CN-121981321-A - Power consumption prediction method and device

CN121981321ACN 121981321 ACN121981321 ACN 121981321ACN-121981321-A

Abstract

The invention provides a power consumption prediction method and device, wherein the method comprises the steps of obtaining power consumption data of a designated area in a target period, preprocessing the power consumption data, processing the preprocessed power consumption data by a target algorithm to obtain target decomposition parameters for power consumption prediction, performing variation modal decomposition on the preprocessed power consumption data based on the target decomposition parameters to obtain a plurality of intrinsic modal components, predicting the power consumption by a bidirectional long-short-term memory network based on the plurality of intrinsic modal components to obtain a candidate prediction result of the power consumption, determining influence factors influencing the power consumption, constructing an error prediction model by combining the influence factors, wherein the error prediction model is used for predicting an error sequence in the candidate prediction result, predicting the candidate prediction result by the error prediction model to obtain a target error sequence, and correcting the candidate prediction result based on the target error sequence to obtain a final power consumption prediction result.

Inventors

CHEN RUIZHI
YANG XIAOYUN
CHENG HAO
LIU JIANTING
YAN HAN
ZHU PENG
ZHANG WEI
LIU YUETING
ZENG YU
LI CHONGYANG
HE YIWEI
ZHAO XING

Assignees

国网重庆市电力公司璧山供电分公司

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (10)

1.A method for predicting power consumption, comprising: Acquiring electricity consumption data of a designated area in a target period, and preprocessing the electricity consumption data; Processing the preprocessed electricity consumption data by using a target algorithm to obtain target decomposition parameters for predicting the applied electricity consumption, wherein the target algorithm is obtained based on a gray wolf algorithm; Performing variation modal decomposition on the preprocessed electricity consumption data based on the target decomposition parameters to obtain a plurality of intrinsic modal components; Based on a plurality of the intrinsic mode components, the power consumption is predicted by utilizing a two-way long-short-term memory network, and a candidate prediction result of the power consumption is obtained; Determining influencing factors influencing the electricity consumption; Constructing an error prediction model by combining the influence factors, wherein the error prediction model is used for predicting an error sequence in the candidate prediction result; predicting the candidate prediction result by using the error prediction model to obtain a target error sequence; And correcting the candidate prediction result based on the target error sequence to obtain a final power consumption prediction result.
2. The power consumption prediction method according to claim 1, wherein the preprocessing the power consumption data includes: Normalizing and encoding the electricity consumption data; Filling missing data into the power consumption data after normalization and encoding, wherein the filling comprises weighting filling or average filling; And identifying and correcting abnormal values of the power consumption data after filling the missing values by using an isolated forest algorithm and a horizontal smoothing algorithm to obtain target power consumption data.
3. The power consumption prediction method according to claim 1, wherein constructing a target algorithm based on the wolf algorithm comprises: Performing a random grouping of the wolf population, the random grouping comprising performing the random grouping both before and after growth of the wolf population; establishing a mutual mechanism between the wolf populations, wherein the mutual mechanism is used for determining optimal individuals in the populations Adopting a simulated annealing strategy to reserve different solutions on the strategy of reserving the solutions; An adaptive mechanism is applied to the optimal individual reservation strategy to adjust the reserved optimal solution; And in the disturbance process of the optimal individuals in the wolf population, carrying out disturbance on the variables of the appointed number of dimensions of the optimal individuals, wherein the appointed number is smaller than the total number of dimensions contained in the optimal individuals.
4. The electricity consumption prediction method according to claim 1, characterized in that the method further comprises: Constructing a decomposition parameter evaluation standard for electricity consumption prediction; The application target algorithm processes the preprocessed electricity consumption data to obtain target decomposition parameters for application electricity consumption prediction, and the method comprises the following steps: and processing the preprocessed electricity consumption data by using the target algorithm based on the decomposition parameter evaluation standard to obtain target decomposition parameters for application electricity consumption prediction.
5. The power consumption prediction method according to claim 4, wherein the decomposition parameter evaluation criterion includes: wherein the Loss is the decomposition parameter evaluation criterion, and represents an average absolute error between the reconstructed signal and the original signal, and the smaller the average absolute error is, the higher the prediction accuracy is, the more the average absolute error is, the more the prediction accuracy is, and the prediction accuracy is As the original signal, the And the T is the time length.
6. The method for predicting electricity consumption according to claim 1, wherein the predicting electricity consumption by using a bidirectional long-short-term memory network based on a plurality of the eigen-mode components, to obtain a candidate prediction result of electricity consumption, includes: Constructing a component prediction model formed by a long-term and short-term memory network corresponding to each intrinsic mode component to obtain a power consumption prediction value corresponding to each intrinsic mode component; And summarizing the power consumption prediction values to obtain the candidate prediction results of the power consumption.
7. The electricity consumption amount prediction method according to claim 1, wherein the determining an influence factor that influences the electricity consumption amount includes: in the target period, the influence of the economic growth rate in the designated area on the electricity consumption; the influence of workdays and rest days on the electricity consumption; influence of air temperature and seasons on electricity consumption; Holiday effects on electricity usage.
8. The method of claim 7, wherein said constructing an error prediction model in combination with said influencing factors comprises: establishing an initial error prediction model based on BILSTM neural networks; And obtaining a plurality of historical eigenmode components, namely the influence factors, as input, obtaining a difference sequence between a historical candidate prediction result and a power consumption sequence formed by corresponding historical power consumption data as output, and training the initial error prediction model to obtain the error prediction model.
9. The method for predicting power consumption according to claim 8, wherein predicting the candidate prediction result by using the error prediction model to obtain a target error sequence includes: Inputting the influence factors and a plurality of eigenvalue components into the error prediction model to obtain a target error sequence aiming at the candidate prediction result; the step of correcting the candidate prediction result based on the target error sequence to obtain a final power consumption prediction result comprises the following steps: and eliminating the target error sequence from the candidate prediction results to obtain final power consumption prediction results.
10. A power consumption prediction apparatus, comprising: The first acquisition module is used for acquiring electricity consumption data of a designated area in a target period and preprocessing the electricity consumption data; the first processing module is used for processing the preprocessed electricity consumption data by applying a target algorithm to obtain target decomposition parameters for predicting the applied electricity consumption, and the target algorithm is obtained based on a wolf algorithm; The decomposition module is used for carrying out variation modal decomposition on the preprocessed electricity consumption data based on the target decomposition parameters to obtain a plurality of intrinsic modal components; the first prediction module is used for predicting the electricity consumption by utilizing a two-way long-short-term memory network based on a plurality of the intrinsic mode components to obtain a candidate prediction result of the electricity consumption; the determining module is used for determining influence factors influencing the electricity consumption; the first construction module is used for constructing an error prediction model by combining the influence factors, and the error prediction model is used for predicting an error sequence in the candidate prediction result; the second prediction module is used for predicting the candidate prediction result by using the error prediction model to obtain a target error sequence; And the correction module is used for correcting the candidate prediction result based on the target error sequence to obtain a final power consumption prediction result.

Description

Power consumption prediction method and device Technical Field The invention belongs to the technical field of electric quantity prediction, and particularly relates to an electric quantity prediction method and device. Background The electricity consumption prediction refers to the prediction and calculation of future electricity consumption by using corresponding technologies and methods after the collection, integration and analysis of past data. The accuracy of electricity consumption prediction is used as a key index by national grid companies for intra-industry peer-to-peer calibration, and meanwhile, the accuracy of electricity consumption prediction is also a key component of an electricity marketing campaign in the current competitive electricity market. The accuracy of electricity consumption prediction is improved, and the main influence aspects are as follows: (1) Meets the examination requirements. Many grid/utility companies have explicitly adopted the accuracy of electricity consumption prediction as a criterion for performance assessment. Specifically, the prediction accuracy is obtained by comparing the predicted data of the electricity consumption of the current month to the predicted data of the actual electricity consumption of the next month. At present, the requirement of the power grid/power company on the prediction precision is more than or equal to 96%, but many power companies cannot reach the index, so that a proper prediction method is needed. (2) Controlling company profit. Ensuring the balance of profit is central to the financial robustness of the utility company in terms of economic management strategies. The utility company can translate into accurate revenue budgets through accurate electricity prediction to optimize the payout architecture. (3) Potential users are discovered. When marketing schemes are formulated for different power users, the power utilization modes of the power users are analyzed firstly, future power utilization trends of the power users are determined qualitatively or quantitatively, and then the users with high development potential are identified by using the acquired real-time information. (4) And making an effective decision. The electric company performs a reliable evaluation on the total sales power in the future through predictive analysis of the monthly power consumption, so that an effective strategy is formulated and fine adjustment is performed in the electric power purchase. In summary, accurate prediction of electricity consumption plays a vital supporting role in performance evaluation and execution of power marketing strategies for electric power companies. In addition, the accuracy of the predictions is closely related to the economic development at the enterprise level. Therefore, the method has great practical significance in scientific prediction and research of the electricity consumption and wide engineering application value. Disclosure of Invention The invention aims to provide a power consumption prediction method, which comprises the following steps: Acquiring electricity consumption data of a designated area in a target period, and preprocessing the electricity consumption data; Processing the preprocessed electricity consumption data by using a target algorithm to obtain target decomposition parameters for predicting the applied electricity consumption, wherein the target algorithm is obtained based on a gray wolf algorithm; Performing variation modal decomposition on the preprocessed electricity consumption data based on the target decomposition parameters to obtain a plurality of intrinsic modal components; Based on a plurality of the intrinsic mode components, the power consumption is predicted by utilizing a two-way long-short-term memory network, and a candidate prediction result of the power consumption is obtained; Determining influencing factors influencing the electricity consumption; Constructing an error prediction model by combining the influence factors, wherein the error prediction model is used for predicting an error sequence in the candidate prediction result; predicting the candidate prediction result by using the error prediction model to obtain a target error sequence; And correcting the candidate prediction result based on the target error sequence to obtain a final power consumption prediction result. In an embodiment, the preprocessing the electricity consumption data includes: Normalizing and encoding the electricity consumption data; Filling missing data into the power consumption data after normalization and encoding, wherein the filling comprises weighting filling or average filling; And identifying and correcting abnormal values of the power consumption data after filling the missing values by using an isolated forest algorithm and a horizontal smoothing algorithm to obtain target power consumption data. In one embodiment, constructing the target algorithm based on the wolf algorithm includes: Per