Data-Driven Power System Operation: Exploring the Balance Between Cost and Risk

Abstract

Supervised machine learning has been successfully used in the past to infer a system's security boundary by training classifiers (also referred to as security rules) on a large number of simulated operating conditions. Although significant research has been carried out on using classifiers for the detection of critical operating points, using classifiers for the subsequent identification of suitable preventive/corrective control actions remains underdeveloped. This paper focuses on addressing the challenges that arise when utilizing security rules for control purposes. Illustrative examples and case studies are used to show how even very accurate security rules can lead to prohibitively high risk exposure when used to identify optimal control actions. Subsequently, the inherent tradeoff between operating cost and security risk is explored in detail. To optimally navigate this tradeoff, a novel approach is proposed that uses an ensemble learning method (AdaBoost) to infer a probabilistic description of a system's security boundary. Bias in predictions is compensated by the Platt Calibration method. Subsequently, a general-purpose framework for building probabilistic and disjunctive security rules of a system's secure operating domain is developed that can be embedded within classic operation formulations. Through case studies on the IEEE 39-bus system, it is showcased how security rules derived from supervised learning can be efficiently utilized to optimally operate the system under multiple uncertainties while respecting a user-defined balance between cost and risk. This is a fundamental step toward embedding data-driven models within classic optimisation approaches.

Recent Developments in Machine Learning for Energy Systems Reliability Management

“…In their paper, the authors present the challenges of this approach, which are the computational complexity to build the database and the accuracy of the such sample-derived rules. This approach is further developed in [119], where learnt condition-specific safety margins are proposed to be incorporated in a decision-making program. These margins allow, according to the authors, to improve the risk/cost balance.…”

Wide-Area Measurement System-Based Optimal Multi-Stage Under-Frequency Load-Shedding in Interconnected Smart Power Systems Using Evolutionary Computing Techniques

“…In the control center of modern power systems, a suitable frequency model can be created based on the data provided from WAMS. From (2), the frequency response model which can be used to design an optimal UFLS scheme, can be set up based on the swing equation considering the classical model of each generator [3]. The basic idea on analyzing the power system dynamics is based on the swing equation of each synchronous generating unit which is given in (3).…”

“…where α, β, γ and ε are the constant cost coefficients of transient frequency deviations (∆w ts ), steady-state frequency deviations (∆w ss ), load-shed ∆P j and integral of squared error (ISE) of the frequency 1 2 ∆w(t) 2 , respectively. The function f (Y) is a further term that can be used to describe an economic dispatch problem.…”

“…Modern power systems have become so sophisticated that research in the area of protection and control remains of major interest. Due to the ever-increasing electrical demand, large electrical power systems (EPS) are usually loaded close to their steady-state stability limit and thus are more vulnerable to network disturbances [1,2]. The operation at points close to the steady-state stability limit is a huge threat to power system security, such that in the event of a disturbance, the probability of cascading events, which might lead to total system collapse, is inevitable if not rectified in time.…”

An Imbalance Fault Detection Algorithm for Variable-Speed Wind Turbines: A Deep Learning Approach

“…Since 2006, DL has appeared as a new research field in machine learning research field [19]. DL can be used to extract the features of a large number of data [20][21][22]. Due to the high calculation costs and the features being difficult to be extracted, it is not applicable to obtain the features by using traditional mathematical methods [21].…”