FORECASTING ACTIVE AND REACTIVE POWER AT A SUBSTATION TRANSFORMER IN DISTRIBUTION NETWORK

FORECASTING ACTIVE AND REACTIVE POWER AT A SUBSTATION TRANSFORMER IN DISTRIBUTION NETWORK

ABSTRACT

This work addressed the problem of forecasting active and reactive power at a substation transformer in a distribution system. Accurate power forecast is of great importance in power distribution planning, reactive power support control and intelligent power management. Due to the complexity of the power system, an intelligent and adaptive forecast algorithm based on the Adaptive Neuro-fuzzy Inference System (ANFIS) was modeled for the power forecast. For the proposed ANFIS forecast model training and validation, historical data of active and reactive power from the Abakpa Enugu Nigeria distribution network was used. The case study power system is modeled in MATLAB SIMULINK with the proposed neuro-fuzzy forecast model integrated. Simulation is carried out to obtain the time series of one hour ahead and three hour ahead forecast of the active and reactive power. Graphical output shows that the forecasted active and reactive power time series follow the signal profile of the actual (measured) system active and reactive power. The evaluation of coefficient of multiple determination was used to determine the accuracy of the forecast model. Result evaluation carried out determined the coefficient of determination to be 0.98 and 0.72 for the one hour ahead and the three hour ahead active power forecast respectively. Similarly, the one hour ahead and three hour ahead reactive power forecast gave 0.82 and 0.71 respectively. For the one year ahead (long term) forecast obtained, the coefficients of multiple determination are 0.54 and 0.62 for active and reactive power respectively. The results indicate very strong degree of correlation between the actual power time series and the forecasted time series. However these values show that the near real-time forecast of one hour ahead and three hour ahead, are more accurate than the long term forecast. This shows the high degree of accuracy of the proposed neuro-fuzzy forecast model.