異質性蜂巢式網路下再生能源小型基地台之容量最大化研究

Abstract

對汲取來的能源做有效率的利用在綠能蜂巢式網路中是一個重要的設計目標。在這篇論文裡，我們研究了使用再生能源的小型基地台在異質性蜂巢式網路的容量最大化問題。具體來說，在同時滿足能源因果與電池容量的限制下，我們嘗試最大化每一個從環境汲取能量(例如:太陽能與風能)的小型基地台的平均功率。在有傘細胞基地台的涵蓋保證下，我們研究藉由小型基地台的睡醒排程運作與功率控制達到容量提高的潛在可能。我們在對於這個混合整數最佳化問題做線性規劃鬆弛後提供一個容量最大化的緊湊上界並且設計一個能夠從線性規劃鬆弛後的結果找出可行解的方法。對於這個NP困難的混合整數最佳化問題，我們在基於推導出的最佳睡醒行程與功率分配的重要特性下，提出一個能得到近似最佳解的多項式時間啟發式演算法，動態能量延遲排程(DEDS)。只要小型基地台能夠配備足夠大容量的電池，由我們提出的演算法就能找到最佳的網路容量。此外，在電池容量小的情況下，線性轉二元方法也有很好的表現。在模擬結果下，與一個採用傾向保持醒來排程的基地台相比，我們呈現出我們所提出來的演算法與設計的方法能夠增加大約25%的系統容量。我們也發現甚至在有採用最佳功率分配的情況下，對於想要最大化網路容量，總是傾向將一個小型基地台維持在開啟的狀態可能不會總是一個好的策略，特別是對於一個在異構蜂窩網路下受限於干擾的小型基地台。睡醒機制排程能夠有效率的增加基地台容量。

Efficient energy utilization of harvested energy is a key design goal for green cellular networks. In this thesis, we investigate the capacity maximization problem for heterogeneous cellular networks with renewable energy harvested by small cells. Specifically, we try to maximize the average capacity of each small cell with the harvested energy from the environments (e.g., solar and wind) while satisfying the constraints on energy causality and battery capacity. With the coverage preserved by the umbrella cell, we investigate the potential of capacity improvement by joint sleep-awake scheduling operations and power control of small cells. We provide a tight upper bound for the maximum capacity by linear programming relaxation of the mixed-integer problem and design a method to derive feasible approximate solution from the results of linear programming relaxation. Based on the key properties derived for optimal sleep-awake schedules and power allocation patterns, we propose a heuristic polynomial-time near-optimal algorithm, Dynamic Energy Deferment Scheduling (DEDS), for this mixed-integer optimization problem which is NP-hard. The capacity obtained by our proposed heuristic algorithm can approach the maximum capacity as long as the small cell can be equipped with a battery with sufficiently large capacity. Also, the Linear-to-Binary method has good performance when the battery capacity is small. Compared with cells applying greedy-on schedules, we demonstrate by simulations that our proposed algorithm and designed method can increase the system capacity by 25%. We also find that even when the optimal power allocation is applied, always attempting to keep a small cell in active state may not always be a good strategy for capacity maximization, especially for interference-limited cells in HetNets. The sleep-awake scheduling can effectively enhance cell capacity.