ELECTRICAL LOAD FORECASTING IN NIGERIA USING ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM (ANFIS) TECHNIQUE

Abstract

Load forecasting is essential to the operations of any power utility company. It allows optimum usage of the available resources. Many load forecasting techniques are available. This research article presents a soft computing technique based on the learning abilities of an adaptive network and the cognitive skills of the fuzzy inference system. The Adaptive Neuro-Fuzzy Inference System (ANFIS) is to be used for this research purpose ANFIS belongs to a class of adaptive nerworks and is functionally equivalent to a Fuzzy Inference System.

Selection of significant inputs and identification of a structure that correctly captures the information contained in the system to be modeled, are the two basic steps in developing any model. This model forecasts the load demand by extrapolating the time of the day and the latest load behaviour. The short term historical load data is used as the variable input as well as time for short term load forecast, while the population impact is considered for long term load forecast. In carrying out this research, model is developed on a MATLAB platform for this purpose.

The ANFIS model performed excellently when applied to the Nigerian electrical system obtained quite promising results with the root mean square error evaluated as 0.51%. ANFIS proved to be a valid and promising technique for forecasting electricity consumption.