Department of Electrical and Electronics Engineering, Universiti Teknologi PETRONAS, 31750 Tronoh, Perak, Malaysia

---

Abstract—Load forecasting plays a very vital role for efficient and reliable operation of the power system. Often uncertainties significantly decrease the prediction accuracy of load forecasting which affect the operational cost dramatically. In this paper, comparison of Back Propagation (BP) and Levenberg Marquardt (LM) neural network (NN) forecast model for 24 hours ahead is presented. The impact of lagged load data, calendar events and weather variables on load demand are analyzed in order to select the best forecast model inputs. The mean absolute percentage errors (MAPE), Daily peak error and regression analysis of NN training are used to measure the NN performance. The Forecast results demonstrate that, LM based forecast model outperform than BP NN model for performance matrices. This model is used to predict the load of ISO-New England grid.

Index Terms—Short Term Load Forecasting (STLF), Neural Network (NN), Back Propagation (BP), Levenberg- Marquardt (LM), Mean Absolute Percentage Error (MAPE), Regression Analysis (RA).

[PDF]

Cite: Muhammad Qamar Raza, Zuhairi Baharudin, Badar-Ul-Islam, and Perumal Nallagownden, "A Comparative Analysis of Neural Network Based Short Term Load Forecast Models for Anomalous Days Load Prediction," Journal of Computers vol. 9, no. 7, pp. 1519-1524, 2014.