Predict-then-optimise based day-ahead scheduling towards demand response and hybrid renewable generation for wastewater treatment

Abstract

Promoting a 100% renewable energy system requires intelligent scheduling strategies, yet the challenge remains on the prediction and optimisation of variable renewable energy supply and demand. This study proposes a Predict-then-optimise paradigm to explore day-ahead scheduling strategies for high renewable energy systems and demonstrates its application in a grid-connected biogas–solar–wind-storage system with load shifting for wastewater treatment plants. The scheduling strategy aims to maximise energy prosumption and minimise operation costs. Demand response is enabled by the wastewater pre-treatment reservoir, battery storage, and biogas storage, all mathematically modelled in this study. The Temporal Convolutional Network-based Transformer model is applied to forecast uncertain variable renewable energy generation and wastewater flow for the upcoming day. Then budget uncertainty sets are constructed based on forecast errors for robust optimisation. A case from Sichuan, China is analysed to explore the practicality and effectiveness of the proposed framework. The results indicate that the robustness of the model increases the day-head scheduling operational cost and decreases the self-sufficiency ratio. Wastewater pre-treatment reservoir scheduling can effectively shift the demand load, promoting cost reduction and system prosumption; besides, pre-treatment reservoir, battery storage and biogas storage have substitution and combination effects on demand response, can reduce daily operating costs by 20%–50%. The influence of a defined allowable sale ratio, seasons, and weather conditions are also discussed. Overall, the proposed predict-then-optimise framework is an effective solution for the upcoming day’s decision-making.