Performance of exponential smoothing, a neural network and a hybrid algorithm to the short term load forecasting of batch and continuous loads

摘要

Short term load forecasting (STLF) is an important part of generation scheduling and optimal energy management. Small island power systems, however, often have consumers that individually represent a relatively large percentage of the systems overall load and need to specifically forecast these individual loads. This paper, using data from two large industries in Trinidad and Tobago, evaluates the performance of Exponential Smoothing, Neural Network and Hybrid (Artificial Neural Network and Exponential Smoothing) algorithms in predicting the demands of two differing load types, batch and continuous processes. Results indicate similar performance for all techniques with less than 1.5% mean average percentage error for continuous loads while none was able to accurately predict the erratic behavior of the batch process due to fundamental limitations in prediction algorithms.