Neural Network Handover Module for Incomplete Grid State Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for controlling power grids require high computational effort, are inaccurate under low measurement density, and suffer from error propagation, especially in critical grid conditions, making them unsuitable for edge nodes with limited resources.
Innovation Solution
A training method for an artificial neural network that uses synthetic training data to map network states and determine control commands without state estimation, enabling efficient and reliable control with minimal measurements, using a grid transfer module with a communication, memory, and computing unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional state estimation and load flow parameter calculation methods are used, then control accuracy can be maintained under normal conditions, but computational effort becomes prohibitively high for edge nodes with limited resources
Solution Approach 1:
The patent replaces the conventional mechanical/computational system of state estimation and load flow calculations with an artificial neural network-based system. The neural network is trained offline to learn the complex relationships between measurements and grid states, enabling real-time control decisions at edge nodes without requiring intensive online computational resources, thus resolving the contradiction between accuracy and computational effort
Solution Approach 2:
The neural network is trained in advance using synthetic training data that maps various grid states and control scenarios. This preliminary training phase performs the computationally intensive work offline, allowing the deployed network to make rapid control decisions at edge nodes with minimal real-time computational burden, thereby maintaining accuracy while reducing online computational effort
2Productivity
If state estimation is performed with insufficient measurements, then control can continue operation, but estimation quality and reliability deteriorate
Solution Approach 1:
The patent uses synthetic training data that copies and represents various grid states and measurement scenarios. During training, the neural network learns to infer complete grid states from partial measurements by pattern recognition, enabling reliable control decisions even when actual measurements are insufficient, thus maintaining both operational continuity and estimation quality
3Reliability
If multiple control methods (pseudo-value calculation, state estimation, load flow optimization) are coupled, then comprehensive control coverage is achieved, but error propagation increases and system complexity rises
Solution Approach 1:
The patent merges the functionality of multiple separate control methods (pseudo-value calculation, state estimation, load flow optimization) into a single unified neural network model. The network is trained to simultaneously perform all these functions, eliminating the need for separate implementation of each method and their complex coupling, thus reducing system complexity while maintaining comprehensive control coverage
Solution Approach 2:
The neural network is designed as a universal controller that can handle various grid conditions and control scenarios through a single model. It performs state estimation, parameter calculation, and control optimization simultaneously, replacing the need for multiple specialized methods and reducing overall system complexity while maintaining broad control coverage
4Ease of operation
If conventional control methods are used with incomplete measured value implementation, then control can be implemented with available data, but error propagation becomes difficult to estimate and control reliability decreases
Solution Approach 1:
The neural network incorporates feedback mechanisms through its training process, where it learns from synthetic data representing various measurement completeness scenarios. The network adjusts its internal parameters to compensate for missing or incomplete measurements, providing robust control decisions even when measured value implementation is incomplete, thus maintaining both ease of operation and control reliability
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a training method for training an artificial neural network (ANN) for a network transfer module (10) for use in a network station (100), in particular in the form of a local network station or a transformer substation, wherein the artificial neural network (ANN) is trained using synthetic training data (SD) which serve to map possible network states during operation of the network section (A) and to determine control commands (SB) in the network section (A), wherein only selected parts of the synthetic training data (SD) are used to train the artificial neural network (ANN) in order to simulate the control commands (SB), in particular despite incomplete state data, preferably without a state estimate (SE) in the network section (A), and thus to avoid critical network states, comprising current and/or voltage deviations, in particular overcurrent, overvoltage and/or undervoltage.