A distribution network short-term load forecasting method based on multi-mode fusion

Abstract

The invention discloses a distribution network short-term load forecasting method based on multi-mode fusion, which mainly comprises the following steps: 1) collecting power network historical load sequence data X. 2) performing STL time series decomposition on that historical load time serie data X. 3) The LSTM neural network model with N kinds of structure of trend term serie, Xtrend and LSTM neural network model with N kinds of structure of residual term series Xremainder and an ensemble forecasting model are obtained. 4) that prediction result Os of the periodic term is obtained. 5) obtaining a prediction sample. 6) inputting that prediction sample into the prediction model, thereby obtaining a trend item prediction result Ot and a residual item prediction result Or. 7) integrating that prediction results Os of the cycle item, the trend item prediction result Ot and the remaining item prediction result Or, and using integrated forecasting to obtain a final forecasting result (shownin the description). The present invention contributes to improving the forecasting accuracy of the model while improving the robustness and generalization performance of the load forecasting model force.

Description

technical field [0001] The invention relates to the field of power system load forecasting, in particular to a short-term load forecasting method for distribution networks based on multi-mode fusion. Background technique [0002] Accurate load forecasting is of great significance for operation and maintenance personnel to master the safe and stable operation of the distribution network system. The existence of prediction error directly increases the additional cost of power system operation, which is not conducive to the improvement of economy. In the past few decades, many scholars at home and abroad have proposed many algorithm models for the problem of short-term load forecasting to improve its accuracy. However, because the load sequence is greatly affected by external factors, and the sequence change is nonlinear, random and uncertain, it is very difficult to improve its accuracy. [0003] At present, load forecasting technology has gradually shifted from traditional ...

AI Technical Summary

Problems solved by technology

Although the hybrid model can effectively improve the accuracy of short-term load forecasting, the forecasting part of the model is single, which is greatly affected by parameter adjustment, and is not adaptable to multi-type characteristic data.

Images