Techno-financial evaluation of a hybrid renewable solution for supplying the predicted power outages by machine learning methods in rural areas

Abstract

From a worldwide perspective, increasing grid reliability by applying renewable energies is one of the most affordable and environmentally-friendly options available for governments. In this study, the possible numbers of future rural power outages in Iran were predicted using machine learning methods based on nine different prediction models such as PLS Regression, Basian Ridge, and LARS Lasso. Then, the predicted blackouts were imported into the grid tool of HOMER software to analyze a common grid-connected HRES that includes PV, biodiesel generator, and battery bank under three power outage scenarios, including peak-time outages, planned outages, and random outages. The analyzed results until 2040 showed that using renewable sources to electrify possible annual blackouts under a day-hours planned power outage scheme can be an affordable solution with energy costs ranging from 0.066 to 0.070 $/kWh. With less than a 5% increase over the national energy tariff, more than 15% of the supplied electricity was renewable and without any blackouts. Furthermore, the sensitivity analysis showed that the government must keep the power outage's mean repair time to less than 2 h to ensure the effective use of renewables and solve the increasing rural blackouts problem in the hot months of the year.