Simulated annealing based optimal chiller loading for saving energy

doi:10.1016/j.enconman.2005.12.022

Energy Conversion and Management

Volume 47, Issues 15–16, September 2006, Pages 2044-2058

https://doi.org/10.1016/j.enconman.2005.12.022 Get rights and content

Abstract

This study employs simulated annealing (SA) to solve the optimal chiller loading (OCL) problem. SA overcomes the flaw that the Lagrangian method is not adaptable for solving the OCL problem as the power consumption models or the kW–PLR (kilo Watt–partial load ratio) curves include non-convex functions. This study uses the chilled water supply temperature as the variable to be solved for the decoupled system. After analysis and comparison of the case study, it has been concluded that this method not only solves the problem of the Lagrangian method but also produces results with high accuracy within a rapid time frame. It can be perfectly applied to the operation of air conditioning systems.

Introduction

Taiwan is located in the subtropical region where the summer is hot and humid. The tremendous usage of air conditioning equipment during the summer has caused a great need for power consumption. According to research [1], air conditioning uses 30% of the total power and up to 40% during peak periods. Of all the air conditioning users, the semiconductor industry has the greatest need. Commercial usage of air conditioning rarely reaches 10,000 RT (refrigeration tons), whereas in a semiconductor plant, it usually exceeds 10,000 RT. The output of IC (integrated circuit) foundries in Taiwan is about 70% of the world wide output. The packing testing industry of Taiwan owns 30% of the global market share. Taiwan stands in the number one place in the world for both industries. The IC R&D (research and development), design and innovation industry in Taiwan has about 20% of the global market share, just after the US and Japan. Taiwan has successfully integrated the supply chain of the semiconductor industry. According to a study of IT (information technology) suppliers by ITRI (Industrial Technology Research Institute) Electronics Research and Service Organization, the large investment plans of the Taiwan semiconductor manufacturers to be realized from 2003 to 2010 include 7 eight-inch wafer foundries and 36 twelve-inch wafer foundries in Hsinchu Science Based Park, Tainan Science Based Park and Luchu Science Based Park. One IC foundry with a monthly output of 35,000 six-inch wafers needs 23,000 W of electricity every day, 35,000 W per day for an eight-inch wafer foundry and 70,000 W per day for a twelve inch wafer foundry. As seen, the demand for power is significant.

On average, the facility system consumes 56.6% and the production equipment consumes 40.4% of all the power consumption in a semiconductor plant. Furthermore, chiller units use most of the power, about 27% of all power consumption [1]. If there is a way to reduce the chiller unit power consumption, the result can be extremely advantageous to the manufacturers, power supply safety and to the semiconductor industry as a whole. To maintain efficiency, centralized air conditioning systems often utilize centrifugal chiller units. Each unit has different features. The longer the machine runs, the more apparent the differences are. Inappropriate operation will waste power. When the usage is substantial, like the semiconductor industries, the effects will be profound.

The common way to allocate the chiller unit capacity in air conditioning systems has been rather simple (setting equal chilled water temperature (ECHWT) to all on line units). To solve the OCL problem, the Lagrangian method has been adopted in Refs. [2], [3] based on the convex function of the power consumption model [2] or the kW–PLR curve [3]. However, the Lagrangian method is not adaptable for solving the OCL problem as the power consumption model or the kW–PLR curves include non-convex functions [4]. This study solves the problem by using SA to overcome this shortcoming. After extensive experiments, SA has been modified to provide results with high accuracy within a rapid time frame.

Section snippets

System structure

The designers of air conditioning systems often develop multiple chiller systems because they provide operational flexibility, standby capacity and less disruption by maintenance. Such a system has a reduced starting current, reduced power cost under partial load conditions and a set of chillers can be operated at the best efficiency [5]. Fig. 1 depicts the structure of a decoupled chilled water system (called a decoupled system), which includes multiple chillers [5]. A decoupled system offers

Conventional OCL methods

ASHRAE proposed the concept of OCL in Chapter 40 of a 1999 Handbook [2] in which the power consumption model of the chiller is shown with the cooling water return temperature (T_cwr,i), chilled water supply temperature (T_chws,i) and chiller load (Q_ch,i) as independent variables as $P_{ch, i} = k_{0, i} + k_{1, i} (T_{cwr, i} - T_{chws, i}) + k_{2, i} (T_{cwr, i} - T_{chws, i})^{2} + k_{3, i} Q_{ch, i} + k_{4, i} Q_{ch, i}^{2} + k_{5, i} (T_{cwr, i} - T_{chws, i}) Q_{ch, i}$ where k₀–k₅ are regression coefficients.

Then, the OCL problem is to find a set of chiller outputs that do not violate

Simulated annealing technique

The Metropolis algorithm was first put forward in 1953 [10], and it was not until 1983 that the SA technique was employed by Kirkpatrick et al. [11] to seek an optimal combination based on the Metropolis algorithm. It borrowed its theory from the gradual cooling process of metal annealing after having been heated to the melting point. At high temperature, the atoms are embodied with great amounts of energy, and move arbitrarily in any state. As the temperature drops, the energy of the atoms

Conclusion

With more traditional centralized air conditioning systems (such as in office buildings, hotels and hospitals), the total freezing capacity is not very substantial, the number of units is low, the operation method is simple and the OCL is not well thought out. As the requirement increases due to the development of the semiconductor industry, however, the freezing capacity and the number of units multiply. The OCL of a chiller unit has to be taken seriously. Each unit has to run with highest

Acknowledgement

The authors would like to thank the National Science Council of the Republic of China for financially supporting this research under Contract No. NSC 91-2212-E-027-009.

References (16)

S.C. Hu et al.
Power consumption of semiconductor fabs in Taiwan area
Energy
(2003)
Y.C. Chang
A novel energy conservation method-optimal chiller loading
Electric Power Syst Res
(2004)
Y.C. Chang
Genetic algorithm based optimal chiller loading for energy conservation
Appl Therm Eng
(2005)
ASHRAE Handbook, Supervisory control strategies and optimization, CH. 40,...
W.L. Winston
Operations research: Applications and algorithms
(1987)
ASHRAE Handbook, Liquid chilling system, CH. 38,...
Roy J. Dossat
Principles of refrigeration
(1991)
M.J. Bitondo et al.
Chiller plant control
(1999)

There are more references available in the full text version of this article.

Cited by (44)

A new approach for optimal chiller loading using an improved imperialist competitive algorithm
2023, Energy and Buildings
A new optimization algorithm based on an improved imperialist competitive algorithm (ICA-DE) is proposed to reduce the energy consumption of a multi-chiller system. The assimilation process is very important in ICA. Colonies of empires move randomly toward imperialists during the assimilation process in ICA, which decreases population diversity and can lead to premature convergence. Therefore, to create more new locations for colonies and increase population diversity, the idea of differential mutation proposed by differential evolution (DE) was applied to ICA. The novel ICA-DE method was used to distribute the partial load rate (PLR) of chillers to solve the optimal chiller loading (OCL) problem, with the goal of minimizing the energy consumption of multiple chillers. The established model was experimentally verified in an actual multi-chiller system in a building, and the results showed that the ICA-DE method can achieve good results. Finally, the ICA-DE method was compared with GA, DE, PSO, DCSA and ICA in two different test systems. The results show that the ICA-DE method has superior performance.
A novel data-driven optimal chiller loading regulator based on backward modeling approach
2022, Applied Energy
Citation Excerpt :
From the experimental results, it is evident that the operation and power consumption model established using the big data-driven OCLR has very high performance. We have also carried out a comparative evaluation of energy saving capability of the proposed OCLR with Equal Chilled Water Temperature (ECHWT) method in this work [29]. ECHWT is commonly used for chiller loading distribution in many factories.
This paper proposes a new method termed backward modeling approach (BMA) to achieve optimal chiller loading (OCL) for reducing energy consumption in industries running multiple-chillers with different efficiency. The developed OCL regulator (OCLR) based on novel BMA approach is composed of conditional generative network (cGAN) and deep neural network (DNN). Most works on the optimal chiller loading problem are to find out the setting of partial load rate (PLR) for each chiller. However, PLR for each chiller cannot be controlled directly and can only be achieved through setting chilled water supply temperature. A novel feedback control framework was developed to identify the relationship between chilled water supply temperature and the PLR. In light of this, the control instruction for chilled water supply temperature can be set to achieve the desired energy saving. The practical feasibility of loading optimization based on developed OCLR was evaluated by conducting field validation for 1 year in a reputed panel manufacturing factory based in Taiwan running multiple-chiller system. From the experimental results, it is evident that the developed data-driven OCLR based on BMA has very high performance and was able to conserve significant energy in the range of 81.9 MWh to 198 MWh per year.
Optimizing multi-chiller dispatch in HVAC system using equilibrium optimization algorithm
2021, Energy Reports
The HVAC (heating, ventilation and air-conditioning) system accounts for almost half of the total electricity consumption in buildings, and the chillers in HVAC system are the major energy consumers. Hence, it is significant to reasonably control the chillers for satisfying the thermal comfort demand of occupancy according to indoor environment change, and avoiding excessive cooling output and saving energy consumption. Equilibrium optimization (EO) algorithm is a novel physics-based meta-heuristics algorithm which originates from the dynamic mass balance equation. This paper attempts to optimize chiller load dispatch of HVAC system for energy conversation with EO algorithm. To validate the effectiveness of EO algorithm in solving optimal chiller loading (OCL) problem, three well-known cases are utilized to implement simulation experiments. Moreover, the effects of different particle number and parameter setting of EO algorithm on the optimization results are discussed. The experiment results demonstrate that EO algorithm can find out the optimal solution in solving OCL problem with less particles (the minimum equals 10). The optimal solutions are obtained in less than 200 iterations in case 1, 150 iterations in case 2, 50 iterations in case 3, respectively. Comparing with traditional optimization algorithms, the optimum power consumption using EO algorithm can obtain the best solutions, and simultaneously is better than AVL (Average Loading) and SA (Simulated Annealing) algorithm in case 1, AVL and GA (Genetic Algorithm) algorithm in case 2, AVL and GA algorithm in case 3, respectively. Moreover, Friedman test and Wilcoxon signed ranks test are also performed to prove efficiency of EO algorithm in solving OCL problem. In conclusion, EO algorithm is suitable for solving OCL problem and has the good performance.
Artificial intelligence techniques in refrigeration system modelling and optimization: A multi-disciplinary review
2021, Sustainable Energy Technologies and Assessments
Citation Excerpt :
Fig. 5 represents the major steps of SA algorithm. The SA-based optimal chiller loading (OCL) problem is solved by Chang et al. [101] where they minimized energy consumption and considered chilled water supply temperatures as a decision variable. Their designed SA overcomes the problem of the Lagrangian method in the case of solving the non-convex kilo Watt–partial load ratio (kW–PLR) model and provides almost similar results obtained by GA.
Refrigeration systems currently utilize 17% of total electric energy, and this consumption is expected to increase by more than 30% by 2050. Moreover, these systems significantly contribute to global warming (10%) and environmental degradation. Optimization of these systems can overcome these critical issues, increase thermal efficiency, and decrease the total cost. Refrigeration system optimization problems are complex, multi-modal, non-linear, and time-consuming. Both classic and non-classic (computational intelligence (CI)) methods are successfully applied to overcome these challenges. This comprehensive review presents state-of-the-art theory and application of the most widely used CI techniques such as GA, PSO, SA, DE, HTS, CRO, MOGA, and NSGA II in the optimization of various refrigeration systems. The properties of these algorithms, their computational efficiency, robustness, and applications are highlighted most. Additionally, the authors discuss several surrogate modelling techniques and their applications in refrigeration systems, including ANN, RSM, and regression analysis. According to trend analysis, the major cost function to optimize is the COP, followed by total cost, exergetic efficiency, energy consumption, and cooling capacity. Subsequently, the application of CI approaches has increased dramatically in the optimization of refrigeration systems, where GA and its variants are typically used, and authors are more interested in multi-objective optimization.
Optimal consumption modeling of multi–chiller system using a robust optimization algorithm with considering the measurement, control and threshold uncertainties
2020, Journal of Building Engineering
Citation Excerpt :
In Ref. [21], the researcher analyzed the effect of gradient method (GM) in finding the better functioning points of the chillers considering less power usage than those defined through the Lagrangian optimization procedure in less calculation time. The new algorithm is named simulated annealing (SA) algorithm which is worked un two-step, tries to heat the material and after it is started to cool the material to reduce defects, removes the limitations of the Lagrangian algorithm in solving the non-convex OCL issues [22,23]. Augmented group search optimization method [24], improved ripple be swarm optimization method [25], and invasive weed optimization (IWO) [26] algorithm have been introduced via scholars to optimize using of electricity in OCL procedure.
Recently, one of the best options for reducing energy consumption is optimizing energy management of multi-chiller systems in large scale and industrial plants. Different control methods have been implemented to achieve the best performance of the chiller plant but these methods not supply the cooling adequacy for indoor thermal comfort. In order to solve this problem, an optimal chiller loading is controlled in this paper, by applying the conventional and Lagrange methods. Since multi-chiller systems have some uncertainty elements, achieving a precise result is impossible. Three important parameters that cause uncertainty are measurement, threshold, and cooling demand. Using some various load characteristics, a novel hybrid sequential control method is proposed. With employ the historical data, in order to prove the superiority of the proposed hybrid control, its results have been compared with other control methods in the presence of different uncertainties. In the following, to achieve more accurate and optimal results, the controller parameters using a robust method have been improved. Since chiller plant optimization is a mixed-integer nonlinear problem. Then, a new algorithm, known as the camel traveling behavior algorithm, is used to obtain the best results. By analyzing the results, it is seen that the suggested method has the ability to compute in the inexact location.
Energy efficient hourly scheduling of multi-chiller systems using imperialistic competitive algorithm
2020, Computers and Electrical Engineering
The hour-ahead optimization of the multi-chiller systems is necessary in order to minimize their power consumption and save energy. In this paper, an imperialistic competitive algorithm based optimal chiller loading problem is solved to find the partial load ratio, cooling flux, and the electricity requirement of the chillers and minimize the energy consumption of the whole system. The proposed approach simulates the socio-political competition using the initial population composed of the colonies and imperialists. The empires with their relative colonies and imperialists compete with each other to collapse the weakest empire and dominate the most powerful one achieving the minimum power consumption. The cooling capacity of the chillers and balance constraint are satisfied in hourly economic dispatch process. The optimization problem is tested on three standard systems. The imperialistic competitive algorithm based optimal chiller loading method finds better operating points for chillers than those obtained by recently published strategies.

View all citing articles on Scopus

View full text

Simulated annealing based optimal chiller loading for saving energy

Abstract

Introduction

Section snippets

System structure

Conventional OCL methods

Simulated annealing technique

Conclusion

Acknowledgement

Energy

Electric Power Syst Res

Appl Therm Eng

Operations research: Applications and algorithms

Principles of refrigeration

Chiller plant control