| Popis: |
Development of Battery Energy Storage System projects and their subsequent installation and connection to electrical grids throughout the world is expected to increase over the coming years. Two key concerns are at the forefront of entities undertaking these installations. The first is determining the optimal BESS size for a given application. The second is whether this optimal BESS size reflects the goals (referred to as planning objectives in this dissertation) set out in the BESS project planning phase. Recognising these two concerns, it is determined that BESS sizing approaches must be fit for purpose, can be used adequality as a planning tool and capable of modelling important planning objectives. The Front-End Planning framework was utilised in this dissertation as a means to assess if existing BESS sizing approaches are suitable for modelling planning objectives as part of BESS project planning. In total, 32 of the most-cited articles from the BESS sizing literature were reviewed for their inclusiveness of scoping elements set out by the Front-End Planning framework. The results of this review showed that existing BESS sizing approaches are lacking in three key planning objectives called Investment Scale, Investment Timing and Dispatch Adaptability. This research sought to answer the following questions: 1) Is it possible to form the planning objectives Investment Scale, Investment Timing and Dispatch Adaptability as part of optimising energy capacity size for new BESS installations seeking maximum profit? 2) Are there any circumstances where the inclusion of the three planning objectives as part of BESS sizing helps overcome shortcomings of existing sizing approaches? To incorporate the planning objective Investment Scale as part of BESS sizing, maximisation of opposing financial objective functions using two different multi-objective optimisation methods called Rating Method and Paired Comparison was used. These approaches were tested on a simple microgrid under various electricity price scenarios. The results show that the Rating Method performed best when selecting BESS size in significant knee regions near maximum daily worth. The Rating Method can also select optimal BESS size at maximum daily worth when less-significant knee regions are present. This approach gives an appropriate balance between forming the planning objective Investment Scale and maximising profit. To incorporate the planning objective Investment Timing as part of BESS sizing, two different models were used, referred to as the operational model (controlling operational decisions i.e. BESS dispatch) and the planning model (controlling BESS size at different yearly intervals). Reinforcement learning was used as the operational model solution method, while global optimisation was used as the solution method for planning model. This approach was tested on data from the Integrated Single Electricity Market Day-Ahead Market. It was found that splitting BESS operational decisions and BESS planning decisions into two different models is an effective technique. To incorporate the planning objective Dispatch Adaptability as part of BESS sizing, model-based and model-free stochastic optimisation methods are used. This was done for model-free optimisation by utilising deep reinforcement learning methods, while stochastic programming was used as the solution method for model-based approach. Both approaches were tested on historical Day-Ahead and Intraday Markets electricity clearing prices from the Integrated Single Electricity Market. It was found that the model-based approach outperformed the model-free approach. However, it is not clear that such a broad statement can be made about model-free and model-based approaches in general based on the results gained through this thesis. The significance of this study’s results is that BESS sizing is now more functional and adaptable for project planning purposes. It is now possible to size BESS without suffering scale issues resulting from ever-diminishing returns of larger BESS sizes, where the timing of the investment can be chosen optimally rather than assuming “here and now” investment, and where the operational strategy employed to simulate BESS dispatch is more reflective of actual BESS use and adaptability. |
