Search

Substation equipment inspection is the fundamental of equipment management and security. However, the substation equipment inspection is facing the prominent contradiction between the surge in workload and the relative shortage of personnel. Therefore, this paper proposes an improved image-based method for detecting defects in substation equipment. This method is based on Faster R-CNN algorithm and proposes a series of improvement strategies according to the actual application scenarios, including inter-class sampling, center-guided NMS, and category-adaptive threshold. Experimental results demonstrate that this method can solve the problems of uneven sample distribution, redundant bounding-boxes, missed or false detections, etc., and improve the accuracy and practical effect of defect detection. In addition, this paper tests on a lightweight network framework, which proves the engineering applicability of this method.

The generative adversarial network (GAN) is a kind of unsupervised learning approach with the capacity of dealing with the challenge of limited labeled data in the supervised learning world. However, the limited data bandwidth, as well as the performance gap between processing units and memory of the conventional computing platform becomes a major obstacle in the GAN based applications. In this paper, a memristor-based convolution neural network (CNN) unit synthesized with the spintronic memristor crossbar circuit and a general sigmoid activation function circuit is designed for the implementation of convolutional calculation. Notably, multiple memristor-based CNN units can be utilized for the construction of the generator and discriminator respectively. Based on this, a compact GAN architecture composed of the generator block and discriminator block is presented. For verification, the presented memristor-based GAN is applied to the single image super-resolution (SR). The experimental results demonstrate the validity and effectiveness of the entire scheme.

In this paper, an effective optimize method based on adaptive PID neural network controller is presented. Chaotic Particle Swarm Optimization (CPSO) is introduced to initialize the parameters of neural network for improving the convergent speed and preventing weights trapping into local optima. In order to realize the hardware platform of PID neural network controller and facilitate weight update algorithm, the nanoscale memristor is utilized to analog electronic synapse. Meanwhile, the stability of closed-loop system is further proved by Lyapunov theory. Based on several comparative experiments on automatic voltage regulation system, the final results illustrate that the proposed controller can obtain better precision in a shorter time.

The memristor was originally defined as one of the fundamental electrical elements which provided the vacant connection between charge and flux. So far, most of the research has concentrated on the unique features of the individual devices. And the overall behavior of the multiple memristive systems has not been fully studied. Especially, the lack of corresponding fault diagnosis method for complex memristor circuits makes all the existing applications based on the multiple memristor circuits unstable and shaky. In this paper, the composite properties of multiple memristor circuits are further investigated. Then a neotype parametric fault diagnosis approach for memristive networks is presented using the doublet generator, sensitivity method and fuzzy math method, which offers huge benefits in terms of accuracy and time consumption.