With the development of smart grid information physical systems, some of the data processing functions gradually approach the edge layer of end-users. To better realize the energy theft detection function at the edge, we proposed an energy theft detection method based on the power consumption information acquisition system of power enterprises. The method involves the following steps. In the centralized data center, K-means is used to decompose a large amount of data into small data and then input and train neural network parameters to realize feature extraction. We design a neural network named DWMCNN, which can extract features from the day, week, and month and can extract more accurate features. In the edge data center, the random forest (RF) algorithm is used to classify the extracted features. The experimental results show that the clustering method accords with the idea of edge computing-distributed processing and improves the operation speed and that the feature extractor has good convergence performance. In addition, compared with the methods based on various classifiers, this method has higher accuracy and lower computational complexity, which is suitable for the deployment of edge data centers.

No abstract

No abstract

No abstract

Mechanical parameter monitoring based on optical mode detection benefits from its low cross sensitivity and inexpensive instrument. The key to improving detection accuracy is to generate high-quality detection light and use efficient algorithms. We present a strain-independent torsion sensor based on acoustically-induced fiber grating (AIFG) in the dual-mode fiber (DMF) and use the enhanced self-integration algorithm to improve the sensing accuracy. By tuning the radio frequency of driving signal, the LP11 mode generated by the AIFG can be exploited to measure the dynamic torsion variations. Without the complex device such as fiber interferometers and photonic crystal fibers (PCFs), the simple structure built by mode converter and charge coupled device (CCD) can track the dynamic variations and has less cross sensitivity of strain along the transmission direction. The AIFG driven by a radio frequency as a mode converter at specific wavelength does not participate in sensing but generates the high-purity LP11 mode that accounts for more than 90% of total power. With the twist from the rotator stage, the DMF keeps rotating and CCD records the spatial distribution of mode profiles. The features of optical mode is enhanced based on matrix analysis and then the relationship between twist angle and mode features is obtained. Based on image processing, the dynamic variation of spatial beam detected by CCD can be easily tracked and quantified. In experiment, the rotation angle can be obtained by calculating the feature value of the optical mode. Our image detection algorithm is specially designed for the optical fiber mode. Compared with traditional image recognition based on feature learning, it is simple and fast because it is needless to use image segmentation and stochastic processing. Through a series of experiments on angle rotation and parallel strain, we verify the correctness of the enhanced self-integration model and analyse the computational uncertainties that influence the stability of experiment. In the 0 to 180 measurement range, the maximum range of measurement error is less than 11%. When the axial strain is between 100 and 1500 , the sensor is strain-independent. Thus, it is verified that the torsion sensor based on AIFG has high sensitivity and can overcome the cross sensitivity of strain along a certain direction. The pertinent results have significant guidance in designing the multi-parameter sensor. The optical mode detection, instead of the traditional spectrum measurement, enables the whole structure to have the potential to be rebuilt by inexpensive devices that work in visible wavelengths.