Consumer Load Prediction and Theft Detection on Distribution Network Using Autoregressive Model

Abstract

Load prediction is essential for the planning and management of electric power system and this has been an area of research interest recently. Various load forecasting techniques have been proposed to predict consumer load which represents the activities of the consumer on the distribution network. Commonly, these techniques use cumulative energy consumption data of various consumers connected to the power system to predict consumer load. However, this data fails to reveal the activities of individual consumers as related to energy consumption and stealing of electricity. A new approach of predicting consumer load and detecting electricity theft based on autoregressive model technique is proposed in this paper. The objective is to evaluate the relationship between the consumer load consumption vis-a-vis the model coefficients and model order selection. Such evaluation will facilitate effective monitoring of the individual consumer behaviour, which will be indicated in the changes in model parameters and invariably lead to detection of electricity theft on the part of the consumer. The study used the data acquired from consumer load prototype which represents a typical individual consumer connected to the distribution network. Average energy consumption obtained over 24 hours was used for the modelling and 5-minute step ahead load prediction based on model order 20 of minimum description length criterion technique was achieved. Electricity theft activities were detected whenever there are disparities in the model coefficients and consumer load data.