Artificial intelligence modeling for power system planning

Abstract This paper presents artificial neural network (ANN) models for prediction of energy consumption compared with autoregressive integrated moving average (ARIMA) modeland adaptive neuro fuzzy inference system (ANFIS) models for thermal power plant production, hydro power plant production as well as CO₂ emission.These models were developed to streamline power systems planning and minimize errors in planning. The model's accurate functionality is ensured by incorporating the collective expertise of power system planning professionals accumulated over the years. By adhering to these calculations, human errors are eliminated, the benefits of human knowledge are maximized, and a precise framework is established for planning the functionality of large power systems. ANN model for prediction were compared with autoregressive integrated moving average (ARIMA) model. Prediction of power plants production is an effective way to organize power systems for future time intervals so there will be less problems in the growing complexity of power systems. These are problems with a great number of data whose connections are not easy to define. In this paper the solution is presented in the form of ANNas part of supervised machine learning (SML) and ANFISmethodology. Calculations with ANN model was done simultaneously with standard forecasting ARIMA model in order to facilitate a comparison. Databases that are used for training ANN, ANFIS and ARIMA models contain hours of yearly data from Serbian power systems that are related to production of different energy sources, energy consumption and parameters that influence them. Trained ANN models are used to predict energy consumption and ANFIS models are used for prediction of energy production from different energy sources and CO₂ emission on yearly or daily basis. Considering the existence of great numbers of data from previous years from which experts made conclusions for production scheduling this data is used to train ANFIS models that will help experts in optimal production scheduling. Using models like these human errors would be minimized and planning will be optimized and improved for experts.