ANN Based Current Controller For Hybrid Electric Vehicles

E3S Web of Conferences 309, 01065 (2021)ICMED 2021https://doi.org/10.1051/e3sconf/202130901065 ANN Based Current Controller for Hybrid Electric VehiclesKavati.Nagendar1,*, and V.Vijaya Rama Raju21PGStudent, Department of EEE, GRIET, Hyderabad, India.Professor, Department of EEE, GRIET, Hyderabad, India.2AssociateAbstract. The use of Hybrid Electric Vehicles (HEVs) across the world is growing enormously everyday. The single-phase bi-directional convertors are presented in this study for HEVs on-boardcharging(OBC). In HEVs, we use power electronics converters for the converting and invertingoperations. Artificial Neural Network(ANN) is presented in this study for simple operation and highoptimization approaches. ANN control technique regulates the system's THD and enhances chargingsystem optimization, enables two-way power delivery that is from the grid to vehicle and the vehicle togrid. An ANN based current controller model that achieves fast-dynamic reaction and that improves gridcurrent harmonic characteristics is proposed in this study. The system's THD is reduced by the ANNcontroller being suggested. The results prove the validity and feasibility of design and control technique ofthe proposed integrated charging system.Keywords: DC-DC bi-directional converters (BDCs), Artificial Neural Network algorithms, Electric Vehicles.1. IntroductionHybrid Electric Vehicles (EVs) and Electric Vehicles(EVs) are generally the vehicles that must meet thestandards for urban transportation emissions during thenext 10 years. HEVs are a hybrid between traditionalcombustion engine vehicles and electric vehicles(EVs)[1], a compromised alternative that addresses issuessuch as air pollution, petroleum combustion, and limiteddriving range[2]. HEV [3] is made up of variouscomponents such as machines, gearbox, battery, charger,traction engine, and its driving inverter. A second powergenerator and its drive inverter are also included insystem. The starter-generator system's goal is to startengine from a standstill and transform vehicle's kineticenergy into electric energy [4-5]. In this study, startermechanism is modified to act as battery charger, rather ofOBC, which is found in most HEVs (OBC). As result,upgraded start-generator system employs two distinctmodes. The first is a motor drive that acts as a typicalstarter-generator. The charging battery is the other.Because this integrated charging method substitutestraditional OBC, it is possible to eliminate traditional OBCfrom HEVs. As a result, the redesigned circuit reduces thevolume and weight of HEVs while also increasing powerdensity. A feedback-based current control technology,such as proportional-integral controller, has emphasisedthe power converter's performance in recent decades[5-6].The PI current controller achieves fine control results bytranslating time variable on synchronous frame into timevariable. Some of the flaws in this control strategy,however, are due to the controller's system parameterdesign. Furthermore, designing acontroller for a sophisticated system is difficult. ModelPredictive Current Control (MPCC) is introduced in thisstudy as method for controlling 1-phase full-bridgeinverter when integrated circuit is operating in batterycharging mode. For high total harmonic distortion (THD)in output current, the typical MPCC produces eightseparate voltage vectors because single phase two-levelinverter only makes three distinct output voltages. THD inoutput current can be reduced by increasing samplingduration and reducing calculation time. Harmonic featuresof output current are limited and increased using anadvanced MPCC method. Compared to typical feedbackbased control method, the MPCC methodology has afaster dynamic response time. Output current has superiorharmonic characteristics when compared to MPCC'snormal method. Simulation results confirm the accuracyof proposed integrated charging design and controlstrategy of the system. There are several industries thatlook to bidirectional DC-DC converters (BDCs) as energysources to improve the transmission and distribution ofenergy. As a result of system losses, efficiency is reducedand circuit elements' lifespan is shortened. Deshalb solltenin order to avoid power loss, better control systems beimplemented whenever possible. A bidirectional powerflow's efficiency is crucial to obtaining high-qualityelectrical energy. To improve the efficiency of theconverter, a variety of control mechanisms are currentlybeing explored. One of these control approaches is theartificial neural network (ANN), which is based on avariety of algorithms. Vector support (SVM) andautonomous map (SOM) approaches are used in ANN tofeed forward information (FF). They are divided into two The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0(http://creativecommons.org/licenses/by/4.0/).

E3S Web of Conferences 309, 01065 (2021)ICMED 2021categories, supervised and unsupervised. To determinethe most efficient input voltage for the converter, variousways are utilised, and one of these techniques is anacceptable algorithm. ANNs are not often used inelectronic equipment for power regulation, according tosome study. Matlab simulates a single-phase Full-BridgeIsolated BDC with input voltages ranging from 100V to400V and an output voltage of 160V. Parallel to theIGBTs, this type of converter also incorporatescondensers, which help to improve the efficiency of thesystem. In supervised or uncontrolled learning, SOM,SVM, and FF identify the data set of the converter thathas been taught to provide low error levels. In the thirdsection, the ANN method is discussed, the optimal inputvoltage is found, and input voltage classification isdisplayed. This research presents a new method forforecasting adaptive dc connection voltage using theSOM algorithm. In this approach, artificial neuralnetworks (ANNs) are utilised to forecast and adapt thereference dc connection voltage quickly. Further, thesystem will feature both supervised and unsupervisedcontrol options. In the performance analysis, both theproposed controllers and alternative control methods ofANN algorithms were examined. To conclude,MATLAB simulation software is used to verify whetheror not the new control strategy has been effective.2.Starter-Generator BasedIntegrated Charging System.A standard HEV is composed of a generator, a drivinginverter, a battery, and an OBC. Thus, higher powerdensity is one of major significant obstacles todevelopment of HEVs. This study discusses integratedcircuit's battery charging and motor driving capabilities.As seen in Fig1, proposed circuit consists of two-levelinverter, a generator, and seven power relays. Theadvantage is that starter-generator drive system's powerrating enables charging of battery. When this systemdesign is applied to traction motor circuit, the high powerrating of traction motor circuit can be used to 2130901065 B.Bi-Steer OBC ModeAs illustrated in Fig. 1(b), integral circuit is modified toprovide a single-phase, two-way OBC circuit. Grid filterreactor, full-bridge ac-dc converter, and battery compriseOBC method circuit. Relay 05 and 07 will be triggered inorder to employ starter generator winding as filter reactor forDC-DC conversion. Then, the grid will be connected to fullbridge conversion unit via relays 01 and 02, which will beON. The corresponding induction equivalent of the filter is1.5 times that of generator's single winding. On this OBCcircuit, two alternate transmission statuses are provided. Whenpower is delivered to the car from the grid, the battery ischarged and the state is referred to as G2V mode. Rather thanthat, V2G mode uses the battery's energy as a standard energystorage device.3.ANN-BASED CONTROLLER.The ANN controller has two outputs one for voltage levelcontrol via the grid corrector and another for current controlvia the Buck/Boost conversion to maintain load current. Gridtension and DC-link tension control are used to controltension controlled system. As result, some parameters areestimated as critical inputs to controller. To begin, there is anerror between DC- estimated link's voltage and establishedreference voltage. Calculated error is then communicated toANN control unit by establishing corresponding shift angle.Second, from immediate grid voltage, direct and quadraturevoltages are generated. Real voltage, in comparison to desireddirect voltage, an error that specifies amount of modulationsignal required between two as input to ANN controller forcontrolling DC/DC converter. Critical operating equations forbucking and boosting activities can be calculated by (1) to(4) for bucking and (5) to(9) for boosting operations, respectively.(1)(2)(3)(4)Drive Mode Starter GeneratorThe mode in which the starting generator is driven isdepicted in Fig. 1. (a). In this mode, relay 06 is enabledto connect the battery to the three-phase inverter. DC linkand battery share same electrical node, and batterypowers three-phase inverter that drives starters. Whenrelays 03 and 04 are active, inverter is able to supplygenerator set with three phase electricity. This mode isidentical to oldest starter-generator, which drives orbreaks vehicle using regenerative energy.2

E3S Web of Conferences 309, 01065 (2021)ICMED 2021https://doi.org/10.1051/e3sconf/202130901065 (a)(b)Fig. 1. Integrated charging system construction: (a) drive mode of the starter generator. (b) single-phase OBC mode.Fig.2 ANN controller block diagram(5)(6)(7)(8)*Corresponding author: kavati.nagendar@gmail.com3The current controlled loop establishes the workingmode of two-way buckboost converter by utilisingbattery reference current. The calculated batterycurrent inaccuracy is used to calculate the duty ratioof the power output of ANN control unit. The tariffratio is communicated to PWM modulator, whichmodulates switching signal of IGBT. Low errorbetween objective result and network reference isrequired for ANN training. This training is carried outin this study utilising a well-established techniquebased on Levenberge- Marquadt algorithm. For thischallenge, device created here has two inputs and twooutputs. The error equation for particular node isequated by taking into consideration the pth layer of

E3S Web of Conferences 309, 01065 (2021)ICMED 2021https://doi.org/10.1051/e3sconf/202130901065 the kth neuron connected to the nth neuron of theprevious layer (9). Equation illustrates summativeerror Equ(10).(9)(10)The algorithm for updating the weight of any object isgiven by:(11)the parameters are LM TrainingWhereresults demonstrate a perfect fit with low errors for theANN network.(a)Results :MPCC Method of controlling.(b)Fig.4. Integrated charging system simulation waveforms(a) voltage control DC-link (b) Battery voltage control(a)(b)Fig. 3. MPCC method simulation results: (a) currentconventional MPCC control method. (b) current MPCCcontrol method.(a)4

E3S Web of Conferences 309, 01065 (2021)ICMED 2021https://doi.org/10.1051/e3sconf/202130901065 (a)(b)Fig.5. The simulation results shows current THD is 22.84% and3.25% in the conventional MPCC method ofcontrol.Proposed ANN Controller :(b)Fig.7. Simulation waveforms of the integrated charging system (a)control voltage of DC-link (b) Battery voltagecontroller(a)(a)(b)Fig.6.Grid currents (a) starter-generator ANN controlcurrent (b)OBC ANN control current monitoring.5