Search

Serverless computing is an emerging computing paradigm in cloud computing. Serverless users benefit from the features of serverless computing (e.g., auto scaling and pay-as-you-go model), and widely adopt it. Besides, ARM servers are becoming increasingly popular nowadays. It is of great importance to explore the performance of serverless computing on ARM platforms, as well as X86 platforms. This paper provides in-depth study of serverless computing on X86 and ARM platforms. We evaluate typical characteristics (e.g., startup latency, auto scaling, and performance isolation) of popular open source serverless platforms: Knative and Open-FaaS. We gain some valuable insights in different platforms. We believe our findings can guide serverless providers to design more efficient and robust serverless systems, thereby improving user experience.

Large amounts of data collected by power company has laid the foundation for load forecasting at different time scales. In this paper, a Dynamic Bayesian Network is established as a short-term load prediction model for distribution transformer supply area. The parallel computational operators of Apache Spark are used to calculate the model parameters based on Bayesian algorithm and a large number of history data of the transformer. At the same time, the forward backward algorithm is parallelized by using the Pregel calculation model to obtain the predicted results. The experimental results show that the proposed method has better performance and higher efficiency than traditional methods.

Through the development of machine learning and computer vision, image scene classification has made immense progress over the last decade. Remote sensing land use analysis remains a topic of great interest. Using deep learning methods from computer vision, we develop a novel approach that combines a convolutional structure and gated recurrent unit layers with a fully connected neural network to solve land use classification tasks. Simulation studies confirm the proposed method can more accurately classify and recognize remote sensing images in the EuroSAT dataset than current state-of-the-art algorithms.

The total transfer capability (TTC) of flowgate is an important concern for operator during the power system operation. To provide fast and accurate TTC evaluation, this paper presents a TTC evaluation methods using the stacking ensemble learning method. Firstly, the repeated power flow is applied to calculate the TTC value under different scenarios. Then, the steady variables, including the active and reactive power, voltage and angle of generators, the active and reactive power of loads and the active power flow on transmission lines, are used as the input features. Finally, the stacking ensemble learning based on the XGBoost, RF, GBDT, MLP and SVM is used to obtain the final TTC evaluation model. The simulation results illustrate the effectiveness of the proposed methods.

Traditional diagnosis algorithms generally only diagnose switch faults in rectifiers and inverters, respectively. However, in the traction converter containing both a rectifier and an inverter, the combination of two diagnostic algorithms increases the complexity of fault diagnosis and the compatibility between diagnostic algorithms needs to be fully considered. Therefore, an individual detection variable is expected to detect faults of the traction converter. Based on the above considerations, in this paper, a unified diagnostic algorithm using dc-link voltage residuals as an individual detection variable is proposed to identify single and multiple switch faults that occur simultaneously in both rectifier and inverter. The proposed method can complete the fault detection and location within one line cycle for the single or multiple switch faults, where no additional sensors are required and all the parameters required are available from the control system. The simulation results have verified the effectiveness of the proposed method.

With China's power grid gradually showing the characteristics of long distance, heavy load, large-area interconnection, and introduction of market competition, the situation that power flow calculation does not converge appears more and more frequently. Therefore, solving the problem of power flow non-convergence will become the focus and difficulty of power flow calculation. At present, analysis and calculation personnel often can only make adjustments based on experience. This traditional adjustment method is inefficient, which has high work intensity, and the effects is also not obvious. This paper proposes a artificial intelligence adjustment method of large-scale power flow based on knowledge-driven. This method expresses the knowledge and experience accumulated in power flow adjustment process in the form of knowledge graph, and builds knowledge base on this basis. And then, this paper takes the knowledge base rules as the criterion of power flow intelligent adjustment, which can provides the reasoning and deduction of logical and possible about the adjustment direction and adjustment measures So as to make correct judgments and decisions. Finally this paper realizes the artificial intelligence adjustment of the large-scale power flow.

With the development of UHV AC / DC hybrid power grid, transient stability is gradually becoming a key factor that affects the stability of power grid. In recent years, due to its superior performance, methods based on artificial intelligence are widely used in dynamic security assessment. Analysing key features of transient stability can effectively reduce the input dimension of the model and therefore promote their performance, besides, the dispatchers can also get more control guidance from the analysis results. In this paper, we propose a two-stage branch and bound method to analyse the key features of transient stability. This method can reduce the cost in time and space, and reduce the redundancy between key features at the same time. The proposed method is validated on New England 39-bus system.

To improve both system reliability and voltage profiles in distribution networks requires the simultaneous installation of distributed generations (DGs) and sectionalizing switches (SSs). This paper builds a bi-level programming model for the coordinated configuration of DGs and SSs. The upper-level problem is to minimize the total investment cost of DGs and SSs and the penalty cost of power shortages, and the lower-level problem is to minimize the voltage deviation during typical operation mode. To explicitly represent the relationship between the penalty cost of power shortages and installation locations of DGs and SSs, an analytical expression of the failure rates of load nodes with respect to installation locations of DGs and SSs is derived. The bi-level model can be transformed to a mixed-integer nonlinear programming problem by applying the Karush-Kuhn-Tucker theorem because the lower-level problem is a quadratically constrainted linear programming problem. Case studies on an IEEE 33-bus distribution system is used to verify the effectiveness of the proposed method.

Cascaded H-bridge multilevel converters (CHBMCs) have been applied in various high-power topologies. Due to the multiple modular structures, each switch in every module is a potential source of failure. In order to quickly identify faulty switches of CHBMCs for improving maintenance efficiency, a simpler diagnosis approach is proposed by only extracting the change rate of DC-link voltages. Combined with switching states of the bridge arms and H-bridges, the specific current paths under different switch faults are analyzed to summarize the change of DC-link voltage. And then, the individual diagnosis variables are established to detect single or multiple faulty switches for single or multiple modules. The proposed diagnostic method can complete the different fault identification process in less than 3/4 current cycle, which only needs a fault counter to avoid the false alarm without the other detection thresholds. The proposed method is confirmed by the hardware-in-the-loop (HIL) test.

Lack of sparseness is one of the main problems of least square support vector machine (LSSVM); it results in slower arithmetic speed and larger requirement for machine memory. This paper aims at an optimal sparseness algorithm for LSSVM. Firstly, Global Representative Indicator is defined in feature space, in fact it is constructed through density and dispersion. Then the sparseness of LSSVM is formulated as an optimization problem, where the training sample amount and root-mean-square error are taken as optimization variable and objective function, respectively. A PSO based technique is further proposed to solve the optimization problem. Finally, a soft measuring LSSVM model for carbon content in fly ash is taken as a study case to verify the proposed algorithm. With the data collected from a 1000MW utility boiler, the verification results are presented. Moreover, neighborhood value, a sensitive parameter for the algorithm, is studied and remarked.

Nowadays, portable notebook computers have made mobile offices very common. At the same time, as a piece of common office equipment, printers have not made a breakthrough in portable performance. For those who need to travel frequently, often the printer cannot be found anytime, anywhere. We have designed a new type of intelligent portable printer, which can be easily carried in a bag. Our main contributions include letting the printer enter the mobile office field for the first time, and greatly expanding the functions of traditional printers. In terms of function, it can simultaneously cover a lot of existing technical inventions such as traditional household printers, copiers, Polaroids, clothing pattern printing and dyeing machines, large-scale banner poster printings machines, mobile phone shell printing machines.

With the scarcity of fossil energy, the development of renewable energy is becoming more and more rapid. However, when new energy is integrated into the power grid, the inertia of the system is reduced and the stability of the system is lowered. This also poses a new challenge to frequency control, and energy storage as a technology that can quickly respond to system frequency changes is now widely concerned. This paper summarizes the domestic and international research status of energy storage and AGG capacity allocation and coordinated control through the influence of renewable energy on AGC. Finally, the development prospects of energy storage and participation in new energy integration into the power grid are prospected.

In order to improve the equipment utilization of distributed photovoltaic power generation system, the harmonic and reactive currents generated by grid-connected photovoltaic system can be accurately detected and compensated. A phase-locked loop based on second-order generalized integrator (SOGI-PLL) is introduced into the traditional FBD current detection method to separate the positive and negative sequence components of unbalanced voltage and detect the phase and frequency of reference voltage more accurately. The simulation results show that the algorithm can reduce the computational complexity and increase the detection accuracy and steady-state performance.

This paper studies the optical axis stable tracking of high-precision photoelectric stability platform system, and proposes fractional-order compound control proposes. Firstly, a fractional-order terminal sliding mode control (FO-TSMC) method is designed to enhance the tracking accuracy of the optical axis. Secondly, for the tracking accuracy affected by friction torque and nonlinear disturbance, the method of disturbance compensation is studied from the level of refinement anti-disturbance, and the compound fractional-order disturbance compensation control strategy is adopted. A fractional-order disturbance observer based on velocity (FO-VDOB) is designed with the equivalent disturbance of the systems. The disturbance is effectively estimated and compensated. A fractional-order sliding mode compensator (FO-SMCO) is designed for the unestimated state of the disturbance, further compensating for the unestimated disturbance. Finally, the finite-time stability is proved by Lyapunov and the finite-time control theory, and the effective validation of the protocol is in the MATLAB simulation environment.

With the fact that most wind resource-intensive areas are far away from load centers in China, the rapid increase of wind power generation (WPG) integration is hindered. Ultra-high voltage (UHV) DC transmission can help improve this situation. However, the uncertainty of wind energy and the random contingency of DC lines pose a great threat to system security. This paper tends to study it based on the risk assessment theory. Firstly, according to the characteristic of power exchange between UHV DC line and AC system, a DC power injection model is proposed, and a five-state fault model is constructed considering the protection logic of UHV DC lines. Secondly, the whole risk assessment process is established based on the improved scattered sampling Monte-Carlo method (ISS-MC). Comprehensive risk indicator system is proposed from the aspects of transient frequency, steady state, system structure and economic loss, and the comprehensive risk indicator is further carried out by entropy weight method. Steady state and economic loss risk indicators are calculated by AC power flow (AC-PF) and AC optimal power flow (AC-OPF) to improve the accuracy of assessment results. Simulation tests on modified IEEE-RTS79 system validate the effectiveness of the proposed model and method.

Based on a pure electric logistics vehicle with a single reducer, the paper matches a new electric motor, takes the electricity consumption in NEDC working condition as the optimization objective, designs the two-speed transmission gear ratios to replace the fixed gear ratio and gives a new shift schedule matching the electric motor. Then, the two gear ratios are optimized according to the new shift schedule. The optimization results indicate that the electricity consumption(Kwh/100km) in NEDC working condition decreased by 1.21%. The maximum motor working rotation speed decreased by 44%. The maximum inclination increased by 11.68%. The acceleration time of 0~50km/h is the same as single reducer vehicle, which can meets the requirements. Compare with the single reducer, the two-speed transmission has a better performance.

Aiming at the higher priority to detect and protect against fault conditions for the single-phase cascaded H-bridge rectifiers (SPCHBR) in the power electronic traction transformer (PETT), this paper has proposed a voltage-based open-circuit fault diagnosis approach to detect and locate failure modules and faulty switches of the SPCHBR. According to the control features of the model predictive control (MPC), the measured and estimated values of the input-side voltage for H-bridge modules are obtained to generate voltage residuals, which include fault information of faulty modules. And then, combining the voltage residuals with switching states, the count comparisons are designed to detect faulty modules. What’s more, faulty switches can be accurately located by matching fault features with specific switching signals after the failure modules are detected. The proposed diagnostic approach needs no additional sensors. Meanwhile, multiple faulty switches can also be isolated within one line cycle of the grid current. The simulation results have verified the availability of the proposed diagnostic algorithm.

In order to solve the lifetime prediction problem of DC-link capacitors, an optimized lifetime model considering power module faults is proposed. In terms of electrical stress, four power module faults models are established, and the actual multi-operation data in the traction drive system is introduced as the mission profile to collect the electrical parameters of the DC-link capacitor. In terms of thermal stress, Newton's law of cooling is introduced to analyze the hot spot temperature through actual testing. Finally, the existing lifetime model is optimized in the traction drive system based on the lifetime loss rate of the faults. This conclusion can provide a reference for studying the bottleneck of capacitor reliability.

Line-of-Sight(LOS) stabilization is the primary condition for ensuring stable platform tracking accuracy. Factors affecting the stability of the LOS include wind resistance torque, friction torque, mass imbalance torque, etc. In order to overcome the disturbance and ensure the stability of the LOS, in this paper the method of disturbance compensation is studied from the level of refined anti-disturbance, and the composite high-order sliding mode interference compensation control strategy is adopted. A high-order sliding mode disturbance observer is designed for the equivalent disturbance of the system and combined with the finite time convergence theory. The disturbance is effectively estimated and compensated. A high-order sliding mode compensator is designed to compensate for the unestimated state of the interference, further compensating for the unestimated state of the interference. Finally, the stability of the system is proved by Lyapunov theory, and the effectiveness of the method is verified in the MATLAB simulation environment.

The increasing importance of cascaded H-bridge multilevel converters (CHBMC) in supplying electrical energy to loads or grids places a higher priority to detect and protect against fault conditions. This paper proposes a simple open-circuit fault diagnosis method for CHBMC by analyzing the current residuals in the output side. Comparing the estimated output current with the calculated current, the change features of current residuals are extracted to detect and isolate single or multiple faulty switches. Compared to the existing approaches, the proposed method can detect and identify multiple faulty switches in the different faulty cells. The fault feature extraction of different cells is independent, which ensures that the fault diagnosis between cells is not affected each other. Furthermore, all of specified parameters can be obtained directly. And the diagnosis time is within three control periods under the appearance of fault features. Finally, simulation and experiment results have verified the applicability and effectiveness of the proposed method.

This paper presents a fast and reliable fault diagnosis and fault-tolerant approach for grid current and dc-link voltages sensor faults in the single-phase cascaded H-bridge rectifiers (SPCHBR). Firstly, a state observer with the mixed logic dynamic (MLD) model is established for obtaining the observed sensor signals and system residuals. And then normalized residuals are generated to diagnose and isolate single or multiple sensor faults online by comparing different residuals with thresholds. Moreover, fault-tolerant control is realized by redesigning the observer gain and substituting the observed values for the measured values. The proposed approach can accomplish the accurate location of the faulty sensors and system reconfiguration within half of the current fundamental period. The effectiveness and availability of the proposed approach has been verified by MATLAB/Simulink.

This paper proposes an intelligent multi-agent approach in a real-time strategy game, StarCraft, based on the deep deterministic policy gradients (DDPG) techniques. An actor and a critic network are established to estimate the optimal control actions and corresponding value functions, respectively. A special reward function is designed based on the agents’ own condition and enemies’ information to help agents make intelligent control in the game. Furthermore, in order to accelerate the learning process, the transfer learning techniques are integrated into the training process. Specifically, the agents are trained initially in a simple task to learn the basic concept for the combat, such as detouring moving, avoiding and joining attacking. Then, we transfer this experience to the target task with a complex and difficult scenario. From the experiment, it is shown that our proposed algorithm with transfer learning can achieve better performance.

A fast open-circuit fault diagnosis method based on residual analysis is proposed for a dual active bridge (DAB) DC-DC converter in this paper. The proposed method uses the magnetic component voltage and the primary-side current for fault diagnosis. A state estimator of the primary voltage and current is established by the switching function, and then the residual characteristics are extracted by comparing the estimated values with the measured values. Furthermore, after the specific switch signals are injected, the change trend of the current residual is extracted to locate the faulty switch. The magnetic component voltage can be obtained by the auxiliary winding in the magnetic core, and the measurement of current needs to add a sensor. The proposed method can accomplish the accurate location of the faulty switch within 3 switching periods by increasing a small number of hardware devices. The hardware-in-the-loop (HIL) test results have verified the validity and feasibility of the proposed diagnostic method.

The labor intensity and danger can be reduced by using humanoid manipulator instead of manual operation. For better use of a humanoid manipulator, a design method of tactile sensor of humanoid manipulator is proposed in this paper. The method takes PVDF piezoelectric film as the sensing material of the sensor, and the output pressure signal of PVDF is processed by a signal conditioning circuit to realize piezoelectric conversion, filtering and amplitude adjustment, then different morphological characteristics of the tactile and slip signals are analyzed through data processing. The experimental results show that the tactile sensor designed in this paper can effectively analyze the tactile signal and the slip signal in three-dimensional space.

Aimed at promoting the engineering applications of risk-based maintenance, a design scheme of weak link identification system for transmission grid is proposed. The equipment-grid relevance model established by CIM data analysis and association set decomposition is a major innovation of the system. This model can coordinate the functional dependence among different devices, which helps to formulate more efficient overhaul schedules. Operational risk caused by association set outage is evaluated based on outage model, contingency screening and load shedding optimization. The evaluation results are used to identify weak links and guide maintenance schedules. It is proved in the application case that the proposed system not only helps the maintenance department, but also reduces the burdens of operation dispatching.

In the conventional CNC servo feed system, using traditional PID control method is difficult to control the tracking error reduced effectively and quickly, so its result of processing geometric accuracy is not satisfactory. The human simulated intelligent PID control method is introduced to the servo feed control system in the paper. The transfer function of the servo driver system model is derived and the human simulated intelligent PID control algorithm is designed. Simulation results show that the human simulated intelligent PID control method has the advantages of good stability and robustness in the servo driver control system, which can effectively inhibit the influence of the tracking error.

The single-phase power electronic traction trans­former (PETT) is a power device to realize the higher efficiency and reduction of weight in railway traction system. The single-phase cascaded H-bridge rectifiers (CHBR) are an important part of the single-phase PETT. In order to ensure its reliability, an open-circuit fault diagnosis method based on the mixed logical dynamic (MLD) model and the state estimator is built for the open-circuit fault of a single IGBT. The MLD model and the state estimator of the CHBR are constructed under both normal and faulty conditions, and the residual characteristics are analyzed in the two states. Fault diagnosis is achieved according to the residuals including fault information. The proposed method requires no additional sensor. Experimental results verify the availability and accuracy of the proposed method, and the location of the fault can be accomplished in a fundamental cycle.

With the improvement of computer performance, particle filter as a method of state estimation, is widely used for nonlinear non-Gaussian dynamic systems. In order to reduce the partical degeneracy in the state estimation process of high-dimensional stochastic dynamic systems which commonly occurs in the conventional particle filter, an improved particle filter method based on adding more accurate direct observations of some dimensions of the system state is proposed in this paper. By arranging a plurality of sensors in the stochastic dynamic system which are able to directly and precisely measure some dimensions of the state, lots of accurate state information can be obtained. Combined the information with the inverse process of the system measurement, the particles move towards the high observation likelihood region, so that the effectiveness and accuracy of resampling process can be greatly enhenced. In the case study, a stochastic dynamic system of near-surface wind farm is established by the theory of computational fluid mechanics. Numerical results show that the proposed PF method is able to enhance the accuracy of the estimation for high dimensional stochastic dynamic system.

A robust and effective evaluation method for the safety of optimal generation scheduling has great potential on optimizing operation of electrical power and energy system. Based on the concept of transmission capacity margin, a robust model of transmission capacity margin evaluation (TCME) is established to assess the overload risk of transmission lines when loads fluctuate. This TCME method can provide the system operators with useful information such as possibly overloaded lines and time periods when lines are potentially overloaded under the uncertainties of nodal loads. However, TCME formulation has a multi-level (min-max) structure and the solution methods are usually complicated. In order to overcome the computational complexity of original problem, an All-Scenario-Feasible optimization is utilized. It is established based on very limited scenarios while the solution is feasible to all scenarios. The proposed formulation is single-level rather than multi-level. So no iterative algorithms are needed and commercial solver like CPLEX could be directly adopted. In order to use the proposed formulation for large-scale power grid, considering that the transmission capacity constraint is the most important one in all problem constraints, we further reduce the number of scenarios. Numerical testing is also performed and the results show that the proposed formulation is effective.

This paper introduces the composition of continuous emission monitoring system, and the basic concepts of the artificial intelligent technologies. According to the engineering practice and the basic concepts, authors found that the sampling handing system is prone to malfunction. Furtherly, the sampling probe unit is more vulnerable, so PLC control system of the probe is detailed analyzed. In addition, the authors try to realize fault diagnosis based on expert system. Therefore, based on field experience, most likely failure points were discussed in this paper. As the core of expert system for fault diagnosis, this pre-diagnosis method is very helpful to the acquisition of knowledge. In the end, the desirable properties of CEMS control and diagnosis frameworks were discussed for further research and development.

Currently, big data and its applications have become one of the emergent topics. knowledge can be extracted from a high volume of information by using big data technologies. In practice, MapReduce framework and its different extensions are the most popular approaches for big data. Among the different approaches, those models based on fuzzy systems stand out for many applications. Fuzzy set theory allows for the inclusion of variety and veracity in big data computing problems. However, when a given observation has no overlap with antecedent values, no rule can be invoked in classical fuzzy inference can also appear in big data environment, and therefore no consequence can be derived. Fortunately, fuzzy rule interpolation techniques can support inference in such cases. Combining traditional fuzzy reasoning technique and fuzzy interpolation method may promote the accuracy of inference conclusion. Therefore, in this paper, an initial investigation into the framework of MapReduce with dynamic fuzzy inference/interpolation for big data applications (BigData-DFRI) is reported. The results of an experimental investigation of this method are represented, demonstrating the potential and efficacy of the proposed approach.

Multi grid-connected inverters islanding detection methods include two categories: passive islanding detection and active islanding detection. Firstly, an active islanding detection method based on negative sequence current disturbance is analyzed. Aiming at the shortcomings of this method, a novel passive islanding detection method is proposed in this paper. The proposed islanding detection method uses wavelet transform to extract signal features, and then according to the extracted signal features adopts the neural network to judge whether or not the islanding occurs. The simulation results show that the passive multi grid-connected inverters islanding detection method has good islanding detection performance, and does not affect the stability and power quality of the system.

To meet the requirements of ultra-high speed protection on UHV transmission lines, this paper proposes a fault phase selection based on wavelet analysis theory, which can detect the mutations' time and size of singular signal in fault signal. Firstly, this algorithm transforms the transient fault current into model fault current to eliminate the impact of the various phases' coupling. Then modulus maximum is extracted from model fault current using discrete wavelet transform (DWT) and is treated as the criterion factors after normalizing. Finally, the relationship among criterion factors is summarized according to the characteristics of mode fault current, while constructing a new UHV fault phase selection algorithm. Through simulation studies by PSCAD and MATLAB, this method is proved that it is disturbed less by amplitude of the transient fault signal, and it has a good adaptive capacity and a strong ability for fault resistance. Also its speed is fast and accurate, so the proposed algorithm can be implemented simply and effectively.

As location-based services (LBSs) become popular, location-dependent queries have raised serious privacy concerns since they may disclose sensitive information in query processing. Among typical queries supported by LBSs, shortest path queries may reveal information about not only current locations of the clients, but also their potential destinations and travel plans. Unfortunately, existing methods for private shortest path computation suffer from issues of weak privacy property, low performance or poor scalability. In this paper, we aim at a strong privacy guarantee, where the adversary cannot infer almost any information about the queries, with better performance and scalability. To achieve this goal, we introduce a general system model based on the concept of Oblivious Storage (OS), which can deal with queries requiring strong privacy properties. Furthermore, we propose a new oblivious shuffle algorithm to optimize an existing OS scheme. By making trade-offs between query performance, scalability and privacy properties, we design different schemes for private shortest path computation. Eventually, we comprehensively evaluate our schemes upon real road networks in a practical environment and show their efficiency.