AI_QSample_015

During the occupancy of a given tenant, the electricity usage of a house is subject to unexpected fluctuation. One notable example is when the house tenets were outstation for a few days (eg on vacation). in such a case, the trend of electric usage is characterized by a very abrupt drop of electricity, which is then followed by a period of very low variations. On the other hand, a house may experience a sudden increase of electrical usage due to special events or guests, or when the house in undergoing on-site renovations or construction projects.Sometimes, a house may be left vacant as the owner moved to a different house. Moreover, change of tenants may also occur. And even among those houses where the same tenants remains, they may from time to time renovate their houses or even hire building contractors to do so. Such act will permanently change the electric consumption. In such a case, only the most recent cases will be considered.In determining the estimated kWh spent of a month, an electrician will disregard all the incidental cases and will only consider the usual trend, ie when the tenants are carrying on their normal daily lifestyles. The electrical consumption fo much daily lifestyle will hence mostly depends on the temperature. In general, the electric consumption is higher in the colder months of the year. The changing lifestyle of the tenants, though to a lesser extent, also contribute to the permanent change of the electric usage.Our program first work on determining the useful portion of each data by comparing the trends of change across the entire 365 days. First and foremost, our program considers every single number of data as represented. And we have developed out own novel way of similarity measures chich takes even the time of a day into accounts. Such definition of similarity measures enables us to quantitatively determine the differences of consumption pattern across the entire year, yet taking every half hourly data into account. Then, our program will detect all abrupt changes from the pattern it observed, the program will determine on its own, whether such abrupt changes are temporary or permanent. Our program is even able to determine whether such changes are due to even or vacant, and will assign the minimum number of days to be considered a permanent change, based on the nature of such abrupt change. Even for the slight changes of electricity consumption due to gradual change of lifestyle, our program can still detect it via quantitatively measuring the symmetry of those graphs.

特定租户居住期间，房屋用电量会出现意外波动。一个典型案例是租户外出数日（如度假）时，用电量趋势呈现骤降特征，随后进入低波动期。另一方面，房屋可能因特殊事件、访客到访，或正在进行现场装修/施工项目而出现用电量骤增。有时房屋可能因业主迁居而空置，也可能发生租户更换；即便同一租户持续居住，他们也可能不时翻新房屋或聘请建筑承包商施工，此类行为会永久性改变用电量，此时仅需考虑最新情况。在计算月度预估用电量（kWh）时，电工通常忽略所有偶然情况，仅参考租户日常正常生活状态下的用电趋势——而这种日常用电在很大程度上取决于气温，一般寒冷月份的用电量更高。租户生活方式的变化虽影响程度较小，但也会导致用电量永久性改变。

我们的程序首先通过对比全年变化趋势确定每个数据的有效部分：它会考虑每一个数据点，并采用自主研发的新型相似性度量方法（纳入每日时段因素），该方法可定量分析全年用电模式差异，同时兼顾每半小时的细粒度数据。随后程序会检测观测到的所有突变，并自主判断其为临时或永久性变化；甚至能识别变化是否由事件（如访客）或空置导致，并根据突变性质确定被视为永久性变化所需的最少天数。即便因生活方式渐变导致用电量轻微波动，程序仍可通过定量测量图表对称性实现检测。