Systematic Security Analysis of Sensors and Controls in PV Inverters: Threat Validation and Countermeasures

Fengchen Yang, Kaikai Pan, Chen Yan, Xiaoyu Ji, Wenyuan XuSensors 2025 · DOI

TL;DR

This paper reveals a critical security vulnerability in photovoltaic (PV) inverters: their internal sensors are susceptible to intentional electromagnetic interference (IEMI), even with existing protections. The authors demonstrate that this can lead to inverter shutdowns (DoS), physical damage, or reduced power output, posing a significant threat to grid stability as renewable energy sources grow. They also propose and evaluate three novel detection methods to counter these attacks.

What Problem Does This Paper Address?

As renewable energy sources (RES) like solar power become increasingly integrated into our grids, power inverters, which convert DC to AC, are becoming ubiquitous and essential. However, the security of these inverters, particularly their internal sensors that are crucial for reliable operation, has not been thoroughly investigated. Traditional electromagnetic compatibility (EMC) standards are designed to handle unintentional interference, but they may not be sufficient against intentional, targeted electromagnetic interference (IEMI). This gap leaves PV inverters vulnerable to attacks that could disrupt power generation, damage equipment, and destabilize the grid. The authors aim to systematically analyze these vulnerabilities, validate the threats on real-world systems, and propose effective countermeasures.

Methods

The core of this research involves a multi-pronged approach:

  • Vulnerability Analysis: The authors delve into the internal workings of PV inverters, focusing on how IEMI can affect analog voltage and current sensors. They explain mechanisms like parasitic capacitance, nonlinear rectification by op-amps, and asymmetric circuit designs that can be exploited.
  • IEMI Attack Scenarios: Three primary attack outcomes are defined and investigated:
  • Denial of Service (DoS): Causing the inverter to shut down.
  • Damage: Physically harming the inverter, potentially leading to burnout.
  • Damping: Reducing the inverter's output power below its capability.
  • Experimental Validation: The proposed attacks were tested on a development kit, six commercial single-phase and three-phase PV inverters, and a real-world microgrid. IEMI signals were emitted from distances of 100-150 cm with power levels up to 20 W.
  • Countermeasure Development: Three detection methods were proposed and evaluated:
  • Sensor Level: Leveraging the distributed effect of IEMI, where the noise appears differently before and after the transducer.
  • Model Level: Based on the energy conservation law of the inverter, detecting deviations from expected efficiency.
  • Combination Level: Utilizing a lightweight Convolutional Neural Network (CNN) to analyze sensor data and detect anomalies.

Datasets

  • Newly Introduced Benchmark: The paper introduces a new dataset for IEMI attack detection on PV inverters. This dataset was built using data from a Ti C2000 micro solar inverter under various normal operating conditions and simulated IEMI attack scenarios. It contains 5000 samples, with 29.5% labeled as positive (IEMI attack) and 70.5% as negative (no attack), split into training, testing, and validation sets.

Key Insights and Ideas

  • IEMI Bypasses EMC: Existing EMC standards, which typically cover frequencies up to 1 GHz, are insufficient against IEMI signals at higher frequencies (1 GHz and above), allowing them to bypass protections.
  • Sensors as the Weak Link: The analog sensors within PV inverters, despite being critical for operation, are surprisingly vulnerable to IEMI. Even low-power signals can cause significant deviations.
  • Controllable Attacks: The authors demonstrate that IEMI attacks are not just disruptive but can be precisely controlled using modulation techniques (like AM) to achieve specific outcomes (DoS, Damage, Damping).
  • Universality of Threat: The vulnerability extends to various types of Hall sensors (analog and digital) and across different commercial PV inverter models, suggesting a widespread issue.
  • Microgrid Impact: The research validates that these attacks can indeed destabilize a real-world microgrid, highlighting the broader implications for grid stability.
  • Detection is Crucial: While passive defenses are limited, proactive detection methods are essential. The proposed methods offer different trade-offs in terms of coverage, robustness, and computational cost.

Key Results

  • Successful Attacks: IEMI attacks were successfully demonstrated on six commercial PV inverters and a real-world microgrid.
  • Attack Parameters: A 20 W IEMI signal was effective at distances of 1 to 1.5 meters.
  • DoS: Inverters could be shut down by manipulating DC bus voltage sensors. For example, a 735 MHz signal at 10 W could trigger DoS on the DC side, and specific frequencies around 1000 MHz could trigger DoS on the AC side.
  • Damage: The DC bus capacitor could be made to break down, leading to physical damage, by manipulating the DC bus voltage.
  • Damping: PV inverter power output could be reduced by up to 68.75% (e.g., from 80W to 25W) by interfering with the MPPT algorithm.
  • Detection Performance:
    • Sensor Level (STD): Achieved a detection capability where STD exceeded normal conditions during IEMI attacks.
  • Model Level (Efficiency): Successfully detected Damping attacks, with the efficiency (σ) deviating significantly from the normal range (0.9-1).
  • Combination Level (CNN): Achieved high accuracy (94%), precision (96%), recall (97%), and F1 score (97%) in detecting IEMI attacks on the test set.

Limitations and Future Work

  • Attack Parameters: The effectiveness of IEMI attacks is subject to power and distance. Higher power allows for greater working distances.
  • Limited Scale: The current attack's impact is primarily limited to individual PV inverters or local microgrids. Large-scale grid outages are not directly achievable with this method alone.
  • No Feedback for Closed-Loop Attacks: The current work focuses on open-loop attacks. Future work could explore closed-loop attacks that utilize sensor feedback for more precise power manipulation.
  • Computational Cost: While the CNN model is computationally efficient, further optimization for resource-constrained microcontrollers is suggested.
  • Hybrid Defenses: The authors suggest that a hybrid approach combining passive and active defenses might offer the best protection.
  • Broader Applications: The findings and methodologies could be extended to analyze the security of other power electronic devices and inverters in different domains (e.g., electric vehicles, industrial machinery).

Summary updated

Abstract

As renewable energy sources (RES) continue to expand and the use of power inverters has surged, inverters have become crucial for converting direct current (DC) from RES into alternating current (AC) for the grid, and their security is vital for maintaining stable grid operations. This paper investigates the security vulnerabilities of photovoltaic (PV) inverters, specifically focusing on their internal sensors, which are critical for reliable power conversion. It is found that both current and voltage sensors are susceptible to intentional electromagnetic interference (IEMI) at frequencies of 1 GHz or higher, even with electromagnetic compatibility (EMC) protections in place. These vulnerabilities can lead to incorrect sensor readings, disrupting control algorithms. We propose an IEMI attack that results in three potential outcomes: Denial of Service (DoS), physical damage to the inverter, and power output reduction. These effects were demonstrated on six commercial single-phase and three-phase PV inverters, as well as in a real-world microgrid, by emitting IEMI signals from 100 to 150 cm away with up to 20 W of power. This study highlights the growing security risks of power electronics in RES, which represent an emerging target for cyber-physical attacks in future RES-dominated grids. Finally, to cope with such threats, three detection methods that are adaptable to diverse threat scenarios are proposed and their advantages and disadvantages are discussed.
BibTeX
@inproceedings{yang2025systematic,
  author = {Fengchen Yang and Kaikai Pan and Chen Yan and Xiaoyu Ji and Wenyuan Xu},
  title = {{Systematic Security Analysis of Sensors and Controls in PV Inverters: Threat Validation and Countermeasures}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25051493},
}

Loading related content...

Figure