基于贝叶斯迭代的非侵入式负荷事件检测方法

An approach for non-intrusive load event detection based on Bayesian iteration

无监督型的非侵入式负荷监测技术是负荷监测领域的发展趋势,而负荷事件检测是其中的重要环节。目前的负荷事件检测方法存在超参数复杂不易调整的问题,且对不同类型的负荷事件自适应能力差。为此,提出一种基于贝叶斯迭代的负荷事件检测方法。该方法构建了事件检测模型,基于贝叶斯迭代对模型进行求解,最后结合负荷监测的应用实际优化求解速度。采集典型家用负荷的实际运行数据对所提方法进行测试,测试结果表明,文中方法对于不同类型的负荷事件有较强的自适应能力,相比于GLR方法与CUSUM方法,其具有更简单、更易于调整的超参数,且能获得更高的检测准确率,为非侵入式负荷事件检测提供了新思路,同时可以更方便地应用于工程实践。

Unsupervised non-intrusive load monitoring technology is a new developing trend in load monitoring field,in which the fundamental part is load event detection.The current load event detection methods have the problems of hyperparameter complexity,difficulty in tuning and poor adaptability for different types of load events.On this basis,a load event detection method based on Bayesian iteration was proposed in this paper.The methodology firstly constructed a load event detection model,and then,solved the model based on Bayesian iteration.Finally,the model solving process was accelerated considering the application condition of non-intrusive load monitoring.Some real data of typical household load was collected to test the presented method.The results show that this approach has strong adaptability for different types of load events.Compared with GLR method and CUSUM method,it can obtain higher detection accuracy and has simpler hyperparameters which are easier to tune.Accordingly,it provides a novel idea for non-intrusive load event detection,and can be more conveniently applied to engineering practice as well.