Prediction method of monthly power consumption

Abstract

The invention discloses a prediction method of monthly power consumption. The method comprises the following specific steps: step one, screening an industry index according to Pearson's correlation coefficient; step two, random forest training and predicting. The industry index with strong relevancy relation with the power consumption is screened by combining with the Pearson's correlation coefficient according to the upstream-downstream industry chain condition of each industry by use of each important industry as the point of view; and then the random forest training and modelling are performed on the screened industry index and the industry power consumption, thereby predicting the monthly power consumption of each industry of the important industries; the prediction method comprises thinning a prediction object to each important industry and thinning the prediction time to month, thereby contributing to sufficiently master the monthly power consumption condition of each industry by a forecaster; and meanwhile, the economy and the industry factors are effectively integrated into the power consumption prediction under the fast-changing economic situation, the influence over the power consumption by the economic structure change is sufficiently considered, and the power consumption development trend is immediately and accurately mastered.

Description

technical field [0001] The invention relates to the research field of medium-term load forecasting of electric power systems, in particular to a forecasting method of monthly electricity consumption. Background technique [0002] Monthly power consumption forecasting is an important task for power planning departments, power consumption and marketing departments. Its purpose is to reasonably arrange the medium-term operation plan of the power system, reduce operating costs, and improve power supply reliability. Monthly electricity consumption forecasting methods include conventional methods and intelligent methods. Conventional methods mainly include time series method, trend extrapolation method, regression analysis method, etc.; intelligent methods mainly include artificial neural network, gray system model, support vector machine, wavelet analysis, etc. [0003] In the conventional forecasting method, the time series method only associates the power consumption with the ...

AI Technical Summary

Problems solved by technology

However, some intelligent methods also have certain defects; artificial neural networks require a large number of parameters, such as network topology, initial values ​​of weights and thresholds, and the learning time is too long, and may even fail to achieve the purpose of learning; the gray system model is Historical data dependence is strong, and the connection between various factors is not considered; in wavelet analysis, the selection of wavelet base has a greater impact on the prediction results

[0005] The random forest algorithm can effectively avoid the above defects in many intelligent decision-making algorithms and achieve more accurate load forecasting, but people have not yet applied the random forest algorithm to the field of power forecasting

Images