Abstract
This paper presents a confidence interval estimating method for load forecasting with consideration of error causes. In recent years, it has been planned to introduce the new balancing rules by Ministry of Economy, Trade and Industry (METI). It will be required to submit the planned values of power generation amount and load amount. Thus, it is necessary to evaluate the confidence interval for load forecasting in terms of power producer and supplier. The proposed method uses the decision tree to analyze error causes and the beta distribution to estimate proper confidence interval. The former has functions to classify data into terminal nodes and extract rules from each terminal node with consideration of error causes. The latter expresses a variety of distribution shape by parameters estimated maximum likelihood for obtained error distribution. The effectiveness of the proposed method is demonstrated using actual data of considerable local electric power utility. The proposed confidence interval estimating method succeeded in reducing 5.0%pt of Prediction Interval Coverage Probability (PICP) and 0.9%pt of Normalized Mean Prediction Interval Length (NMPIL) for the conventional method without consideration of error causes. Therefore the proposed method can estimate more accurately than conventional method.
