Optimisation study of geographical allocation of solar PV generation capacity for grid support

[摘要] ENGLISH ABSTRACT: Due to growing concerns regarding environmental impacts and energy security, the renewableenergy sector and its role in large-scale power generation is becoming increasingly signi cant, both locally and internationally. While it reduces reliance on fossil fuels, however,grid-integration of renewable energy generation also introduces considerable risks at high penetrationlevels due to the intermittent nature of sources such as photovoltaic (PV) solar andwind energy. The development of methodologies for addressing these concerns has become aprominenteld of research, which includes aspects such as smart-grid technologies and optimalsiting and sizing of renewable energy plants with regard to the grid infrastructure and demandpro le. This project investigates one such aspect, namely optimisation of the allocation of solarPV generation capacity in the context of grid-related metrics such as the seasonal and dailypeak demand periods and energy supply variability. The main research objectives focus ondetermining the impacts of optimal geographical distribution of solar PV sources in the SouthAfrican context, using grid support metrics de ned in terms of diurnal and seasonal trends asperformance criteria.The development and implementation of the optimisation strategy required the identi cationand speci cation of appropriate optimisation parameters, such as sets of potential locationsthat reect the seasonal and diurnal diversity of the local solar resource, a comprehensive rangeof objective functions that is representative potential grid support metrics and a range of optimisationalgorithms. Three groups of locations, referred to as site groups, were selected basedon geographical signi cance with regard to diurnal and seasonal solar cycles. A set of optimisationobjectives were de ned based on grid-support considerations such as maximisingaverage energy delivery, prioritising delivery during peak demand periods and minimising theday-to-day variability of the solar PV renewable energy contribution. A selection of candidatemetaheuristic optimisation algorithms were identi ed for evaluation, namely the genetic algorithm,two standard variations of the pattern search technique and an additional pattern searchvariation that incorporates a genetic algorithm for a hybrid approach. The various site groups,diurnal and seasonal speci cations, optimisation objectives and optimisation algorithms wereconsolidated to de ne a comprehensive set of optimisation problem cases.The evaluation and subsequent analysis of the problem cases was implemented via an integratedsoftware platform that forms part of an ongoing software application developmentproject. Based on the designated optimisation strategy, a solar PV optimisation module wasdeveloped and integrated with an established database-driven user interface. The corresponding relational database structure was optimised for all current and future applications. Thesimulation software for implementing the optimisation problem cases on an external simulationplatform was developed and amalgamated with the solar PV optimisation module.The results of the optimisation study confirm the seasonal and diurnal significance of thegeographical distribution of solar PV generation, with seasonal variation occurring along a north-south axis and diurnal variation occurring along an east-west axis. The results for thesame optimisation objectives evaluated for different seasonal and diurnal periods often showdisparities. This indicates that solar PV distributions that exhibit the best performance characteristicsoverall are not necessarily ideal for supporting peak demand periods. This supports the notion that variable feed-in tariffs rather than the commonly used at tariffis should apply to renewable energy generation, since this would encourage development that supports thegrid without sacrificing plant profitability. The results for the different optimisation objectivesalso show a clear trade-off between maximising the annual cumulative energy yield and minimising variability of supply. In general, the daily variability throughout a seasonal period decreases as geographical dispersion increases, often at the cost of lowering the cumulativeannual energy yield. With regard to the optimisation algorithms investigated, the technique combining pattern search and the genetic algorithm proved to be the most robust. Due to its non-deterministic nature, however, the algorithm generally necessitated multiple evaluations toensure high quality solutions.The quantitative impacts of the results achieved in the study are, to a degree, limited due to the geographical parameters of the South African case study considered in the investigation.The proposed optimisation strategy, however, has excellent potential in the context of largerinterconnected grids, such as the European grid, United States mainland and the Southern African power pool region where an increased range of diurnal and seasonal characteristics applies. With regard to future work, it is recommended that a similar optimisation strategyshould be investigated for combined wind and solar PV generation, since the contrastingcharacteristics of these sources could produce much more optimal aggregated power profiles.

[发布机构] Stellenbosch University