Power System Load Forecasting Using Machine Learning Algorithms: Optimal Approach

Accurate forecasting of long-term electricity demand has a prominent role in demand side management and electricity system planning and operation. Demand over-estimation gives rise to over-investment in system assets, driving up the electric power prices, while demand under-estimation may give on to under-investment results in unreliable and insecure electricity. The electrical load on a station varies more dynamically and hence forecasting of load values is more significant. In long-term load forecasting (LF), economic factors, weather conditions, time factors and random effects plays a prominent role. Machine Learning (ML) algorithms often used to estimate the future values constructed on the inferences drawn from the existing values. For these algorithms, the input data is supplied for training and as well testing the algorithm efficacy. Now a days, large amount of data is available everywhere. Therefore, it is essential to analyze this data in order to extract some profitable information and to implement an algorithm based on this analysis. These algorithms works based on historical relationship between the various factors considered in the problem. In this paper, some popular ML algorithms such as Linear develop the Forecasting model for standard 14 and 30 bus systems. The efficiency of the forecasting procedure is examined based on RMSE, Accuracy and time taken to generate the model. Further, the optimization results encourages for implementing these algorithms for real time operations in order to minimize number of loss violation instances in a given system