The study, published in Energy Reports by Xiangxu Wang, Junjie Sun, Xinwei Li, Xiaoheng Zhang, Shubo Hu, Jifeng Cheng, Chao Wang, and Qiang Zhang, addresses a fundamental gap in how engineers assess frequency security. Frequency is one of the key indicators for monitoring and operating alternating-current power systems, and dispatching agencies must accurately assess frequency conditions and develop adequate control measures to keep fluctuations within the nominal range. Several large-scale power incidents in recent years, including major blackouts, have demonstrated that wide-range frequency fluctuations are a primary cause of cascading failures. Frequency response, the collective reaction of generation and load to a sudden power imbalance, serves as the first line of defense, working alongside automatic generation control to arrest frequency decline before protective relays begin shedding load.

Existing analytical approaches to frequency response analysis trace back to a landmark low-order System Frequency Response model that assumed frequency behaves identically everywhere in the grid, an assumption reasonable when synchronous machines dominate. But modern grids violate that assumption in two profound ways. First, the resources providing frequency response have diversified. Renewable generators and aggregated virtual power plants now offer so-called fast frequency response through virtual inertia and droop control implemented in power-electronic converters, while battery storage, flywheels, and industrial interruptible loads such as electrolytic aluminum plants inject power in rapid step changes. Second, because these resources are geographically uneven, the frequency spatial distribution characteristics across interconnected regions have become pronounced, meaning each region experiences a distinct rate of change of frequency, a distinct frequency nadir, and a distinct recovery trajectory after a disturbance.

Ignoring this spatial variation, the authors argue, can lead to misjudgments of frequency stability and inaccurate control, causing local frequency instability even when the system-wide average appears secure. The problem with capturing it is mathematical: once the transmission network is retained in the model, power angle coupling between generators causes the model order to balloon exponentially, making analytical solutions infeasible for realistic systems. Earlier regional frequency response models either worked only for two regions, required Taylor series expansions running to ninety-first order, relied on empirical parameter estimation, or omitted fast frequency response resources altogether. None of the previous closed-form regional models, the researchers note, simultaneously captured the frequency spatial distribution characteristics, incorporated fast frequency response resources, and achieved a rigorous closed-form solution.

The new model does all three for a three-region interconnection, which the authors consider a fundamental and representative form sufficient for engineering practice. Each region is modeled as an equivalent generator obeying a simplified swing equation, with three categories of frequency response resources represented: synchronous generators with their dominant reheat governor dynamics retained and faster time constants discarded; proportional or derivative response resources such as wind, photovoltaic, and virtual power plants that respond to both frequency deviation and its rate of change; and step response resources such as batteries and flywheels that inject a preset power increment for a specified duration. Between regions, a DC power flow approximation, which linearizes the nonlinear power flow equations by neglecting reactive power and voltage dynamics, describes how power transfers across tie lines in response to rotor angle differences, preserving the crucial network coupling while keeping the equations tractable.

The mathematical centerpiece of the paper is a striking analogy: the regional frequency response problem is shown to be isomorphic to the forced vibration of a three-degree-of-freedom mass-damper-spring system. The inertia matrix of the power system plays the role of mass, the damping coefficients play the role of the vibration damper, the network admittance plays the role of stiffness, and the power deficit plays the role of external excitation. This correspondence allows the researchers to import the well-established modal analysis method from vibration mechanics. Solving the eigenvalue problem of the undamped free vibration equation reveals that a three-region grid possesses three modes. The first, with a zero eigenvalue and a uniform eigenvector, represents the rigid-body motion of the whole system, which corresponds exactly to the frequency of the system’s center of inertia, the weighted average frequency that all regions eventually converge toward. The remaining two modes describe damped inter-regional oscillations, the relative swinging of one region’s frequency against the others.

The resulting closed-form solution for any region’s frequency is therefore the linear superposition of three terms: the identical center-of-inertia component shared by all regions, and two proportional damped-sinusoidal oscillation components driven by the other two regions. Because the three regional systems are weakly interconnected, a forced decoupling approximation renders the modal equations independent, allowing each to be solved analytically and transformed back through the inverse Laplace transform. The final expressions are concise, symmetric, and require no iteration: frequency at any instant in any region can be computed directly from grid parameters and disturbance data, with the computational load equivalent to a single step of a conventional iterative solver.

The validation is where the model proves its worth. On a modified IEEE Western System Coordinating Council nine-bus test system, reconfigured with 120 megawatts of wind power and 70 megawatts of energy storage so that fast frequency response resources hold 30 percent of capacity, a sudden 100-megawatt load increase exposed dramatic discrepancies in the older system-wide model. In one region the true maximum rate of change of frequency was 1.926 hertz per second, while the single-frequency model predicted minus 0.988, and the time to frequency nadir in another region was off by nearly a full second, deviations that could lead to drastically different control strategies. The new three-region model tracked the reference curves closely.

More compelling still is the test against a real provincial power system in China: 48 buses, 123 transmission lines, 313 loads, 44 synchronous generators across nine conventional plants, and 73 doubly fed induction generators across five wind farms, with renewables supplying 24 percent of generation. Benchmarked against PSASP, a commercial full time-domain simulation package using sixth-order generator models, the new analytical model computed all frequency response indicators in 23 milliseconds, roughly 480 times faster than the 11-second simulation, with mean absolute percentage errors of just 3.06 percent for maximum rate of change of frequency, 6.26 percent for maximum frequency deviation, and 15.73 percent for time to nadir. The older system-average model, by contrast, posted errors as high as 38 and 152 percent and took no less time. The study also mapped the model’s boundaries: as long as spinning reserve stays within the standard engineering range of 6 to 12 percent of load, generator output saturation never activates and accuracy holds; only under abnormally thin reserves, where generators hit output limits during transients, do the linear assumptions break down.

The practical implications reach well beyond a faster calculator. Millisecond-scale analytical frequency assessment makes online situational awareness, emergency control, economic dispatch, reserve planning, and unit commitment with explicit frequency constraints genuinely feasible on modern, low-inertia grids, letting operators coordinate scarce conventional frequency response with rapidly growing fast frequency response from renewables and storage. The authors acknowledge the model’s simplifications, including neglected dead zones, limiters, and reactive power coupling, and propose piecewise linearization, superposition of active and reactive components, and hybrid data-physics methods as remedies. Future work will incorporate the uncertainties of renewable output, customer behavior, and weather to capture the probabilistic distribution of frequency, an essential step as grids worldwide race toward decarbonization without sacrificing the split-second stability that modern life depends on.

Subject of Research: Analytical modeling of regional frequency response in three-region interconnected power systems with fast frequency response resources

Article Title: Regional frequency response analytical model for three-region interconnected power systems considering fast frequency response resources

Article References: Wang, X., Sun, J., Li, X., Zhang, X., Hu, S., Cheng, J., Wang, C., & Zhang, Q. (2026). Regional frequency response analytical model for three-region interconnected power systems considering fast frequency response resources. Energy Reports, 16, Article 109670. https://doi.org/10.1016/j.egyr.2026.109670

Image Credits: AI Generated

DOI: 10.1016/j.egyr.2026.109670

Keywords: frequency response, fast frequency response, interconnected power systems, frequency spatial distribution, low-inertia grids, modal analysis, closed-form solution, renewable energy, RoCoF, frequency nadir, power system stability, Energy Reports