Bayesian Inference for Power Grid Situational Awareness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Extended blackouts due to cyberattacks and denial-of-service events compromise power grid operations, rendering traditional SCADA systems unreliable, making it difficult to maintain situational awareness and restore power effectively.
Innovation Solution
A method and system that utilize a mesh network with strategically deployed sensors and communication devices to generate a loop-free Bayesian inference model, providing situational awareness by inferring missing states and locating cyberattacks or faults, even when traditional systems are unavailable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional SCADA systems are used for monitoring power grid, then complete situational awareness can be achieved, but the system becomes vulnerable to cyberattacks and denial-of-service events
Solution Approach 1:
The patent segments the monitoring function by deploying distributed sensors across the power grid network rather than relying on a centralized SCADA system. Each sensor independently monitors its local node, and the collective data from segmented sensor nodes provides comprehensive situational awareness while eliminating the single point of failure vulnerability of traditional SCADA systems
Solution Approach 2:
The patent introduces Bayesian inference as an intermediary computational layer that processes sensor measurements and infers the states of unmonitored components. This intermediary enables the system to achieve complete situational awareness from partial observations, maintaining reliability while reducing the attack surface by not requiring direct monitoring of all grid components
2Loss of information
If sensors are deployed on all network nodes to achieve complete monitoring, then full situational awareness is obtained, but the cost and complexity of the system increases significantly
Solution Approach 1:
The patent uses Bayesian inference to create virtual copies of physical sensor measurements. By mathematically inferring the states of unmonitored nodes based on measurements from a subset of sensors, the system creates virtual measurement data that complements physical sensor readings, achieving complete information coverage without deploying sensors everywhere
Solution Approach 2:
The patent changes the monitoring approach from direct physical measurement at all nodes to indirect statistical inference. By transforming the problem from a physical sensor deployment challenge to a computational inference problem, the system achieves complete monitoring with fewer sensors by changing the fundamental parameter of how monitoring is accomplished
3Reliability
If a mesh network with multiple loops is used for sensor deployment, then system redundancy is improved, but the complexity of deriving situational awareness increases due to interdependencies
Solution Approach 1:
The patent dynamically adapts the Bayesian inference model to the specific mesh network topology. Rather than using a fixed complex model, the system dynamically constructs the inference relationships based on the actual network structure, allowing it to handle redundant loops efficiently by adjusting the inference calculations to match the dynamic topology
Data Source
AI summary
A computer-program product, a system, and a computer-implemented method include a processor(s) obtaining a configuration of a network including configurations of multiple network nodes and configurations of the network communication devices. The program code automatically models the network to generate a system model. The program code derives, from the system model, a loop-free Bayesian inference model, by generating a loop-free Bayesian network from the network.