Search

In this paper, we study distributed-observer-based fault diagnosis and propose a fault-tolerant control approach for a class of discrete interconnected systems. The distributed fault observers are designed to estimate faults based on the improved fast adaptive fault estimation (FAFE) algorithm. As a result of the improved FAFE algorithm, the constraints which are necessary to the general FAFE algorithm can be reduced while the fault estimation accuracy can be maintained. Based on the online fault estimates, the distributed output feedback controllers are developed to accommodate faults. To solve the observers and controllers, the corresponding algorithms are proposed. Finally, various fault situations are considered in detail in a simulation, and the results verify the accuracy of the theory and method.

The Modular Multilevel Converter (MMC) is one of the most attractive converter topologies in the High Voltage Direct Current (HVDC) transmission field. The latest widely used sorting method has a low algorithm complexity. It can effectively balance the sub-module (SM) capacitor voltages, but it would cause relatively high switching frequency and power loss. Aiming at the problem that the sub-module (SM) capacitor voltage sorting algorithm has a large switching loss due to the high switching frequency of the device, this paper proposes an auto-optimized capacitor voltage balancing control strategy. Firstly, the topology and operation principle of MMC are analyzed. Secondly, a SM capacitor voltage control method based on the dynamic deviation threshold is proposed. Considering the switch switching state of the SM and the difference between the voltages of each SM, the algorithm can obtain the dynamic deviation valve using the closed-loop control. The method can avoid the unnecessary repeated switching of the Insulated Gate Bipolar Transistor (IGBT) under the premise of ensuring that the capacitance voltages of the SMs are basically the same, which effectively result in reducing the switching frequency of the MMC SM and reducing the switching loss, thereby improving the operating efficiency of the system. Finally aiming at the proposed control strategy, the simulation and experimental verification are carried out which shows that the proposed algorithm can better control the system voltage deviation, reduce the switching loss of the system and improve the stability of the system.

This paper proposes a fault diagnosis and fault-tolerant control method for a system with a fast time-varying delay and time-varying parameters. A fault observer is designed to estimate faults, and an improved fast adaptive fault estimation (FAFE) algorithm is developed to reduce the relevant constraints in the general form of this algorithm. With newly introduced relaxation matrices, this study estimates faults in a system exhibiting a fast time-varying delay. Based on the estimated faults, an output feedback controller is designed to accommodate the faults. The fault-tolerant control is realized using the introduced relaxation matrices. An algorithm is derived to solve for the observer and controller. Finally, the theory and method are validated using a real example of a helicopter system.

Aiming at the function and technical requirements of high-power photovoltaic cell simulation, high-performance programmable logic power supply and dc motor simulation, a high frequency isolation hybrid topology and control strategy based on current-source/voltage-source converter was studied and proposed. Firstly, according to the performance requirements of photovoltaic cell analog power supply, the control strategy requirements of the high-precision wide-range hybrid topology were proposed. Secondly, the working principle of the new hybrid topology was analyzed. At the same time, the equivalent model of the new hybrid topology was simplified and established, and the overall control strategy of the hybrid topology based on current compensation and sliding mode variable structure was proposed. Finally, simulation and experimental research on the hybrid topology was carried out, and the experimental test of photovoltaic cell simulation was completed. The simulation and experimental results show that the hybrid topology and control strategy proposed in this paper has the characteristics of wide-range output regulation, fast dynamic response, high efficiency and high power factor, and can be used for high performance photovoltaic cell simulation, programmable logic power supply and DC motor simulation.

Gradient descent optimization algorithm is very important in deep learning. In order to obtain a more stable convergence process and reduce overfitting in multiple epochs, we propose an improved Adagrad gradient descent optimization algorithm in this paper. Our approach is tested both on the Reuters dataset and the IMDB dataset with many gradient descent optimization algorithms. The results show that our approach has a more stable convergence process and can reduce overfitting in multiple epochs.

LED visible light communication indoor positioning method based on the mobile phone at any horizontal orientation angle. The system includes a single LED downlight, a mobile phone orientation sensor, a mobile phone front single camera and a positioning system. The steps include: the orientation sensor of the handset is used to measure the orientation of the handset in the horizontal direction; the developed camera is used to call the front camera of the handset to record the LED downlight image; and the image is processed and decoded to obtain the ID information of the LED downlight, which is the three-dimensional position coordinates information of the downlight in the indoor space. Take the previously obtained information including the handset horizontal orientation and downlight ID information into the developed indoor positioning method to obtain the real-time accurate position of the mobile phone. The positioning method is simple and easy to operate. The mobile phone can be accurately positioned in real time at any horizontal orientation angle just by using the front camera of mobile phone in the case of just a single LED downlight. This method overcomes the problem that even if the position of the camera is kept unchanged, the positioning result of the mobile phone is different due to the difference of the horizontal orientation of the mobile phone. This method is more feasible and accurate.

As technology advances, wireless charging system of electric vehicles(EVs) based on magnetic resonance becomes prevalent. A novel parameter identification method of primary side compensation capacitance and coil inductance in wireless charging system is proposed. According to the typical resonance equivalent circuit of wireless charging system, current effective value equation of primary side is established. And then, based on the current effective value equation, input signal frequency constraint equation of current extremum is established. By detecting primary side real-time current and using Fourier transform on primary side transient current waveform, two characteristic quantities are extracted. Finally, unknown parameters (primary side capacitance and inductance) in both equations are identified utilizing the adaptive particle swarm optimization. The simulation results show that the method proposed is feasible and the identification accuracy is high.

Most state-of-the-art anomaly detection methods are specific to detecting anomaly for pedestrians and cannot work without adequate normal training videos. Recently, there is a growing demand for detecting anomalous vehicles in traffic surveillance videos. However, the biggest challenge in this task is the lack of labeled datasets for training supervised models. By examining the resemblances of anomalous vehicles, we find it reasonable to label a vehicle as anomaly if it stays still in the video for a relatively long time. Utilizing this property, in this paper we introduce a novel unsupervised anomaly detection method for traffic surveilliance based on background modeling, which shows great potentials in handling heterogeneous scenes as well as extremely low resolution videos recordings without the dependence on labeled data. In the proposed system, we first employ background modeling using MOG2 to remove the moving vehicles as foreground while keeping the stopped vehicles as part of the background. Then we use Faster R-CNN to detect vehicles in the extracted background and decide if they are new anomalies under certain conditions. All information is updated on a frame basis until the end of the video which contains the final results. In this way, we make full use of the characteristics that abnormal vehicles stay in the scene for a relatively long time and reduce the difficulty of vehicle anomaly detection. Eventually, we can detect almost every anomaly in the NVIDIA AI CITY CHALLENGE track-2 dataset except for several extremely complex cases with a 81.08% F1-score and 10.2369 RMSE.

The three-phase current source inverter has the characteristics of accurate current control, high dynamic response speed and strong current limiting ability, and the DC side voltage can be adjusted from zero. In this paper, the three-phase current source inverter topology is studied, the mathematical model of the converter is established in the three-phase static coordinate system and two-phase rotating coordinate system, and get the equivalent circuit in two coordinates. The current modulation method including SPWM and SVPWM is studied. A new type of system control strategy based on two-phase rotating coordinate system is proposed. The simulation model of the whole system is established by PSIM. The results show that the proposed new control strategy is correct and feasible.

Taking double stage type three phase photovoltaic grid-connected generation system as the research object, the equivalent model of photovoltaic cells, the maximum power point tracking (MPPT) technology and the control strategy of three-phase grid connected inverter are analyzed in this paper. On the basis of the theoretical analysis, the double closed-loop vector control of the double stage type three-phase grid-connected inverter and the MPPT method based on T-S fuzzy neural network are studied deeply, and the relevant simulation and experimental research are carried out. The experimental results show that the new grid-connected control strategy can make the grid-connected system run with high power factor, high efficiency and low total harmonic distortion, and the grid-to-network voltage and current distortion is low, which proves the effectiveness and feasibility of the control strategy.

With the popularity of the concept of green energy, reducing the harmonic pollution of power equipment to power grid has become an important research topic in the field of power electronics. This paper takes three-phase current-source PWM rectifier system as the research object, and it presents a current-source PWM pulse signal tri-logic control method, and puts forward a pre-modulation method. Then the topology of the system is studied and a new equivalent model is proposed. After that the closed-loop control strategy of the system is studied, and a PI tuning method is obtained. Finally, the control strategy is simulated by PSIM and the experimental platform is set up for experimental verification.

The visualization of large networks is a challenging problem due to the limitation of human visual perception. Sampling technology is an essential strategy in most of large network visualization and analysis. Moreover, when we consider large heterogeneous networks consisting various types of nodes and links, which has attracted much attention from multiple disciplines, it is much more difficult for users to explore and mine some useful information in such complex networks. Thus to design sampling methods for large heterogeneous network visualization has become a challenge. However, most of existing sampling methods ignore the types of nodes and links, hence they could not be used for heterogeneous network directly. In this paper, we propose an eigenvector centrality based sampling method considering both multiple types of nodes and links, supporting sampling the heterogeneous networks directly and efficiently. Furthermore, we develop HeteVis, a prototype visualization system with the proposed sampling algorithm built-in. The parameters of HeteVis can be dynamically personalized through user interactions and such system can help users to explore real large heterogeneous networks easily and flexibly. Extensive case study using the UCI KDD Movie network demonstrates how users can apply HeteVis to explore the heterogeneous networks.

In this paper, an adaptive robust drive control system for an axial flux permanent magnet synchronous motor of an electric medium-sized bus based on the optimal torque distribution method is studied. The drive control system is mainly divided into two parts. First, a torque distribution method is proposed. The optimal torque distribution method based on particle swarm optimization algorithm is used to increase the high efficiency interval of the system and apply it to the energy feedback braking. Secondly, in order to reduce the nonlinear disturbance of the system and improve the accuracy of the unified control, this paper models and studies the vector system based on adaptive robust control. Finally, the whole drive control system is modeled, simulated and experimented. The simulation and experimental results show that the torque distribution method proposed in this paper can effectively increase the high-efficiency running time of the electric medium bus, and improve the shortcomings of insufficient mileage of the electric medium-sized bus. The use of a current controller based on adaptive robust control improves the control accuracy of the drive system and can effectively suppress the disturbances generated by it.

In order to enhance the accuracy of building load prediction. This paper proposed a hybrid model based on time-indexed autoregressive with exogenous meteorologically inputs (ARX) in which the coefficients are estimated through a second order weighted least squares support vector regression(LS-SVM). This model is combined with Support Vector Machine Regression (SVMR) developed with Particle Swarm Optimization (PSO) and Differential Evolution Algorithm (DE). Hybrid methods named PSO-LSSVM and DE-LSSVM, both meteorological and historical information into consideration fundamentally. These proposed methods are tested on a medium-sized office-building. As a result, the accuracy of hybrid prediction models has the better performance compared to that of other prediction methods in predicting Cooling load and Heating load.

Rapid increases in building energy consumption in recent decades have also resulted in the popular research of energy conservation. Air temperature distribution is taken into consideration for indoor environments when carrying out energy simulations. In this paper, the CFD software FLUENT combined with a network model was used to calculate the temperature distribution of an air-conditioned office room, and it was compared with the measured data. The results showed good consistency of measured data and simulated data in both steady and transient simulations of all the four working conditions. In conclusion, the FLUENT simulation model may help developing a proper real-time building energy performance management system in future study.

With the progress of voltage source converter (VSC) technology, VSC-HVDC system is becoming a promising solution to renewable energy access. To address the problem of power control when renewable energy parallels in the power grid through VSC-HVDC, a direct power control method is proposed for VSC-HVDC system. First, the architecture of VSC-HVDC system is introduced. Then a direct power control method is presented to meet the characteristics of renewable energy (i.e. randomness, volatility and uncontrollability). Moreover, the operational characteristics of the HVDC system under different conditions were analyzed by numerical simulation based on PSIM. Finally, a 10kw VSC-HVDC prototype system was developed, and the experimental results validate the effectiveness of the proposed power control method.

In this paper, a sliding mode vector control system based on collaborative optimization of an axial flux permanent magnet synchronous motor (AFPMSM) for an electric vehicle is proposed. In order to increase the high efficiency range of electric vehicles and improve the cruising range, a collaborative optimization control strategy is firstly proposed. Due to the use of a dual stator-single rotor AFPMSM, the multi-motor efficiency optimization map and torque cooperative control are used to realize the working mode conversion of single stator and double stator, and the torque ripple caused by the working mode conversion is improved by fuzzy control. In order to improve the torque tracking capability, speed limiting characteristics, and operating characteristics, a speed limit and current vector control strategy based on a sliding mode controller is proposed and studied. The dynamic performance of electric vehicles is improved by a sliding mode vector control. Finally, a drive control system was developed for the proposed control strategy, and the complete vehicle test was carried out. The collaborative optimization control experiment and torque tracking and speed limiting experiments verify the correctness and effectiveness of the proposed control strategy. The acceleration performance and endurance experiments show that the proposed control strategy can effectively improve the cruising range and the acceleration performance of electric vehicles.

A hybrid method which combines time series model and artificial intelligence method is proposed in this paper to improve the prediction accuracy of building thermal load. Firstly, a simple auto regressive (AR) model is utilized to predict present load using previous loads, the order and the parameters of AR model are identified by the data produced by DeST. Then, a 3-layer back-propagation neural network optimized by particle swarm optimization (PSO) neural network (PNN) is set up to predict the error which is derived by comparing the precious AR predicting load. The error and its corresponding meteorological data generate the training sample data. At last, the hybrid model, named autoregressive and particle swarm neural network (APNN), is obtained. It uses historical load information and real-time meteorological data as input to predict a refined real-time load by adding error to preparative load. To evaluate the prediction accuracy, this hybrid model APNN is compared with several common methods via different statistical indicators, the result show the APNN hybrid method has higher accuracy in thermal load forecasting.

Accurate thermal load prediction is beneficial for minimizing cost and energy consumption of buildings. Indoor thermal comfort and the air quality greatly influence the health and work efficiency of building occupants. For energy efficiency and thermal comfort is significantly impacted by the choice of an appropriate reference air temperature, indoor air temperature distribution analysis is very important. In this study, the temperature distribution of an air conditioned room was calculated by FLUENT under different air supply parameters, and the calculated temperatures agreed well with the measured data in both cooling and heating modes of simulations. Then, proper simulation models was determined and used to seek the characteristic point that can represent the room reference air temperature directly. The outlet point of the air conditioner was verified to be the approximation of the characteristic point in the tested room. This provides a convenient approach to obtain the reference temperature used to predict real-time thermal load which is the key factor of building energy performance improving.

Due to the curse of the dimensionality, high-dimensional data visualization has always been a difficult and hot problem in the field of visualization. Some of the existing works mainly use dimensionality reduction methods to generate latent dimensions and visualize the transformed data. However, these latent dimensions often have no good interpretability with user knowledge. Therefore, this paper introduces a high-dimensional data visualization method based on user knowledge. This method can derive dimensions aligned with user knowledge to reorganize data, then it uses scatter-pie plot matrix, an extension of scatter plot matrix, to visualize the reorganized data. This method enables users to explore the relationship between the known and unknown data as well as the relationship between the unknown data and the derived dimensions. The effectiveness of the method is validated by the experiments.