define the heat load L at startup or over a time span Δ of the heat source device 100 on the predicted day as given by the user, such as the person in charge of the operation. inf Make a prediction. That is, set the heat load L. inf The average thermal load represents the time span Δ.
[0078] It should be noted that although the time width Δ can be set to any time width such as 15 minutes or 30 minutes, it is typically considered to be set to Δ = t3 - t1. When Δ = t3 - t1 is set, the average value of the heat load at the start-up of the heat source device 100 is predicted.
[0079] Method 1
[0080] Method 1 is a simplified prediction method. The peak heat load during the startup of the heat source device 100 is mainly the load generated to bring the outlet temperature T2 to the set temperature. Generally, this load is large, and it often becomes the rated output of the heat source device 100 (that is, the maximum output under startup conditions). Therefore, in Method 1, L is set... inf =Rated output.
[0081] Here, the moment when the peak occurs is the predicted heat load L. inf The generation time can be set to the start time of the predicted day in the operation schedule contained in the calendar information (the start time of the predicted day for heat source device 100). Alternatively, if there is no operation schedule, the time when a peak occurs in the load performance of heat source device 100 over the past few days (e.g., a specified past period such as the past week) can be used, and that time in the predicted day can be set as the predicted heat load L. inf The moment it is generated is sufficient.
[0082] Therefore, it is possible to easily and with high accuracy predict the daily heat load L. inf It also predicts the time of its occurrence.
[0083] Method 2
[0084] Method 2 is a more rigorous prediction method compared to Method 1. In Method 2, the set temperature is set to T. s The heat load L is calculated using the following formula. inf Make predictions.
[0085] L inf =|T s -T1|×P×C
[0086] It should be noted that T1 and P also represent the average value within the time width Δ.
[0087] Here, since the inlet temperature T1 and flow rate P at the predicted date and time (i.e., the time when the peak occurs in the predicted day) cannot be obtained, T1 is set, for example, as the actual value of the inlet temperature at the predicted time point, or as the average of the actual values of the inlet temperature at the same time as the predicted time point among the actual values of the inlet temperature over the past few days (e.g., the past week, etc., a specified past period). Similarly, P is set, for example, as the rated value of duct 320 (i.e., the maximum flow rate under startup conditions). However, similar to T1, P can be set, for example, as the actual value of the flow rate at the predicted time point, or as the average of the actual values of the flow rate at the same time as the predicted time point among the actual values of the flow rate over the past few days.
[0088] It should be noted that the predicted time of the midday peak (that is, the predicted heat load L) is... inf The time of its generation can be determined using the same method as in Method 1.
[0089] Therefore, compared with method 1, it is possible to predict the daily heat load L with higher accuracy. inf It also predicts the time of its occurrence.
[0090] Method 3
[0091] With Δ = t3 - t1 set, in both Method 1 and Method 2 described above, the predicted heat load L at startup can be obtained. inf While the exact time of its generation is unknown, by setting Δ more precisely, it is possible to predict the heat load with high accuracy.
[0092] That is, for example, based on setting Δ1=((t3-t1) / 2)-t1 and Δ2=t3-((t3-t1) / 2), the heat load L is adjusted at Δ1 and Δ2 respectively using the above method 1 or method 2. inf It also predicts the time of its occurrence.
[0093] In addition, more generally, N can be set to a predetermined integer of 2 or higher, and the values can be set at Δ1, ..., Δ... N The heat load L is processed using either method 1 or method 2 described above. inf It also predicts the time of its occurrence.
[0094] Therefore, using method 1 or method 2, it is possible to predict the daytime heat load L. inf And to make more detailed predictions about the time of its occurrence (i.e., to make predictions with higher accuracy).
[0095] <Hardware Components of a Load Prediction Device>
[0096] Figure 5 The hardware configuration of the load prediction device 500 of this embodiment is shown in the figure. Figure 5 As shown, the load prediction device 500 of this embodiment is implemented using general computer or computer system hardware, such as an input device 901, a display device 902, an external I / O panel 903, a communication I / O panel 904, a processor 905, and a memory device 906. These hardware components are connected in a manner that allows them to communicate via a bus 907. It should be noted that since the database server 600 and the terminal 700 can also be implemented using the same hardware, the description of the hardware configuration of the database server 600 and the terminal 700 is omitted.
[0097] Input device 901 may be, for example, a keyboard, mouse, touch panel, or physical buttons. Display device 902 may be, for example, a monitor. It should be noted that the load prediction device 500 may not have at least one of the input device 901 and display device 902.
[0098] The external I / F 903 is an interface to external devices such as the recording medium 903a. The load prediction device 500 can read from and write to the recording medium 903a via the external I / F 903. It should be noted that, for example, the recording medium 903a can be a CD (Compact Disc), DVD (Digital Versatile Disk), SD memory card (Secure Digital Memory Card), or USB (Universal Serial Bus) memory card.
[0099] Communication I / F 904 is an interface used to connect the load prediction device 500 to the communication network 800. Processor 905 is, for example, a CPU, GPU (Graphics Processing Unit), or various other computing devices. Memory device 906 is, for example, a HDD (Hard Disk Drive), SSD (Solid State Drive), flash memory, RAM (Random Access Memory), ROM (Read Only Memory), or various other storage devices.
[0100] The load prediction device 500 of this embodiment has Figure 5 The hardware configuration shown enables the aforementioned load prediction processing. It should be noted that... Figure 5 The hardware configuration shown is an example; the load prediction device 500 can have various hardware configurations. For example, the load prediction device 500 can have multiple processors 905 and multiple memory devices 906.
[0101] This invention is not limited to the specific embodiments disclosed above, and various modifications, alterations, and combinations with known technologies can be made without exceeding the scope of the claims.
Claims
1. A load forecasting device for forecasting the load of a heat supply device used to supply heat to heat-demanding devices. The load prediction device has a first prediction unit that, with reference to a database storing past load performance values and operating schedules of the heat supply device, predicts the peak load that occurs when the heat supply device starts up, and the time when the peak load occurs on the target day of the prediction. The first prediction unit sets the average load of the heat supply device within a predetermined time span Δ as the predicted value of the peak-shaped load. The aforementioned spike-shaped load corresponds to a heat load consisting of the following heat: The heat required by the aforementioned heat-demanding device; This represents the heat loss due to heat dissipation when the heat is supplied from the aforementioned heat supply device to the aforementioned heat demand device; and Heat used to bring the heat medium supplied from the heat supply device to the heat demand device to a predetermined set temperature.
2. The load prediction device according to claim 1, wherein, The first prediction unit sets the rated output of the heat supply device to the average value of the load of the heat supply device in the time width Δ.
3. The load prediction device according to claim 1, wherein, The first prediction unit sets the average value of the inlet temperature of the heat supply device over the time width Δ to T1, sets the specific heat of the heat medium to C, and sets the set temperature of the outlet side of the heat supply device to T. s Let P be the average value of the flow rate of the heat medium between the heat supply device and the heat demand device over the time width Δ, and then pass through |T s -T1|×P×C, calculate the average load of the heat supply device in the time width Δ.
4. The load prediction device according to any one of claims 1 to 3, wherein, The first prediction unit will predict the time when the peak load value occurs in the past load performance values or the start time of the heat supply device in the operation schedule as the time when the peak load occurs in the target day.
5. The load prediction device according to any one of claims 1 to 3, wherein, The first prediction unit sets the start-up time of the aforementioned heat supply device as the first moment, the start-up time of the aforementioned heat supply device as the second moment, and the difference between the aforementioned second moment and the aforementioned first moment as the aforementioned time width Δ. The above-mentioned normal operation refers to the operating state exhibited by the load of the above-mentioned heat supply device as a load corresponding to the heat constituted by the following: The heat required by the aforementioned heat-demanding device; and This refers to the heat loss during the supply of heat from the aforementioned heat supply device to the aforementioned heat demand device.
6. The load prediction device according to any one of claims 1 to 3, wherein, The first prediction unit sets the start-up time of the aforementioned heat supply device as the first moment, the time when the aforementioned heat supply device becomes normally operational as the second moment, sets a predetermined integer of 2 or more as N, and sets the time width Δ as the time width that divides the difference between the aforementioned second moment and the aforementioned first moment into N parts. Within each time span Δ from the first time point to the second time point, the predicted value of the peak load and the time of occurrence of the peak load within the predicted target day are predicted. The above-mentioned normal operation refers to the operating state exhibited by the load of the above-mentioned heat supply device as a load corresponding to the heat constituted by the following: The heat required by the aforementioned heat-demanding device; and This refers to the heat loss during the supply of heat from the aforementioned heat supply device to the aforementioned heat demand device.
7. The load prediction device according to any one of claims 1 to 3, wherein, It has a second prediction unit that, referring to the aforementioned database, uses a pre-learned prediction model to predict the load value of the aforementioned heat supply device during normal operation. The above-mentioned normal operation refers to the operating state exhibited by the load of the above-mentioned heat supply device as a load corresponding to the heat constituted by the following: The heat required by the aforementioned heat-demanding device; and This refers to the heat loss during the supply of heat from the aforementioned heat supply device to the aforementioned heat demand device.
8. A load forecasting method, wherein, The load forecasting device performs the following steps for forecasting the load of a heat supply device used to supply heat to heat demand devices: Referring to a database storing past load performance values and operating schedules of the aforementioned heat supply devices, predictions are made of the peak load values that occur during the startup of the aforementioned heat supply devices, as well as the times when the peak loads occur on the target days for prediction. In the above prediction step, the average load of the heat supply device within a predetermined time width Δ is set as the predicted value of the peak-shaped load. The aforementioned spike-shaped load corresponds to a heat load consisting of the following heat: The heat required by the aforementioned heat-demanding device; This refers to the heat loss during the supply of heat from the aforementioned heat supply device to the aforementioned heat demand device. as well as Heat used to bring the heat medium supplied from the heat supply device to the heat demand device to a predetermined set temperature.
9. A program product that causes a load forecasting device for forecasting the load on a heat supply device used to supply heat to a heat demand device to perform the following steps: Referring to a database storing past load performance values and operating schedules of the aforementioned heat supply devices, predictions are made of the peak load values that occur during the startup of the aforementioned heat supply devices, as well as the times when the peak loads occur on the target days for prediction. In the above prediction step, the average load of the heat supply device within a predetermined time width Δ is set as the predicted value of the peak-shaped load. The aforementioned spike-shaped load corresponds to a heat load consisting of the following heat: The heat required by the aforementioned heat-demanding device; This represents the heat loss due to heat dissipation when the heat is supplied from the aforementioned heat supply device to the aforementioned heat demand device; and Heat used to bring the heat medium supplied from the heat supply device to the heat demand device to a predetermined set temperature.