Machine Learning-Incorporated Transient Stability Prediction and Preventive Dispatch for Power Systems with High Wind Power Penetration
Historically, transient instability has been the most severe stability challenge for most systems. Transient stability prediction and preventive dispatch are two important measures against instability. The former measure refers to the rapid prediction of impending system stability issues in case of a contingency using real-time measurements, and the latter enhances the system stability against preconceived contingencies leveraging power dispatch. Over the last decade, large-scale renewable energy generation has been integrated into power systems, with wind power being the largest single source of increased renewable energy globally. The continuous evolution of the power system poses more challenges to transient stability. Specifically, the integration of wind power can decrease system inertia, affect system dynamics, and change the dispatch and power flow pattern frequently. As a result, the effectiveness of conventional stability prediction and preventive dispatch approaches is challenged. In response, a novel transient stability prediction method is proposed. First, a stability index (SI) that calculates the stability margin of a wind power-integrated power system is developed. In this method, wind power plants (WPPs) are represented as variable admittances to be integrated into an equivalent network during transients, whereby all WPP nodes are eliminated from the system, while their transient effects on each synchronous generator are retained. Next, the calculation of the kinetic and potential energies of a system is derived, and accordingly, a novel SI is put forward. The novel approach is then proposed taking advantage of the machine learning (ML) technique and the newly defined SI. In case of a contingency, the developed SI is calculated in parallel for all possible instability modes (IMs). The SIs are then formed as a vector and applied to an ensemble learning-trained model for transient stability prediction. Compared with the features used in other studies, the SI vector is more informative and discriminative, thus lead to a more accurate and reliable prediction. The proposed approach is validated on two IEEE test systems with various wind power penetration levels and compared to the existing methods, followed by a discussion of results. In addition, to address the issues existing in preventive dispatch for high wind power-integrated electrical systems, an hour-ahead probabilistic transient stability-constrained power dispatching method is proposed. First, to avoid massive transient stability simulations in each dispatching operation, an ML-based model is trained to predict the critical clearing time (CCT) and IM for all preconceived fault scenarios. Next, a set of IM-categorized probabilistic transient stability constraints (PTSCs) are constructed. Based on the predictions, the system operation plan is assessed with respect to the PTSCs. Then, the sensitivity of the probabilistic level of CCT is calculated with respect to the active power generated from the critical generators for each IM category. Accordingly, the implicit PTSCs are converted into explicit dispatching constraints, and the dispatch is rescheduled to ensure the probabilistic stability requirements of the system are met at an economical operating cost. The proposed approach is validated on modified IEEE 68- and 300-bus test systems, wherein the wind power installed capacity accounts for 40% and 50% of the total load, respectively, reporting high computational efficiency and high-quality solutions. The ML-incorporated transient stability prediction and preventive dispatch methods proposed in this research work can help to maintain the transient stability of the system and avoid the widespread blackouts.
Critical clearing time, decision tree, extended equal-area criterion, machine learning, phasor measurement units, transient stability, stability index, power dispatch, wind power.
Doctor of Philosophy (Ph.D.)
Electrical and Computer Engineering