Data-driven sizing and control of energy storage for wind-powered offshore platforms – Energy Management and Control of Offshore Platforms Integrating Renewable Energy

Abstract

One of the key aspects of facing the global environmental crisis and achieving sustainable development for future societies is re-thinking the way energy is produced and managed to serve the demand. This perspective calls for new technologies and methods for shaping how future power systems will operate for all residential, commercial, and industrial sectors. However, such efforts would not be complete without considering the particular case of isolated power systems, where additional challenges emerge. Those become even more critical for isolated industrial power systems where the restrictive environmental policies and regulations to be adopted may not be favorable to strict case-specific technical requirements. A characteristic example of this comes from the Oil & Gas (O&G) industry, where the next-generation isolated offshore platforms will integrate renewable energy, profoundly affecting the way those should be operated. This thesis considers the significant environmental impact of offshore O&G operations, and explores various concepts toward decarbonizing this sector and dealing with the open technical challenges associated with the integration of renewable sources in a cost-effective way. The ideas presented throughout the thesis target this goal and are framed into a hierarchical structure where the use of energy storage is investigated as a potential solution from various perspectives and considering different time scales. The first idea being presented is an optimal techno-economical analysis of energy storage that considers the uncertainty arising from both loading conditions and renewable (wind) power generation. This analysis enables decision makers to identify proper storage sizing and system configurations that maximize wind penetration and minimize fuel consumption. Nevertheless, modeling the uncertainty when considering the operation of such systems is not trivial and this can heavily impact the sizing results. Then, a new method is developed to capture the effects of combined uncertainties more accurately by generalizing the underlying patterns from the available datasets using statistical learning. This allows more accurate estimation of the potential techno-economical benefits created by the energy storage integration. We then note that the above analyses refer to the daily operation of an O&G platform using historical information for longer reference periods (i.e., years) and even though this is useful for the investment decision process, this does not capture the effects of real-time operation. To address that, in the following chapters, this thesis revisits the problem of optimal operation under uncertainty fromthe real-time dispatch time scale (sub-hourly). One of the major challenges related to this time scale is the presence of sudden variations and non-smooth transitions in both load and wind power generation. Those further complicate the problem of deciding the optimal commitment and dispatch of the conventional power generating units, in the presence of additional degrees of freedom that come from the energy storage, which acts as a connecting rod across time. This thesis thus presents a method devised for that purpose that exploits available historical data and machine learning to better quantify such non-anticipated events, and that leads to better energy management decisions while considering several objectives (environmental, economical, and technical) simultaneously. The potential of this energy management algorithm is demonstrated through case studies where fuel consumption, operational costs, and switching of the conventional units are reduced compared to the corresponding deterministic state-of-the-art-method. The challenges of renewable integration for isolated systems are, however, distributed in various time scales, where the optimal decisions taken on one time scale may affect the system in another (shorter) one. One of the most notorious problems in isolated power systems is that of real-time active power balancing and frequency regulation. This becomes even more important for the case of isolated offshore O&G platforms that integrate renewable energy, and this not only because of the reduced system inertia but also because of the sudden yet large scale active power variations resulting from various operations. Narrowing to such time scales and further deeper in this thesis, new methods are developed that leverage historical data with probabilistic machine learning and risk to optimally coordinate the use of conventional units and energy storage to regulate the grid frequency while simultaneously tracking a techno-economical optimal operation under uncertain conditions. Such methods are focused on finding optimal control laws and allocating optimal storage capacity and primary reserves, providing frequency support with bounded and pre-defined deviation from the system’s optimal operating point, while ensuring frequency stability. In this way, decisions satisfying objectives at higher time scales do not heavily interfere with technical requirements from lower ones, alleviating the impact of optimal energy management decisions, under user-defined risk acceptance and corresponding theoretical guarantees, paving the way toward fully autonomous and sustainable O&G platforms.