This paper presents a quantitative, multi-method approach to tackle power system reliability problems affecting sustainable energy supply. Through case studies and analysis, it underscores the urgency for a shift in asset management paradigms in the developing world where cost of un-served energy may account for up to 27.7% of energy sales. It applies systems thinking philosophy to analyse root causes of problems, then uses statistical and probabilistic inferences to solve them. Failure statistics for transmission and distribution transformers are processed using least squares method, maximum likelihood estimation and method of moments in order to determine Weibull life modelling parameters. Modelling of reliability functions and costs associated with planned and run-to-failure strategies were implemented using MATLAB algorithms based on the parameter estimates. Results show that the runto-failure strategy achieves short term cost savings, but it eventually leads to unsustainable energy supply. They further show how problems cascade down from the generation to the distribution business units. The study advances use of Loss of Energy Expectation and Loss of Load Expected as metrics to augment traditional use of availability in order to measure the loss margin. The model developed can be used in risk management, in power asset management planning and in developing strategies on industrial and commercial use of energy in power utilities.