Date of Award

Spring 5-2017

Embargo Period


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Engineering and Public Policy


Paulina Jaramillo


Expanding access to electricity is central to development in East Africa but massive increases in investment are required to achieve universal access. Private sector participation in electrification is essential to meeting electricity access targets. Policy makers have acknowledged that grid extension in many remote rural areas is not as cost effective as decentralized alternatives such as microgrids. Microgrid companies have been unable to scale beyond pilot projects due in part to challenges in raising capital for a business model that is perceived to be risky. This thesis aims to identify and quantify the primary sources of investment risk in microgrid utilities and study ways to mitigate these risks to make these businesses more viable. Two modeling tools have been developed to this end. The Stochastic Techno-Economic Microgrid Model (STEMM) models the technical and financial performance of microgrid utilities using uncertain and dynamic inputs to permit explicit modeling of financial risk. This model is applied in an investment risk assessment and case study in Rwanda. Key findings suggest that the most important drivers of risk are fuel prices, foreign exchange rates, demand for electricity, and price elasticity of demand for electricity. The relative importance of these factors is technology dependent with demand uncertainty figuring stronger for solar and high solar penetration hybrid systems and fuel prices driving risk in diesel power and low solar penetration hybrid systems. Considering uncertainty in system sizing presents a tradeoff whereby a decrease in expected equity return decreases downside risk. High solar penetration systems are also found to be more attractive to lenders. The second modeling tool leverages electricity consumption and demographic data from four microgrids in Tanzania to forecast demand for electricity in newly electrified communities. Using statistical learning techniques, improvements in prediction performance was achieved over the historical mean baseline. I have also identified important predictors in estimating electricity consumption of newly connected customers. These include tariff structures and prices, preconnection sources of electricity and lighting, levels of spending on electricity services and airtime, and pre-connection appliance ownership. Prior exposure to electricity, disposable income, and price are dominant factors in estimating demand.