Short-term load prediction method and device based on BP neural network, and storage medium

Abstract

The invention discloses a short-term load prediction method based on a BP neural network, and relates to the field of power supply and distribution, in particular to a short-term load prediction method and device based on the BP neural network. According to the method, nonlinear fitting is carried out by using the neural network, and the load state of the day is doubly positioned by combining the abnormal day state characteristics constructed by calendar information and the load capacity of the concentrated load period of each day, so that the accuracy of abnormal day load prediction is improved; abnormal day identification is carried out through analysis of the same type of days, the influence of abnormal days on normal day load prediction is eliminated, the maximum temperature of each day is extracted to carry out barrel dividing processing to construct effective temperature features, and the load prediction accuracy of the model on the normal day is improved.

Description

technical field [0001] The invention relates to the field of power supply and distribution, in particular to a short-term load forecasting method, equipment and storage medium based on a BP neural network. Background technique [0002] Among the load forecasting methods in the prior art, the forecasting method for normal days has become mature, but the load forecasting on abnormal days is affected by various random fluctuation factors and potential interference factors, resulting in low accuracy, so how to improve the accuracy? The accuracy of abnormal daily load forecast has become the focus of forecasters in the power industry. [0003] For the load forecasting of abnormal days in the prior art, due to the relative scarcity of data, it is difficult to select a suitable model for forecasting; when the load forecasting is carried out directly based on the calendar state, due to the different implementation methods of national statutory holidays by different units, the actual...

AI Technical Summary

Problems solved by technology

[0002] Among the load forecasting methods in the prior art, the forecasting method on normal days has become mature, but the load forecasting on abnormal days is affected by various random fluctuation factors and potential interference factors, resulting in low accuracy, so how to improve The accuracy of abnormal daily load forecasting has become the focus of attention of forecasters in the power industry

[0003] Due to the relative scarcity of data, it is difficult to select a suitable model for load forecasting on abnormal days in the prior art; when load forecasting is performed directly based on the calendar status, due to the different implementation methods of different units for national statutory holidays, the actual arrangement Often deviates significantly from the calendar schedule, resulting in less accurate load forecasting on unusual days

Images