Optimization Algorithm Evaluation Function for Long-Term Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optimization algorithms for technical systems, such as water networks, struggle to effectively incorporate long-term constraints into shorter optimization periods without significantly increasing computation time or compromising solution quality.
Innovation Solution
The method involves setting up evaluation functions for long-term constraints, using penalty functions integrated into the objective function, allowing the optimization algorithm to consider long-term constraints without extending the optimization period, and using these functions to determine optimized control variables that satisfy both short-term and long-term conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the optimization period is extended to incorporate long-term constraints, then the ability to satisfy long-term constraints is improved, but the execution time of the optimization algorithm increases significantly
Solution Approach 1:
The patent divides the long-term constraint satisfaction problem into multiple short-term optimization periods. Each period optimizes control variables for a brief duration while incorporating evaluation functions that assess long-term constraint compliance. This segmentation allows the system to maintain reliability for long-term constraints without requiring a single lengthy optimization run, thus reducing execution time while preserving constraint satisfaction capability.
2Productivity
If conventional boundary conditions are used for short-term optimization, then the computation speed is improved, but the ability to model long-term constraints is lost
Solution Approach 1:
The patent introduces evaluation functions as intermediary mechanisms between short-term optimization and long-term constraint satisfaction. These evaluation functions are incorporated into the objective function of each short-term optimization period, acting as mediators that translate long-term constraint requirements into short-term optimization criteria. This allows conventional fast optimization algorithms to remain usable while gaining the ability to model and satisfy long-term constraints through the intermediary evaluation functions.
3Device complexity
If simple arithmetic operations are used to transform long-term constraints, then the computational effort is reduced, but the solution quality for the entire control period deteriorates
Solution Approach 1:
The patent transforms long-term constraints into evaluation functions that dynamically adjust parameters within the objective function. Rather than using simple static arithmetic transformations, the evaluation functions compute constraint violation metrics based on current system state and projected trajectories. This parameter change approach maintains relatively low computational effort while significantly improving solution quality by providing more accurate and state-dependent constraint assessments compared to simple arithmetic transformations.
Data Source
Figure 1~3
Figure 4~5
Figure 6
AI summary
A method is described for operating a system comprising components controllable by means of control variables, comprising the determining of at least one optimized control variable at an optimization period for minimizing or maximizing a target function of the system in such a way that at least one system variable fulfills predetermined secondary conditions with respect to the control period, wherein the determining comprises the establishment of an evaluation function for each of the secondary conditions to evaluate a change of the system variable at the end of the optimization period based on a current value of the system variable at the beginning of the optimization period, based on a gradient of the system variable relating to the secondary condition and maximally achievable over the control period, and based on a gradient relating to the secondary condition and minimally achievable over the control period, and the application of an optimization algorithm for minimizing or maximizing the target function using the established evaluation functions, and the adjustment of the at least one determined optimized control variable for operating the controllable components.