Study of Unit Commitment with Load Forecasting Through Neural Network

Abstract

Unit commitment scheduling of power system depends upon the prediction of the load demand, load demand trend, availability of generating units, minimum and maximum generating capacity of the units, minimum up and down time of the generators and initial status of units. As per the past experiences of various power system utilities, different commitment schedules of units‟ can lead to huge difference in total operating cost incurred. Prediction of future load trends is quite essential for optimal decisiveness in power system operation and planning. Accurate hourly and daily load demand prediction holds an important role for appropriate scheduling of units. The medium term load forecasting is applied for the scheduling of annual maintenance, scheduling of fuel supplies, load dispatch, planning of generation shifting etc. This work presents dynamic programming forward approach to perform the unit commitment with medium-term load forecast, obtained through training of neural network. Both structure learning and parameter learning procedures are applied to train the neural network. The input data is constituted of historic weather sensitive parametric quantities i.e. temperature, humidity, wind speed, hour of the day, day type (weekday, weekend, holiday), month of the year and hourly load demand data. For structure learning, a relative study on the multi-layer feed forward networks and recurrent networks has been executed. The performance of the network configurations is judged based on the mean square error and training time. For the optimally chosen network, parameter learning is carried out using supervised learning and the results attained are reported. Unit commitment is carried out on all thermal units. Conventional forward approach dynamic programming technique is implemented on different test cases with forecasted load demand to generate optimal solutions. The integration of neural network based load forecasting with unit commitment scheduling is carried out with an objective to improve the quality of solution of unit commitment generation.