A Method of Power Load Forecasting Based on Sparse Representation

Abstract

The present invention relates to a power load forecasting method based on sparse representation, comprising the following steps: obtaining real power load data through a power company, and preprocessing the data, including specific operations such as data missing, data abnormality and data normalization, etc. ;By concatenating the basic dictionary, unit matrix dictionary and external factor dictionary to form an overcomplete dictionary, the overcomplete dictionary includes two parts: a training dictionary for solving sparse coefficient vectors and a test dictionary for predicting future power loads; according to the preprocessing The power load data and the training dictionary use the orthogonal matching pursuit OMP algorithm to solve the coefficient vector; combine the test dictionary and the sparse coefficient vector to predict the future power load. Based on the real electric load data set, this method can effectively predict the future electric load, and can significantly improve the accuracy of load forecasting under the condition of adding external factors.

Description

technical field [0001] The present invention relates to the technical field of power load forecasting, in particular to a power load forecasting method based on sparse representation. Background technique [0002] With the development of national economy and technology and the improvement of people's living standards, electric energy has become an indispensable secondary energy in people's daily production and life, bringing endless convenience to people's production and life. Power load forecasting is one of the important research topics and important development directions of power grid operation. At present, a lot of forecasting research has been done based on short-term load forecasting theories and methods, and many methods have been proposed, which can be roughly divided into two categories: one is traditional methods represented by time series methods, such as time series methods, etc. These methods The algorithm is simple, fast and widely used, but because it is ess...

AI Technical Summary

Problems solved by technology

At present, based on short-term load forecasting research theory and methods, a lot of forecasting research has been done, and many methods have been proposed, which can be roughly divided into two categories: one is the traditional method represented by time series method, such as time series method, etc. These methods The algorithm is simple, fast, and widely used, but because it is essentially a linear model method, there are many shortcomings and limitations, and it cannot truly reflect the nonlinear characteristics of different load models of power companies; the other is based on artificial neural networks. The new artificial intelligence method represented by the network, the neural network has the ability of parallel distribution of information and self-learning and arbitrarily approximating continuous functions, and can capture various trends of short-term power loads

The BP network requires a large amount of historical data for training, and its ability to learn and deal with uncertainty and artificial information is poor

FUZZY forecasting is a forecasting method that has been emerging in power company load forecasting in recent years, but from the perspective of practical application, the accuracy of FUZZY method for load forecasting is often unsatisfactory

In addition, with the expansion of the grid scale, there are more and more factors involved in various aspects, and a large amount of uncertain information will inevitably appear. Therefore, the current commonly used load forecasting methods cannot adapt well to this aspect.

Images