Future-OPF-SC-CA

GO Competition Challenge 4: Optimal Power Flow with Stability Constraints and Control Algorithms

We propose a Grid Optimization (GO) Competition Challenge 4 focusing on a combined optimization and control problem of optimal power flow (OPF) with stability constraints (SC) and control algorithms (CA). That is, in addition to the traditional OPF constraints, our proposed OPF-SC-CA problem considers constraints of voltage and frequency stability, and in addition to the traditional OPF decision variables, we consider the feedback control algorithms as part of the decision space. At best, these models use conservative heuristic approaches such as surrogate constraints or additional reserve requirements to ensure stability. Accurately modeling the dynamic response and fast controls will enable the OPF solution to be less conservative while still ensuring stability, thus unlocking substantial economic benefits such as allowing greater use of available wind and solar energy. OPF with stability constraints and control algorithms is a very hard problem, and the lack of capable solvers prevents the electricity industry from even considering its possible benefits. GO Competition Challenge 4 will spur needed research on this problem and accelerate its commercial impact.

In a traditional OPF problem, many of the decision variables represent quantities that are periodically updated at time intervals ranging from five minutes, to hourly, to perhaps daily. These include continuous quantities such as generator MW dispatch levels and “PV-bus” voltage magnitude values, as well as binary quantities such as breaker positions for capacitor banks that determine reactive compensation.  More advanced problems may also include binary decision variables associated with grid topology changes and unit commitment. These optimizations are based on sinusoidal steady state models of grid behavior. They assume that any feasible decision variable selection yields an operating point having sufficient margin of stability and that the grid dynamics allow a stable transition from the prior operating point to this new equilibrium. While OPF problems sometimes include heuristics-based steady state constraints that seek to ensure stability assumptions are valid, rapidly evolving grid technology and controls make these dynamic issues increasingly important and challenging. The proliferation of inverter coupled resources creates a much wider range of possible dynamic behavior, both advantageous and disadvantageous, than was possible when grid dynamics were almost entirely dominated by the behavior of traditional synchronous machines. Moreover, extreme weather and cyber events impose more stringent requirements to ensure adequate margins of grid stability.

These emerging issues in grid operation motivate more careful treatment of the interplay of slower time scale updates of setpoints and grid configuration/topology with faster time scale dynamic controls. To address this, we are considering a combined optimization-control problem that would treat grid operator actions and controls from the steady state time scale of about 5 minutes down to the transient stability time scale of tens of milliseconds.  As noted above, actions that can be taken at the steady state time scale include generator redispatch and voltage set point revision, adjustment of voltage and flow control devices such as transformer taps and capacitor bank switching, and topology reconfiguration such as bus splitting and line switching. Some of these actions might seem to offer great benefit in a steady state optimization model yet have deleterious stability effects that are not adequately treated by heuristic, “surrogate” OPF constraints. In particular, topology switching has the potential to resolve transmission congestion that would otherwise prevent the fullest use of available variable renewable energy resources, but such switching actions are currently underutilized because their effects on grid stability have only been adequately studied in an offline context. As a first step, it is critical to ensure that switching actions yield stable performance on the dynamic time scale.  In a more sophisticated formulation, one may wish to enhance dynamic controls specifically to enable switching actions that are beneficial over longer time scales, while maintaining a safe margin of stability in grid dynamic response.

Another such action is microgrid islanding and other possible responses to extreme weather conditions. To date, no one has articulated clear action plans to be followed during extreme weather events, even though that is the real resource adequacy question of the 21^st century. It is known that forced outage rates are higher during times of extreme cold and heat, and that such conditions can pose real threats to human life as well as property. Utilities are also starting to build resilience centers equipped with microgrid, battery energy storage, and other technologies in various metropolitan areas across the U.S. A formulation of this problem could involve islanding, line switching, supply dispatch, and demand dispatch variables with solutions evaluated on amount of critical load served, AC feasibility, dynamic stability, and cost.

We propose to investigate these issues through a software competition in which a competitor provides a solution comprising both steady state decision variable values (e.g., setpoint, commitment, and topology updates) and dynamic control algorithms for multiple components to ensure stable dynamic response for the overall system. A competitor could ensure stability in their solution by including a dynamic simulation and a control algorithm in their solver, or they could use more conservative traditional heuristic approaches such as surrogate constraints and additional reserve requirements. Our solution evaluation would then check for stability by running a dynamic simulation.

The evaluation engine of such a competition would be an enhancement to what is often termed a “long term transient stability simulator.” While many details will benefit from further consideration and debate, as a strawman we envision a positive-sequence dynamic model for this evaluation engine, mainly based on phasor representation of electrical quantities.  This choice would allow adequate computational efficiency to treat large-scale networks (up to ten thousand buses), with accurate modeling of dynamic phenomena down to several 10’s of milliseconds, and simulation horizons out to several hours.  However, it is also true that some inverter behavior and distribution-level disturbances may benefit from even faster time scale models of instantaneous voltage and current waveforms (“point-on-wave” in power systems terminology). Considerable progress has been made in recent years in simulation techniques and software that mix phasor representations in most of the network with point-on-wave representations in select subsections of the network. The work proposed here may draw from such methods.