A novel approach for performance prediction analysis of big data application

Abstract

newline Smart grid technology and energy storage systems are fascinating due to considerable attention on energy crisis. A reliable and accurate electricity prediction model is a crucial factor to be considered for a suitable energy management policy. At present, electricity consumption is rapidly increasing due to the rise in human population, technological development, and modern living standards. Therefore, a two-stage methodology is established for electricity load prediction in this research. The two stages are: In the first stage, the raw data of electricity consumption are mined using incremental and progressive data mining for effective training and the second step includes a hybrid model with the integration of Artificial Neural Network (ANN) and Fuzzy Inference System (ANFIS) is developed. ANFIS, as a hybrid intelligent system, has the capability of fast and precise learning, extending its capacity, adopting data and exiting expert knowledge, and having explanation facilities in the form of semantically meaningful fuzzy rules. Due to the various limitations in the existing methodologies, it is motivated to propose a new energy forecasting method based on ANFIS to deal with encountered uncertainties in an energy management system with perfect learning and prediction capabilities. The ANFIS model is evaluated using the R-squared error, the Root Mean Square Error (RMSE), Mean Square Error (MSE), and Mean Absolute Error (MAE) metrics. Finally, our models are assessed over benchmark datasets that exhibited an extensive drop in the error rate in comparison to other techniques. The results indicated that the ANFIS model had reduced errors over the ARIMA, ANN, RBF, and SVM (i.e., RMSE (0.4076), MAE (0.9049), and MAPE (0.0373), respectively. iv Household electric energy consumption is the amount of energy consumed per unit of time. A prediction model that can handle time series data has to be developed to forecast energy consumption. In deep learning, the time-series data are dealt with recurrent neural n