Multi-layer Granular Computing for Steel Gas Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting long-term gas generation and consumption in the steel industry face challenges with low accuracy and high computing costs due to the use of one-layer structures and iteration mechanisms, which are inefficient for handling large datasets and unstable performance.
Innovation Solution
A multi-layer granular computing structure is established with adaptive parameter determination using the Monte-Carlo method and parallel computing to construct long-term prediction intervals, optimizing information granularity and employing fuzzy modeling for improved accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Duration of action of stationary object
If a one-layer structure for assigning information granularity is used, then prediction intervals can be provided for a long period of time, but the determination of massive quantities of parameters results in unstable performance on computing cost and accuracy
Solution Approach 1:
The patent divides the one-layer granular computing structure into multiple layers, where each layer processes and determines parameters for a specific time horizon segment. This segmentation reduces the complexity of determining massive parameters at once, improving performance stability while maintaining long-term prediction capability.
Solution Approach 2:
The patent employs preliminary action by determining parameters for earlier time horizons before proceeding to later horizons. The multi-layer structure allows parameters to be determined in a sequential manner, where each layer builds upon the previous layer's results, ensuring stable and reliable parameter determination throughout the entire prediction period.
2Measurement precision
If iteration mechanisms are employed in SVM and statistical-based models, then prediction intervals can be constructed, but the models perform not well for long-term prediction and can only give accurate results in less than 60 data points
Solution Approach 1:
The patent changes the fundamental parameter of the modeling approach by transitioning from iterative methods (SVM, statistical models) to a non-iterative granular computing-based optimization model. This parameter change enables the model to achieve high accuracy for long-term predictions beyond the 60 data point limitation of traditional iterative methods.
Solution Approach 2:
The patent substitutes the mechanical iteration mechanism of traditional models with a direct optimization approach using granular computing. Instead of iteratively adjusting parameters to converge on a solution, the method uses optimization algorithms to directly determine the optimal granular structure and parameters, enabling efficient long-term prediction.
3Measurement precision
If a multi-layer granular computing model is used, then prediction accuracy can be improved, but the structure is directly related to the accuracy of the results and it is highly demanded to design a learning approach for efficiently and reasonably obtain the optimal structural parameters
Solution Approach 1:
The patent implements self-service by designing a learning approach where the multi-layer granular computing model automatically determines its own optimal structural parameters. The system uses reinforcement learning to adaptively adjust the number of layers, granularity levels, and other structural parameters based on prediction performance, eliminating the need for manual tuning and reducing structural complexity.
Solution Approach 2:
The patent introduces dynamics by making the granular computing structure adaptive rather than fixed. The number of layers, granularity levels, and parameter determination methods are dynamically adjusted based on the prediction task requirements and historical performance, allowing the model to optimize its structure for each specific prediction scenario.
Data Source
AI summary
The present invention belongs to the field of information technology, involving the techniques of fuzzy modeling, reinforcement learning, parallel computing, etc. It is a method combining granular computing and reinforcement learning for construction of long-term prediction interval and determination of its structure. Adopting real industrial data, the present invention constructs multi-layer structure for assigning information granularity in unequal length and establishes corresponding optimization model at first. Then considering the importance of the structure on prediction accuracy, Monte-Carlo method is deployed to learn the structural parameters. Based on the optimal multi-layer granular computing structure along with implementing parallel computing strategy, the long-term prediction intervals of gaseous generation and consumption are finally obtained. The proposed method exhibits superiority on accuracy and computing efficiency which satisfies the demand of real-world application. It can be also generalized to apply on other energy systems in steel industry.