CN-122000861-A - Power load prediction method, device, electronic equipment and storage medium

CN122000861ACN 122000861 ACN122000861 ACN 122000861ACN-122000861-A

Abstract

One or more embodiments of the present disclosure provide a power load prediction method, apparatus, electronic device, and storage medium. The method comprises the steps of obtaining a first power load corresponding to a preset number of continuous historical moments, obtaining power load characteristics corresponding to the first power load based on the first power load, predicting a second power load at the current moment by using a pre-trained power load prediction model based on the power load characteristics, wherein the power load prediction model is constructed based on a selective state space model, the selective state space model is used for describing a dynamic process of the power load changing along with time so as to obtain a state at the current moment, and the state represents accumulated dynamic information of the power load in a time dimension.

Inventors

GAO JIANBO
SU LIANCAI
ZHANG XIRUN
YU QIAN
LIU XIAODAN
GAN LIQING
YU LONG

Assignees

北京中电飞华通信有限公司

Dates

Publication Date: 20260508
Application Date: 20251205

Claims (10)

1. A method of predicting an electrical load, comprising: acquiring first power loads corresponding to a preset number of continuous historical moments; based on the first power load, obtaining a power load characteristic corresponding to the first power load; Predicting a second electrical load at a current time using a pre-trained electrical load prediction model based on the electrical load characteristics; the power load prediction model is constructed based on a selective state space model, and the selective state space model is used for describing a dynamic process of power load change along with time so as to obtain a state at the current moment, wherein the state represents accumulated dynamic information of the power load in a time dimension.
2. The method of claim 1, wherein predicting a second electrical load at a current time using a pre-trained electrical load prediction model based on the electrical load signature comprises: Inputting the power load characteristics into the selective state space model to output first intermediate characteristics corresponding to the first power load, wherein the first intermediate characteristics bear the association relation between the first power load at the historical moment and the second power load at the current moment; screening second intermediate features in the first intermediate features by using a gating mechanism, wherein the second intermediate features represent valid features in the first intermediate features; And carrying out linear mapping on the second intermediate characteristic to obtain the second power load.
3. The method of claim 2, wherein inputting the electrical load characteristic into the selectively enabled spatial model to output a first intermediate characteristic corresponding to the first electrical load comprises: Multiplying the power load characteristic by a control matrix to obtain an action item of the power load characteristic on the state; Multiplying the current state by a state matrix to obtain an evolution item of the state; integrating the sum of the action item and the variation item to obtain an updated state; multiplying the updated state by an output matrix to obtain a contribution of the state to an output result; Multiplying the power load characteristics by an instruction matrix to obtain a main item of an output result; and adding the contribution item and the main body item to obtain the first intermediate feature.
4. A method according to claim 3, further comprising: And configuring the updated state as a state corresponding to the input data of the next round.
5. The method of claim 1, wherein deriving the electrical load signature based on the historical time-of-day electrical load data comprises: Dividing the power load data of the historical moment into a plurality of data fragments in a time sequence dimension; and carrying out convolution processing on the plurality of data fragments to obtain the power load characteristic.
6. The method as recited in claim 1, further comprising: The following steps are iteratively executed until the first parameter meets the preset condition: Acquiring a first parameter of a power load prediction model sent by a central server, and initializing the power load prediction model based on the first parameter; And in response to determining that the data amount of the first power load is greater than a preset threshold, training the power load prediction model based on the first power load to obtain a second parameter of the power load prediction model, wherein the second parameter is used for integrating the central server to obtain the first parameter.
7. The method as recited in claim 6, further comprising: Acquiring a first parameter sent by the central server in response to determining that the data amount of the first power load is less than or equal to the preset threshold; The power load prediction model is deployed based on the first parameter.
8. An electrical load prediction apparatus, comprising: The acquisition module is configured to acquire first power loads corresponding to a preset number of continuous historical moments; the extraction module is configured to obtain a power load characteristic corresponding to the first power load based on the first power load; A prediction module configured to predict a second electrical load at a current time using a pre-trained electrical load prediction model based on the electrical load characteristics; the power load prediction model is constructed based on a selective state space model, and the selective state space model is used for describing a dynamic process of power load change along with time so as to obtain a state at the current moment, wherein the state represents accumulated dynamic information of the power load in a time dimension.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and run by the processor, wherein the processor implements the method of any of claims 1 to 7 when executing the computer program.
10. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Description

Power load prediction method, device, electronic equipment and storage medium Technical Field One or more embodiments of the present disclosure relate to the field of power technology, and in particular, to a power load prediction method, a device, an electronic apparatus, and a storage medium. Background It should be noted that the foregoing description of the background art is only for the purpose of providing a clear and complete description of the technical solution of the present invention and is presented for the convenience of understanding by those skilled in the art. The above-described solutions are not considered to be known to the person skilled in the art simply because they are set forth in the background of the invention section. The electric power industry is a key basic industry of national development, is closely related to the development of various fields, and plays an important supporting role for the stable operation of socioeconomic. The accurate short-term power load prediction is a core link for guaranteeing the stable operation and the maintenance of the social ordered operation of the power system. In recent years, with the continuous development of artificial intelligence technology, methods such as machine learning, deep learning and the like are gradually introduced into the field of power load prediction, and the technology can identify key elements influencing the power load through deep analysis of historical data, so that a technical foundation is provided for improving prediction accuracy. The main technical schemes in the current power load prediction field are mainly divided into two types. The method comprises the steps of taking a cyclic neural network as a core model, predicting the current power load based on power load data at the last historical moment, and capturing global time sequence association through a self-attention mechanism based on power load data at a plurality of historical moments to realize prediction. However, both the two technical schemes have obvious limitations that the former depends on recurrence logic of single-moment historical data, deep dependency features in a time sequence are difficult to fully extract, so that prediction accuracy is poor, and the latter can capture global association, but the model parameters are large in quantity, high in calculation complexity, easy to generate higher calculation power and storage cost, and limited in suitability. Disclosure of Invention In view of this, it is an object of one or more embodiments of the present disclosure to provide a power load prediction method, apparatus, electronic device and storage medium, so as to solve the problems set forth in the background art. Based on the above object, the present disclosure provides, in a first aspect, a power load prediction method, including: acquiring first power loads corresponding to a preset number of continuous historical moments; based on the first power load, obtaining a power load characteristic corresponding to the first power load; Predicting a second electrical load at a current time using a pre-trained electrical load prediction model based on the electrical load characteristics; the power load prediction model is constructed based on a selective state space model, and the selective state space model is used for describing a dynamic process of power load change along with time so as to obtain a state at the current moment, wherein the state represents accumulated dynamic information of the power load in a time dimension. Optionally, predicting the second electrical load at the current time using a pre-trained electrical load prediction model based on the electrical load characteristics, comprising: Inputting the power load characteristics into the selective state space model to output first intermediate characteristics corresponding to the first power load, wherein the first intermediate characteristics bear the association relation between the first power load at the historical moment and the second power load at the current moment; screening second intermediate features in the first intermediate features by using a gating mechanism, wherein the second intermediate features represent valid features in the first intermediate features; And carrying out linear mapping on the second intermediate characteristic to obtain the second power load. Optionally, inputting the electrical load characteristic into the selective state space model to output a first intermediate characteristic corresponding to the first electrical load, including: Multiplying the power load characteristic by a control matrix to obtain an action item of the power load characteristic on the state; Multiplying the current state by a state matrix to obtain an evolution item of the state; integrating the sum of the action item and the variation item to obtain an updated state; multiplying the updated state by an output matrix to obtain a contribution of the state to an output r