利用電力線拓撲資訊改善非侵入式負載監測技術

Abstract

非侵入式負載監控系統藉由分析迴路之總電壓與總電流變化來了解耗電情形，利用每一電器所產生的電氣特性來判別個別電器狀態，而每種電器的電器特性由不同的特徵所組成，稱為特徵集。在以往的研究中多半使用固定之特徵集來辨識電器，而環境可能導致特徵的改變，進而影響判斷的準確性。我們經由實驗找出影響特徵改變的原因，其中最關鍵因素之一為電力傳輸線並非理想傳輸導線，而是存在微小阻抗，這樣的特性使得每個插座電源會獲得不同的電壓，使電器特徵受到影響。因此在本研究中，我們提出一種檢測裝置和方法來建立電力線的拓撲，使用者可以用最少的人為動作來完成拓撲的學習和建置。獲得之電力線拓譜資訊則可以用來協助我們修正電器特徵，我們提出了兩種修正機制，一為電器特徵隨其插座位置的修正機制，高功率電器會因其位置不同而使特徵值改變；其次，我們提出改善電器辨識的機制，利用電力線拓譜資訊再經由計算獲取個別插座電壓，找出最相近區間之電器，改善僅利用單一特徵集辨識電器的機制。

Nonintrusive appliance load monitoring (NIALM) system detects the appliance and appliance states and can disaggregate the power consumption of individual appliances from the total power consumption by analyzing the changes of voltage and current of the main power breaker. Each appliance has its own electrical characteristics so that an NIALM system can determine the state of individual appliances. The electrical characteristic of each appliance is composed of different features, called signature set. Previous studies usually used a fixed set of signatures for detecting appliance. However, environmental factors may lead to distortions of features, thereby affecting the accuracy of appliance detection. In this study, we conduct a number of experiments and find one of the key issues causing distortions of features, i.e., unideal power transmission line. The power transmission line contains tiny impedance so that each power socket supplies different voltages to appliances. Therefore, appliance’s feature is thus influenced by the environmental factors such as the location of the power socket. We then propose a testing device and an efficient method to establish entire power-line topology with minimal manual operations. We adjust the appliances’ features by using the power-line topology information. We propose two adjustment methods. The first one is the adjustment based on location. The second one is to recalculate the voltage of individual socket to identify the most similar signature of appliance.