Peer-to-peer energy trading in a distribution network considering the impact of short-term load forecasting

Demand forecasting and electricity trading are two major fields of study for power system operators. These two issues have not been integrated by scholars for analysing performance of power system operator. The main motivation behind this manuscript is to integrate load forecasting with energy trading. Emergence of robust infrastructure in decentralized energy trading among power system operators and development in recent trends in artificial intelligence techniques to forecast load has helped to rethink the strategies behind operation of power systems. In this manuscript, an objective for peer-to-peer (P2P) energy trading is designed in a distribution network with integrated short-term load forecasting (STLF). The solution for the designed objective is proposed using mid-market rate (MMR) method for P2P trading with stochastic integrated STLF. Stochastic methods used for STLF are tree, linear regression, support vector machine, ensemble, Gaussian process regression and neural network. After comparative analyses, the best method is found, i.e. tree, and it is further optimized using three optimization techniques, viz. grid search, random search and Bayesian optimization, with an objective of reducing mean square error (MSE) for hyperparameters tuning. Among the three optimization methods, grid search has a reduced MSE of 0.0066 and leaf size of 3. The hyperparameter-tuned STLF is used for P2P energy trading. The energy trading price is decided using MMR method, and it optimized the total energy cost. The whole algorithm is implemented in Python and MATLAB platforms. It is observed that total energy cost in P2P trading is reduced by 36.114% with STLF which validates the proposed algorithm.