Cost Optimization of a Hybrid Off-Grid Power System for Remote Localities: A Statistical Model

Abstract

Renewable energy sources like solar radiation, wind energy, geothermal energy, bioenergy are known to well integrate with off-grid stand-alone power systems. The choice of a renewable energy source depends on the geographical location and climatic condition of the region, available resources, and the economics of the power system. Thus, the evaluation and optimization of hybrid power systems in terms of cost of energy (CoE) with various alternative energy sources considering the demand of the location and available resources is highly pertinent for configuring a hybrid power system for remote localities. The present work describes a case study on development of a response surface (RS) model for predicting the optimum CoE of a hybrid off-grid power system (HOPS). RS model is developed using Box-Behnken experimental design (BBD) technique. Athree-factor three-level BBD is used to describe the optimum CoE. The three process variables under consideration in BBD model are size of photovoltaic arrays (PV), diesel generator (DG) capacity, and number of battery (BAT) cells. The analysis of variance (ANOVA) is conducted on the response (CoE) for each factor level settings of a three-factor three-level BBD in order to evaluate a full quadratic factor space. The results of ANOVA established significant linear, quadratic, and interaction terms in the RS model representing the factor space. The coefficients of the RS-model are determined using multiple regression analysis technique at 95 % level of confidence. The RS-model is checked for error in prediction by residual analysis technique and validated against experimental data to confirm the accuracy of the model. The results confirmed that the model has 97.5 % accuracy in prediction of CoE for a HOPS. The RS-model is also allowed to identify the optimal configuration of hybrid power system for minimum CoE. A minimum CoE predicted by the model is comparable with that reported in the literature fromearlier studies. The model presented in this study can be a useful tool for cost comparison among similar architectures of varying capacities for HOPS.