Improving Energy Consumption Monitoring in Food and Drinks Manufacturing and Storage Systems with Predictive Machine Learning

在能源生产增长相对较慢的情况下，预计全球人口激增和对粮食的需求增加是一个重大的可持续性挑战。食品和饮料行业是联合王国最大的制造业，由于其巨大的能源消耗和温室气体排放，面临着巨大的可持续性压力。该博士学位旨在通过利用预测性机器学习技术来改善食品和饮料制造和储存系统中的能源消耗监控，以应对这些挑战。这一博士学位包括三项相互关联的研究，每项研究都侧重于食品和饮料制造和冷藏环境中的能源消耗和预测方面。基于天气数据，第一项研究旨在开发可解释的机器学习技术，用于预测电力消耗、室内温度和室内湿度。这项研究调查了当前模型在这一专业领域提出的挑战，并探索了季节性因素和人类行为在这些预测模型中的整合。准确预测用电量可以通过改进监测和规划大大有助于减少能源使用。这种优化可以帮助简化此类环境中的运营调度，促进更可持续的能源管理实践，并最终减少二氧化碳排放。此外，在这样的环境中监测温度和湿度对于确保食品安全和保存易腐烂物品的质量至关重要。在第一项研究的基础上，第二项研究旨在开发基于生产需求和天气条件的食品制造环境中电力和天然气消耗的预测模型。通过将生产计划与天气条件相结合，这些模型具有优化资源利用和降低成本的潜力。第三项研究的重点是开发食品制造环境中气体和电力消耗的实时异常检测系统，旨在最大限度地减少能源浪费，提高运营效率。通过应用科学的方法，这项研究旨在为食品和饮料行业提供有价值的见解，最终提高运营效率。这不仅将导致成本降低，还将有助于全球减缓气候变化的努力。这项研究的预期结果具有很好的潜力，为食品和饮料企业带来实实在在的好处，使它们能够做出明智的决定，并有效地将能源消耗降至最低。

The projected surge in the global population and the increasing demand for food present a significant sustainability challenge in the face of relatively slower growth in energy production. The food and drink industry, being the largest manufacturing sector in the United Kingdom, faces substantial sustainability strains due to its significant energy consumption and greenhouse gas emissions. This PhD aims to address these challenges by leveraging predictive machine learning techniques to improve energy consumption monitoring in food and drinks manufacturing and storage systems. This PhD comprises three interconnected studies, each focusing on aspects of energy consumption and prediction within the food and drinks manufacturing and cold storage environments. Based on weather data, the first study aims to develop explainable machine learning techniques for predicting electricity consumption, indoor temperature, and indoor humidity. This research investigates the challenges presented by current models in this specialised domain and explores the integration of seasonal elements and human behaviour into these predictive models. Accurate prediction of electricity consumption can greatly contribute to reducing energy usage through improved monitoring and planning. This optimisation can help streamline operational scheduling in such environments, facilitating more sustainable energy management practices and ultimately reducing CO2 emissions. Additionally, monitoring temperature and humidity in such settings is crucial to ensure food safety and preserve the quality of perishable items. Building upon the first study, the second study aims to develop predictive models for not only electricity but also gas consumption in a food manufacturing environment based on production demand and weather conditions. By integrating production schedules with weather conditions, these models have the potential to optimise resource utilisation and reduce costs. The third study focuses on developing a real-time anomaly detection system for gas and electricity consumption in food manufacturing environments, aiming to minimise energy waste and improve operational efficiency. Through an applied science approach, this research aims to provide valuable insights for the food and drinks industry, ultimately improving operational efficiency. This will not only result in cost reduction but also contribute to global climate change mitigation efforts. The anticipated outcomes of this study hold good potential for tangible benefits to food and drink businesses, empowering them to make informed decisions and effectively minimise their energy consumption.