Analysis and Design of Demand Side Management Model in Smart Grids

Abstract

Electricity load and price prediction (ELPP) plays a newlinesignificant role in demand side management that helps customers modify their power newlineconsumption based on the prediction. A short term load and price prediction model electricity newlineload and price prediction model using mutli layer cascaded feed forward neural network is newlinedesigned and simulated. The performance parameters like prediction accuracy, Mean Absolute newlinePercentage error (MAPE), Mean Square error (MSE) of the proposed system are evaluated and newlinecompared with similar existing works with improvement in design constraints. Peak load newlinemanagement (PLM) of residential buildings is proposed in this work by adopting two methods of newlineDemand side management like peak clipping and load shifting. A controller is developed using newlinemultilayer cascaded feed forward neural network that is trained to generate control signals that newline newlineix newline newlinecontrols the loads of residential buildings so as to reduce peak demand and relieve the burden on newlinethe utilities to meet the demand with limited supply. The proposed MLCFNN is simulated and newlinethe performance metrics like peak to average ratio (PAR), billing cost of the consumers and newlinediscomfort of the customers adopting DSM are evaluated and compared with the existing newlinesystems. The proposed controller is integrated with green energy (solar) to manage the peak newlinepower demand in conjunction with the controller; this method highlights the amalgamation of newlinegreen energy into the grid, which is the future of smart grids in India. newlineThe electricity load and price prediction model and the peak load management are modeled and newlinesimulated using latest version of Matlab and Simulink R2021a.The MLCFNN model for load newlineprediction utilised 5 cascaded hidden layers with one input node and 100 output nodes and newlinemultilayer FFNN model for tariff prediction utilised 5 hidden layers with one input node and 1 newlineoutput node. The MLCFNN controller model for peak load management utilised 6 cascaded newlinehidden layers with single input and 7 output nodes. Green energy scheduling for peak load newlinemanagement was proposed using simulator model of solar panel available in Matlab that consists newlineof 7 series connected module with 88 parallel strings that could manage the peak demand. newlineThe performance evaluation of ELPP model and PLM model is analysed using various newlineperformance parameters. A prediction accuracy of 99.85% with MAPE of lt1% and MSE of newline0.7297 was achieved through the ELPP model. An average peak demand reduction of 1.25kW, newlinePAR reduction of 6.5%, and a bill cost reduction of 2.5 % was achieved with the peak clipping newlineDSM controller and average peak demand reduction of 2.9kW, PAR reduction of 11.35% was newlineachieved using load shifting DSM controller. The integrated green energy scheduling peak load newlinemanagement model could achieve a PAR reduction of 2.2%. newline