Die IEEE Konferenz Innovative Smart Grid Technologies (ISGT 2016) fand in Melbourne, Australien vom 28.11. bis 1.12.2016 statt.
Mein erster Vortrag „Input Data Analysis for Optimized Short Term Load Forecasts“ (DOI) konnte ich in der Session „Demand Forecast“ am Vormittag des 1.12. vorstellen.
Accurate electrical short term load forecasts play an important role for grid operation, power plant scheduling and power trading. The need for precise forecasts rises as energy markets are in a phase of transition due to severe changes in the energy system for European countries. This transition is mainly caused by rising shares of renewable energies, increasing energy efficiency and consumption pattern changes.This paper presents a case study based on publicly available data, on how selection of weather and economic data, the historic availability of training data and regional aspects affect the quality of short term load forecasts. By using evaluation metrics, such as the normalized rooted mean square error (NRMSE) and the mean average percentage error (MAPE), forecasting results may be compared to other case studies. Furthermore, the case study is based on a novel model framework utilizing e. g. artificial neuronal network (ANN), support vector regression (SVR) and similar day models. By selecting optimized training data sets we increase forecasting accuracy for supported models up to 11%. Forecasting results of supported models for a whole year, including holidays and weekends, range from 2.1% to 2.38% MAPE.
Der zweite Vortrag bezog sich ebenfalls auf die Verbesserung von Lastprognosen und wurde in der Nachmittagssession präsentiert: „SAWing on Short Term Load Forecasting Errors: Increasing the Accuracy with Self Adaptive Weighting“ (DOI)
Die Leitung der Session „Demand Forecast“ habe ich übernommen, Prof. Mori von der Meiji University in Japan hat mich unterstützt.
Accurate electrical load forecasts are of vital interest to power companies. Short term load forecasts for next hours in particular are important for power dispatch, power trading and system operation. This paper analyzes the conjectures that a self-adaptive weighting algorithm (SAW), blending different standard load forecasting approaches, such as a dynamic standard load profile model, a linear regression model and an artificial neuronal network model, can increase forecasting performance on micro grids for one hour intraday to 24 hours day ahead forecasts.The SAW methodology and forecasting models are applied to a publicly available smart meter data set. Common evaluation metrics such as the mean average percentage error (MAPE) and the normalized rooted mean square error (NRMSE) are used to evaluate the performance of this new hybrid approach and allow a comparison to other studies. Self-adaptive weighing leads to a significant improvement of intraday and day ahead forecasts from 50% – 54% and 30% – 35% (MAPE improvement for 1h and 24h compared to input forecasts). The resulting intraday and day ahead load SAW forecasts range from 3.19% 1h MAPE to 4.50% 24h MAPE in this case study.