CN-119582181-B - Power consumption demand determining method and device, storage medium and electronic equipment

CN119582181BCN 119582181 BCN119582181 BCN 119582181BCN-119582181-B

Abstract

The invention discloses a power consumption demand determining method, a device, a storage medium and electronic equipment. The method comprises the steps of obtaining current load data of an electricity account in a current period, decomposing the current load data to obtain a plurality of current load components, obtaining a plurality of predicted load components by adopting an electricity demand prediction model based on the plurality of current load components, and obtaining an electricity demand prediction result of the electricity account in a predicted period based on the plurality of predicted load components, wherein the predicted period is the next sampling period of the current period. The method solves the technical problem that in the related art, the electricity demand prediction is directly carried out by taking the whole time period as a unit, the information of the whole time domain cannot be fully utilized, and the electricity demand prediction accuracy is low.

Inventors

JIAO JIANLIN
WEI TIANYING
DONG NING
LIU HUI
SUN HELIN
ZHANG JUNYI
NAN LIN
LIU WEI
LU ZHONGYAN
XI SHAOQING

Assignees

国网北京市电力公司
国家电网有限公司

Dates

Publication Date: 20260512
Application Date: 20241125

Claims (9)

1. A method of determining electricity demand, comprising: Acquiring current load data of an electricity consumption account in a current period, wherein the current period is one day, and the current load data comprises load data respectively acquired at a plurality of sampling moments in the current period; decomposing the current load data to obtain a plurality of current load components; Based on the current load components, a power demand prediction model is adopted to obtain a plurality of predicted load components, wherein the power demand prediction model comprises a first bidirectional long-short-term memory network layer, an attention layer, a second bidirectional long-short-term memory network layer and a full-connection layer, hidden features in the current load components are extracted by the first bidirectional long-short-term memory network layer to obtain a plurality of groups of initial hidden features, the attention mechanism in the attention layer is adopted to carry out weighted processing on the plurality of groups of initial hidden features to obtain a plurality of groups of weighted hidden features, the second bidirectional long-term memory network layer is adopted to carry out feature extraction on the plurality of groups of weighted hidden features to obtain a plurality of groups of feature extraction results, and the full-connection layer is adopted to predict the plurality of predicted load components based on the plurality of groups of feature extraction results; and obtaining a power consumption demand prediction result of the power consumption account in a prediction period based on the plurality of predicted load components, wherein the prediction period is the next sampling period of the current period.
2. The method of claim 1, wherein the decomposing the current load data to obtain a plurality of current load components comprises: adding multiple Gaussian white noise to the current load data to obtain multiple groups of current noise data, wherein the multiple Gaussian white noise obeys standard front-to-front distribution; Respectively carrying out empirical mode decomposition on the plurality of groups of current noise data to obtain a plurality of groups of eigenmode functions, wherein the plurality of groups of current noise data are in one-to-one correspondence with the plurality of groups of eigenmode functions, each group of eigenmode functions in the plurality of groups of eigenmode functions comprises a plurality of eigenmode functions, and the plurality of eigenmode functions are obtained by carrying out multiple empirical mode decomposition on the corresponding current noise data; And obtaining the plurality of current load components based on the plurality of groups of eigenmode functions.
3. The method of claim 1, wherein said weighting the plurality of sets of initial hidden features using an attention mechanism in the attention layer to obtain a plurality of sets of weighted hidden features, comprises: Any one set of weighted hidden features in the plurality of sets of weighted hidden features is obtained by: determining, for any one of the plurality of sets of initial hidden features, score values respectively corresponding to multidimensional hidden features included in the any one set of initial hidden features; Normalizing the scoring values corresponding to the multidimensional hidden features respectively to obtain standard scoring values corresponding to the multidimensional hidden features respectively; Determining the attention weights respectively corresponding to the multidimensional hidden features; Weighting calculation is carried out based on the attention weights respectively corresponding to the multi-dimensional hidden features and the standard grading values respectively corresponding to the multi-dimensional hidden features, so that any group of weighted hidden features are obtained; And obtaining the multiple groups of weighted hidden features by adopting a mode of obtaining the arbitrary group of weighted hidden features.
4. The method of claim 1, wherein prior to said deriving a plurality of predicted load components based on said plurality of current load components using a power demand prediction model, the method further comprises: Acquiring historical load data acquired respectively by a plurality of historical time periods; Preprocessing the historical load data respectively acquired by the plurality of historical time periods to obtain preprocessed load data respectively corresponding to the plurality of historical time periods, wherein the preprocessing comprises at least one of missing value filling and abnormal value correction; normalizing the preprocessing load data corresponding to the plurality of history time periods respectively to obtain normalized load data corresponding to the plurality of history time periods respectively; And training an initial model based on the normalized load data respectively corresponding to the plurality of historical time periods to obtain the power demand prediction model.
5. The method of claim 4, wherein preprocessing the historical load data collected during each of the plurality of historical time periods comprises: Detecting whether missing values exist in the historical load data respectively collected by the plurality of historical time periods; Filling the missing value under the condition that the missing value exists in the historical load data respectively collected by the plurality of historical time periods; detecting whether abnormal values exist in the historical load data respectively collected by the plurality of historical time periods; And when the abnormal value exists in the historical load data acquired by the plurality of historical time periods respectively, correcting the abnormal value.
6. The method of claim 5, wherein detecting whether an outlier exists in the historical load data collected during each of the plurality of historical periods comprises: judging whether the historical load data respectively collected by the plurality of historical time periods meet the following preset conditions or not: ; Wherein, the Representing load data corresponding to an ith sampling time in an nth historical period in the plurality of historical periods, wherein each historical number segment in the plurality of historical periods comprises a plurality of sampling times; represents any one of the plurality of history periods, N represents a total number of the plurality of history periods; representing an average value of load data corresponding to an ith sampling time in the plurality of history periods; representing the variance of the load data at the i-th sampling instant; a standard deviation of the load data representing the i-th sampling time; Representing a preset threshold value; Determining that the abnormal value does not exist in the load data corresponding to the plurality of history periods when the load data corresponding to the plurality of history periods respectively do not meet the predetermined condition; And determining any load data as the abnormal value under the condition that any load data meets the preset condition in the load data corresponding to the historical periods respectively.
7. An electricity demand determining apparatus, comprising: the system comprises an acquisition module, a power consumption account generation module and a power consumption account generation module, wherein the acquisition module is used for acquiring current load data of the power consumption account in a current period, the current period is one day, and the current load data comprises load data respectively acquired at a plurality of sampling moments in the current period; The decomposition module is used for decomposing the current load data to obtain a plurality of current load components; The prediction module is used for obtaining a plurality of predicted load components by adopting a power demand prediction model based on the plurality of current load components, and comprises the steps of adopting the first bidirectional long-short-term memory network layer to extract hidden features in the plurality of current load components to obtain a plurality of groups of initial hidden features, adopting an attention mechanism in the attention layer to carry out weighted processing on the plurality of groups of initial hidden features to obtain a plurality of groups of weighted hidden features, adopting the second bidirectional long-short-term memory network layer to carry out feature extraction on the plurality of groups of weighted hidden features to obtain a plurality of groups of feature extraction results, and adopting the full-connection layer to predict the plurality of predicted load components based on the plurality of groups of feature extraction results; And the result determining module is used for obtaining a power consumption demand prediction result of the power consumption account in a prediction period based on the plurality of predicted load components, wherein the prediction period is the next sampling period of the current period.
8. A non-volatile storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the electricity demand determining method of any one of claims 1 to 6.
9. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the electricity demand determination method of any of claims 1-6.

Description

Power consumption demand determining method and device, storage medium and electronic equipment Technical Field The invention relates to the field of smart grids, in particular to a method and a device for determining electricity demand, a storage medium and electronic equipment. Background In the field of electricity demand prediction, a prediction model plays a key role in load prediction. In recent years, deep learning techniques typified by deep neural networks have been widely developed, and have become a popular field of study for students. The long-term and short-term memory network is a variant of the cyclic neural network, solves the gradient disappearance problem of the cyclic neural network, can learn long-distance dependence in time sequence data, and is the most popular deep neural network in the current load prediction field. However, the long-term and short-term memory network can only process data according to the time sequence order, only considers past information, and cannot fully utilize the information of the whole time domain. Aiming at the extraction problem of time sequence characteristic information, the influence of the information of different time steps of power load data on a prediction result is different, a long-term and short-term memory network cannot identify key time sequence characteristic information, and high prediction accuracy is difficult to obtain. In view of the above problems, no effective solution has been proposed at present. Disclosure of Invention The embodiment of the invention provides a power consumption demand determining method, a device, a storage medium and electronic equipment, which at least solve the technical problem that the power consumption demand prediction accuracy is low because the power consumption demand prediction is directly carried out by taking the whole time period as a unit in the related technology and the information of the whole time domain cannot be fully utilized. According to one aspect of the embodiment of the invention, a power consumption demand determining method is provided, which comprises the steps of obtaining current load data of a power consumption account in a current period, carrying out decomposition processing on the current load data to obtain a plurality of current load components, obtaining a plurality of predicted load components by adopting a power demand prediction model based on the plurality of current load components, and obtaining a power consumption demand prediction result of the power consumption account in a prediction period based on the plurality of predicted load components, wherein the prediction period is a next sampling period of the current period. According to another aspect of the embodiment of the invention, the power consumption demand determining device further provides a power consumption demand determining device, which comprises an acquisition module, a decomposition module, a prediction module and a result determining module, wherein the acquisition module is used for acquiring current load data of a power consumption account in a current period, the decomposition module is used for carrying out decomposition processing on the current load data to obtain a plurality of current load components, the prediction module is used for adopting a power demand prediction model to obtain a plurality of prediction load components based on the plurality of current load components, and the result determining module is used for obtaining a power consumption demand prediction result of the power consumption account in a prediction period based on the plurality of prediction load components, wherein the prediction period is the next sampling period of the current period. According to another aspect of an embodiment of the present invention, there is also provided a nonvolatile storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform any one of the electricity demand determining methods. According to another aspect of the embodiment of the present invention, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any one of the electricity demand determination methods. According to the embodiment of the invention, the current load data of the electricity consumption account in the current period is obtained, the current load data is decomposed to obtain a plurality of current load components, a power demand prediction model is adopted to obtain a plurality of predicted load components based on the plurality of current load components, and the electricity consumption demand prediction result of the electricity consumption account in the prediction period is obtained based on the plurality of predicted load components, wherein the prediction period is the next sampling pe