Strengthening the Security of Electric Power Systems during the Energy Transition: Predicting Cascading Failures and Reinforcing the Network Topology

Abstract

The reliable, affordable, environmentally friendly supply of electricity is crucial for the current and future well-being and success of modern society. A stable yet efficient operation of electric power systems is vital in this context. However, recent developments such as the scheduled phase out of dispatchable nuclear and conventional power plants in favor of weather-dependent renewable sources increasingly pose challenges to further improving or even maintaining the current level of security-of-supply. Simulation-based power system investigations play a key role in meeting those challenges, requiring suitable methods of examination, as well as accurate computer models that are ideally fast to evaluate. The present work is aimed at advancing the status quo in this area of research with the objective of enhancing the reliability and performance of electric power systems. To do so, first of all three cascading failure simulation models for power system analysis (PSA) are compared, namely the DC power flow-based OPA model, the AC power flow-based Manchester model, and the proprietary AC power flow-based Cascades model. These models are extended to incorporate temperature dependent dynamic transmission line ratings. The investigation is based on power flow studies of the IEEE 24-Bus reliability test system (RTS) and is performed for a variety of different operating conditions, i.e., demand levels and ambient temperatures. Several metrics are considered in the comparison, such as the demand not served (DNS), the most frequent transmission line overloads, or the overall risk of operations. The results highlight that, when the power grid is subjected to elevated temperature and demand levels, the numerical deviations between the Manchester/ Cascades model and the OPA model can be large. However, all models show the same general trends, namely that the DNS generally increases with increasing temperature and demand. Further similarities are found in terms of the most critical lines and the most heavily loaded generators. Regarding the overall risk of operations the OPA model indicates elevated risk for a larger part of the input space that also includes almost the entire area found by the Manchester and the Cascades model. Its reasonably accurate behavior makes the fast, easy-to-adapt OPA model perfectly suited for methodology development. Investigations that rely on power system behavior and control being recreated as realistically as possible, however, should be conducted using the Cascades model as it captures relevant physical features of the operations of power systems and is the most up-to-date. Having determined the applicability of suitable computer models for PSA, a novel method for the identification and mitigation of critical states in power systems is developed and tested in a second step. The integral part of this method is the reconstruction of the limit state surface of a power system through thousands of cascading failure simulations at varying operating points, e.g., weather conditions or demand levels. “Limit state surface” in this context describes the interface separating the set of operating conditions leading to non-critical operations from those causing detrimental operating states that require load shedding or other control actions by the transmission system operator (TSO) to prevent component or even system failure. The methodology is exemplified on a modified version of the North Italian 380kV transmission grid, whose limit state surface is determined as a function of ambient temperature, wind speed and demand volume using the OPA model. Based on the reconstructed limit state surface an online criticality indication approach is presented that determines the closeness of a power system to detrimental states of operation by computing the equivalence of a spatial distance between the current operating conditions and the limit surface. Its application is tested within the scope of a 24-h trajectory analysis, with the indicator correctly identifying those times when the North- Italian power system is approaching unsafe states of operation. Upon the occurrence of such states, two mitigation strategies are applied, i.e., line switching or distributed generation (DG). For the analyzed system and operating conditions, DG more effectively mitigates the critical states as compared to line switching. Aside from that, the study indicates the benefits of considering dynamic line ratings instead of traditional static or seasonal line ratings. Due to the exponential increase in computational cost afforded by increasing the dimensionality of a limit state surface reconstruction, the third part of this work is focused on improving the performance of high-dimensional power system assessments. In particular, the efficacy of surrogate modelling, specifically by polynomial chaos expansion (PCE), is investigated in various case studies based on the Swiss high-voltage transmission grid. Corresponding power flow simulations are carried out as a function of 22 correlated load- and weather parameters using the Cascades model. Despite the large variability and non-smoothness observed in the Cascades model responses, which is generally a poor prerequisite for the successful application of surrogate modeling techniques, the PCE metamodel performed very well aside from a slight tendency of underestimating the criticality of system operations. Also general trends, such as a clustering of critical conditions whenever there is high demand, high insolation, low wind speeds, high temperatures, high electricity exports and low imports, are reproduced flawlessly. As for the computational cost, even evaluating tens of thousands of data points only lasts a few seconds compared to multiple hours with the Cascades model. Creating the PCE surrogate, however, takes disproportionally longer, especially with large experimental designs. Striking the right balance between PCE build time and accuracy is, therefore, crucial. Finally, pursuing the objective of optimally guiding future grid expansions a proprietary state-of-the-art risk-informed transmission expansion planning (TEP) tool is presented. The underlying algorithm simultaneously minimizes the cost of expansion and the risk of systemic failures, with the latter objective being evaluated using the Cascades model. Testing of the TEP tool is performed on the IEEE 118-bus system. The results demonstrate the robustness of the method in providing adequate expansion planning solutions, while a comparison with the literature indicates that the presented TEP method is superior to previously published works on the TEP problem.