應用進化策略演算法於冰水主機負載分配之最佳化

Optimal Chiller Loading Using Evolution Strategy

陳建進(Cheng-Chien Chen)

指導教授 ： 張永宗

國立臺北科技大學/機電學院/冷凍空調工程系所/碩士(2006年)

摘要

中央空調系統之耗電量與冰水主機運轉效率之高低關係密切，若能在滿足負載需求下使每部冰水主機依其特性曲線運轉於最佳點，則將消耗最少電力，此即OCL（Optimal Chiller Loading）。目前冰水主機的負載分配法有平均負載法、拉格蘭傑法、基因演算法，這些方法皆有其缺點。例如平均負載法並不是最佳運轉點、拉格蘭傑法λ值初值設定不當會有發散的問題。
雖然基因演算法（Genetic Algorithm , GA）可用來克服此缺點，但其求解過程需應用到複製、交配、突變等機制及編碼、解碼等運算，使程式的編寫相當麻煩，故本文提出進化策略(Evolutionary Strategy , ES)演算法來簡化其求解過程。由於ES 只採用突變及競爭技巧，省去了GA 方法的編碼、解碼及交配等操作，不僅可避免誤差產生，且程式編寫更加容易。另外，因為一個空調系統冰水主機數量不大，以（l+l）−ES來執行OCL 時，收斂速度已足以滿足所需，又節省了(μ +λ)− ES 中的選取及競爭的運算，故本文使用（l+l）−ES 方法。文以三次多項式來表現冰水主機之kW-PLR性能曲線，並以系統總耗電為目標函數，並在相關限制條件及滿足系統負荷需求下，應用進化策略演算法進行冰水主機負載分配最佳化使系統總耗電量為最小。

關鍵字

基因演算法 ； 進化策略演算法 ； 主機最佳負載分配 ； 拉格蘭傑乘數法

並列摘要

This indicates that overall functioning efficiency of air-conditioning system is closely related to the efficiency of Chiller unit. The purpose of the Optimal Chiller Loading (OCL) is to meet the system load and to decide the chillers’ optimal part load ratios (PLR) to reduce the system power consumption. The optimal chiller loading methods include Average Loading (AVL) method, Largrangian Multiplier (LGM) method and Genetic Algorithm (GA) at present. These methods have some shortcomings. Such as AVL method being not optimal. LGM method will diverge if the initial condition isn’t suitable.
Although the GA method overcomes the shortcoming of Lagrangian method and produces results with high accuracy, the process of evolution is very complicated and makes the coding of program be difficult. This paper presets a method by using Evolution Strategy （ES） to improve these defects. Since ES only use the mechanisms of mutation and competition, leaving out all other encoding, decoding and crossover operations in the GA approach, it thus can not only reduce error, but also enable easy implementation. Besides, because the number of chillers in an air-conditioning system is not large, the convergence speed is enough to meet the requirement and saves us from the computation of selection and competition in (μ +λ) −ES when using (1+1)-ES to conduct OCL. Thus, this research will use the (1+1)-ES method. This thesis uses a cubic equation to simulate the chiller’s kW-PLR curve and to find a set of chiller output which doesn’t violate the operating limits while minimizing the objective function. The ES is adopted to find the near optimal solution of the function.

並列關鍵字

Genetic Algorithm ； Evolution Strategy ； Optimal Chiller Loading ； Lagrangian Multiplier Method

參考文獻

[8] 黃博興，空調主機節約能源方法，中國冷凍空調雜誌，Jun，1992。

[15] 江建樺，應用模擬退火演算法於冰水主機負載分配之最佳化，碩士論文，台北科技大學，2005。

[19] 劉青龍，變頻離心式冰水主機之節能分析，碩士論文，台北科技大學，2005。

[3] S.C., Hu and Y.K., Chuah, "Power Consumption of Semiconductor Fabs in Taiwan Area," Energy, 28,2003, pp. 895-907.

[7] 李良梧，漫談建築、空調系統與節約能源，冷凍空調雜誌，Jun，1991，第101-104頁。