Analyzing eDrive Performance in Electric Vehicles Using MATLAB and Simulink - MATLAB & Simulink
Video Player is loading.
Current Time 0:00
Duration 30:53
Loaded: 0.53%
Stream Type LIVE
Remaining Time 30:53
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected
    Video length is 30:53

    Analyzing eDrive Performance in Electric Vehicles Using MATLAB and Simulink

    Alex Edwinson Davidson, Bosch Global Software Technology Pvt. Ltd.

    Electric vehicle (EV) systems consist of components that interconnect in a coordinated manner. The HV battery, traction inverter, electric motor, and transmission system are the major components in an EV. Among these, the traction inverter and the electric machine together are known as the eDrive. The performance of the eDrive in EVs is crucial for several reasons. First, it directly impacts the acceleration and speed capabilities of the vehicle. A powerful and efficient eDrive system ensures that EVs can achieve adequate acceleration and maintain higher speeds, providing a more enjoyable driving experience for users. Second, the performance of the eDrive influences the overall range and efficiency of the EV. A well-designed eDrive system can maximize the energy utilization, enabling the vehicle to travel longer distances on a single charge.

    This session presents a solution using MATLAB® and Simulink® to analyze the performance of the eDrive. The thermal (power module temperature and electric machine temperature), electrical (voltage and current), and mechanical (torque and speed) parameters are chosen to analyze the eDrive performance.

    Published: 22 Dec 2024

    Yeah, once again, thanks for your presence here. And before I start with, I would like to thank MathWorks for giving us a great opportunity to present and discuss our work here. So thanks.

    So without any further delay, let me jump into the topic. So today's discussion going to be focusing on how overall performance of an eDrive system, as has been analyzed, and how the MathWorks plays a vital role in here.

    And looking into the agenda, the main discussion has been classified into five subtopics. So one is the eDrive system. Since the overall focus is going to revolve around the eDrive system, I would like to give some macro level information of what exactly an eDrive system and what are its the system components looks like. And it's followed by an eDrive simulation model. So here I will be discussing upon how the overall eDrive system has been virtualized and how the MathWorks solutions has comes into the picture and how it has been used for the virtualization.

    Then thirdly, we have eDrive system analysis, where, with the help of the model, what are the different analysis that are capable and being performed in the eDrive level that are being done in the Bosch and followed by-- we have use cases where I'll be taking upon a couple of use cases in the actual system development process. And I will be explaining in detail how system analysis, along with the eDrive model, has been useful in order to attain the necessary outcome. And finally, summarization of the overall discussion.

    So eDrive system-- so eDrive, it is nothing but briefly called as electric drive. And if you take any powertrain architecture, it is one of the prominent part where it is the major role is to convert the electrical energy into mechanical energy, which will be in the form of the torque and the rotational speed of the shaft, which will be translated to the wheels through the gearbox and transmission, which will finally lead to the propulsion of the vehicle. And apart from that, if you take generally, we call, somewhat one or two subcomponents or major components, if they independently and they combinedly work together in order to attain one specific or one or more outcome, we generally call this a system.

    In similar way, the eDrive system also consists of several major components, which functions together in order to attain its necessary functionality. So the major components of the eDrive systems are e-machine, which is nothing but electric machine, which is just the actual real work of converting the electrical energy into mechanical energy. And we have the inverter where it converts the DC power into an AC power.

    And followed by, we have software and algorithm, which is nothing but the controller, which is necessary for the optimal functioning of the system with efficient manner. And finally, we have the sensor, which are necessary in order to sense the real environmental physical domain information and convert them into a digital form, and which will be used for the different controller perspective for the proper functioning of the system.

    So, so far we discussed like what is an eDrive system, and what is the important role in the overall powertrain architecture, and what are the different components it looks like. Now let us see how the overall eDrive system seems to be virtualized and how MathWorks seems to be used for that particular model development.

    So here, the whatever they discussed, eDrive system is completely virtualized and modeled in the MATLAB environment. And the completely modeling has been done by using the MATLAB Simulink and system components. And here I also, from this slide, we can see an overview of how we have developed the overall eDrive system by virtualizing.

    And if we go from our right to left, we have eMachine block, which is where the actual eMachine has been modeled here. And the modeling, we have completely modeled using the custom Simscape language instead of using an actual eMachine block set because it has its own necessity. So based upon that.

    But I will just go through-- in upcoming slides, I will discuss like what's the necessity, why we went with the Simscape language. And followed by, we have the inverter block where the complete inverter has been modeled using the Simscape components, predominantly electrical Simscape components. And followed by, we have the in-house-developed Bosch software. And finally, we have the sensor blocks and the battery model.

    And for the overall proper functioning of this eDrive model, we basically needed three important inputs. One is the desired torque and the desired speed and the battery voltage. And one additional information is like whatever the sensor blocks and battery model we are looking here. It is modeled in a very abstract level.

    For example, in the battery model, we didn't consider any cell-based model because the necessary information, the battery voltage, is already getting as an input. So no necessary parallel electrical parameters, which is needed for the network of the overall system that has been implemented in the battery model. And additional features of our eDrive model is we can take these input information in from drive cycle inputs as well. On top of that, we also have the capability to directly feed through us the necessary torque, speed, and voltage inputs as well.

    So let's have a little deeper look into how the specific eMachine model and inverter model has been developed from our end. And from this slide, you can see the complete-- one of the example, which we have showcases the PMSM model, which has been developed in the Simscape. And other features by using the Simscape language is we can able to develop our own custom component.

    One of the reasons is, what are the blockset? Yes, MATLAB has its own machine blockset. But the necessity of having the customly-built eMachine, which is have the capability to adapt not only from our own standard way of incorporating the electrical parameters, on top of that, it also enables to take up the OEM-m specific parameters which will be converted into our own standard way of implementing it.

    And another feature is we have model not only for PMSM, we also have the similar of Simscape modeling of the ASM as well. So it has a capability to select the different variants, whether it is ASM or PMSM-- depends upon the project requirements.

    And in the right hand side, you can also see we have also validated the Simscape model of the machine. For example, here we have considered the PMSM machine here. And from the results, we can able to see the behavior of the torque, speed, and necessary voltages are very much close to the measurements here.

    Coming to the inverter model-- so here the complete inverter model has been modeled using the Simscape electrical components, as specifically the power electronics switches here. And thanks to the capabilities of the MathWorks Simscape electrical components, specifically the powertrain switches, which has the capability to incorporate the actual characteristics of the real hardware like IGBT or MOSFET for the powertrain perspective. So because of these capabilities, which makes us ease to use those information and model it into our separate inverter model, and it enables us to do the analysis, which gives in the whatever the behavior, or the analysis we are doing, it will be in a line with respect to the real environment conditions.

    And on top of that, we also have the modeling from IGBT, as well as MOSFET perspective. And depending upon the project requirements, we have the capability to select which variants needs to be considered. And here in the right hand side, I also showing one of the mapping, taking one of the IGBT as an example, and how the necessary information from the datasheet can be incorporated into our Simscape electrical component of the electronic switches, and how it can be used for our further analysis.

    And we also have validated if we are incorporating necessary datasheet information into our model, how the behavior looks like, and whether it is close to the actual data sheet of the OEMs or not. And here I just showcased a simple model, which we have considered for the validation of the necessary characteristics. And previously mentioned IGBT as an example we took, and here we can see in the right hand side, the comparison of the-- characteristics of the powertrain switches between the simulation and the datasheet for the output casting and the transfer for a specific boundary conditions. So, so far, we have discussed what exactly in the eDrive system, and how we have totally, completely modeled or visualized the complete eDrive system, and how MathWorks seems to be very much useful in order to incorporate and make up those virtualization.

    Now that we can see how these models can be used for the overall analysis. And what are the different analysis that are capable and performed within BGSW? So I have classified the overall several-- we are capable of doing many analysis. And I have categorized all those analysis separately in four categories.

    So one is the system analysis. So right now, we have a complete eDrive model, which is fully well-validated with the actual measurement. On top of that, it also has the capability to incorporate the necessary actual characteristics of the hardware. And with the help of those virtualized model, in case if you want to have any change in the requirement or any change in the design, we can able to analyze what will have an influence in the overall system level so that what we shortly call the cause/effect analysis.

    And followed by-- since we have the overall plan model, which is inverter and emission, and it is also well-validated-- and anything which we want to incorporate new controller algorithm or new advanced controller algorithm. And if we want to see how optimized it is, how efficiently it working upon, we can make use of such models to evaluate it.

    And finally, we can do the sensitivity and tolerance analysis of eDrive, which means in the eDrive model, we are not considering only the software. We are also model the hardware. And model-- the hardware also has its own tolerances. And these tolerances also will have influences in the overall working of the system. And with the help of the model, which is very much well-validated, we can depend upon the results. And we can also see how those tolerances will have the influence in the overall system level or also the component level.

    And coming to the next, which is the performance analysis. And here on top of the complete eDrive model, we also have the power loss model and thermal loss model, which is very much in-house developed models. And with the help of those models, we have the capability to analyze the component-level and the system-level power losses.

    And along with that, with those models, with the drive cycles as an input as well, we have the capability to analyze the overall efficiency of the eDrive system in the system level, as well the component level losses. On top of that, since we are using the powertrain switches, and since it has the capability to take the actual characteristics of the hardware, which makes us to ease to analyze the impact of different modulation strategy. On top of that, what will be the impact of the different switching frequency on the overall system level, which can be related to the several power losses in the overall system and how the overall efficiency of the system looks like?

    And thirdly, we have the fault analysis. So since we have the modeling of the plant model with the Simscape, it gives us an ease to do several fault analysis, like the complete open circuit, short circuit analysis, and also the [INAUDIBLE] voltage network failure analysis. And on top of that, we can also able to analyze different overvoltage or overcurrent analysis.

    For example, if you are taking a solver, which has a sensor to sense the angle information, and any sudden change or a sudden oscillation happens in the angle, that definitely will act as a-- will cause imbalance in the overall system behavior. And we have the ability to analyze how the small change in the shift in the angle will have the impact in the overall system level or from the torque ripple perspective. Or also it can lead to the overall overcurrent-- also if you are trying to maintain a peak operating points.

    And finally, thanks to the use of the Simscape component of the powertrain switches, we have the capability to analyze what will happen if any one of the inverter switches go on failure, or what happens when one of the gate driver go on failure, and how that will have influence on the overall system level, whether it leads to the sudden rise in the current, or how the torque profile looks like, and how long it takes for the system to come to the safe state due to if any overcurrent happens. Those kind of analysis from the hardware failure, also it is capable with the use of this model.

    And finally, we have thermal analysis. And as I mentioned previously, the thermal models, which we have considered in the model, it is an in-house developed model. And with the influence, with the use of those models, we have the capability to analyze what is the overall steady state and transient state, thermal rise in the overall system.

    And on top of that, we have the capability to come up to develop a direct strategy of the system. We know order to maintain the overall performance of the system for a longer period of time, we also need to decide, based upon the temperature, as well as the current constraints. So maintaining a temperature for a certain operating point and maintaining that particular operating point for a certain longer period of time, it is very much important.

    So with considering those constraints, we can able to develop a directing strategy. Like, if the overall temperature system rises higher or lower, then we need to reduce the overall working temperature so that the performance of the system is not violated. So those kind of strategies also we can able to develop. And for those models, the thermal models, and the overall eDrive models have been very much helpful here.

    So, so far, we have discussed about how the overall eDrive system has been virtualized, and what are the several system analysis that are being capable with the use of those models. So let me take upon a couple of use cases where it has been done in the actual system development process, how the system analysis and the eDrive model, which we have virtualized, are very much useful here.

    And first use cases I would like to discuss upon is the capacitancing of the dieseling capacitor. We know if you want to design an inverter system, one of the important, prominent component is the capacitor. As we know, we need a capacitance. We have the DC source as input.

    And in order to convert it into AC through the powertrain switches, we will be using the powertrain switches for it. And here we'll be using a capacitor, which acts as a bridge between those things. And on top of that, we also need those capacitors. Why? Because we know, in order to convert the DC to AC, we'll try to use the powertrain switches to convert them. And they will be trying to do on and off of the actions during the switching frequency.

    And these turn on off of the actions, which leads to the ripple voltage and current ripples across the capacitor, which can be translated to the emission as well. And it is very much needed to have those ripples as minimum as possible. Why? Because these kind of ripples will lead to overall unstability in the controller. On top of that, it also leads to the creation or increase in the overall EMI. And also in the presence of parasitic components in the HP network and use due to this effects of the ripples, it also leads to the overall increase in the temperature as well.

    So it is very much needed. If you are selecting one dieseling capacitor, the constraints for what the selection will be the voltage ripples. And the capacitance, the dieseling capacitor, which we are trying to size up, needs to have the capability to suppress all these kind of effects.

    So, so far we have got to know what is the dieseling capacitor and why its need in the inverter designs. So let's see how these capacitors are being sized based upon the different factors. Yeah. So basically, as I mentioned before, the selection of the capacitance of the dieseling capacitor is mainly due to the voltage ripples. So if you are trying to size it up based upon the ripple criteria or based upon some standards, then we need to see what all the different factors that influences the overall voltage ripple.

    So here we have four prominent factors which influence the overall voltage ripples across the capacitor due to the switching actions. So the first one is basically the capacitance itself. So if you look from the basic voltage equation of the capacitor, we can build up a relationship between the capacitance and the voltage ripple. As we can see from the equation, if the voltage ripple is inversely proportional to the capacitance of the capacitor. So which means if you want to reduce the voltage ripple, definitely we need to have the higher capacitance. But another catch here is if you are having a more capacitance, the size of the capacitor increases, which also leads to the cost. So these factors also considered when you are selecting a capacitor based upon the voltage constraint.

    So another prominent factor is the switching frequency. So the basic reason for having the voltage ripple is due to the action of on and off of the powertrain switches. And these on and off happens in certain frequency, which we call switching frequency.

    So if we try to correlate the switching frequency in terms of voltage ripple-- and we can have a relationship, something like the sync frequency is inversely proportional to the voltage ripple. And from that, we can come to a conclusion-- OK, if we want to have a reduction in the voltage ripple, definitely we need to increase the switching frequency.

    But similar way, here is another catch. If you are trying to increase the overall switching frequency, definitely the losses across the inverter switches will be increases. That leads to the overall losses in the system, which will also lead to the efficiency of the eDrive as well. So these factors also needs to be considered when we are trying to select a capacitor based upon switching frequency.

    Next comes the stator current. And as we know, if you are trying to demand more load or less load, definitely we'll be trying to take up more current from the source. And these translated to the more current of charging and discharging across the capacitor.

    And we can have a correlation between how these load current correlated to the capacitor current. And from that, we can also correlate how that will have influences on the voltage ripple. So here we have an equation, which has been expanded in terms of modulation index and stator current.

    And as we can see from the relationship, if you are trying to increase the mode demand, then definitely the stator current will be increases. And as we increase the stator current, definitely the capacitor current also increases. And if we just try to correlate with the voltage equation of the capacitor, we can see increase in the capacitor current definitely increases the voltage ripple. So in order to-- if you are trying to select capacitance based upon the voltage ripple constraint, from the stator current point of view, this also needs to be considered.

    And finally, we have the modulation index. So modulation index is nothing but-- when you are-- as we know, in order to achieve the DC to AC, we'll be trying to operate the switches with certain switching frequency along with certain modulation. It could like sine PWM or [INAUDIBLE] PWM or any advanced modulation method.

    And with the different usage of the modulation method, the overall modulation index also can be differed. In general, the modulation index is nothing but the-- it's a ratio between what's the voltage that we got through the modulation and what's the available voltage.

    And if you are trying to-- for example, if you're trying to maintain the same torque or the same speed for the different voltage levels based upon the battery capability, the overall modulation index also gets varied. And we can also have the correlation of the capacitor currents with respect to modulation index and the stator current. So here, if you are having a difference in the modulation index, then definitely that will have influences on the capacitor current, and that will have directly influences on the overall voltage ripple.

    So finally, what the bottom line is like-- if you are trying to select a capacitance for reselling capacitor, we need to see not only from the cost perspective or the size perspective, which can accommodate the design, also these other factors also needs to be considered. And for those consideration, all the system level analysis and the eDrive model, which here models, very much comes in handy.

    So the next use case, which I would like to discuss upon the sensitivity analysis of an eDrive system. So as I mentioned before, the overall eDrive system has been modeled. And here we have considered not only the software components, we also have the hardware components. And those hardware components also has been modeled with the help of the MATLAB.

    And here I would like to touch upon one peculiar hardware component, which is the sensor. And we know the necessity of the sensor. So sensors are necessary in order to sense the real environment physical domain information. And that needs to be translated into digital information, which is used by the controller for the proper and efficient optimized working of the system.

    And it is, in the real world, not every hardware component or ideal in nature. It has its own tolerances. And those tolerances have been incorporated into the system. And if those informations are processed by the controller, that will have the overall impact in the overall system level. So it is very much needed.

    We also need to see what will have the effect due to those tolerances. This can be seen in two different ways. So assume that you have some emission, and you have some inverter, and you need to select a sensor. And for every product development, you have certain criteria like, the accuracy. From the torque point of view, the accuracy of the torque needs to be maintained within certain limit.

    And for that, sensor is also one of the contribution where its influences will be translated finally to the torque perspective. So if we want to select a sensor based upon certain requirement, this kind of analysis also comes into the picture and very much helpful.

    On in other sense, we already have a sensor. And we already have a machine which already has been selected for the production. And we want to see how already considered sensor those tolerances will have the overall information from the overall eDrive system. So that's the background of why we need to do this kind of a sensitivity analysis.

    Let's see how it has been analyzed by using a simulation. So here, as I mentioned, I'll be trying to focus how the tolerances of the sensors will have the impact in the overall system, predominantly on the emission torque profile perspective. So as we know, each sensors has its own tolerances due to the way they are manufactured and due to the aging or due to the environmental condition. Those are some of the causes why it has those tolerances.

    And in the simulation environment, we'll be trying to get those necessary tolerances for the specific sensors. For example, from between inverter and emission, we have the current sensor, which is used to measure the current that flows to the [INAUDIBLE] machine. And also we have the voltage to measure. And also we have the temperature sensor, NTC's temperature sensor, which are used to sense the junction temperature of the power module.

    And also from the emission point of view, we have the torque, speed, and temperature. Those things also we need to [INAUDIBLE] All these sensor informations, in the simulation environment, we'll be trying to take it up with respect to the hardware component profiles. And we will try to distribute those tolerances in terms of Gaussian distribution.

    So why the Gaussian distribution has been considered? For example, if I'm trying to compute the height of everyone presenting here, and if I am trying to tabulate occurrence the number of occurrence of the height, which I am populated, and if we try to plot it across a histogram, then the mean, which is the most or the average number of height will be around the center part where the number of occurrence happens, and the very higher height or the lower height will be the number of occurrences will be very much at the lower end. So which is basically kind of an actual real environment behavior. That's what the Gaussian distribution looks like.

    So considering those ideas, we also consider the same distribution for distribute those tolerances of the current sensor. So once we have those distribution, these distributions will be provided into the eDrive model. And because of these tolerances, this will be influenced into the controller perspective. And any influences in the controller perspective will lead to any setting of the IDIQ improperly. And this improper setting of IDIQ will be providing an improper voltage to the eMachine, which will be translated to the torque, which will lead to the oscillation in the torque, or also offset from the torque perspective.

    So if you are trying to see the accuracy-- like, what's my set torque, and what's the final torque I have delivered-- it will have been shifted or have some differences. And these kind of analysis can be done with the help of our model. And once we have this-- tolerance has been injected our model-- we can able to get the torque profile for each operating point for the overall operating region of the emission.

    And then we have the mapping here. And we also will try to compare with respect to the requirement for the specific emission, how much the total deviation in the torque can be allowed. And we'll also see, with the influence of the sensor tolerances, how the overall torque profile looks like. And from that, we have the capability to see which of the operating points seems to be well within the requirement and which are not-- it is violating the requirement.

    And on top of that, one of the prominent additional functionality, which our model is capable is like, it also have the tendency to see which prominent sensor tolerances give rise to most of the torque error, whether it is due to the tolerance of the angle sensor, or it is due to the tolerances of the voltage sensor or current sensor, or in which specific phase of the current sensor it leads to the overall violating of the torque accuracy. Everything can be analyzed with the help of our model.

    So sum up our overall discussion-- something like the overall eDrive system, with the help of the MathWorks, we have the capability to model it up and virtualized up. And on top of that, since the use of the MATLAB components, since it has the capability to incorporate the necessary characteristics of the actual hardware, it makes us ease to use those information and analyze it so that whatever the analysis which you are doing will be incorporate or in relevance with respect to the actual real environment.

    And on top of that, since we have the actual real characteristic hardware components and also the overall system, it also helps us to analyze the cost, the different cost of analysis of the system level and also the component level. On top of that, it helps the overall eDrive model also can be used by the other models. Like, it can be combined in the vehicle model or from the overall system, powertrain model, everything. So it's kind of a cross-collaboration. And on top of that, for collaboration as well, such models can be considered because it's already a well-validated model.

    And yeah, so that's the overall discussion regarding the eDrive system and how it's been very much used with the help of the MathWorks. Thanks.

    [APPLAUSE]

    So any questions you would like to ask?

    Yeah, I think this is on. Thanks, Alex, for providing insights into analyzing eDrive systems here. We have two questions from audience. So let me quickly ask those. We have a question from-- I think if you can pop up the question.

    So we have a question from Anand from Tata Motors. Have you considered a thermal losses in PMSM motor in Simscape? If yes, how did you consider the volumetric heat loss of copper wires in 2D environment?

    Yeah. So regarding the thermal losses, as I mentioned, it is very much the in-house developed model. So I could not able to explain very much in detail, but I can draw upon some information. For example, if you take an IGBT-- just for discussion purpose, I'm taking as an IGBT.

    Those IGBT model can be converted into a similar electrical components with respect to-- since we know if you take a MOSFET, we know there is a capacitance between the gate and the source. And we have the capacitance between the [INAUDIBLE] and [INAUDIBLE], which we call as the input capacitance and the output capacitance. And those kind of characteristics can be modeled for different nodes.

    And completely, all other nodes for the overall component level and system level can be modeled. And if you take from the node perspective, there are so much of networks will be there. And everything will be constrained to some few number of node model, which we shortly called as reduced order model as well. So those reduced order model will be also considered as to finally-- considered as the thermal model, as well as the power loss model for eDrives. So with the help of those models, we have the capability to analyze.

    Thank you. I think there is another question from Vijay from John Deere. There are different applications of the eDrive plant models. Were the single plant model was used to cater all needs? Or multiple plant models developed based on the need?

    Yeah. So here I am mostly focusing on from the eDrive perspective. So for example, if you are trying to build up a cellular environment with respect to the model, we can make use of this particular plant model because, from the eDrive perspective, we need inverter and the emission. So we can make use of this particular eDrive model since it's already a well-validated one.

    And even the similar way for the Hill, so we can make use of similar platform. But still the space is used for the mostly preferred for The Hill, but still we can make use of such from the eDrive level, not from the [INAUDIBLE], from the eDrive level.

    Thank you so much, Alex. That's about it with questions.