Search

High-quality short-term load forecasting, particularly day-ahead, is essential to enable the demand response aggregator’s participation in the electricity market. The accuracy of load forecasting depends on many factors, including the size and quality of historical data, selection of the forecasting model, availability of weather data, and types of business sectors. This paper implements three state-of-the-art regression models, ridge regression (RR), random forests (RF), and gradient boosting (GB) to capture intricate variations in three UK cities (Newcastle, Peterborough, and Sheffield) in five business sectors (retail, entertainment, social, industrial, and other) from the UK non-domestic electricity load profiles and provide accurate day-ahead load forecasting. The models are implemented on a historical dataset that contains 7527 UK businesses with geographical postal codes, 30-min electricity consumption, and weather metrics. The performance is evaluated using the coefficient of determination R-squared. The presented results show that GB outperforms RF and RR as it provides the most accurate forecasting results, with limited improvement in forecasting results by including weather data. The aggregated business sectors’ forecasting accuracy is higher than individual business sectors’ forecasts.

Public cloud providers offer ready-to-use, easily scalable servers on demand for a variety of applications. Storing and processing private and sensitive data in the cloud brings multiple security issues and indeed these concerns currently prevent many users from utilizing cloud resources. Improving both security and trust for users is increasingly important for cloud providers. In this context, we first investigate the classes of security threats encountered by cloud applications. We then analyze various software- and hardware-based solutions to handle these security challenges and provide the user with a chain of trust. Our analysis shows that pure software-based solutions do not sufficiently mitigate the challenges of the cloud environment. Hardware-based solutions utilizing the Trusted Platform Module (TPM) alleviate the issues; however, it is challenging to implement in the public cloud environment. Finally, we introduce our TPM-SGX-based approach that utilizes software TPM and Software Guard Extension (SGX) to provide similar security as the hardware-based approach.

In alignment with the Paris Agreement, the city of Oxford in the UK aims to become carbon neutral by 2040. Renewable energy help achieve this target by reducing the reliance on carbon-intensive grid electricity. This research seeks to optimally size solar photovoltaic and lithium battery storage systems, reducing Oxford's grid electricity reliance in buildings. The analysis starts with modeling the electricity demand. The model uses Elexon electricity settlement profiles, and assembles them into the demand profile according to the quantity and types of buildings in Oxford. Then, solar generation is modeled using Pfenninger and Staffell's method. Solar photovoltaic and lithium storage systems are sized using a hybridized analytical and iterative method. First, the method calculates the solar system size search range, then iterates through the range. At each solar size, the method calculates and iterates through the storage system size search range. Within each iteration, the renewable system is simulated using demand and generation data with a simplified system set-up and the conventional operation strategy. The method outputs combinations of solar system capacity, storage system capacity, and grid electricity import. Each combination's levelized cost of electricity is calculated, and the lowest cost combination is the optimal sizing. Solar and storage system costs are projected from 2019 to 2100, and the optimal sizing is calculated for each year. The result shows that solar photovoltaic is economically competitive, but lithium storage cost is still too high. As solar and storage prices continue to drop, they will take up greater portions of the energy system. However, there will always be a need for the grid, as it provides flexibility and can meet demands that are too costly for solar and storage, Comment: 10 pages, 8 figures, published in the conference of ECEEE 2022 Summer Study on energy efficiency: agents of change

Industrial decarbonisation is one of the most significant hurdles as we move towards a global net-zero CO2 emissions target. Processes requiring high temperatures and heavy-duty transport are major complications, with cost and feasibility of implementation of zero carbon methods often significant barriers, especially for medium sized or smaller businesses. This study shows that decarbonisation of business processes through electrification as well as on-site green ammonia generation, is possible and can be profitable for this scale of company. Mühlendorfer Kreidefabrik, used as a case study, is an Austrian chalk processing factory that emits approximately 4680 tonnes of CO2 per year whilst spending EUR664,000 on energy. Different decarbonisation scenarios were investigated, based on the general concepts of electrification and zero-carbon fuel (ZCF) generation. The study leveraged simulation algorithms in conjunction with past usage data of the company in order to model supply and demand in future years. 4.7 MW of solar PV panels as well as one 3 MW wind turbine are required in a pure electrification scenario. Including the replacement cost of fossil-fuel by electrified equipment and an estimated lifetime of 20 years, this system is expected to cost ~EUR512,000 annually. A larger electricity generating system of 13.7 MW solar PV panels and one 3 MW wind turbine is required in a ZCF scenario, along with a 3.9 MW alkaline electrolyser connected to an air separation unit and Haber-Bosch equipment. This system will produce ~1887 t of green ammonia per year, sufficient to replace both natural gas and Diesel currently used in the factory, requiring only adaptation of current plant and machinery instead of replacement. Costing ~EUR714,000 annually, this option is more expensive, however has further advantages in minimising interference in factory operations. This pathway seems also preferable due to the uncertainty in availability of a pure electrified solution even with the extra costs.

The evolution in mobile wireless communication generations, continues to gift the world with new telecommunication capabilities towards how people live and work. For instance, the roll out of the fifth generation (5G) promises a much-enhanced cloud-native application service, through improved communication speed, higher bandwidth, improved capacity for more connected devices and many more. These exciting new 5G features have given numerous vertical and horizontal industries a reason to explore different ways of delivering value, through emerging cloud-native applications that run on access devices (i.e. the Internet of Things (IoTs) and mobile devices). These cloud-native applications such Virtual Reality, Vehicle to everything communication (V2X), artificial intelligence, video analytics and so on however have strict performance requirements for extremely low latency (10 milliseconds and below) and higher bandwidth, that 5G alone cannot deliver. Unfortunately, the traditional cloud computing and radio access network setups do not sufficiently address these key communication needs which are crucial to the performance of these cloud-native applications. There is therefore an urgency to enhance and optimize the traditional mobile telecommunication network architecture, to meet these performance needs. This thesis seeks to demonstrate how we have developed a facility called Multi-Access Edge Computing with Cloud Radio Access Networks (MECRAN) to address this issue. MECRAN is based on the edge computing paradigm and uses machine learning to optimize the round-trip delivery of data at low latency, by dynamically scheduling cloud-native applications, to run in close proximity to a mobile user, at the edge of the radio access network.