The urban gas consumption time series was analyzed with phase space reconstruction based on chaos theory. The chaotic characters of urban gas consumption were identified by calculating the correlation dimension and largest Lyapunov exponent. Then, several methods including weighted one-rank local-region method, largest Lyapunov exponent method and Bayesian regularization neural network model were applied on forecasting of daily urban gas consumption. The test results indicate that the chaotic time series analysis method is feasible to be used in urban gas consumption forecasting. Combined with the advantages of chaos theory, neural network and Bayesian regularization method, the forecasting performance of Bayesian regularization neural network model based on phase space reconstruction is especially good.
