Enhanced estimation and short-term forecasting of output power of a solar photovoltaic plant using adaptive neuro-fuzzy inference system (ANFIS)

Abstract Estimation and forecasting the power of the PV system is very important to the stability of the grid and the management of the energy. The most accurate and least computational methods are those that researchers aspire to develop for predicting the power. In this paper, the adaptive fuzzy inference system (ANFIS) is proposed for the estimation of the output power of solar PV system. The proposed system is composed by using the temperature of PV system surface and the solar radiation as its inputs and the output power is the output. The ANFIS is architectured by using two and three membership functions per input (ANFIS-2MF and ANFIS-3MF). Dataset is collected from a real solar PV system in Egypt. The collected data is splitted into two parts; large part for the ANFIS training and the small part for the ANFIS testing and verification. The training of the proposed ANFIS systems is carried out using the grid partitioning algorithm. The results obtained from training and testing stages prove that the proposed ANFIS systems are well-established methods for the estimation of the solar PV power. The results from training ANFIS-3MF (mean squared error = 0.014803) is better compared with the corresponding ones by ANFIS-2MF (mean squared error = 0.031425), whereas in the testing stage which is the most important stage the results by ANFIS-2MF (mean squared error = 0.59352) is the best. After that, the trained ANFIS-2MF which is the most preferable is used to conduct short-term forecasting of the solar PV power. In this case, the trained ANFIS-2MF is used to forecast power for the next two days. The results from this stage show the efficiency of the proposed ANFIS-2MF (mean absoluter error = 0.3980). Comparison with other related works is presented at the end of this study to show the high accuracy of the proposed ANFIS systems.