Data Fitting Method of Customer Electric Energy Consumption Based on Orthogonal Polynomial Neural Network

The electric energy consumption data of marketoriented customers must be perfectly metered or accurately fitted, that is essential for the construction of electricity spot market. The electricity consumption behavior of market-oriented customers is complex and variable, to address the issue that the patterns of the electricity consumption data are difficult to be accurately characterized, an orthogonal polynomial neural network-based data fitting model of electric energy consumption is established. The model is implemented using Chebyshev orthogonal polynomials, Hermite orthogonal polynomials, Legendre orthogonal polynomials, and Laguerre orthogonal polynomials respectively, while the neural network weight coefficients are trained by gradient descent algorithm. The results show that the fitting effect of the same type of electricity consumption data differs significantly among different implementation methods. The neural network models using Hermite orthogonal polynomials and Laguerre orthogonal polynomials have higher fitting accuracy than other models. It is an effective way to achieve accurate data fitting of customer electricity consumption by selecting the corresponding fitting method according to the type of electricity consumption behavior.