Household load prediction method and system based on load characteristics and power consumption behavior mode

Abstract

The invention discloses a household load prediction method and system based on load characteristics and a power consumption behavior mode. The method comprises the steps: firstly collecting a household original power consumption load time sequence to obtain an original data set, and carrying out the preprocessing of the original data set to obtain a complete original data set; then collecting power utilization data characteristics from the complete original data set by adopting a PCA algorithm based on load characteristics and a power utilization behavior mode, completing characteristic extraction based on an SVD decomposition covariance matrix, realizing data dimension reduction, obtaining a stable sequence through differential transformation and Exponential motion averge processing, capturing a more complex nonlinear relationship of the sequence by utilizing a combined model, and obtaining a power utilization characteristic matrix. A data input structure is optimized, so that the model is combined with a historical state during prediction, the fitting precision is improved through a multivariable model, and the precision of an existing power load prediction method is improved.

Description

technical field [0001] The invention belongs to the field of electric load forecasting, and relates to a family load forecasting method and system based on load characteristics and power consumption behavior patterns. Background technique [0002] The level of power load forecasting has become a key symbol of the modernization of power system operation and management. It is related to the economic interests of all parties involved in power market transactions, and is related to the safe and economic operation of power systems with increasing uncertainties. Due to the many factors related to short-term load changes (holidays, climate, electricity price, population), significant non-stationary stochastic process in time series, complex load change law, high randomness and uncertainty of load, etc., the short-term load Predictions become more challenging to predict. Short-term power load forecasting methods can be simply divided into three categories: the first category is met...

AI Technical Summary

Problems solved by technology

The high-frequency components obtained after load decomposition usually have a large component of noise, which is difficult to predict

Since this part of the component is also an important part of the original load, if the noise reduction or filtering process is directly adopted, it will obviously reduce the accuracy of the prediction

At the same time, the current short-term power load forecasting model lacks the analysis and processing of data characteristics in power scenarios, which makes the model lack of heterogeneity

The input data of the model usually only includes load data, weather data, temperature data, etc., lacking more information in the load time series based on the operating characteristics of the power system and residents' electricity consumption behavior characteristics, and the prediction of household load cannot be accurately realized

Images