Closed-Loop Planning Under Uncertainty for Industrial Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive systems for industrial planning and scheduling under uncertainty, particularly in high-dimensional domains, face challenges in accurately generating optimized plans that meet sustainability goals like net zero carbon emissions, due to complex relationships between input variables and the presence of uncertain data, leading to inefficiencies and ineffective recommendations.
Innovation Solution
A computer-implemented method and system that generates optimized plans using a non-linear model predictive control approach, incorporating uncertainty quantification and feedback mechanisms to ensure accurate predictions and automatic reconfiguration of industrial plant operations, thereby addressing uncertainty and improving predictive accuracy and resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional predictive systems are used for industrial planning under uncertainty, then the system structure is simple, but the prediction accuracy and reliability deteriorate due to complex relationships between input variables and uncertain data
Solution Approach 1:
The patent introduces an optimization model as an intermediary between uncertain input data and planning predictions. This model processes uncertain data through structured optimization algorithms, transforming complex uncertain inputs into reliable optimized plans. The intermediary model handles the complexity of uncertain relationships while maintaining prediction accuracy.
Solution Approach 2:
The patent implements feedback mechanisms where prediction results and actual outcomes are compared, and the differences are used to refine future predictions. This closed-loop feedback system continuously improves prediction accuracy by learning from past performance and adjusting the optimization model parameters accordingly.
2Measurement precision
If comprehensive optimization is performed for all input variables, then the prediction accuracy improves, but the computational time and processing resources increase significantly
Solution Approach 1:
The patent segments the optimization process by dividing input variables into different groups or categories that can be optimized separately or in priority order. This segmentation allows the system to focus computational resources on the most critical variables first, achieving good prediction accuracy without processing every variable to the same level of detail, thus reducing overall computational time.
Solution Approach 2:
The patent applies partial optimization by focusing computational effort on the most influential uncertain input variables rather than uniformly optimizing all variables. By identifying and prioritizing key variables that have the greatest impact on prediction accuracy, the system achieves satisfactory results with reduced computational time and resources.
3Reliability
If uncertainty quantification is incorporated into the optimization model, then the reliability of predictions improves, but the computational complexity and resource requirements increase
Solution Approach 1:
The patent changes the parameter representation of uncertain data by using statistical parameters (such as mean and variance) to characterize uncertainty instead of requiring detailed probabilistic models for every variable. This parameter-based approach maintains prediction reliability by capturing essential uncertainty characteristics while reducing model complexity and computational requirements.
4Loss of information
If detailed planning predictions are generated for all scenarios, then the completeness of the plan improves, but the storage requirements and data processing needs increase
Solution Approach 1:
The patent extracts only the essential and most relevant information from comprehensive planning scenarios rather than storing and processing all possible details. By identifying and retaining key planning elements and uncertainties that significantly impact decision-making, the system maintains information completeness for critical aspects while reducing overall data storage requirements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present disclosure provide for improved optimized plan predictions. Such embodiments, utilize optimized model(s) that accounts for uncertainty in input data to the model. Some example embodiments, receive input data associated with one or more industrial plants. At least a portion of the input data may include uncertain input data. The noted example embodiments, generate uncertainty-based modification data, and apply the uncertainty-based modification data to the input data to generate updated input data. The noted example embodiments generate, based at least in part on applying the updated input data to an optimization model, predicted optimized plan. The predicted optimized plan comprises optimized plan data. Further, the noted example embodiments initiate the performance of one or more prediction-based actions based at least in part on the predicted optimized plan.