Study on applying interactive multi-objective optimization method for power plant system design

AbstractIn projects of social infrastructure systems such as energy and environment, water supply and sewerage, roads, railways, airports, urban and building, and information communication where experts from different fields participate, various opinions and ideas are given at the upstream design phase, and the system design often becomes increasingly complicated. This study was conducted to overcome such problems; that is, an optimization method that can be considered for the overall system design was established to improve the system performance and achieve appropriate balance among various evaluation indices while satisfying safety requirements and avoiding over-engineered of the entire system. Moreover, in this study, we considered that it is possible to create a completely new technical innovation by accepting diverse system requirements.For example, thermal and nuclear power plants can be described as large-scale systems having a complicated hierarchical structure as they are composed of many systems, machines, and components. In the system design of such a power plant, naturally, the technical development of individual machine and component is important. Furthermore, a system design that captures the overall characteristics of the plant system and its design optimization is essential. The evaluation indices in the system design of power plants are diverse. Specifically, on the premise of ensuring safety of life and property, the evaluation indices have multiple objectives taking many factors into consideration such as maintainability, environmental feasibility, economic efficiency, construction workability, operational controllability, and reliability. These evaluation indices are treated as multi-objective optimization problems. Although the capabilities of computers and optimization tools have improved, we are often confronted with the difficulty of solving problems in actual plant system design. The reasons for the difficulty include the mathematical modeling of the design object; however, the problem of decision-making by the designer is equally important. In engineering design tasks for a power plant, design engineers with a variety of expertise such as mechanical, electrical, chemical, control, information, architectural, and civil work together in the project. In addition, engineering design, from basic planning to construction, has many processes and takes several years to complete. Design engineers are required to make appropriate value judgment according to design processes taking the design progress and market changes into consideration.The system design of a power plant must satisfy safety requirements and take into consideration a plurality of evaluation indices. In addition, features such as the need for flexible decision-making according to design progress and market changes should also be considered. Such advanced engineering designs are mostly dependent on the expertise and intuition of skilled designers, which may increase the project cost and risk. Consequently, the problem of how to carry out the overall system optimal design has not been addressed.Therefore, the aim of this doctoral research is to develop an interactive multi-objective optimization system that supports multiple-criteria decision-making for the system design of a power plant through interaction between a designer and a computer. This system consists of a simulation model that calculates various evaluation indices of the plant system design and an interactive multiple-criteria decision-making system that implements a multi-objective optimization algorithm. The basic planning and design of the plant system can be categorized into three major design problems: process design, where heat balance calculations of the process flowsheet are carried out; operation design, where a control strategy is developed; and layout design, where a plot plan (general arrangement drawing) is created. Since the process design can be mostly evaluated objectively (non-subjectively), we focused on the remaining two design problems in this research, namely operation design and layout design, which are likely to reflect the designers’ (or operators’) subjective decision. Further, since we intend to use this system in the upstream design phase (for instance, basic planning and design), which has high degree of design freedom for decision-making, the system should be able to examine numerous design candidates in a limited time frame using simple design data. We verified the usefulness of the proposed system by evaluating some design problems faced in actual power plants. The proposed method does not require the effort of registering design knowledge and rules on a computer as in the conventional method, and it has a feature that easily reflects the thinking and value judgment of designers in the resulting optimum design. Moreover, there is no need to comprehensively calculate Pareto-optimal solutions beforehand as in the conventional method; the proposed method is superior in terms of its responsiveness to changes in the market environment and plant characteristics.

Abstract In projects of social infrastructure systems such as energy and environment, water supply and sewerage, roads, railways, airports, urban and building, and information communication where experts from different fields participate, various opinions and ideas are given at the upstream design phase, and the system design often becomes increasingly complicated. This study was conducted to overcome such problems; that is, an optimization method that can be considered for the overall system design was established to improve the system performance and achieve appropriate balance among various evaluation indices while satisfying safety requirements and avoiding over-engineered of the entire system. Moreover, in this study, we considered that it is possible to create a completely new technical innovation by accepting diverse system requirements. For example, thermal and nuclear power plants can be described as large-scale systems having a complicated hierarchical structure as they are composed of many systems, machines, and components. In the system design of such a power plant, naturally, the technical development of individual machine and component is important. Furthermore, a system design that captures the overall characteristics of the plant system and its design optimization is essential. The evaluation indices in the system design of power plants are diverse. Specifically, on the premise of ensuring safety of life and property, the evaluation indices have multiple objectives taking many factors into consideration such as maintainability, environmental feasibility, economic efficiency, construction workability, operational controllability, and reliability. These evaluation indices are treated as multi-objective optimization problems. Although the capabilities of computers and optimization tools have improved, we are often confronted with the difficulty of solving problems in actual plant system design. The reasons for the difficulty include the mathematical modeling of the design object; however, the problem of decision-making by the designer is equally important. In engineering design tasks for a power plant, design engineers with a variety of expertise such as mechanical, electrical, chemical, control, information, architectural, and civil work together in the project. In addition, engineering design, from basic planning to construction, has many processes and takes several years to complete. Design engineers are required to make appropriate value judgment according to design processes taking the design progress and market changes into consideration. The system design of a power plant must satisfy safety requirements and take into consideration a plurality of evaluation indices. In addition, features such as the need for flexible decision-making according to design progress and market changes should also be considered. Such advanced engineering designs are mostly dependent on the expertise and intuition of skilled designers, which may increase the project cost and risk. Consequently, the problem of how to carry out the overall system optimal design has not been addressed. Therefore, the aim of this doctoral research is to develop an interactive multi-objective optimization system that supports multiple-criteria decision-making for the system design of a power plant through interaction between a designer and a computer. This system consists of a simulation model that calculates various evaluation indices of the plant system design and an interactive multiple-criteria decision-making system that implements a multi-objective optimization algorithm. The basic planning and design of the plant system can be categorized into three major design problems: process design, where heat balance calculations of the process flowsheet are carried out; operation design, where a control strategy is developed; and layout design, where a plot plan (general arrangement drawing) is created. Since the process design can be mostly evaluated objectively (non-subjectively), we focused on the remaining two design problems in this research, namely operation design and layout design, which are likely to reflect the designers’ (or operators’) subjective decision. Further, since we intend to use this system in the upstream design phase (for instance, basic planning and design), which has high degree of design freedom for decision-making, the system should be able to examine numerous design candidates in a limited time frame using simple design data. We verified the usefulness of the proposed system by evaluating some design problems faced in actual power plants. The proposed method does not require the effort of registering design knowledge and rules on a computer as in the conventional method, and it has a feature that easily reflects the thinking and value judgment of designers in the resulting optimum design. Moreover, there is no need to comprehensively calculate Pareto-optimal solutions beforehand as in the conventional method; the proposed method is superior in terms of its responsiveness to changes in the market environment and plant characteristics.