TGCN-GRU ultra-short-term load prediction method and device based on VMD and electronic equipment

Abstract

The invention relates to a TGCN-GRU ultra-short-term load prediction method based on VMD. The method comprises the steps of obtaining first historical load data before a to-be-predicted time point; performing variational mode decomposition on the first historical load data to obtain a plurality of mode components and residual errors; a graph adjacency matrix is constructed, the graph adjacency matrix comprises graph nodes and connected nodes, the graph nodes are used for representing modal components and residual errors, and the connected nodes are used for representing an original load sequence; splicing the time sequence data of the first historical load data to form a first input matrix; and inputting the first input matrix and the graph adjacency matrix into a trained TGCN-GRU model to obtain load prediction data after the to-be-predicted time point, the TGCN-GRU model comprising a layer of TGCN network and a layer of GRU network which are connected with each other. According to the method, the modal component and the residual error can be analyzed through the TGCN-GRU model, and error accumulation caused by establishment of a plurality of models can be solved. And the original load sequence is used as input, so that the method can fully extract the relation on each graph node space.

Description

technical field [0001] The invention relates to the technical field of power systems, and in particular, to a method, device and electronic equipment for ultra-short-term load prediction of TGCN-GRU based on VMD. Background technique [0002] The short-term load forecasting of the power system is a scientific forecast of the load in the next few days or hours according to the historical load variation law, combined with meteorological, economic and other factors. Accurate load forecasting is an important decision-making basis for arranging power production scheduling and equipment maintenance plans. Therefore, it is necessary to study new methods and technologies of load forecasting to improve the accuracy and reliability of load forecasting and meet the requirements of engineering technology. [0003] At present, load forecasting methods mainly include methods based on mathematical statistics, machine learning methods and methods based on modal decomposition. Among them, ...

AI Technical Summary

Problems solved by technology

[0004] However, the load forecasting method based on modal decomposition needs to build a model for each IMF, and each model cannot predict its corresponding IMF completely correctly, which will inevitably cause errors, and it is impossible to judge whether these errors can offset each other, so It will cause the accumulation of errors and affect the accuracy of load forecasting

In addition, the existing load forecasting methods based on modal decomposition often ignore the role of the original load data, and after the modal decomposition, multiple IMFs and residuals are formed, and there are certain errors between these reconstructed data and the original load, which also have a certain negative impact on the prediction accuracy

Images