Neural Model and Simulation of the Polish Power System Development in MA TLAB and Simulink Environment

Abstract

The paper presents selected results of research related to the design and implementation of a neural model of the Polish Power System (EPS) [Krajowy System Energetyczny] development. A perceptron Artificial Neural Network in MATLAB and Simulink environment using Neural Network Toolbox library was designed. Numeric data for the period of 40 years (1968–2007), i.e. 14 input quantities and 4 output quantities were used for training the ANN with one hidden layer containing 10 neurons. The tansigt() function was selected as the activation function for both the hidden and the input layer, and the trainlm() function, based on the Levenberg-Marquardt training method was used for training. After teaching the ANN the model of the EPS, the ANN was generated in the form of a block in order to use the implemented ANN as the model of the neural EPS system for research in Simulink environment. For that purpose, a block diagram was generated in Simulink, which was then used for conducting comparative research. Relatively insignificant discrepancies were obtained between the output from the model and the output from the system, e.g. for output Y l (t), i.e. for the total maximum power achievable in power plants, the average error value did not exceed 0.12%.