Performance Evaluation of distributed Solar PV Installations

Sammanfattning: Utilization of photovoltaic (PV) cells to generate electricity from solar energy is becoming an increasingly popular source of renewable energy in this era of energy transitions. Adoption of ingenious techniques to increase PV systems’ performance and with the arrival of data analytics for efficient management of energy systems, an opportunity of using statistical models to develop mechanisms for performance evaluation of distributed Solar PV systems is present. Therefore, making use of available data and various statistical techniques, two models for performance evaluation of solar installations were developed. First model estimated PV Power outputs using neighboring PV panels whereas the second model estimated Global Horizontal Irradiance (GHI) using AC PV power outputs. The aim of these modules was to serve as the basis for fault detection and solar forecasting, respectively. With solar forecasting information available insight of future energy production and assistance in smart grid solutions could be carried out. Furthermore, with anomaly detection mechanism in place one can highlight energy reductions in the systems. Sensitivity analysis for PV nowcasting methodology was carried out to optimize characteristics of the model. Increase in the number of neighbors did not have any significant effect, whereas large radius was required for clear sky days and shorter radius were needed for cloudy conditions to cater the rapid change in weather conditions. Overall, nowcasting methodology resulted in Mean Absolute Percentage Error (MAPE) of less than 5%. GHI Estimation model was benchmarked with Nespoli, et al. (2017) method and compared with satellite data also. Results for GHI Estimation Model were comparable to Nespoli et al. (2017) method and better than satellite data. Overall, for test sites under the supervision of CheckWatt AB, MAPE of less than 10% was observed and the results were significantly better than SMHI STRANG estimates which had MAPE of 46%. Sensitivity Analysis of number days for estimation of GHI was carried out and use of 120 days for estimation of GHI was found to give the minimum MAPE. GHI Estimation Model was also used to generate solar map where variation in GHI of 5 sites within Stockholm county was portrayed. These two modules combined serve towards performance monitoring of PV installations.