New methods and results in the optimisation of solar power tower plants

Resumen

Renewable energy technology has seen great advances in recent decades, combined with an ever increasing interest in the literature. Solar Power Tower (SPT) plants are a form of Concentrating Solar Power (CSP) technology which continue to be developed around the world, and are formed of subsystems that are open to optimisation. This thesis is concerned with the development of new methods and results in the optimisation of SPT plants, with particular focus on operational optimisation. Chapter 1 provides background information on the energy sector, before describing the design and modelling of an SPT plant. Here, the optical theory behind the transfer of incident radiation in the system is developed and the relevant equations presented. In Chapter 2, the cleaning operations of the heliostat field are optimised for a fixed schedule length using Binary Integer Linear Programming (BILP). Problem dimensionality is addressed by a clustering algorithm, before an initial solution is found for the allocation problem. Finally, a novel local search heuristic is presented that treats the so-called route ``attractiveness'' through the use of a sequential pair-wise optimisation procedure that minimises a weighted attractiveness measure whilst penalising for overall energy loss. Chapters 3-6 investigate the aiming strategy utilised by the heliostat field when considering a desired flux distribution profile and operational constraints. In Chapter 3, a BILP model was developed, where a pre-defined set of aiming points on the receiver surface was chosen. The linear objective function was constrained with linear equalities that related to distribution smoothing (to protect receiver components from abnormal flux loads) via the use of penalisation. Chapter 4 extended this model by instead considering continuous variables with no fixed grid of aiming points. This led to an optimisation problem with a non-linear, non-convex objective function, with non-linear constraints. In this case, a gradient ascent algorithm was developed, utilising a non-standard step-size selection technique. Chapter 5 further extended the aiming point optimisation topic to consider the dynamic case. In this sense, the aiming strategy across a period of time could be optimised, taking into account SPT plant technological limitations. Two algorithms were considered, Penalisation and Augmented Lagrangian, where theoretical properties for optimality and solution existence were presented. Finally Chapter 6 considered the effects of inclement weather on the optimisation model presented in Chapter 3. Stochastic processes were investigated to determine optimal aiming strategies at a fixed point in time when weather data could not be known for certain. All research presented in this thesis is illustrated using real-world data for an SPT plant, and conclusions and recommendations for future work are presented.