Multi-model fusion electricity sales prediction method based on empirical mode decomposition

Abstract

The invention discloses a multi-model fusion electricity sales prediction method based on empirical mode decomposition, and relates to the technical field of electricity marketing. The method comprises the following steps: 1, obtaining a plurality of different sub-sequences by adopting an EMD decomposition algorithm; 2, sequentially dividing the sequences into a high-frequency sequence, a medium-frequency sequence and a low-frequency sequence; 3, predicting the high-frequency sequence, the medium-frequency sequence and the low-frequency sequence by respectively using a long-short-term memory network model, a random forest model and an XGBoost model based on an attention mechanism to respectively obtain prediction results of each sub-sequence; 4, superposing and reconstructing the prediction result of each sub-sequence to obtain an actual prediction result. Compared with a conventional single model, the fusion model provided by the invention has higher prediction precision, and comparedwith an existing prediction algorithm, the fusion model is more accurate and stable in prediction.

Description

technical field [0001] The invention relates to the technical field of electric power marketing, in particular to a multi-model fusion electricity sales forecasting method based on empirical mode decomposition. Background technique [0002] With the gradual deepening of the new round of electricity reform, the electricity trading market has been further liberalized, and the electricity sales market has been gradually liberalized, resulting in many electricity sales companies. However, due to the fact that a mature electricity spot market mechanism has not yet been established in China, the assessment of deviation electricity has become a An important factor affecting the profit of electricity sales companies[. At present, the methods to reduce the deviation power mostly adopt the method of accurately sensing user behavior, adjusting controllable load and based on market transactions. However, the adjustment of controllable load and the way based on market transactions are p...

AI Technical Summary

Problems solved by technology

However, the effects of each model on the prediction results are different, and the method of taking the average cannot reflect this

[0004] The above method can be used for day-ahead or month-ahead forecasting, but it cannot meet the forecast needs of different time dimensions for annual bilateral negotiation transactions and monthly centralized bidding transactions in the electricity sales market

Images