Electric Fleets with On-site Renewable Energy Sources (EFORES): Data-driven Dynamic Dispatching and Charging under Uncertainties

点击翻译按钮获取中文摘要

In line with the UK's target to reach net zero by 2050, Electrical Vehicles (EV) charged by renewable energy are one of the solutions towards carbon-neutral road transport, which is the 2nd largest carbon emission both nationally in the UK and locally in Newcastle city (it contributed about 33% of total emission in 2020). The electrification of business fleets (either commercial or for public service) has recently emerged as one the key factors in reducing transportation related CO2 emissions. However, according to the Global Covenant of Mayors's (GCoM) guidance the electrification of fleets leads to the reduction of direction emission, it does not imply reduction of overall emission nationally or globally if the electricity charged for EV is still sourced from the fossil fuels (see also the NU et al.'s recent policy report: https://www.seev4-city.eu/wp-content/uploads/2020/09/SEEV4-City-Policy-Recommendations-and-Roadmap-1.pdf).A recent trend in Renewable Energy Sources (RES) is an increasing amount of small-scale RES installed on-site , referred to as ORES. For instance, in March 2021, Newcastle City Council announced a £27M plan to install solar panels, energy storage etc. at schools, leisure centres, cultural venues, depots and offices to decarbonise public buildings and transport. Likewise, Gateshead Council has approved in Nov. 2020 plans to develop two significant-scale urban solar farms, and furthermore installing solar PV canopies above car parking bays in sites like Gateshead Civic Centre, and furthermore are including rooftop solar PV on new developments such as the Gateshead Quay Arena and the proposed Gateshead Quays multi-storey carpark (construction of both commencing in 2021). This provides good opportunities for EV to use more on-site generation renewable electricity to actually reduce the overall emission for road transport. The key issue is the efficient use of ORES.Using battery as a electricity storage can alleviate this, but at significant investment and operation cost. V2G is proposed to reduce static battery storage, but causes battery degradation. And smart charging is needed to avoid or reduce the operation cost of battery degradation. Most existing EV smart charging studies focus on the EV charging only to reduce charging cost and/or peak-shaving, under the assumption of EVs' electracy demand are given and non-adjustable (either constant or statistical model, e.g. Poisson distribution). This is reasonable for non-collaborative individual EVs. However, for a electric fleet (EF) consisting of collaborative EVs, in addition to the optimal EV charging, the electricity demand can be optimized by EF dispatching, i.e. adjusting EF's travel plan by assigning the right EV to the right service to maintain the right state of charge of the battery, and allocating to the right charging station at a right time window, such that a better marginal benefits can be achieved in terms of better efficiency and utilization of on-site renewable energy. However, the power generation of ORES is highly variable - resulting in an undesired fluctuation at the supply side. On the demand side, EVs' charging demand also comes with uncertainties, to meet various tasks with dynamic travelling and charging demands. In shifting EV energy from less variable fossil electricity (imported from the grid) to high variable on-site ORES, the main challenge is the charging strategy of maximizing self-consumption of own ORES under uncertainties, whilst meeting the variable EV demands, at minimized cost in energy storage and less impact on grid's peak load. This project is to investigate the possibility to intelligently integrate the dynamic charging demand of electric fleets with the high variable on-site renewable energy by developing a data-driven reinforcement learning (RL) decision support tool.