Power Device Digital Twin Hybrid Modeling With Knowledge Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power grid device simulation modeling methods face challenges such as low modeling precision, low success rate, and limited practical application range due to difficulties in determining intermediate values, failing to learn underlying dynamics, and lacking a priori knowledge base, leading to inaccurate real-time simulation results.
Innovation Solution
A method involving generating a density vector from a knowledge graph of power system device associations, embedding it into a ConvGRU network for initial training, optimizing a convolutional base and classifier with historical data, and operating the data-driven model in parallel with a mechanism model to construct a digital twin hybrid model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a serial form hybrid model is constructed, then the modeling process can be simplified, but the modeling precision deteriorates due to difficulty in determining intermediate values
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary component that bridges the mechanism model and data-driven model. The knowledge graph encodes domain knowledge about power system devices, fault patterns, and relationships between operating states and faults, providing intermediate representations that facilitate information exchange between the two modeling approaches without requiring direct determination of difficult intermediate values.
Solution Approach 2:
The patent replaces the traditional mechanical serial processing approach with a parallel hybrid architecture. Instead of sequentially processing data through mechanism model then data-driven model (or vice versa), both models operate in parallel and their outputs are fused through the knowledge graph, eliminating the bottleneck of intermediate value determination while maintaining process simplicity.
2Measurement precision
If a parallel form hybrid model is constructed, then the degree of fitting is improved, but the adaptability deteriorates due to limitation to simple systems
Solution Approach 1:
The patent designs the knowledge graph with universal structures and relationships that can accommodate various power system devices (transformers, generators, transmission lines, etc.) and different fault types. The knowledge graph serves multiple functions: storing domain knowledge, facilitating inference, and enabling the hybrid model to adapt to complex systems beyond its initial training scope, thereby expanding practical application range.
Solution Approach 2:
The patent implements dynamic knowledge graph construction that adapts to different operating conditions and fault scenarios. The knowledge graph can be updated and extended based on new data and emerging fault patterns, allowing the model to dynamically adjust to complex systems and varying operational contexts rather than being static and limited to simple predefined scenarios.
3Device complexity
If a tight-coupling form hybrid model is constructed, then the modeling structure is simplified, but the reliability deteriorates due to difficulty in converging during training
Solution Approach 1:
The patent segments the training process into distinct phases: knowledge graph construction, mechanism model training, data-driven model training, and integration. This segmentation allows each component to be trained and validated independently before integration, reducing the complexity of the overall training process and improving convergence reliability by avoiding the need to train all components simultaneously.
Solution Approach 2:
The patent performs preliminary construction of the knowledge graph and preliminary training of individual models before integrating them into the final hybrid system. This preliminary action ensures that each component is properly prepared and validated, reducing training convergence difficulties and improving overall modeling success rate when the complete system is assembled.
Data Source
AI summary
Disclosed in the present invention are a method and apparatus for constructing a digital twin hybrid model of a main device of a power system. The method comprises: performing information coding on an operation state and fault association knowledge graph of a main device of a power system to generate a density vector; embedding the density vector into a first layer of a ConvGRU neural network for performing initial training of a data-driven model to obtain a data-driven initial model; optimizing a convolutional base and a classifier of the data-driven initial model by means of historical collected data to obtain an optimized data-driven model; and enabling the data-driven model and a mechanism model of the main device of the power system to cooperatively operate in parallel, and performing gradient descent of a loss function of the neural network by means of a solver, so as to construct a digital twin hybrid model of the main device of the power system.