Previsão de carga e demanda de energia elétrica através de sistemas inteligentes

Abstract

Intelligent systems of future forecasts are fundamentals for the operation and management of electrical systems where the planning and decision making are performed to meet future energy needs, according to the current availability which depends on the previous events. The different types of consumers use electrical energy in different ways and each segment of them impacts directly the entire system. The demand forecast horizon is essential to define the strategies of planning and operation, fuel purchasing, sales and purchasing of energy, growth and expansion and further negotiations of contracts. This work deals with the development and implementation of an intelligent systems of demand forecast through forecasts neural systems. The network used for training and validation processes was the Multi-layer Perceptron with Time-Delay Neural Network (PMC-TDNN), because systems of electric energy have distribution features which vary with time and thus behave in a similar way with dynamics systems. The system was trained and validated through the historical real data, supplied by a distributor agent of electricity and the data were treated and manipulated so that the system could be more flexible to changes in the analyzes as possible. This flexibility allows that the analyzes are directed to the individual feeders and the forecasts are made for specific classes of consumers. The initial results of the work did not address nor the weather factors and neither the consumer classes segmentation, however the results were satisfactory and indicate that the forecasting system has a great potential for improvement and increase in accuracy from correlating these new variables.