Why weak grids generate flicker: the impedance-PLL feedback trap

Flicker from grid-tied inverters in weak grid conditions originates from a closed feedback loop: current injected by the inverter flows through high grid impedance, creates a voltage disturbance at the point of common coupling (PCC), and that disturbance propagates back through the Phase-Locked Loop (PLL) and current control loop — sustaining oscillations that are perceptible as flicker by connected loads. This mechanism, demonstrated by Hunan University (2020), is the primary oscillation driver in weak grids, and it becomes more severe as the short circuit ratio (SCR) at the PCC decreases.

The problem is compounded in multi-inverter parallel configurations. As analysed by Aalborg University (2017), unknown and unpredictable grid impedance leads to variable resonance frequencies, which challenges the robust design of LCL filters and produces harmonic magnification from resonance circuits formed by the interaction between uncertain grid impedance and the converter. A filter designed for one grid impedance condition may become a resonance amplifier under a different network topology state — a particularly dangerous scenario for flicker-sensitive loads on the same bus.

Low-frequency oscillations driven by PLL-outer loop coupling are identified as the most challenging stability threat. Research from Harbin Institute of Technology (2021) demonstrates that coupling interaction between the DC link voltage controller, AC voltage controller, and PLL creates dominant low-frequency oscillations when an inverter is connected to a weak grid. Their proposed remedy — using a DC link voltage error to modulate the reactive current reference — introduces low-frequency damping that directly decouples the controller interactions responsible for slow-timescale voltage fluctuations at the PCC.

A 2022 review on harmonic resonances in multi-parallel grid-connected inverters further confirms that control algorithm dynamics introduce fast-changing impedance characteristics across wide frequency bands, rendering traditional grid analysis methods insufficient for predicting where resonance will occur — a core design challenge for any engineer targeting flicker compliance in high-renewable-penetration networks.

Control-domain remedies: impedance reshaping, PLL redesign, and feedforward compensation

The most widely adopted family of solutions operates at the control level, reshaping the inverter’s output impedance to suppress the oscillatory modes that drive flicker — without adding physical hardware. Virtual impedance techniques are the foundational approach, and the design principle is precise: the inverter output impedance should be designed relatively high at the harmonic oscillation frequency while remaining relatively low at other frequencies.

Hunan University (2018) establishes this principle concretely by adding two virtual impedances — one in parallel and one in series with the original output impedance — implemented via notch filters. This dual-notch architecture effectively increases system damping at targeted oscillation frequencies without degrading fundamental-frequency current quality. The method has been extended by Zhongyuan University of Technology (2023), which combines a notch filter and a resonator in a biquad structure for LLCL-type inverters, placing the resonance and notch points at appointed frequencies to transfer the resonant peak into the stable region through phase transformation — providing higher control bandwidth and stronger robustness in weak power grids.

PLL redesign is a second critical lever. For single-phase systems, Shenyang University of Technology (2022) demonstrates that the coupling between the all-pass-filter PLL (APF-PLL) and the current control loop is a principal instability mechanism. Their proposed complex coefficient filter modified first-order PLL (CCF-MFOF-PLL) with parameter optimisation reshapes the small-signal impedance and decouples dual d-q frame interactions. State Grid (2018) similarly confirms that changing the PLL design can stabilise an otherwise unstable parallel PV inverter system by modifying the effective output impedance seen by the grid.

“A filter designed for one grid impedance condition may become a resonance amplifier under a different network topology state — a particularly dangerous scenario for flicker-sensitive loads supplied from the same bus.”

Grid voltage feedforward requires careful implementation in weak grids due to digital control delay effects. Wuhan Second Ship Design and Research Institute (2020) identifies that uncompensated digital delay introduces negative phase shift in the output impedance, which deteriorates stability. Their quasi-resonant feedforward items compensate for delay-induced phase errors while boosting output impedance at specific background harmonic frequencies — simultaneously improving harmonic rejection and weak-grid stability margins. Power China Chengdu Engineering (2021) addresses the positive feedback channel coupling problem by adding resonant links selective to background harmonic frequencies, preventing the feedforward path from creating additional instability while still rejecting PCC voltage disturbances.

For systems requiring robustness over wide impedance variation ranges, Yanshan University (2020) proposes combining a capacitor current inner loop with a grid current inner loop alongside voltage feedforward, demonstrating that this combination prevents the excessive damping coefficient that would otherwise cause instability as grid impedance increases. For new inverter designs, State Grid Henan (2019) proposes a parameter normalisation scheme that jointly optimises LCL filter parameters and controller parameters simultaneously — guaranteeing stability and robustness without requiring any compensation network or additional hardware.

Filter-based and hardware-assisted suppression techniques

Beyond pure control-domain approaches, filter design and hardware-level strategies provide complementary suppression of flicker-driving disturbances. Active damping using virtual resistors is the dominant hardware-emulation technique — it absorbs resonant energy at the LCL filter resonance frequency without the thermal losses of a physical resistor.

State Grid Shandong (2023) constructs a virtual resistor via active damping to absorb resonant components near the LCL filter resonance frequency, while simultaneously using an adaptive modulation voltage control to address low-order harmonics of the 6k±1 order caused by nonlinear switching characteristics. Both categories contribute to PCC voltage distortion and flicker, making the coordinated wideband approach particularly effective for photovoltaic grid-connected inverters.

For multi-inverter systems, resonance must first be detected before it can be suppressed. State Grid Jiangsu (2019) proposes cascaded second-order generalised integrators (SOGI) with a normalised frequency-locked loop (FLL) to extract the resonant frequency from severely distorted PCC voltage signals. This resonance detection feeds a resonance damper that virtualises a resistor at the detected frequency — a crucial capability when resonance frequency shifts unpredictably due to changing grid impedance. Nanjing Institute of Technology (2022) complements this with a self-tuning filter for voltage resonance component extraction that constructs a virtual impedance branch modifying grid impedance characteristics at the resonance frequency, improving impedance stability criterion compliance.

Henan Polytechnic University (2019) introduces a digital notch filter into the capacitance current active damping scheme specifically to address the resonance generated by interactive current between parallel LCL inverters. For three-phase four-wire topologies — which introduce an additional zero-sequence current loop — Naval University of Engineering (2023) derives positive, negative, and zero sequence admittance models under multi-perturbation variables and proposes an admittance remodelling strategy based on a negative third-order differential element to suppress high-frequency oscillation. This topology-specific challenge is directly relevant to distributed weak-grid connection scenarios such as microgrids and building-integrated PV systems.

North China University of Technology (2023) takes a hardware-partitioned approach for high-power applications with low pulse ratios, separating power inversion and harmonic elimination into two parallel inverter units — a power inversion unit (PIU) and an active harmonic elimination unit (AHEU) — reducing switching losses while maintaining harmonic suppression capability through feedforward compensation.

System-level coordination: parallel IBR interleaving and SSCI damping

At the system level, coordinating multiple inverter-based resources (IBRs) connected to the same weak-grid PCC provides a powerful complementary mechanism that operates independently of individual inverter control design. Two active patents from General Electric represent the most direct industry-facing protection of this approach.

General Electric Renovables’ 2025 patent (KR pending) discloses a method in which pulse patterns of at least two IBRs connected at a common PCC are interleaved based on measured electrical signals. By establishing a synchronised timing reference that offsets the switching events of parallel units, the method reduces voltage distortion at the PCC directly — mitigating the high-frequency switching harmonic content that couples through grid impedance to create flicker-inducing PCC voltage perturbations. The approach requires only local controller coordination rather than physical hardware modifications, making it deployable via firmware update to existing installations.

Sub-synchronous control interactions (SSCI) represent a distinct low-frequency threat in high-impedance grids that contributes to power oscillations in the flicker-perceptible frequency range — below the fundamental grid frequency of 50/60 Hz. General Electric Company’s 2024 patent (EP active) addresses this by rotating the current feedback signal to a sub-synchronous reference frame, extracting the sub-synchronous component, and generating a damping voltage command proportional to a virtual resistance setting. This explicitly targets the sub-synchronous frequency range where SSCI oscillations occur when grid-forming inverters are connected through high-impedance transmission paths.

State Grid Shanxi (2023) takes a system-level analytical approach, proposing improved weighted average current control (WACC) for background harmonic suppression alongside modal-analysis-based resonance impedance reshaping to address both low- and high-frequency instability simultaneously. The triple-decomposition conductance model enables isolation of harmonic source contributions, allowing targeted suppression without destabilising the fundamental-frequency power flow — a key requirement for grid operators managing high-penetration renewable portfolios in accordance with standards from IEC and IEEE.

Adaptive and voltage-sensorless control for changing grid conditions

A persistent practical challenge is that voltage sensors at the PCC introduce measurement errors and delays that degrade feedforward effectiveness under weak grid conditions — and in some topologies, the sensors themselves become a source of instability. Voltage-sensorless active damping, proposed by Seoul National University of Science and Technology (2022), eliminates the need for PCC voltage sensing entirely by using integral variables to achieve resonance frequency damping despite uncertain grid impedance and distorted harmonics.

The measurement error sensitivity problem is confirmed by Tel Aviv University (2021), which shows that certain synchronverter sensitivity functions exhibit high gains at relevant frequencies, leading to distorted grid currents — a fundamental power quality issue that worsens as grid impedance increases. This finding underscores why sensorless or sensor-independent approaches are gaining traction in high-impedance deployment scenarios.

Adaptive control frameworks that track changing grid impedance without requiring explicit impedance measurement are represented by Guizhou University (2020), which applies a particle swarm optimisation-tuned radial basis function (RBF) neural network to identify PI controller parameters and update active damping coefficients in real time based on stability margin constraints. This approach enables the system to maintain stability margins as grid impedance varies — critical for maintaining low flicker emission across changing grid topology states. International Islamic University Islamabad (2022) similarly demonstrates that PLL-amplified low-order harmonics under weak grid conditions can be attenuated through hybrid adaptive control strategies that simultaneously maintain total harmonic distortion (THD) compliance and stability margins, consistent with power quality requirements tracked by organisations such as IEC and reported by IRENA in renewable integration assessments.

For frequency-adaptive operation under grid frequency fluctuation — common in weak isolated grids — Sichuan University (2021) combines a proportional-resonant controller with repetitive control (PRRC) and a Newton-structure fractional delay filter. This combination achieves zero steady-state tracking error and harmonic distortion compensation even when grid frequency deviates from its nominal value — a situation where fixed-frequency notch and resonant controllers would otherwise lose effectiveness. This is particularly relevant for island grids and remote renewable installations where frequency regulation is less precise than in interconnected transmission systems.

Key players and the direction of innovation

The institutions and companies contributing most actively to flicker suppression in weak-grid inverter systems span Chinese universities, national grid operators, and global industrial assignees — with distinct technical emphases reflecting their operational contexts.

Hunan University has established the foundational virtual impedance design principles applicable to weak grids, with multiple papers on parallel inverter oscillation suppression using notch filters and large-signal impedance modelling with limiter analysis (2018, 2023). General Electric and GE Renovables hold the most directly commercial IP, with two active patents targeting system-level suppression via interleaved pulse patterns for multiple IBRs (2025, KR pending) and a dedicated SSCI damping method for grid-forming IBRs (2024, EP active).

Huawei Technologies / Huawei Digital Power holds active patents covering grid stability detection via disturbance current injection (2020) and adaptive resonance frequency adjustment through variable switching frequencies and resonant branches (2025). The State Grid Research Institutes (multiple provincial branches — Shandong, Shanxi, Jiangsu, Henan) contribute both analytical frameworks and practical mitigation methods drawn from operational experience with high-penetration renewable integration.

Aalborg University (Denmark) and the National Renewable Energy Laboratory (NREL, USA) provide the international academic perspective, with contributions on grid impedance harmonic magnification (Aalborg, 2017) and frequency shaping control for weakly-coupled grid-forming IBRs (NREL, 2023) — the latter reflecting the broader shift from grid-following to grid-forming inverter architectures that is reshaping the field, as documented by NREL and tracked by IEA in its power systems transition reports.

The innovation trajectory is clear: the field is moving from single-inverter passive and active damping toward system-level coordination of multiple IBRs with adaptive, model-free, and machine-learning-assisted control, with increasing emphasis on grid-forming strategies over grid-following ones for high-impedance scenarios. This shift is driven by both the increasing share of inverter-connected generation and the recognition that individual inverter optimisation alone cannot address the emergent coupling behaviours of large multi-IBR installations.

For R&D teams and IP professionals tracking this space, the PatSnap platform provides access to more than 2 billion data points across 120+ countries, enabling landscape analysis of the full assignee and citation network around weak-grid inverter control — accessible via PatSnap Eureka and the broader PatSnap Analytics suite.