Multiobjective optimization of temporal processes

Area of Science:

• Engineering
• Computer Science
• Applied Mathematics

Background:

• Nonlinear temporal processes require advanced modeling for efficient operation.
• Optimizing complex industrial processes like power plants presents significant challenges.
• Integrating data mining with optimization algorithms can unlock performance improvements.

Purpose of the Study:

• To present a novel dynamic predictive-optimization framework for nonlinear temporal processes.
• To demonstrate the framework's application in a power plant setting for efficiency and resource optimization.
• To address multiobjective optimization challenges through flexible problem transformation.

Main Methods:

• Development of a framework integrating data-mining (DM) algorithms and evolutionary strategy (ES).
• Utilizing DM algorithms to learn dynamic equations directly from process data.
• Applying ES algorithms to solve optimization problems informed by DM-extracted knowledge.
• Handling multiobjective optimization by transforming into single-objective or Pareto-optimal problems.

Main Results:

• The proposed framework effectively learns process dynamics and guides optimization.
• Demonstrated success in maximizing boiler efficiency and minimizing limestone consumption in a power plant.
• Computational results validate the framework's effectiveness for complex industrial optimization.
• The approach offers flexibility in handling multiobjective optimization scenarios.

Conclusions:

• The dynamic predictive-optimization framework is effective for nonlinear temporal processes.
• The integration of data mining and evolutionary algorithms provides a powerful optimization tool.
• The framework offers a practical solution for improving industrial efficiency and reducing resource consumption.