Leveraging hybrid probabilistic multi-objective evolutionary algorithm for dynamic tariff design

Dynamic tariffs play an important role in demand response, contributing to smoothing power consumption and reducing generation capacity requirement and carbon emission. However, in the existing works, tariffs are usually designed without comprehensive consideration, such as potential user responses to tariffs. Thus, assuming an electricity trading market contains a utility company and multiple residential users, a dynamic tariff design method is proposed in this paper, considering user responses to tariff changes. Leveraging the non-intrusive load monitoring technique, rated power and user preference features for each appliance are acquired by the utility company to quantify user comfort (discomfort) based on derived user appliance usage habits. Then, a bi-level Stackelberg game model is built on the supply side for designing optimal dynamic tariffs and imitating the influence of tariff changes on DR plans for users. The upper level represents the utility company, trying to maximize utility profit, social welfare and carbon emission reduction. While the lower level represents users, aiming to minimize electricity bills and user discomfort. By solving such an optimization problem with multiple objectives, a novel hybrid probabilistic multi-objective evolutionary algorithm balancing evolutionary efficiency and stability is applied where random forest is adopted to boost performance. The proposed model is benchmarked with two state-of-the-art pricing methods and validated on a publicly accessible REFIT dataset, where low-rate power measurements are collected from real houses in the UK. The experimental results show the proposed model generally outperforms benchmarks on dynamic tariff design in achieving peak-shaving and low carbon emission while preserving user satisfaction. Furthermore, a case study is implemented, which verifies the necessity of various objectives employed in the proposed method.