| Summary: | 隨著各行各業的飛速發展、科技的不斷進步，一般的公司行號、工廠及現代化的建築對於電力需求大大增加。但是在有限的電力資源下，有時候一到用電高峰時期，很難滿足各行各業的用電需求，因此難免會出現很多地方在用電高峰期跳電的情況。電力公司為了更加有效的分配電力，提出所謂時間電價的概念，和用戶實現簽訂各自的契約容量，將這個契約容量作為每個月分配給各個用戶的最大電量標準。對於用戶來說，若選擇相對較低的契約容量，其所需要負擔的基本電費會較低。然而，當用電量超過契約容量時，用戶可能需要支付非常高額的罰款；若選擇相對較高的契約容量，雖然其支付高額罰款的機率會降低很多，但是所需要負擔的基本電費會增多。因此，對於電力公司和用戶而言，使用時間電價系統，來選擇一個適當的且最佳化的契約容量，已然成為一個非常重要的課題。本文介紹如何用分形布朗運動的模型，來描述用戶用電量趨勢，同時介紹了如何估計分形布朗運動模型中的各個參數。本文也介紹如何建立每月總電費期望值的估計方程式，並利用估計出來的用電量分形布朗運動模型來搜尋最佳化的契約容量。最後，本文以美國密西根州的安娜堡的居民住宅大樓用電量為數據資料作為研究的實例，進一步的提出並論證了選擇最佳化契約容量的方法。 === Over the last few decades, the advances in technology and industry have significantly increased the need of electric power, while the power resource is usually limited. In order to best control the power usage, a so-called Time-of-Use (TOU) pricing system is recently developed so that different rates over different seasons and/or weekly/daily peak periods are charged (this is different from the traditional pricing system with flat rate contracts). An important feature of the TOU system is that the consumers have to pre-select the power contract capacities (i.e. the maximum power demands claimed by consumers over different pricing periods) so that the electricity tariff can be calculated accordingly. This means that risk is transferred from the retailer side to the consumer side -- one has to pay more if a larger contract capacity is selected but can potentially mitigate the penalty charge placed when the maximum demand exceeds the contract level. In this thesis, a general stochastic modeling framework for consumer's power demand based on which the contract capacities of a Time-of-Use pricing system can be best selected so as to minimize the mean electricity price. Due to the observed nature of self-similarity and time dependence, the power demand over a homogeneous peak period is modeled as a constant mean with the noise described by a scaled fractional Brownian motion. However, the underlying optimization problem involves an intricate mathematical formulation, thus requiring techniques such as Monte Carlo simulation and numerical search so as to estimate the solution. Finally, a real data set from Ann Arbor, Michigan along with two pricing systems are used to illustrate our proposed method.
|
|---|
