Optimal sizing of microgrids: A comparison of herd-behaviour-oriented meta-heuristics

The strategic long-term planning and design optimisation of renewable and sustainable energy systems, and particularly microgrids (MGs), is essential for the most effective use of limited resources, especially as the deployment of localised, distributed energy generation systems increases. The main objective of such integrated resource planning exercises is to minimise total discounted system costs, whilst adhering to a set of interlinked technical constraints, including reliability. This results in a non-deterministic polynomial time-hard (NP-hard) problem, for which no polynomial time solution exists. This has therefore brought to light the importance of utilising meta-heuristic optimisation algorithms.

Meta-heuristics are higher-level general strategies inspired by natural phenomena, which can be adapted to yield a near-globally-optimum solution to NP-hard problems by iteratively improving the position of candidate solutions under a pre-defined measure of quality or time. Given the approximate nature of meta-heuristics, they have been found to have different efficiencies in different applications due to the fundamental differences in the form of the underlying objective functions – and the nonlinearities and non-convexities involved. Accordingly, testing the efficiency of new meta-heuristics in different areas is an active research area.

In this context, a review of the MG sizing literature has identified that the performance of a number of state-of-the-art, herd-behaviour-oriented meta-heuristics has not yet been addressed, namely the wild horse optimiser (WHO), the artificial hummingbird algorithm (AHA), the artificial gorilla troops optimiser (AGTO), the marine predator algorithm (MPA), the equilibrium optimiser (EO), and the moth-flame optimisation algorithm (MFOA). In response, this study carried out a systematic performance comparison of the above-mentioned algorithms by benchmarking them against the well-established meta-heuristic in the literature, namely the particle swarm optimisation (PSO). To this end, two stand-alone battery-supported MGs were modelled, which provide an efficient solution for the electrification of personal passenger and utility fleets, in addition to serving residential and commercial loads. The first MG integrates solar photovoltaic (PV) and wind resources, while the second MG is solely driven by solar PV panels – both backed by battery storage. Moreover, to effectively coordinate the charge scheduling of integrated electric vehicles (EVs) – for improved cost solutions – specific rule-based dispatch strategies were developed. The conceptual MGs were then populated for three communities residing on Aotea–Great Barrier Island, in Aotearoa–New Zealand, who currently suffer from the unreliability of privately purchased, smaller-scale renewable energy systems.

The comparative summary-statistics-based results obtained from the application of the proposed method, parametrised for the two MG configurations of interest, to the three community cases, reveal the important role of newly advanced meta-heuristics in optimising a statistically robust, minimum cost solution to the associated MG asset allocation problem. Comprehensive capital budgeting, cash flow, and energy flow analyses, as well as various univariate sensitivity analyses, have, furthermore, verified the validity and effectiveness of the proposed optimised systems in providing an integrated, reliable, affordable, clean, secure platform for serving residential, commercial, and EV-charging loads in remote and island communities.