Substation electric quantity trend predictive analysis method based on machine learning

Abstract

The invention discloses a substation electric quantity trend predictive analysis method based on machine learning. The method comprises the steps that influence factors of substation electric quantitytrend prediction are analyzed, and a characteristic quantity type needed for model construction is determined; a multidimensional characteristic quantity dataset is constructed based on collected electric quantity data and the characteristic quantity type; and a GBDT and Adaboost integrated prediction model is constructed, and a value of a root-mean-square error is adopted to compare prediction effects of evaluation models. Through the substation electric quantity trend predictive analysis method based on machine learning, characteristic factors possibly influencing the prediction effects arefully considered, so that predictive analysis is more accurate; and by the adoption of the regression-based GBDT and Adaboost integrated learning algorithm, data over-fitting is prevented, and continuous training, analysis and optimization of the prediction model can be realized.

Description

technical field [0001] The invention relates to the field of electric energy meter measurement, in particular to a method for predicting and analyzing substation power trend based on machine learning. Background technique [0002] With the continuous development of the national economy and the continuous improvement of people's living standards, the annual electricity consumption is also increasing steadily. The analysis of substation power data is closely related to the interests of power companies and regional economies. Academics and industries have been trying to analyze and predict substation power. The existing substation power analysis and prediction methods mainly include prediction algorithm analysis based on classical theoretical formulas and time series prediction analysis based on univariate: [0003] 1. Forecasting algorithm analysis based on classical theoretical formulas A set of classical theoretical formulas is obtained mainly based on historical data mathe...

AI Technical Summary

Problems solved by technology

[0006] This type of method can solve the problem of iterating and correcting model parameters using future actual data, and can integrate the jitter variation law of future actual data to realize data prediction. However, this method only considers the single factor of power, and does not consider the factors that cause power changes. Other factors, such as holidays, temperature, etc., therefore also have certain limitations

Images