Neural Network Power System State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing state estimation methods in electric power systems face challenges with high time consumption, sensitivity to noise, and limited robustness, especially in large-scale systems, due to complex calculations and erroneous measurements, which hinder online implementation and accuracy.
Innovation Solution
A neural network-based state estimation method that processes electrical measurements into sequence data, using observable and unobservable state estimators updated in independent threads, with a multi-thread architecture to improve time efficiency and robustness, and extend observability by building connections between observable and unobservable states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional state estimation methods (WLS, WLAV) are used, then estimation accuracy can be achieved, but time consumption increases and online implementation becomes difficult
Solution Approach 1:
The patent segments the state estimation problem into two distinct parts: observable state estimation using PMU data and unobservable state estimation using neural networks. This segmentation allows each part to be optimized independently - the observable part uses efficient linear estimation while the unobservable part uses trained neural networks, avoiding the computational burden of conventional iterative methods on the entire system.
Solution Approach 2:
The patent applies preliminary action by pre-training neural networks offline using historical data and system models. This pre-training establishes the relationship between observable and unobservable states before real-time operation, enabling fast online estimation without requiring complex iterative calculations during actual state estimation.
2Productivity
If PMU data-driven linear state estimation (LSE) is used, then time efficiency is improved, but the method becomes sensitive to erroneous measurements and calculation complexity increases
Solution Approach 1:
The patent introduces neural networks as an intermediary between PMU measurements and final state estimates. The neural networks process the raw measurements and learned relationships to produce robust estimates, filtering out the effects of erroneous measurements while maintaining the time efficiency of direct estimation methods.
Solution Approach 2:
The patent changes the parameter representation by using neural network weights and biases (learned during training) instead of direct measurement-based calculations. This parameter transformation allows the system to maintain accuracy and robustness while achieving fast computation during online operation.
3Adaptability or versatility
If the number of PMU installations is increased to improve system observability, then more synchrophasor data is available, but device complexity and computational burden increase
Solution Approach 1:
The patent segments the estimation task based on observability - using simple linear estimation for observable states and neural network-based estimation for unobservable states. This segmentation allows the system to handle large numbers of PMUs efficiently by applying the appropriate method to each state, avoiding uniform complex processing across all states.
Solution Approach 2:
The patent creates a universal state estimation framework that can handle both observable and unobservable states using a unified neural network architecture. The same neural network structure serves multiple functions - estimating unobservable states directly and providing robust estimates even when measurement quality varies, reducing the need for separate processing systems.
Data Source
AI summary
State estimation in an electric power system includes acquiring electrical measurements from the electric power system at a reporting rate of the electrical measurements, processing the electrical measurements into sequence data including positive sequence data, processing positive sequence data by an observable state estimator to generate a plurality of estimated states including a plurality of estimated observable states, parameters of the observable state estimator being updated by a first training module in a first time thread, processing the plurality of estimated states by an unobservable state estimator to generate a plurality of estimated unobservable states, parameters of the unobservable state estimator being updated by a second training module in a second time thread independent of the first time thread, and outputting a plurality of final estimated states generated by concatenating the plurality of estimated observable states and the plurality of estimated unobservable states.