Guifang Guo

The application of combinatorial optimization by Genetic Algorithm and Neural Network

A optimization model of sizing the storage section in a renewable power generation system was set up, and two methods were used to solve the model: genetic algorithm or combinatorial optimization by genetic algorithm and neural network. The system includes the photovoltaic arrays, the lead-acid battery and a flywheel. The optimal sizing can be considered as a constrained optimization problem: minimization the total capacity of energy storage system, subject to the main constraint of the loss of power supply probability (LPSP). Both of the two optimal algorithm got good results. We can see that, combinatorial optimization by genetic algorithm and neural network can lessen the calculation time, with the results change little.

Torque ripple control of position-sensorless brushless DC motor based on neural network identification

In order to reduce the torque ripple of position-sensorless brushless DC motor (BLDCM), Based on analyzing the commutation process, a novel control system employing back-EMF method was designed, which disconnected the reference point of detection circuit from battery cathode and did the phase-shifting compensation of back-EMF. Moreover, through regulating the terminal voltage of motor, the system made the rising ratio and dropping ratio of the phase currents be approximate so as to keep the amplitude of the total current in the constant. To further suppress the torque ripple, neural network (NN) control algorithm was researched and applied to the system. The controller comprises a back propagation (BP) NN and a radial basis function (RBF) NN. The former is used to adaptively adjust the parameters of the PID controller on-line. The later is used to establish nonlinear prediction model and perform parameter prediction. The experimental results show that the proposed method in this paper could ensure prominent reduction of torque ripple, have good robustness, and achieve position-sensorless commutation control of BLDCM successfully.

Prediction state of charge of Ni-MH battery pack using support vector machines for Hybrid Electric Vehicles

This paper investigates the use of a support vector machine (SVM) to predict the state of charge (SOC) of a large-scale Ni-MH battery pack in hybrid electric vehicles (HEV). Estimate the state of charge (SOC) is very essential for HEVspsila energy monitoring and management systems. The nonlinear SOC dynamics is represented by a nonlinear autoregressive moving average with exogenous variables (NARMAX) model that is implemented using SVM regression model. Accuracy of the presented SVM method has been verified by UDDS and US06, which a composite aggressive driving cycle provided by U.S. Department of Energypsilas Hybrid Electrical Vehicle test program. The results showed that SVM is able to estimate the SOC with high accuracy and high noise tolerating ability.

A novel fore axle whole-turning driving and control system for direct-wheel-driven electric vehicle

Based on comprehensive analysis of differential principle, a novel driving and control system for a direct-wheel-driven electric vehicle (EV) is presented in this paper. Without differential gears, direct-wheel-driven EV has to employ complicated differential algorithm for steering, which remarkably increased the difficulty of the control system of EV. The system proposed in this paper has changed the traditional structure and control method of direct-wheel-driven EV to simplify the control of EV. The propounded system connected two permanent-magnet brushless dc motors (PMBDCMs) directly in series, simplified the differential algorithm thoroughly in velocity control strategy, and adopted the traditional PI controller. By analyzing the dynamics of Four-wheel independent drive electric vehicle and combining the characteristic of the PMBDCM, the operation principle of the propounded system is explicated. Based on the analysis, an equipollent simulation system was developed under Matlab/Simulink platform, and validated the rationality and feasibility of the driving and control system for direct-wheel-driven EV. The particular performance of the propounded system made direct-wheel-driven EV more practical and competitive.

Application of Parallel Chaos Optimization Algorithm for Plug-in Hybrid Electric Vehicle Design

Plug-in hybrid electric vehicles (PHEVs) have been offered as alternatives that could greatly reduce fuel consumption relative to conventional vehicles. A successful PHEV design requires not only optimal component sizes but also proper control strategy. In this paper, a global optimization method, called parallel chaos optimization algorithm (PCOA), is used to optimize simultaneously the PHEV component sizes and control strategy. In order to minimize the cost, energy consumption (EC), and emissions, a multiobjective nonlinear optimization problem is formulated and recast as a single objective optimization problem by weighted aggregation. The driving performance requirements of the PHEV are considered as the constraints. In addition, to evaluate the objective function, the optimization process is performed over three typical driving cycles including Urban Dynamometer Driving Schedule (UDDS), Highway Fuel Economy Test (HWFET), and New European Driving Cycle (NEDC). The simulation results show the effectiveness of the proposed approach for reducing the fuel cost, EC and emissions while ensuring that the vehicle performance has not been sacrificed.

The combinatorial optimization by Genetic Algorithm and Neural Network for energy storage system in Solar Energy Electric Vehicle

We investigated the optimal sizing of the energy storage system in a solar energy electric vehicle (SEEV) system. A model system was constructed for this that includes the photovoltaic power, the lead-acid battery and a flywheel.The optimal sizing can be considered as a constrained optimization problem: minimization the total capital cost of energy storage system in SEEV, subject to the main constraint of the loss of power supply probability (LPSP). The genetic algorithm or combinatorial optimization by genetic algorithm and neural network were used in this paper. And the decision variables are not only the capacity of batteries in traditional methods, but also the capacity of flywheel. Studies have proved that the optimization algorithms used can converge well and they are feasible. Combinatorial optimization by genetic algorithm and neural network can lessen the calculation time, with the results change little.

Optimization of Ni-MH Battery Fast Charging in Electric Vehicles Using Dynamic Data Mining and ANFIS

Fast and efficient charging of Ni-MH battery is a problem which is difficult and often expensive to solve using conventional techniques. This study proposes a method that the integrated data mining algorithm and the Adaptive Network Fuzzy Inference Systems (ANFIS) for discovering the fast charging more efficiently and presenting it more concisely. Because the battery charging is a highly dynamic process, dynamic data mining technique is used for extracting of control rules for effective and fast battery process. The ideal fast charging current has been obtained. The result indicates that the integrated method of adaptive charging current has effectively improved charging efficiency and avoided overcharge and overheating.

Torque coordinated control of independent driving electric vehicles base on BP neural network

Based on the analysis the driving system of motors in series, it indicated that the system simplifies the structure and realizes the effective self-action differential when the electric vehicle(EV) runs on the level road without any variety or at low velocity. When the EV runs on the variational road, especially the friction force is variant, the EV is unstable according to the character of the permanent-magnet brushless DC motors (PMBDCMs). In this paper, the structure was improved and a BP neural network PID control strategy was proposed to control the coordinated driving torque. In order to enhance the rapidity of convergence and avoid trapping in local optimization, the control system improved BP algorithm and momentum factor is adapted to effectively solve the problem. A model was built for the 2-wheel independent driving EV and when the friction force of the two power wheel is different, the simulations .. were performed and the positive results were obtained by the proposed concept. The propose system improve the EVpsilas maneuverability and stabilization.

Estimation the residual capacity of Ni-MH battery pack using NARMAX method for electric vehicles

This paper presents a nonlinear autoregressive moving average with exogenous variables (NARMAX) method to estimate the residual capacity of high-capacity Ni/MH battery pack in electric vehicles. The state of charge (SOC) represents the battery residual capacity. The SOC of battery cannot be measured directly and estimated from measurable battery parameters such as current and voltage. The proposed NARMAX produces accurate SOC estimate, using industry standard Federal Urban Driving Schedule (FUDS) aggressive driving cycle test procedures. The results indicate that the NARMAX can provide an accurate and effective estimation of the SOC, resulting in minimal computation load and suitable for real-time embedded system application. The maximum average relative error of the estimating results is 0.02%.