Abstract
With the rapid development of integrated energy systems, the large-scale and high-permeability access of distributed generations (DGs) is transforming traditional distribution networks into active distribution networks (ADNs). This increased complexity also increases vulnerabilities to cyberattacks. It is a new challenge to evaluate the situation of an ADN to support grid control decision-making under potential attack conditions. This paper proposes a situation assessment method for ADNs considering cyberattacks. The method consists of two parts: 1) An index system including DG stability, security risk under cyberattacks, and traditional safety metrics. 2) An operating situation assessment method based on the entropy weight method and grey correlation analysis. The framework is verified using 12 attack scenarios on a modified IEEE 33-node standard distribution system.
Introduction: The Vulnerability of Modern Grids
Traditional distribution networks were characterized by simple, unidirectional power flow from substations to consumers. Modern ADNs are far more complex, featuring bidirectional flows from Distributed Generations (DGs) like solar and wind, alongside controllable loads (CLs) like smart appliances and EV chargers.
While this complexity improves efficiency, it creates a massive "attack surface." Malicious actors can exploit communication protocols to tamper with DG outputs or manipulate household loads, potentially causing voltage collapses or frequency oscillations. Existing grid assessment tools are often not designed to evaluate operational safety specifically under the lens of an active cyberattack. There is a critical need to translate "network anomalies" into "operational risk levels" to help grid operators make split-second decisions.
Methodology: A Multi-Dimensional Assessment Framework
The authors developed a two-part assessment engine designed to quantify grid health in real-time.
1. The Multi-Dimensional Index System
Instead of just tracking whether the lights are on, the system monitors three categories of data to provide a holistic view of the grid:
- Safety Indexes: Traditional metrics like voltage qualification rates and load rates.
- DG Stability: Metrics for output volatility and penetration rates.
- Cyber-Risk Indexes: A novel addition called "Mutation Severity." This measures sudden, unnatural spikes or drops in power supply and load that typically signal a malicious override rather than a natural fluctuation (like a cloud passing over a solar farm).
2. The Assessment Algorithm
The framework uses the Entropy Weight Method to objectively assign "importance" to each index, ensuring that if a specific metric starts behaving wildly, it automatically carries more weight in the final risk score. The core of the evaluation is Grey Correlation Analysis, which compares the current state of the grid against a "normal operation" reference scenario. This allows the system to assign a risk grade from 0 (Normal) to 5 (Critical).
Experimental Setup: The IEEE 33-Node Testbed
The study utilizes the IEEE 33-node standard distribution system as its testing ground. To modernize it for the ADN era, the authors modified the simulation by connecting Photovoltaic (PV) arrays to nodes 18 and 22 and adding micro gas turbine units to node 33.
Twelve distinct attack scenarios were established, ranging from simple DG removals to coordinated strikes that simultaneously cut power generation while spiking consumer demand.
Key Results and Insights
The framework successfully categorized the 12 attack scenarios into risk levels:
- Critical Threats (Level 5): Scenarios involving the synchronous removal of all DGs or coordinated attacks (cutting DGs while increasing load) reached the highest risk level.
- Location-Dependent Risk: The results highlight that where an attack happens matters. Removing a DG far from the main power source is significantly riskier than removing one close to it, as distant nodes have a more critical task in maintaining local power balance.
- Quantifiable Deviation: The correlation degree ($r$) provided a clear numerical scale. For example, a normal grid has an $r$ of 1.0, while the most severe attack dropped the correlation to 0.357.
Conclusion and Future Directions
This research reinforces that in modern infrastructure, cybersecurity and operational safety are no longer separate silos. While the current model focuses on data integrity attacks, future work will explore "Availability" attacks (like DDoS) and transition from static snapshots to situation projection—predicting where the grid will be in the next 10–30 minutes to provide even earlier warnings.
Summary updated