DISAGGREGATED ACTIVE AND REACTIVE DEMAND FORECASTING USING FIRST DIFFERENCE MEASURED DATA AND NEURAL NETWORKS

This work presents a complete methodology for disaggregated substation peak active and noncompensated reactive power forecasting. The method requires measured active and reactive power from substation transformers, billing data divided by consumer type and global energy forecasting by consumer type. The peak active power forecasting is based on the cointegration between this variable and the energy associated with the substation. The non-compensated peak reactive power is based on artificial neural networks and uses the first difference from the measured data to filter atypical events such as capacitor banks switching. The forecast of peak power of substations permits to support the planning of new investments on grid expansion, reactive support and energy purchasing.