Assessment of two optimisation methods for renewable energy capacity expansion planning

When deciding on a country’s future energy policy, it is essential to accurately estimate the future generation mix and technology costs, as well as future generation site locations in electricity power grids. This estimate needs to allow for changing levels of demand and increasing levels of renewable energy supply, with both having high fluctuations geographically and temporally. This objective leads to high-dimensional mathematical models with high computational complexity that cannot be solved analytically. Various simplifications and heuristics are proposed globally, however, these need to be anchored in terms of an indication of their likely performance. We offer a novel parallel investigation of a near-optimal and heuristic optimisation approach on a country scale. The Linear Programming (LP) optimisation problem finds the global minimum for the chosen set of input variables, this however comes at the cost of limitations regarding model size due to complexity. The alternative heuristic model limits connectivity of model parameters but rather than aggregating variables over space or time, it maintains a high resolution to allow for more granular estimates. We compare the quality and performance of a heuristic and LP optimisation method, using two configurations for the example of Australia and find relevant solutions for both. The solution space requires substantial simplification for the LP to be solvable and leads to overly optimistic capacity factors, installed capacity and cost, due to spatial aggregation. The heuristic approach has a significant performance advantage requiring only 3% of the near-optimal runtime and a fraction of calculation iterations.