Improving the electricity distribution loss forecasting accuracy, by Michael Heeren (with Enexis B.V.)
June 8, 2013
This master thesis describes research for improving the distribution loss forecasting process at Enexis B.V. This research compared two commonly used forecasting methods, namely based on multiple regression and based on the time series approach. The objective was to accurately forecast the distribution losses one year ahead on an hourly basis. It is shown that the simple multiple regression methods is the optimal approach to forecast electricity infeed. The electricity infeed is forecasted using the regression approach, taking into account calendar, meteorological, and historical infeed data as independent variables. Electricity infeed is forecasted for every hour of a day and for every grid of Enexis separately. This approach outperforms the approach of using one model. The importance of the independent variables are described in the thesis. Unfortunately, due to the limited amount of data, no significant effect has been found between economic and demographic data and electricity infeed. It is concluded that electricity infeed can accurately be forecasted using a simple multiple regression approach for every hour of a day. However, the relationship between electricity infeed and electricitydistribution losses is still vague. It is therefore recommended to the company to gain more insight in this complicated part of distribution loss forecasting.