Segmented Plant Control Learning for Faster Parameter Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing plant control systems face challenges in optimizing operation states due to a broad learning range and large number of parameters, leading to unrealistically long times for converging on optimal parameter values or non-convergence during artificial intelligence learning.
Innovation Solution
The plant control supporting apparatus divides the plant into segments, refines parameters by deleting non-effective parameters, and uses a relationship model to facilitate convergence, allowing for efficient learning and optimization of parameter values through artificial intelligence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If artificial intelligence learning is applied to optimize plant operation states, then automation and optimization capability are improved, but the learning time becomes unrealistically long or convergence fails due to broad learning range and large number of parameters
Solution Approach 1:
The plant is divided into multiple segments based on process flow or functional groups. Each segment contains a subset of parameters that are learned independently. This segmentation reduces the overall learning scope from the entire plant-wide parameter set to smaller, manageable segment-level parameter sets, thereby significantly reducing learning time while maintaining optimization capability.
Solution Approach 2:
Parameter refinement is performed in advance by deleting non-effective parameters before the AI learning process begins. This preliminary action reduces the number of parameters that need to be learned, preventing the learning process from being overwhelmed by irrelevant parameters and thus reducing convergence time.
2Manufacturing precision
If all parameters are used for AI learning, then comprehensive optimization is achieved, but the number of parameters becomes too large causing learning non-convergence
Solution Approach 1:
Non-effective parameters are extracted and removed from the parameter set through a refinement process. This extraction eliminates parameters that do not contribute meaningfully to optimization, reducing the overall parameter count while retaining the essential parameters needed for achieving optimization precision.
Solution Approach 2:
The parameter set is segmented into segment-specific parameters rather than treating all parameters globally. Each segment's parameter subset is learned independently, reducing the complexity of the learning task while maintaining comprehensive optimization across all segments through aggregation of segment results.
3Reliability
If plant-wide optimization is performed simultaneously, then global optimality is achieved, but learning convergence fails due to excessive parameter interactions
Solution Approach 1:
The plant optimization problem is segmented into independent or loosely-coupled sub-problems corresponding to different process segments. Each segment is optimized independently through AI learning, avoiding the computational intractability of simultaneous plant-wide optimization while still achieving near-global optimality through coordinated segment optimization.
Solution Approach 2:
Parameter refinement and segment definition are performed in advance before the optimization learning begins. This preliminary structuring of the optimization problem establishes a framework that enables efficient learning convergence while maintaining the capability to achieve global optimality through the coordinated results of segment optimizations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A plant control supporting apparatus includes a segment selector configured to select, from among a plurality of segments defined in a plant, a segment for which learning for acquiring an optimal value of at least one parameter representing an operation state is executed, a reward function definer configured to define a reward function used for the learning, a parameter extractor configured to extract at least one parameter that is a target for the learning in the selected segment on the basis of input and output information of a device used in the plant and segment information representing a configuration of a device included in the selected segment, and a learner configured to perform the learning for acquiring the optimal value for each segment on the basis of the reward function and the at least one parameter.