Abstract:

There are many factors to affect the accuracy of load forecasting. The problem is how to choose the factors. To solve the problem, an attribute reduction algorithm of concept lattice was introduced. We chose a property parameter that has good relativity to forecasting load as the input  parameter of the forecasting model of neural network. It reduces the scope of the input parameters, ensures the rationality of input parameters of the forecasting model, and solves the problem how to determine the input parameters of the neural network model. The actual data of the maximum load in some place of Chongqing City was calculated. The results show the prediction accuracy of neural network model is improved with such a method, and reduction algorithm is reasonable and effective.

Key words: neural network, concept lattice, attribute reduction, load forecasting