Synthetic Inertia in Low-Inertia Grids

• Synthetic inertia is the emulation of synchronous machine inertia using power-electronic devices to inject or absorb energy based on grid frequency and its rate of change.
• Control strategies include fixed-gain RoCoF, model reference control, adaptive and discrete methods that mimic kinetic energy exchange for rapid response.
• Optimization and market integration of synthetic inertia involve strategic placement, sizing under device limits, and economic incentives to maintain frequency stability.

Synthetic inertia refers to the emulation of the inertial response of synchronous machines by power-electronic or non-synchronous devices, implemented in power systems increasingly populated by inverter-based resources. As traditional mechanical inertia from synchronous units is displaced by renewable energy sources (RES) with negligible intrinsic inertia, grid frequency stability is compromised. Synthetic inertia restores part of this lost stabilizing capability by commanding devices—such as wind turbines, batteries, supercapacitors, or even large coordinated loads—to modulate their active power injection or withdrawal in proportion to local grid frequency measurements and, critically, its rate of change. This enables fast system-level response to disturbances and is a cornerstone in the operation and planning of modern low-inertia power grids (Borsche et al., 2017, Badesa et al., 2022, Vaca et al., 2024).

1. Physical Principles and Core Modeling

Synthetic inertia is achieved by measuring the instantaneous grid frequency $\omega$ and/or its time derivative $\dot\omega$ at the connection point, then commanding the inverter or flexible device to inject or absorb active power accordingly. The classical two-pole model for a synthetic-inertia device at bus $v$ is described by the transfer function

$$\tilde P_v(s) = \frac{M_v\,s + K_v}{(T_{1v}s+1)(T_{2v}s+1)}\,\omega_v(s)$$

where $M_v$ (in s) is the inertia constant (proportional to the response to RoCoF), $K_v$ (in p.u.) is the damping coefficient on frequency deviation, and $T_{1v}, T_{2v}$ are filter and PLL time constants. For $M_v > 0$, a high RoCoF triggers power injection in opposition to the frequency deviation, mimicking the kinetic response of synchronous machines. The $K_v$ term provides effective damping torque (Borsche et al., 2017).

In grid-forming converters, the power reference is augmented to

$$P^*_i = P_{i,\text{set}} - m_i(t)\,\dot\omega_i(t) - d_i\,\omega_i(t)$$

where $m_i(t)$ is the (possibly time-dependent) virtual inertia and $d_i$ is the damping coefficient (Fritzsch et al., 2023).

Aggregated at the system level, the swing equation with both synchronous and synthetic inertia is

$$2(H_{\text{sync}} + H_{\text{synt}})f_0^{-1}\,\frac{d\Delta f}{dt} = \text{FR}(t) - P_L - P_{\text{rec}}(t)$$

where $H_{\text{sync}}$ and $H_{\text{synt}}$ are the system synchronous and synthetic inertia in MW·s, $f_0$ is the nominal frequency, $\Delta f$ is the frequency deviation, $\text{FR}(t)$ is frequency response (e.g., governor action), and $P_{\text{rec}}(t)$ models the recovery effect in converter-coupled machines that temporarily exhaust stored energy and then under-produce during recovery (Badesa et al., 2022).

2. Control Architectures: Continuous, Adaptive, and Discrete

Synthetic inertia can be supplied by several control strategies:

• Fixed-gain RoCoF control: Active power is injected in proportion to measured $\dot f$. Dominant implementation in wind turbines, batteries, and grid-forming inverters (Borsche et al., 2017, Chu et al., 2019, Nicolet et al., 2024).
• Model Reference Control (MRC): The desired inertia $M$ is specified in a reference swing equation, and the plant (e.g., diesel-wind hybrid) is controlled to track this model, yielding exact synthetic inertia emulation with bounded $H_\infty$ tracking error (Zhang et al., 2017).
• Adaptive inertia: The inertia provided by virtual synchronous generators (VSGs) is made time-varying, increasing rapidly with $|\dot\omega|$ after a disturbance, then relaxing over a controlled timescale back to a baseline value. The adaptive law

$$\dot m_{i} = \alpha_i |\dot\omega_i| - \beta_i(m_i - m_{\min,i})$$

enables fast recovery from faults while avoiding under-damped low-frequency oscillations (Fritzsch et al., 2023).

• Discrete synthetic inertia: Aggregations of discrete devices (loads or DERs) switch in integer multiples of a power packet $\Delta p$ according to a virtual oscillator model. The devices round the desired virtual machine power $p_e$ to the nearest discrete level, potentially leading to cycling that is mitigated by asynchronous triggering, packet-size diversity, or hysteretic logic (Vaca et al., 2024).

The principle across all implementations is that the controlled device mimics the energy exchange characteristics of a spinning mass, releasing or absorbing energy proportional to the local RoCoF within device power and energy constraints.

3. Optimization, Allocation, and Economic Integration

The spatial and temporal allocation of synthetic inertia is a critical aspect of system operation and planning:

• Placement algorithms: Optimal placement of synthetic inertia and damping aims to minimize metrics such as the worst-case RoCoF, frequency nadir (overshoot), and mode damping ratios, under physical and device constraints. Advanced optimization frameworks formalize the problem as a multi-objective nonlinear program over device gains $\{M_v, K_v\}$ with explicit box or dual-norm constraints from inverter nameplate ratings (Borsche et al., 2017). Sequential linear programming methods, leveraging eigen-sensitivity analysis, enable scalable optimization across large system models.
• Matrix perturbation and H₂-norm analysis: Analytical sensitivities of system frequency performance to synthetic inertia placement may be derived using matrix perturbation theory, enabling ranking heuristics. Uniform (homogeneous) distribution of synthetic inertia across generator buses is near-optimal for global H₂-norm (energy of frequency deviation), while primary control benefits from placement at buses participating most in slow inter-area eigenmodes (Pagnier et al., 2019).
• Co-optimization in scheduling and energy markets: Synthetic inertia can be treated as a co-optimized ancillary service, integrated with unit commitment and dispatch. Constraints include frequency-security requirements on RoCoF, nadir, and steady-state deviation, potentially linearized for tractability in MILPs (Chu et al., 2019, Badesa et al., 2022). Allocation variables include device curtailment headroom, inertia constants per site, and primary/damping headroom.
• Shadow pricing and market incentives: Shadow prices for synthetic inertia are derived from the duals of RoCoF and nadir constraints. These signal the locational and temporal value of providing (or witholding) synthetic inertia, allowing RES owners and operators to make informed investment and operational decisions (Badesa et al., 2022).

4. Performance Metrics, Case Studies, and Empirical Validation

Performance of synthetic inertia schemes is quantified by several key metrics:

• RoCoF peak (maximum rate of change of frequency)
• Frequency overshoot / nadir after a disturbance
• Damping ratio of oscillatory modes
• Resynchronization (settling) time
• Integrated frequency deviation ($L_2$ norm) and inertial energy injected

Empirical studies demonstrate:

• Replacement of synchronous inertia by optimally allocated synthetic inertia achieves comparable or improved RoCoF and frequency nadir at reduced inertia headroom (up to 5$\times$ reduction in some cases) (Borsche et al., 2017).
• Adaptive inertia controllers reduce both frequency deviation and resynchronization time by 15–20% and more than double the RoCoF attenuation, especially when “warm-started” with high initial inertia (Fritzsch et al., 2023).
• Discrete-device-based synthetic inertia with 300,000+ devices limits RoCoF peaks to ~0.02 Hz/s (WSCC 9-bus) and ~0.03 Hz/s (all-island Ireland), keeping frequency deviation and power-balance within tight bounds (Vaca et al., 2024).
• Variable-speed pumped storage units, with Ka tuned to match physical inertia constants, reproduce the inertial response of synchronous machines accurately, with frequency excursions nearly suppressed at sufficient inertia gain (Nicolet et al., 2024).

A representative table (core scenarios from (Borsche et al., 2017)):

Scenario	$\zeta_{\min}$ (%)	$R_{\max}$ (mHz/s)	$S_{\max}$ (mHz)	Total $M$	Total $K$
Full-inertia	18.6	193	56.5	—	—
Low-inertia (after removal)	19.1	396	98.3	—	—
Opt: Min. $R_{\max}$	15.0	94.1	28.4	429	163
Opt: Min. $S_{\max}$	15.0	96.2	27.4	417	238
Opt: Minimal headroom	15.0	100	30.0	87.6	63.3

5. Implementation Challenges and Device Limits

Deployment of synthetic inertia is limited by:

• Converter power, energy, and speed excursion constraints: Fast large RoCoF events may require more power than available from inverter limits or stored kinetic/electric energy, especially in wind or hydro units (Nicolet et al., 2024). Recovery periods (depletion of available energy) must be managed, modeling post-event underproduction (Badesa et al., 2022).
• Device quantization and cycling: In discrete-device approaches, excessive quantization (large packet size) or synchronized operation can induce unwanted cycling—repetitive on-off switching around the required power level. Asynchronous update logic, hysteresis in switching criteria, and diverse packet sizes are essential remedies (Vaca et al., 2024).
• Measurement and communication: High-fidelity local measurement (PLL, frequency estimation) is mandatory. Coordination is required to avoid large clusters of devices switching simultaneously, risking instability or communication bottlenecks (Vaca et al., 2024).
• Small-signal and transient stability: Excess inertia, or poorly tuned synthetic inertia gains, can excite under-damped system modes, induce large mechanical excursions, or violate device thermal constraints. Adaptive and time-domain constraint-driven tuning methodologies are necessary (Borsche et al., 2017, Fritzsch et al., 2023).

6. Market Mechanisms and Future Research Directions

Integration of synthetic inertia into market-based ancillary services frameworks is ongoing:

• Explicit unbundling: Distinction is drawn between synchronous inertia, synthetic inertia, enhanced frequency response (EFR), and primary frequency response (PFR) in both technical modeling and market products (Badesa et al., 2022).
• Shadow pricing and payments: Market designs assign transparent value to synthetic inertia provision, accounting for device-specific limitations and the “recovery effect” penalty (Badesa et al., 2022).
• Dynamic and adaptive scheduling: Real-time adjustment of device inertia settings ($H_i$) and proactive coordination maximize delivered value versus device risk and wear (Chu et al., 2019, Fritzsch et al., 2023).
• Integration with new grid architectures: Ongoing research addresses distributed device coordination, HVDC-embedded inertia, stochastic and robust optimization under uncertainty, and coupling with investment decisions (e.g., location-marginal inertia pricing) (Borsche et al., 2017).

Emerging directions include the extension of synthetic inertia to large aggregations of flexible loads, implementation in grid-topology-aware optimal control, advanced state estimation for decentralized response, and full integration into stochastic scheduling under variable renewables (Vaca et al., 2024, Chu et al., 2019).

---

The practice of synthetic inertia provision is a cornerstone in ensuring the frequency stability of low-inertia grids with high penetration of inverter-based resources. Its implementation leverages advanced control theory, optimization, and market mechanisms for robust, cost-effective, and scalable deployment (Borsche et al., 2017, Fritzsch et al., 2023, Badesa et al., 2022, Vaca et al., 2024, Nicolet et al., 2024, Zhang et al., 2017, Pagnier et al., 2019, Chu et al., 2019).