Plant Control System Reducing Model Input Dimensionality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In complex plants like thermal power plants, the high dimensionality of model inputs and large search spaces in reinforced learning methods make it difficult to perform online correction and re-learning of control algorithms within the control cycle, leading to slow learning speeds and potential biases in teacher data classification.
Innovation Solution
A plant control system that includes an operation signal generation part, a numerical calculation execution part, a simulation model, a learning part, and a pattern generation part to reduce model input dimensionality and enhance learning speed by generating pattern data and selecting optimal learning results, allowing for high-speed re-learning and accurate control algorithm correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reinforced learning method with model is used to automatically learn control algorithm, then control accuracy is improved, but learning time becomes too long to complete within control cycle
Solution Approach 1:
The patent segments the high-dimensional model input space into multiple subspaces based on pattern categories. Each subspace contains a subset of input dimensions relevant to specific control scenarios. This segmentation reduces the effective search space for learning in each category while maintaining comprehensive coverage of all control situations, enabling faster convergence within the control cycle.
Solution Approach 2:
The patent performs preliminary classification of model inputs into pattern categories before the actual learning process. By pre-organizing the input space and identifying relevant patterns, the system prepares the learning structure in advance, reducing the computational burden during the control cycle and enabling faster re-learning when plant conditions change.
2Measurement precision
If model input dimension increases to capture complex plant behavior, then control precision is improved, but search space increases making learning slower
Solution Approach 1:
The patent divides the high-dimensional model input into multiple pattern categories, each representing a specific control scenario or plant state. Within each category, only the relevant input dimensions are considered for learning, effectively segmenting the search space. This maintains high control precision by considering all inputs overall, while improving learning speed by reducing the simultaneous search space for each pattern.
Solution Approach 2:
The patent applies different learning focuses to different pattern categories based on local requirements. Each pattern category has its own optimized search space containing only the input dimensions relevant to that specific control scenario. This local optimization allows the system to maintain high precision for each specific situation while achieving faster overall learning through targeted search.
Data Source
AI summary
A plant control system includes: a numerical calculation execution part which calculates the operation characteristic of the plant; a model for simulating the plant control characteristic according to information on the numerical calculation result; a learning part which learns the plant operation method by using the model; a learning information database which stores learning information data on the learning part; a pattern generation part which generates pattern data expressing a state input based on the learning information data in the learning part with a smaller input number than the model input dimension; a pattern database which stores the pattern data generated in the pattern generation part; and a learning result determination part which selects a learning result having a preferable control effect from the learning result obtained by using a plurality of patterns.