Modelling self-optimised short term load forecasting for medium voltage loads using tunning fuzzy systems and Artificial Neural Networks. (December 2015)

• Record Type: Journal Article
• Authors: Mahmoud, Thair S.; Habibi, Daryoush; Hassan, Mohammed Y.; Bass, Octavian
• Is Part Of: Energy conversion and management. Volume 106(2016)
• Page Start: 1396
• Page End: 1408
• Publication Date: 2015-12
• Subjects: Adaptive model -- Artificial Neural Networks -- Fuzzy Logic -- Generation scheduling -- Optimum tuning -- Short term forecasting; Direct energy conversion -- Periodicals; Energy storage -- Periodicals; Energy transfer -- Periodicals; Énergie -- Conversion directe -- Périodiques; Direct energy conversion; Periodicals; 621.3105
• DOI: 10.1016/j.enconman.2015.10.066
• Languages: English
• ISSNs: 0196-8904

Abstract

Highlights: A novel Short Term Medium Voltage (MV) Load Forecasting (STLF) model is presented. A knowledge-based STLF error control mechanism is implemented. An Artificial Neural Network (ANN)-based optimum tuning is applied on STLF. The relationship between load profiles and operational conditions is analysed. Abstract: This paper presents an intelligent mechanism for Short Term Load Forecasting (STLF) models, which allows self-adaptation with respect to the load operational conditions. Specifically, a knowledge-based FeedBack Tunning Fuzzy System (FBTFS) is proposed to instantaneously correlate the information about the demand profile and its operational conditions to make decisions for controlling the model's forecasting error rate. To maintain minimum forecasting error under various operational scenarios, the FBTFS adaptation was optimised using a Multi-Layer Perceptron Artificial Neural Network (MLPANN), which was trained using Backpropagation algorithm, based on the information about the amount of error and the operational conditions at time of forecasting. For the sake of comparison and performance testing, this mechanism was added to the conventional forecasting methods, i.e. Nonlinear AutoRegressive eXogenous-Artificial Neural Network (NARXANN), Fuzzy Subtractive Clustering Method-based Adaptive Neuro Fuzzy Inference System (FSCMANFIS) and Gaussian-kernel Support Vector Machine (GSVM), and the measured forecasting error reduction average in a 12 month simulation period was 7.83%, 8.5% and 8.32% respectively. The 3.5 MW variable load profile of Edith Cowan University (ECU) in Joondalup, Australia, was used in the modelling and simulations of this model, and the data was provided by Western Power, the transmission and distribution company of the state of Western Australia.