Estimation of Load-Head and Load–Lightning Dependencies in Electric Load Model, Formed with Artificial Neural Network

Abstract

Problems of accurate electric load forecasting is principal for control of electric power system. For this purpose in electric power system control centres widely used artificial neural network electric load model with different structure. If this model used load – heat and load – lightning dependencies, amount of model inputs and learning dates increasing several times, and there are problems with learning processes convergence. In this research is proposed load – head and load – lightning dependencies estimation in electric load forecasting model, using mathematics laws of these processes and adding this with basic part – composite model of this structure is simple. Basic part of this model is made using artificial neural nets. Model of this structure have ability of load forecasting in time of transient heat process, caused changes of temperature. Strategy of model learning and adaptation to date changes are determined. Electric load forecasting with this model was done with Lithuanian electric power system load dates and accuracy of forecast is sufficient for practical application.