Forecasting of Load and Solar PV Power to Assess Demand Response Potential

Abstract

Estimating demand response potential is important for the power utility in planning and management of the energy resources and power infrastructure. In this work, we have developed a machine learning model using a long short-term memory neural network for multivariate time series load forecasting using Python. The response of the load forecasting model with the different network configuration parameters is also analyzed in order to improve the accuracy of the model. Real-time electrical load data of the 11 kV feeder from one of the substations of the Goa state electricity board is used for training and developing the short-term forecasting model. The result yields model capable of accurate electrical load forecasting. Energy forecasting of SPV system of 100 kWp capacity is also done for a similar period by using PV* SOL software. The data of the scheduled power reserved for the 11 kV feeder feeding an industrial area is analyzed. From the forecasted load, SPV energy prediction, and the scheduled power data the demand response potential is estimated. This work will help the state power utility to plan and coordinate demand response programs along with scheduling renewable energy resources for maintaining the reliability and security of the power system network.