Exploring data-driven technique in optimal power flow study through real-time cyber-physical power system under uncertainties

Data-driven techniques have become a significant trend in recent years across various engineering domains, particularly in power systems. The ability to analyse large datasets and process input signals efficiently has positioned these techniques as a powerful alternative to traditional calculation methods. Since power systems evolve and become more complex, there have been increased challenges. One of the challenges is the growing number of control variables that place a heavy burden on conventional computational approaches, leading to extended calculation times. Given the necessities to address this challenge, this paper provides an in-depth analysis of the data-driven approach in the Optimal Power Flow (OPF) studies, highlighting its potential for managing large-scale power systems and its suitability for real-time applications. The key contributions of this paper can be listed as follows: (i) Modelling power system uncertainties by using stochastic models for load demands and the fluctuations of distributed renewable energy resources (DERs) such as solar, or wind power generations, (ii) A comparison study of two common optimisation numerical techniques for OPF, differential evolution, and interior point optimiser, to identify the most well-suited method for data-driven models, (iii) Leveraging the data-driven approach to forecast optimum setpoints based on the power network profiles, and (iv) Implementation of a real-time cyber-physical power system (CPPS) for validating the proposed data-driven approach. This paper helps clarify the growing trend of data-driven technique utilisation in engineering domains, both in industry and research, with a particular focus on power systems, aiming to wide-area monitoring and control in future smart grids.