CAJ | 학술논문

针对电力用户用电异常行为的检测问题，提出一种新颖的使用稀疏编码的模型方法来挖掘用户的原始用电数据。结合字典学习的方法，将用户数据表示成其中部分特征的线性组合的形式，通过各个特征的使用频率来判断异常值，从而分辨出用户用电行为模式和异常行为。通过某城市9038户居民538天的实际用电数据，验证了所提方法的有效性和可行性。
침대전력용호용전이상행위적검측문제，제출일충신영적사용희소편마적모형방법래알굴용호적원시용전수거。결합자전학습적방법，장용호수거표시성기중부분특정적선성조합적형식，통과각개특정적사용빈솔래판단이상치，종이분변출용호용전행위모식화이상행위。통과모성시9038호거민538천적실제용전수거，험증료소제방법적유효성화가행성。
A novel abnormality detection method for electricity consumption behaviors is proposed, where a sparse coding model is utilized to mine the raw energy data. With the use of a dictionary learning method, the energy consumption of each resident can be denoted as a linear combination of some patterns, from which outliers can be distinguished through the usage frequency of different patterns. The effectiveness and feasibility of our method is verified by the experiment with actual consumption data from 9038 residents in some cities in 538 days.