Probabilistic Fault Detection in Power Distribution Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Power distribution networks face challenges in efficiently detecting faults due to noisy sensor data and the computational impracticality of calculating marginal probability density functions at each location, especially in large-scale networks.
Innovation Solution
A system and method that utilize data input devices, a memory device, and a processing device with probabilistic fault detection algorithms to estimate the powered state of nodes and faulted state of distribution lines, assigning visual characteristics based on calculated probabilities, and employing message passing algorithms like the sum-product algorithm to infer the network state from collected data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensors are used to detect the powered state at many locations throughout the network, then fault detection capability is improved, but sensor noise reduces measurement precision
Solution Approach 1:
The patent combines data from multiple noisy sensors across the distribution network and merges this with probabilistic models to produce a unified, more reliable fault detection system. By aggregating information from multiple sources and combining it with prior knowledge through Bayesian inference, the system achieves better fault detection capability while compensating for individual sensor noise.
Solution Approach 2:
The system implements feedback through iterative probabilistic updating, where initial probability estimates are continuously refined based on incoming sensor data. The feedback loop allows the system to adjust probability distributions for node powered states and line fault states based on new measurements, improving measurement precision over time despite noisy individual sensor readings.
2Measurement precision
If marginal probability density functions are computed at each location in the network, then fault detection accuracy is improved, but computational complexity becomes impractical for large-scale networks
Solution Approach 1:
The patent segments the computation by dividing the distribution network into manageable components (nodes and lines) and computing probabilities locally at each segment rather than computing the full joint probability distribution across the entire network. This segmentation allows the system to maintain high fault detection accuracy while reducing computational complexity to practical levels for large-scale networks.
Solution Approach 2:
The system computes only the necessary marginal probability density functions for individual nodes and lines rather than the complete joint distribution. This partial computation approach provides sufficient fault detection accuracy for practical purposes while avoiding the exponential computational burden of calculating the full joint probability distribution across all network components.
3Measurement precision
If probabilistic models are used to account for sensor noise, then measurement precision is improved, but device complexity increases due to computational requirements
Solution Approach 1:
The probabilistic computation is segmented into local calculations at each node and line, where simple Bayesian updates are performed independently. This segmentation maintains measurement precision through proper probabilistic modeling while reducing overall device complexity by avoiding the need for complex centralized computation of the full joint probability distribution.
Solution Approach 2:
The system performs partial probabilistic computation by calculating only the necessary marginal probability distributions for fault detection rather than complete probabilistic characterizations. This partial action provides sufficient measurement precision for fault detection while keeping computational processing requirements at practical levels.
Data Source
AI summary
Systems and methods for detecting faults in a power distribution network are described In an aspect, the systems and methods determine a probability that each node of the network is powered and a probability that each distribution line in the network is faulted. In another aspect, the systems and methods determine the probabilities by transmitting a signal over a power distribution network with an active sounding system. In an additional aspect, the systems and methods determine the probabilities by utilizing collected data coupled to the power distribution network.