Research on Power Grid Loss Prediction Model based on Granger Causality Property of Time Series

J. Wang, W. Yan, H. Xu, X. Wang, and J. Yuan (PRC)


Power grid loss, Granger causality, prediction model, Energy Efficiency


We propose a kind of prediction method for line loss based on Granger causality property of time series. Firstly, the stable property of the time series is analyzed by unit root tests. Then the Granger causality relationship between line loss and other related variables is tested. Then those time series that are Granger caused with line loss is employed for modeling. We thus design three kinds of prediction models, first model based on line loss binomial using electricity sales as predicting variable, second model considering both power sale and grid capacity, and the last model based on Autoregressive Distributed Lag Approach (ARDL) incorporating both power sale and square of power sale as predicting variables. Case study using 1980-2008 data of China power grid indicates that the model precision is satisfactory as 2.7-3.9% measured by mean error rate.

Important Links:

Go Back