Improved wavelet packet power consumption prediction method in consideration with season characteristic

Abstract

The invention discloses an improved wavelet packet power consumption prediction method in consideration with season characteristic. The method comprises the steps of de-noising continuous two-dimensional power consumption matrixes to fit power data having stronger rule; meanwhile, predicting the power consumption based on a radial basis function neural network model, recognizing seasons having obvious power consumption season characteristic by introducing a season index, performing multi-month synthetic prediction on monthly power consumption having obvious season characteristic, and correcting the monthly prediction value according to proportional distribution. By adopting the technical solution, the requirement for power consumption prediction is met, the power consumption prediction accuracy can be effectively improved, the monthly prediction precision in peak and valley seasons can be effectively improved, and the goal of high overall prediction precision is achieved.

Description

technical field [0001] The invention belongs to the technical field of power market demand prediction, and relates to an improved wavelet packet power consumption prediction method considering seasonal characteristics, which can effectively improve the monthly prediction accuracy of peak and valley seasons. Background technique [0002] As my country's power market becomes more reasonable and perfect, the power industry needs to analyze investment demand trends, so as to adjust the company's development strategy, adapt to the complex power trading market environment after the power reform, and thus put forward higher standards for the quality and accuracy of power consumption forecasting In order to strategically arrange investment and financing plans and provide strong support for power system planning, it is required to predict electricity consumption for 12 months of the following year. [0003] Meteorological factors such as temperature, humidity, and wind speed will caus...

AI Technical Summary

Problems solved by technology

Factors such as holidays will have a large-scale impact on monthly electricity consumption, but noise reduction processing will underestimate seasonal peaks and valleys, and electricity consumption correction is required

[0004] Due to the randomness and complexity of short-term and medium-term unit volume changes, but also periodic fluctuations, the horizontal and vertical time points are continuous

Traditional prediction and correction methods such as linear regression method, sliding time prediction method and simulated annealing cannot be adapted, and cannot accurately distinguish the noise in the signal, which affects the correction of the predicted power consumption, so that the final predicted value cannot reach the accuracy rate. Require

Images