Gas load forecasting based on chaotic theory and Volterra adaptive filter

Abstract

Gas is an important source of clean energy, so in the future, gas load forecasting will have a very important meaning for the gas company to develop an appropriate business plan between‘upstream’gas supplier and‘downstream’users. Aiming at the gas short-term load forecasting problem, a prediction model has been put forward based on chaotic theory and Volterra adaptive filter. Firstly, conduct day-to-day correlation analysis for collected gas hour load data, and then calculate delay times and find out the optimal embedding dimension by the approach of mutual information and pseudo-nearest-neighbor. Secondly, on the basis of phase-space reconstruction, carry out chaotic characteristic analysis for collected gas hour load data. Once again, aiming at the current situation of most existing prediction models being subjective, reduce the subjectivity in the process of gas load forecasting by introduction of Volterra adaptive filter prediction model. Finally, the influence of the different order of Volterra adaptive filter on prediction results, and the comparison of accuracy among the Volterra adaptive filter prediction model, ANN (artificial neural network, ANN) prediction model and Fourier series prediction model are discussed through gas load forecasting example. Additionally, the forecasting results of the second order Volterra adaptive filter prediction model showed that: compared with ANN prediction model and fourier series prediction model, the second order Volterra adaptive filter prediction model has higher accuracy, and may provide a useful reference for practical engineering applications of short-term gas load forecasting.