Abstract

Short-term load forecasting plays a major role in the operation of electric power systems to ensure instantaneous balance between electricity generation and demand. The accuracy of the forecast generated by a neural network (NN) has several factors, including but not limited to, the algorithm used to train the network, how much and what kind of data are used in the network's training set, how many hidden layers are in the NN, and the size of the hidden layer(s). We investigate the best combination of these factors to decrease the mean absolute percent error (MAPE) and to give the best forecast possible. Based on system load data from the Electric Reliability Council of Texas (ERCOT), this paper focuses on a comprehensive understanding of the accuracy of forecasts generated by NN's with different algorithms, while varying the length of the network's training set, hidden layer size, number of hidden layers, and the addition of data sets when training to create a forecast with the highest accuracy possible.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call