A method, device, equipment and storage medium for predicting power load
By selecting the sub-prediction model with the highest correlation with the data to be input from the pre-trained power load prediction model and performing weighted calculation and model training, the problem of low power load prediction accuracy in the existing technology is solved, and more refined is achieved electric load analysis and power production scheduling management.
Patent Information
- Application Number
- CN202210186722.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-02-28
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2042-02-28
AI Technical Summary
The existing power load prediction results are not accurate, making it difficult to accurately analyze the distribution of power load, affecting the refined management of power production and scheduling plans.
By obtaining the factor data of the historical day, the mapping is converted into the data to be input, and the sub-prediction model with the highest correlation with the data to be input is selected from the pre-trained power load prediction model, perform weighted calculations and model training, and update the model parameters to improve the Prediction accuracy.
It improves the accuracy of power load prediction, can analyze the distribution of power load more refinedly, and supports lean management of power production and scheduling plans.
Smart Images

Figure CN114565156B_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the technical field of power load forecasting, and in particular, to a power load forecasting method, device, equipment and storage medium. Background Art
[0002] Under the background of the deepening reform of the power system, the whole society has increasingly higher requirements for the safe, energy-saving and efficient operation of the power grid and the entire power industry. When the dispatching department studies the planned operation mode of the future period, it must first conduct demand analysis and analyze the future power load. It not only needs total information, but more importantly, it needs to analyze the distribution of these power loads and conduct load forecasting. Load forecasting provides refined demand analysis data and analysis results for power production, laying the foundation for realizing lean management of the dispatching planning process.
[0003] There are many factors that affect power load (such as social factors, policy factors, meteorological factors, holiday factors, etc.). Power load forecasting is extremely complex, and the existing forecasting results are not very accurate. Summary of the invention
[0004] The present invention provides a power load forecasting method, device, equipment and storage medium to improve the forecasting accuracy of a model.
[0005] In a first aspect, the present invention provides a method for predicting power load, comprising:
[0006] Obtaining data of factors affecting the load of the power system for at least one historical day before the day to be predicted;
[0007] Mapping the factor data into data to be input;
[0008] Selecting a first sub-prediction model having the highest correlation with the input data from a plurality of sub-prediction models of a pre-trained power load prediction model, wherein the power load prediction model includes a plurality of sub-prediction models built based on a plurality of prediction algorithms;
[0009] The data to be input is input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model.
[0010] Optionally, the data to be input includes expression vectors of each factor in the factor data, and selecting a first sub-prediction model having the highest correlation with the expression vector from a plurality of sub-prediction models of the pre-trained power load prediction model comprises:
[0011] Performing weighted calculation on the expression vector of each factor in the factor data to obtain a comprehensive weight value;
[0012] Based on the comprehensive weight value, a first sub-prediction model having the highest correlation with the data to be input is selected from multiple sub-prediction models of the pre-trained power load prediction model.
[0013] Optionally, each of the sub-prediction models is associated with a weight range, and the weight ranges of the multiple sub-prediction models constitute a weight space, and the first sub-prediction model with the highest correlation with the to-be-input data is selected from the multiple sub-prediction models of the pre-trained power load forecasting model based on the comprehensive weight value, including:
[0014] Mapping the comprehensive weight value to the weight space;
[0015] The sub-prediction model corresponding to the weight range within which the comprehensive weight value falls is used as the first sub-prediction model with the highest correlation with the data to be input.
[0016] Optionally, the factor data includes temperature data, humidity data, rainfall data and date data.
[0017] Optionally, the training process of the power load prediction model includes:
[0018] Acquire historical load samples of the power system and factor data corresponding to the historical load samples;
[0019] Mapping the factor data corresponding to the historical load sample into a data sample to be input;
[0020] Selecting a second sub-prediction model having the highest correlation with the data sample to be input from a plurality of sub-prediction models of the power load prediction model;
[0021] The second sub-prediction model is trained using the historical load samples and the data samples to be input as training samples.
[0022] Optionally, training the second sub-prediction model using the historical load samples and the to-be-input data samples as training samples includes:
[0023] Inputting the data sample to be input into the second sub-prediction model to obtain the load prediction result output by the power load prediction model;
[0024] Calculating the loss value of the load forecast result and the historical load sample;
[0025] When the loss value between the load forecast result and the historical load sample is greater than a preset value, updating the model parameters of the second sub-forecast model, and returning to execute the step of obtaining the historical load samples of the power system and the factor data corresponding to the historical load samples;
[0026] When the loss value between the load forecast result and the historical load sample is less than or equal to a preset value, it is determined that the training of the second sub-forecast model is completed.
[0027] Optionally, updating the model parameters of the second sub-prediction model includes:
[0028] The model parameters of the second sub-prediction model are updated by using a point-to-point ratio method, and the selection range of the point-by-point smoothing coefficient of the point-to-point ratio method is 0.1-0.99.
[0029] In a second aspect, the present invention further provides a power load forecasting device, comprising:
[0030] A first data acquisition module is used to acquire data of factors affecting the load of the power system for at least one historical day before the day to be predicted;
[0031] A first data conversion module, used for mapping the factor data into data to be input;
[0032] A first sub-prediction model determination module, used to select a first sub-prediction model having the highest correlation with the data to be input from a plurality of sub-prediction models of a pre-trained power load prediction model, wherein the power load prediction model includes a plurality of sub-prediction models built based on a plurality of prediction algorithms;
[0033] The prediction result determination module is used to input the data to be input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model.
[0034] Optionally, the first sub-prediction model determination module includes:
[0035] A weighting submodule, used for performing weighted calculation on the expression vector of each factor in the factor data to obtain a comprehensive weight value;
[0036] The first sub-prediction model determination submodule is used to select the first sub-prediction model with the highest correlation with the data to be input from multiple sub-prediction models of the pre-trained power load prediction model based on the comprehensive weight value.
[0037] Optionally, each of the sub-prediction models is associated with a weight range, and the weight ranges of the multiple sub-prediction models constitute a weight space, and the first sub-prediction model determination submodule includes:
[0038] A weight value mapping unit, used for mapping the comprehensive weight value to the weight space;
[0039] The first sub-prediction model determination unit is used to use the sub-prediction model corresponding to the weight range within which the comprehensive weight value falls as the first sub-prediction model with the highest correlation with the data to be input.
[0040] Optionally, the factor data includes temperature data, humidity data, rainfall data and date data.
[0041] Optionally, the power load prediction device further includes a power load prediction model training module, which is used to train the power load prediction model, and the power load prediction model training module includes:
[0042] A second data acquisition module, used to acquire historical load samples of the power system and factor data corresponding to the historical load samples;
[0043] A second data conversion module, used for mapping the factor data corresponding to the historical load sample into a data sample to be input;
[0044] A second sub-prediction model determination module, used to select a second sub-prediction model having the highest correlation with the data sample to be input from multiple sub-prediction models of the power load prediction model;
[0045] The model training submodule is used to train the second sub-prediction model using the historical load samples and the data samples to be input as training samples.
[0046] Optionally, the model training submodule includes:
[0047] A prediction result determination unit, used for inputting the data sample to be input into the second sub-prediction model to obtain the load prediction result output by the power load prediction model;
[0048] A loss value calculation unit, used to calculate the loss value of the load forecast result and the historical load sample;
[0049] A model updating unit, configured to update the model parameters of the second sub-prediction model when the loss value between the load prediction result and the historical load sample is greater than a preset value, and return to execute the step of obtaining the historical load sample of the power system and the factor data corresponding to the historical load sample;
[0050] The model determination unit is used to determine that the training of the second sub-prediction model is completed when the loss value between the load prediction result and the historical load sample is less than or equal to a preset value.
[0051] Optionally, the model updating unit uses a point-to-point ratio method to update the model parameters of the second sub-prediction model, and the selection range of the point-by-point smoothing coefficient of the point-to-point ratio method is 0.1-0.99.
[0052] In a third aspect, the present invention further provides a computer device, comprising:
[0053] one or more processors;
[0054] A storage device for storing one or more programs;
[0055] When the one or more programs are executed by the one or more processors, the one or more processors implement the power load forecasting method provided in the first aspect of the present invention.
[0056] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power load forecasting method provided in the first aspect of the present invention.
[0057] The power load forecasting method provided by the present invention includes: obtaining factor data affecting the load of the power system for at least one historical day before the day to be predicted, mapping the factor data into data to be input, selecting a first sub-prediction model with the highest correlation with the data to be input from multiple sub-prediction models of a pre-trained power load forecasting model, wherein the power load forecasting model includes multiple sub-prediction models built based on multiple prediction algorithms, inputting the data to be input into the first sub-prediction model, and obtaining a load forecasting result output by the power load forecasting model. In an embodiment of the present invention, the power load forecasting model includes multiple sub-prediction models built based on multiple prediction algorithms, and selecting the first sub-prediction model with the highest correlation with the data to be input from multiple sub-prediction models of the pre-trained power load forecasting model according to the data to be input, thereby improving the prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] Figure 1 A flowchart of a power load forecasting method provided in Embodiment 1 of the present invention;
[0059] Figure 2 A schematic diagram of the structure of a power load prediction device provided by Embodiment 2 of the present invention;
[0060] Figure 3 A schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION
[0061] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.
[0062] Embodiment 1
[0063] Figure 1This is a flow chart of a power load forecasting method provided in Embodiment 1 of the present invention. This embodiment is applicable to the case of forecasting power load. The method can be executed by a power load forecasting device provided in an embodiment of the present invention. The device can be implemented in software and / or hardware and is usually configured in a computer device, such as Figure 1 As shown, the method specifically comprises the following steps:
[0064] S101. Obtain data on factors affecting the load of the power system for at least one historical day before a day to be predicted.
[0065] In the embodiment of the present invention, the day to be predicted may be one or more days after the current date, and the historical day may be one or more days before the current date.
[0066] There are many factors that affect the load of the power system. In the embodiment of the present invention, the factors that affect the load of the power system include temperature data, humidity data, rainfall data and date data. In the power load forecasting, many factors affect the predicted value of the power load to varying degrees. Some factors vary due to nature, such as weather. Some factors vary according to regional conditions, such as the speed of industrial and agricultural development. Some factors are major events that cannot be estimated, such as serious disasters, and the impact of each factor on the load may be different, and different levels of the same factor have different impacts on the load.
[0067] Meteorological factors including temperature, humidity, rainfall, etc. will directly affect load fluctuations, especially in areas where residential load accounts for a higher proportion, this impact is greater.
[0068] Due to the influence of holidays and special conditions, the load on holidays is generally significantly lower than that on normal working days. Taking the Spring Festival as an example, the load curve during the Spring Festival generally shows a significant downward deformation, and its variation cycle roughly coincides with the holiday cycle.
[0069] Impact of emergencies among large industrial users For areas where the installed capacity of large industrial users occupies a high electricity load, large industrial users also play a relatively large role in load forecast deviations.
[0070] It should be noted that the above-mentioned data on factors affecting the load of the power system is an exemplary description of the present invention. In other embodiments of the present invention, the data on factors affecting the load of the power system may also include other factors, which are not limited in the embodiments of the present invention.
[0071] S102: Map and convert the factor data into data to be input.
[0072] In an embodiment of the present invention, feature extraction (e.g., word vector embedding, data mapping, etc.) can be performed on the factor data, and the factor data mapping can be converted into the data to be input, and the data to be input includes the expression vector of each factor in the factor data. In practical applications, since the dimensions of each factor data (feature quantity) are different, it is necessary to map the values of different dimensions to a specific interval through dimensionless processing, so that the values can be numerically comparable, thereby facilitating the quantitative calculation of similarity and difference.
[0073] S103: Select a first sub-prediction model having the highest correlation with the input data from a plurality of sub-prediction models of the pre-trained power load prediction model.
[0074] In an embodiment of the present invention, the power load forecasting model includes multiple sub-forecasting models built based on multiple forecasting algorithms to adapt to different types of input data. In an embodiment of the present invention, after the factor data is mapped and converted into the data to be input, the sub-model with the highest correlation with the data to be input is selected from the multiple sub-forecasting models of the pre-trained power load forecasting model as the first sub-forecasting model. The so-called highest correlation refers to the sub-forecasting model that is most suitable for processing the data to be input. Exemplarily, a certain sub-forecasting model is most suitable for processing the data to be input with the highest proportion of temperature data. If the temperature data accounts for the highest proportion of the data to be input, the sub-forecasting model can be selected as the sub-model with the highest correlation with the data to be input as the first sub-forecasting model.
[0075] Exemplarily, in some embodiments of the present invention, selecting a first sub-prediction model having the highest correlation with the expression vector from a plurality of sub-prediction models of a pre-trained power load prediction model comprises the following steps:
[0076] 1. Perform weighted calculation on the expression vector of each factor in the factor data to obtain the comprehensive weight value.
[0077] Specifically, the corresponding weight is set according to the influence of each factor on the power load. For example, if the temperature has a greater impact, the corresponding weight of the temperature is relatively large. The expression vector of each factor in the factor data is multiplied by the corresponding weight, and each product is added to obtain a comprehensive weight value.
[0078] 2. Selecting the first sub-prediction model with the highest correlation with the input data from multiple sub-prediction models of the pre-trained power load prediction model based on the comprehensive weight value.
[0079] Exemplarily, after obtaining the comprehensive weight value, the sub-prediction model corresponding to the factor is determined according to the size of the comprehensive weight value, that is, the first sub-prediction model with the highest correlation with the data to be input.
[0080] Exemplarily, each sub-prediction model is associated with a weight range, and the weight ranges of multiple sub-prediction models constitute a weight space. The first sub-prediction model with the highest correlation with the input data is selected from multiple sub-prediction models of the pre-trained power load forecasting model based on the comprehensive weight value, including:
[0081] The comprehensive weight value is mapped to the weight space, and the sub-prediction model corresponding to the weight range in which the comprehensive weight value falls is taken as the first sub-prediction model with the highest correlation with the data to be input.
[0082] S104: input the data to be input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model.
[0083] After determining the first sub-prediction model with the highest correlation with the data to be input, the data to be input is input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model, thereby predicting the load curve formed by the power load for multiple days after the current date.
[0084] The power load forecasting method provided by the embodiment of the present invention includes: obtaining factor data that affects the load of the power system for at least one historical day before the day to be predicted, mapping the factor data into data to be input, selecting a first sub-prediction model with the highest correlation with the data to be input from multiple sub-prediction models of a pre-trained power load forecasting model, wherein the power load forecasting model includes multiple sub-prediction models built based on multiple prediction algorithms, inputting the data to be input into the first sub-prediction model, and obtaining a load forecasting result output by the power load forecasting model. In the embodiment of the present invention, the power load forecasting model includes multiple sub-prediction models built based on multiple prediction algorithms, and selecting the first sub-prediction model with the highest correlation with the data to be input from multiple sub-prediction models of the pre-trained power load forecasting model according to the data to be input, thereby improving the prediction accuracy of the model.
[0085] In some embodiments of the present invention, the training process of the power load forecasting model includes:
[0086] 1. Obtain historical load samples of the power system and factor data corresponding to the historical load samples.
[0087] Specifically, the historical load sample is the power load of the historical day, and the factor data corresponding to the historical load sample is the factor data affecting the historical load sample. As mentioned above, there are many factor data affecting the load of the power system. In the embodiment of the present invention, the factor data affecting the load of the power system includes temperature data, humidity data, rainfall data and date data, which will not be described in detail in the embodiment of the present invention.
[0088] 2. Map the factor data corresponding to the historical load samples into the data samples to be input.
[0089] Exemplarily, feature extraction (e.g., word vector embedding, data mapping, etc.) is performed on the factor data corresponding to the historical load sample, and the factor data corresponding to the historical load sample is mapped and converted into a sample of data to be input, and the sample of data to be input includes the expression vector of each factor in the factor data. In practical applications, since the dimensions of each factor data (feature quantity) are different, it is necessary to map the values of different dimensions to a specific interval through dimensionless processing, so that the values can be numerically comparable, thereby facilitating the quantitative calculation of similarity and difference.
[0090] 3. Select a second sub-prediction model with the highest correlation with the data sample to be input from multiple sub-prediction models of the power load prediction model.
[0091] In an embodiment of the present invention, the electric power load forecasting model includes a plurality of sub-forecasting models based on different forecasting algorithms to adapt to different types of input data. In an embodiment of the present invention, after the factor data corresponding to the historical load sample is mapped and converted into the data sample to be input, the sub-model with the highest correlation with the data to be input is selected from the plurality of sub-forecasting models of the pre-trained electric power load forecasting model as the second sub-forecasting model. Specifically, the process of selecting the sub-model with the highest correlation with the data to be input from the plurality of sub-forecasting models of the pre-trained electric power load forecasting model as the second sub-forecasting model is similar to the process of selecting the sub-model with the highest correlation with the data to be input in the aforementioned embodiment from the plurality of sub-forecasting models of the pre-trained electric power load forecasting model as the first sub-forecasting model, and the embodiments of the present invention will not be repeated here.
[0092] 4. Use the historical load samples and the data samples to be input as training samples to train the second sub-prediction model.
[0093] The historical load samples and the data samples to be input are organized into training samples to train the second sub-prediction model.
[0094] Exemplarily, the training process is as follows:
[0095] 4.1. Input the data sample to be input into the second sub-prediction model to obtain the load prediction result output by the power load prediction model.
[0096] 4.2. Calculate the loss value of load forecast results and historical load samples.
[0097] In the embodiment of the present invention, the square loss and absolute value loss of the load forecast result and the historical load sample may be calculated, and the embodiment of the present invention is not limited here.
[0098] 4.3. When the loss value between the load forecast result and the historical load sample is greater than the preset value, the model parameters of the second sub-forecast model are updated, and the step of obtaining the historical load samples of the power system and the factor data corresponding to the historical load samples is returned.
[0099] Specifically, the model parameters of the second sub-prediction model are updated by using the point-to-point multiple ratio method, and the selection range of the point-by-point smoothing coefficient of the point-to-point multiple ratio method is 0.1-0.99.
[0100] 4.4. When the loss value between the load forecast result and the historical load sample is less than or equal to the preset value, it is determined that the training of the second sub-forecast model is completed.
[0101] In the process of short-term forecasting, in order to obtain accurate forecast results, it is necessary to carefully collect various information and related data, rely on accurate mathematical and physical models to improve accuracy, and thus obtain better results each time.
[0102] However, with the rapid economic development and the tense power supply situation in recent years, the randomness of bus load is relatively large, and the difficulty of bus load prediction is relatively high. The results of the identification of the regularity and stability of bus historical load show that the bus load sequence has different regularities in different regions, different seasons, and different times.
[0103] Traditional prediction methods often rely on experience. After selecting a mathematical model and fixing the model parameters, the prediction results are obtained. The fixed parameter model obviously has different manifestations of the regularity of bus historical load in different regions and at different times. For different regions and seasons, how to select model parameters is the key to the accuracy of bus load prediction.
[0104] The embodiment of the present invention continuously updates the model parameters during the model training process, and finally uses the parameters that obtain the best prediction effect as the final model parameters for future day load prediction, thus greatly improving the prediction accuracy of the model.
[0105] Embodiment 2
[0106] Figure 2 A schematic diagram of the structure of a power load prediction device provided in the second embodiment of the present invention is as follows: Figure 2 As shown, the device comprises:
[0107] The first data acquisition module 201 is used to acquire data of factors affecting the load of the power system for at least one historical day before the day to be predicted;
[0108] A first data conversion module 202, used for mapping the factor data into data to be input;
[0109] A first sub-prediction model determination module 203 is used to select a first sub-prediction model with the highest correlation with the input data from a plurality of sub-prediction models of a pre-trained power load prediction model, wherein the power load prediction model includes a plurality of sub-prediction models built based on a plurality of prediction algorithms;
[0110] The prediction result determination module 204 is used to input the data to be input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model.
[0111] Optionally, the first sub-prediction model determination module 203 includes:
[0112] A weighting submodule, used for performing weighted calculation on the expression vector of each factor in the factor data to obtain a comprehensive weight value;
[0113] The first sub-prediction model determination submodule is used to select the first sub-prediction model with the highest correlation with the data to be input from multiple sub-prediction models of the pre-trained power load prediction model based on the comprehensive weight value.
[0114] Optionally, each of the sub-prediction models is associated with a weight range, and the weight ranges of the multiple sub-prediction models constitute a weight space, and the first sub-prediction model determination submodule includes:
[0115] A weight value mapping unit, used for mapping the comprehensive weight value to the weight space;
[0116] The first sub-prediction model determination unit is used to use the sub-prediction model corresponding to the weight range within which the comprehensive weight value falls as the first sub-prediction model with the highest correlation with the data to be input.
[0117] Optionally, the factor data includes temperature data, humidity data, rainfall data and date data.
[0118] Optionally, the power load prediction device further includes a power load prediction model training module, which is used to train the power load prediction model, and the power load prediction model training module includes:
[0119] A second data acquisition module, used to acquire historical load samples of the power system and factor data corresponding to the historical load samples;
[0120] A second data conversion module, used for mapping the factor data corresponding to the historical load sample into a data sample to be input;
[0121] A second sub-prediction model determination module, used to select a second sub-prediction model having the highest correlation with the data sample to be input from multiple sub-prediction models of the power load prediction model;
[0122] The model training submodule is used to train the second sub-prediction model using the historical load samples and the data samples to be input as training samples.
[0123] Optionally, the model training submodule includes:
[0124] A prediction result determination unit, used for inputting the data sample to be input into the second sub-prediction model to obtain the load prediction result output by the power load prediction model;
[0125] A loss value calculation unit, used to calculate the loss value of the load forecast result and the historical load sample;
[0126] A model updating unit, configured to update the model parameters of the second sub-prediction model when the loss value between the load prediction result and the historical load sample is greater than a preset value, and return to execute the step of obtaining the historical load sample of the power system and the factor data corresponding to the historical load sample;
[0127] The model determination unit is used to determine that the training of the second sub-prediction model is completed when the loss value between the load prediction result and the historical load sample is less than or equal to a preset value.
[0128] Optionally, the model updating unit uses a point-to-point ratio method to update the model parameters of the second sub-prediction model, and the selection range of the point-by-point smoothing coefficient of the point-to-point ratio method is 0.1-0.99.
[0129] The above-mentioned power load prediction device can execute the power load prediction method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the power load prediction method.
[0130] Embodiment 3
[0131] Embodiment 3 of the present invention provides a computer device, Figure 3 A schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention is shown in FIG. Figure 3 As shown, the computer device includes a processor 301, a memory 302, a communication module 303, an input device 304 and an output device 305; the number of processors 301 in the computer device can be one or more. Figure 3 A processor 301 is taken as an example; the processor 301, memory 302, communication module 303, input device 304 and output device 305 in the computer device can be connected by a bus or other means. Figure 3 The above processor 301, memory 302, communication module 303, input device 304 and output device 305 can be integrated on a control mainboard of a computer device.
[0132] The memory 302 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as the module corresponding to the power load forecasting method in this embodiment. The processor 301 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 302, that is, the power load forecasting method provided in the above embodiment is implemented.
[0133] The memory 302 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 302 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 302 may further include a memory remotely arranged relative to the processor 301, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0134] The communication module 303 is used to establish a connection with an external device (such as a smart terminal) and realize data interaction with the external device. The input device 304 can be used to receive input digital or character information and generate key signal input related to user settings and function control of the computer device.
[0135] This embodiment provides a computer device that can execute the power load forecasting method provided by any of the above embodiments of the present invention, and has specific corresponding functions and beneficial effects.
[0136] Embodiment 4
[0137] Embodiment 4 of the present invention provides a storage medium containing computer executable instructions, on which a computer program is stored. When the program is executed by a processor, the power load forecasting method provided in any of the above embodiments of the present invention is implemented. The method includes:
[0138] Obtaining data of factors affecting the load of the power system for at least one historical day before the day to be predicted;
[0139] Mapping the factor data into data to be input;
[0140] Selecting a first sub-prediction model having the highest correlation with the input data from a plurality of sub-prediction models of a pre-trained power load prediction model, wherein the power load prediction model includes a plurality of sub-prediction models built based on a plurality of prediction algorithms;
[0141] The data to be input is input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model.
[0142] Of course, the computer executable instructions of a storage medium including computer executable instructions provided in an embodiment of the present invention are not limited to the operations of the method described above, but can also execute related operations in the power load forecasting method provided in an embodiment of the present invention.
[0143] It should be noted that, for the device, equipment and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0144] Through the above description of the implementation method, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a robot, a personal computer, a server, or a network device, etc.) to perform the power load forecasting method described in any embodiment of the present invention.
[0145] It is worth noting that in the above-mentioned device, the various modules, sub-modules and units included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.
[0146] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution device. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0147] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0148] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims
1. A method for predicting power load, characterized in that: include: Obtaining data of factors affecting the load of the power system for at least one historical day before the day to be predicted; Mapping the factor data into data to be input; Selecting a first sub-prediction model having the highest correlation with the input data from a plurality of sub-prediction models of a pre-trained power load prediction model, wherein the power load prediction model includes a plurality of sub-prediction models built based on a plurality of prediction algorithms; Inputting the data to be input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model; The training process of the power load forecasting model includes: Acquire historical load samples of the power system and factor data corresponding to the historical load samples; Mapping the factor data corresponding to the historical load sample into a data sample to be input; Selecting a second sub-prediction model having the highest correlation with the data sample to be input from a plurality of sub-prediction models of the power load prediction model; Training the second sub-prediction model using the historical load samples and the to-be-input data samples as training samples; The mapping and converting the factor data corresponding to the historical load samples into the data samples to be input includes: extracting features from the factor data corresponding to the historical load samples, and mapping and converting the factor data corresponding to the historical load samples into the data samples to be input; The step of training the second sub-prediction model using the historical load sample and the to-be-input data sample as training samples includes: Inputting the data sample to be input into the second sub-prediction model to obtain the load prediction result output by the power load prediction model; Calculating the loss value of the load forecast result and the historical load sample; When the loss value between the load forecast result and the historical load sample is greater than a preset value, updating the model parameters of the second sub-forecast model, and returning to execute the step of obtaining the historical load samples of the power system and the factor data corresponding to the historical load samples; When the loss value between the load forecast result and the historical load sample is less than or equal to a preset value, it is determined that the training of the second sub-forecast model is completed.
2. The power load forecasting method according to claim 1, characterized in that: The data to be input includes expression vectors of each factor in the factor data, and selecting a first sub-prediction model having the highest correlation with the expression vector from a plurality of sub-prediction models of the pre-trained power load prediction model includes: Performing weighted calculation on the expression vector of each factor in the factor data to obtain a comprehensive weight value; Based on the comprehensive weight value, a first sub-prediction model having the highest correlation with the data to be input is selected from multiple sub-prediction models of the pre-trained power load prediction model.
3. The power load forecasting method according to claim 2, characterized in that: Each of the sub-prediction models is associated with a weight range, and the weight ranges of the multiple sub-prediction models constitute a weight space. Based on the comprehensive weight value, a first sub-prediction model with the highest correlation with the data to be input is selected from the multiple sub-prediction models of the pre-trained power load forecasting model, including: Mapping the comprehensive weight value to the weight space; The sub-prediction model corresponding to the weight range within which the comprehensive weight value falls is used as the first sub-prediction model with the highest correlation with the data to be input.
4. The power load forecasting method according to claim 1, characterized in that: The factor data includes temperature data, humidity data, rainfall data and date data.
5. The power load forecasting method according to claim 1, characterized in that: Updating the model parameters of the second sub-prediction model includes: The model parameters of the second sub-prediction model are updated by using a point-to-point ratio method, and the selection range of the point-by-point smoothing coefficient of the point-to-point ratio method is 0.1-0.
99.
6. A power load forecasting device, characterized in that: include: A first data acquisition module is used to acquire data of factors affecting the load of the power system for at least one historical day before the day to be predicted; A first data conversion module, used for mapping the factor data into data to be input; A first sub-prediction model determination module, used to select a first sub-prediction model having the highest correlation with the data to be input from a plurality of sub-prediction models of a pre-trained power load prediction model, wherein the power load prediction model includes a plurality of sub-prediction models built based on a plurality of prediction algorithms; A prediction result determination module, used for inputting the data to be input into the first sub-prediction model to obtain the load prediction result output by the power load prediction model; The power load prediction device further includes a power load prediction model training module, which is used to train the power load prediction model. The power load prediction model training module includes: A second data acquisition module, used to acquire historical load samples of the power system and factor data corresponding to the historical load samples; A second data conversion module, used for mapping the factor data corresponding to the historical load sample into a data sample to be input; A second sub-prediction model determination module, used to select a second sub-prediction model having the highest correlation with the data sample to be input from multiple sub-prediction models of the power load prediction model; A model training submodule, used for training the second sub-prediction model using the historical load samples and the data samples to be input as training samples; The second data conversion module is specifically used to extract features from the factor data corresponding to the historical load samples, and map the factor data corresponding to the historical load samples into data samples to be input; The model training submodule includes: A prediction result determination unit, used for inputting the data sample to be input into the second sub-prediction model to obtain the load prediction result output by the power load prediction model; A loss value calculation unit, used to calculate the loss value of the load forecast result and the historical load sample; A model updating unit, configured to update the model parameters of the second sub-prediction model when the loss value between the load prediction result and the historical load sample is greater than a preset value, and return to execute the step of obtaining the historical load sample of the power system and the factor data corresponding to the historical load sample; The model determination unit is used to determine that the training of the second sub-prediction model is completed when the loss value between the load prediction result and the historical load sample is less than or equal to a preset value.
7. A computer device, characterized in that: include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the power load forecasting method as described in any one of claims 1-5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the power load forecasting method as described in any one of claims 1 to 5 is implemented.
