An Integrated Approach to Forecasting Intermittent Demand for Electric Power Materials

Forecasting intermittent demand for electric power materials is crucial for electric power companies to ensure the economical supply of electricity in terms of stock and cost management. This paper proposes an integrated approach to forecasting intermittent demand for electric power materials. The approach decomposes the demand time series into two parts: a binary time series representing demand occurrences and a series representing non-zero demands. From the perspective of the origin of demands, we forecast demand occurrences by both intrinsic and extrinsic factors. The probability of demand occurrence due to intrinsic factors is estimated by using the data of ontology failure, while for extrinsic factors is estimated by using the historical demand data and time series data of explanatory variables. The weighted sum of these two probabilities is compared with a critical value to predict demand occurrence. We apply the multivariate nonlinear regression method to forecast non-zero demands, which are combined with the predicted demand occurrences to form the final forecasting demand series. Two performance measures are developed to assess the forecasting methods. By making use of data set of 50 types of electric power materials from the State Grid Shanghai Electric Power Company in China, we show that our approach provides more accuracy forecast than Markov bootstrapping method, Syntetos–Boylan Approximation method and integrated forecasting method method.