Hybrid Optimization for Power Plant Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current system control methods in power plants often get stuck in local optima due to the non-linear and non-convex nature of physical models, making it difficult to find globally optimal operating conditions for efficiency and emission reduction, especially with changing environmental parameters.
Innovation Solution
A hybrid method combining gradient-based optimization with a random-based system, where a random generator is used to generate additional sets of control values, which are then evaluated and compared to find global optima, ensuring quick convergence and adaptation to changing conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If gradient methods are used to optimize control values, then convergence speed is improved, but the ability to find global optima deteriorates due to getting stuck in local optima
Solution Approach 1:
The patent combines gradient-based optimization methods with random-based optimization methods into a hybrid approach. The gradient method provides fast convergence by following the steepest descent direction, while the random method periodically introduces random perturbations to escape local optima. This merging allows the system to maintain fast convergence speed while improving the reliability of finding global optima in non-convex objective functions.
Solution Approach 2:
The optimization approach dynamically switches between gradient-based and random-based methods. During normal operation, the gradient method is used for efficient convergence. When stagnation is detected or at scheduled intervals, random perturbations are introduced to explore new regions of the search space. This dynamic adaptation allows the system to balance speed and reliability based on the current optimization state.
2Reliability
If random-based optimization methods are used, then the ability to find global optima is improved, but computational complexity and time consumption increase
Solution Approach 1:
Instead of using purely random optimization throughout the entire process, the patent applies random-based methods partially - only when needed to escape local optima or at specific intervals. The gradient method handles the majority of the optimization work for efficient convergence, while random perturbations are applied as corrective actions when stagnation is detected. This partial application of random methods reduces time consumption while maintaining the ability to find global optima.
3Reliability
If the number of control value sets is increased to improve global optimization, then optimization quality is improved, but computational effort increases
Solution Approach 1:
The patent applies different optimization strategies to different regions of the search space. In regions where the objective function is smooth and convex, the gradient method efficiently finds optima with minimal computational effort. In regions where local optima are suspected or when global optimization is critical, random-based methods are applied to generate additional control value sets. This localized application of different methods improves optimization quality without uniformly increasing computational effort across the entire search space.
Data Source
AI summary
The invention relates to a method for the installation control in a power plant, wherein a functional value of a target function (7) based on a physical model is generated for a plurality of sets of variables (15), from respectively a set of environment variables on the one hand and the respective set of variables (15) on the other hand, said functional value being allocated to the respective sets. According to the invention, the set of variables (15) is selected to be transmitted to a control device (21) of the power plant, whose allocated functional value complies with a predefined optimization criterion. The inventive method shall allow an improved operation of the power plant with respect to a provided optimization criterion, such as for example an improved efficiency or a reduction of emission with little technical complexity for control purposes. In addition to a starting set and a set determined on the basis of the starting set and the functional value allocated thereto by means of a gradient method, the number of sets of variables (15) further comprises a set selected by a random generator.