Residential electricity load prediction method based on multi-model fusion, medium and equipment

Abstract

The invention relates to a residential electricity consumption load prediction method based on multi-model fusion, a medium and equipment, and the prediction method comprises the following steps: constructing a plurality of prediction models for residential electricity consumption prediction, obtaining residential electricity consumption data xi (i = 1, 2,..., t-1) as a training set, and training each prediction model by using the training set; each prediction model predicts and obtains the predicted electricity consumption at the moment t based on the resident electricity consumption data xi, i = 1, 2,..., t-1, the moment t has known real electricity consumption, and the corresponding relative error of each prediction model is calculated based on the predicted electricity consumption and the real electricity consumption; calculating a confidence factor of each prediction model based on the relative error of each prediction model; each prediction model performs prediction based on the resident electricity consumption data xi, i = 1, 2,..., t to obtain a resident electricity consumption prediction result at the t + 1 moment; and performing fusion processing based on the confidence factor of each prediction model and the resident electricity consumption prediction result to obtain a final prediction result at the t + 1 moment. Compared with the prior art, the method has the advantages of wide application range, high prediction precision and the like.

Description

technical field [0001] The invention relates to the technical field of residential electricity consumption forecasting, in particular to a method, medium and equipment for residential electricity load forecasting based on multi-model fusion. Background technique [0002] Accurate prediction of residential electricity consumption is of great significance to guide the dispatch and optimization of power flow. [0003] Existing residential electricity prediction methods can be divided into methods based on gray scale prediction, autoregressive moving average model, artificial intelligence method, and deep learning model. It has the following defects: the method based on grayscale prediction is to model the historical data of residential electricity consumption with differential equations, so as to make predictions based on differential equations. This method is sensitive to abnormal data. The autoregressive moving average model is a regression analysis of the historical data of...

AI Technical Summary

Problems solved by technology

It has the following defects: the method based on grayscale prediction is to model the historical data of residential electricity consumption with differential equations, so as to make predictions based on differential equations. This method is sensitive to abnormal data

The autoregressive moving average model is a regression analysis of the historical data of residential electricity consumption and its white noise, and predictions are made based on the regression model. This method can only represent the linear relationship between the data

Artificial intelligence methods, such as artificial neural network (single hidden layer model) and support vector machine model, use the historical electricity consumption data of residents for training, so as to use the trained model to predict electricity consumption. This method can represent the relationship between data Complex relationship, but the amount of calculation is large

Deep learning models, such as recurrent neural network and long-term short-term memory network methods, are consistent with the idea of ​​using artificial neural network (single hidden layer model) for residential electricity consumption prediction. Training, and then predicting the electricity consumption of residents based on the trained model. Compared with artificial neural network and support vector machine, the representation ability of deep learning model is stronger, but the amount of calculation is correspondingly larger

Both artificial intelligence methods and deep learning models are black-box models, which lack interpretability for the predicted residential electricity consumption, and are difficult to apply in scenarios with high safety factors

At the same time, due to the complexity and randomness of residents' electricity consumption behavior, the existing forecasting methods are difficult to accurately predict the electricity consumption of different residents in different scenarios

Images