Papers by Matthias Preindl
Energies
Intelligent and pragmatic state-of-health (SOH) estimation is critical for the safe and reliable ... more Intelligent and pragmatic state-of-health (SOH) estimation is critical for the safe and reliable operation of Li-ion batteries, which recently have become ubiquitous for applications such as electrified vehicles, smart grids, smartphones, as well as manned and unmanned aerial vehicles. This paper introduces a convolutional neural network (CNN)-based framework for directly estimating SOH from voltage, current, and temperature measured while the battery is charging. The CNN is trained with data from as many as 28 cells, which were aged at two temperatures using randomized usage profiles. CNNs with between 1 and 6 layers and between 32 and 256 neurons were investigated, and the training data was augmented with noise and error as well to improve accuracy. Importantly, the algorithm was validated for partial charges, as would be common for many applications. Full charges starting between 0 and 95% SOC as well as for multiple ranges ending at less than 100% SOC were tested. The proposed C...
This paper proposes a way to augment the existing machine learning algorithm applied to state-of-... more This paper proposes a way to augment the existing machine learning algorithm applied to state-of-charge estimation by introducing a form of pulse injection to the running battery cells. It is believed that the information contained in the pulse responses can be interpreted by a machine learning algorithm whereas other techniques are difficult to decode due to the nonlinearity. The sensitivity analysis of the amplitude of the current pulse is given through simulation, allowing the researchers to select the appropriate current level with respect to the desired accuracy improvement. A multi-layer feedforward neural networks is trained to acquire the nonlinear relationship between the pulse train and the ground-truth SoC. The experimental data is trained and the results are shown to be promising with less than 2\% SoC estimation error using layer sizes in the range of 10 - 10,000 trained in 0 - 1 million epochs. The testing procedure specifically designed for the proposed technique is e...
IEEE Transactions on Industrial Electronics
Varying the operating frequency helps dual-active bridge topologies increase the system efficienc... more Varying the operating frequency helps dual-active bridge topologies increase the system efficiency since the soft-switching is regained at low-load condition. This paper proposes a conduction-loss-based variable frequency modulation (VFM) that decouples the phase shift from the frequency control that conventional VFM applies. The power losses for the shared-bus battery balancing topology are elaborated as switching frequency varies. The results show that increasing the operating frequency within reasonable boundaries can benefit not only the switching loss but also the conduction losses. The switching loss is nearly constant, whereas the conduction losses decrease significantly. A factor of output power and transformer current, namely power-per-ampere, is derived as a function of operating conditions and phase shift independent of switching frequency. The operating setpoints are selected to minimize the factor using both online and offline optimizations. The proposed control strategy outperforms constant frequency modulation by up to 30% below 30% rated power. Furthermore, compared with conventional VFM, the proposed method improves the efficiency of the system under test by 1.5% below 50% normalized power, with significantly less computational resources needed.
2017 IEEE Transportation Electrification Conference and Expo (ITEC)
This paper proposes a modeling and control approach for the three-level DC-DC converter. The conv... more This paper proposes a modeling and control approach for the three-level DC-DC converter. The converter is described in a sum and difference (Σ∆) framework. It is shown that the formulation is useful to model the inverter and derive design-specific equations. The Σ component is responsible for the inductor current, i.e. the power flow, and the ∆ component is used to balance (or unbalance) the DC-link capacitor voltages. It is shown that there are cross-coupling terms between the Σ and ∆ axes that can be compensated. The proposed model is validated using high fidelity simulations with a proportionalintegral controller. Two-and three-level converter operation is shown and it is proven that the passive components can be reduced by 50% to 75% using three-level operation without affecting the control performance. The control is verified by introducing load current and DC voltage steps.
2018 IEEE 9th International Symposium on Sensorless Control for Electrical Drives (SLED)
This paper deals with an “active flux” model-based approach for state estimation of Permanent Mag... more This paper deals with an “active flux” model-based approach for state estimation of Permanent Magnet Synchronous Motors to build up sensorless drives. The active-flux vector is aligned to the rotor d-axis for all synchronous machines. In this way, the rotor position and speed observer seems more amenable to a wide speed range, with smaller dynamic errors. A Moving Horizon Estimation algorithm, an optimization-based scheme that yields good performance, is applied for the speed and rotor position estimation. Under assumptions, an optimal problem of Equality Constrained Quadratic Programming type has been solved each iteration. The algorithm has been efficiently implemented and tested for both Surface and Interior Permanent Magnet Synchronous Motors, demonstrating the real-time feasibility the proposed approach at 10 kHz sampling rate.
2018 IEEE 9th International Symposium on Sensorless Control for Electrical Drives (SLED)
This research studies a discrete time optimization-based position estimation strategy. The method... more This research studies a discrete time optimization-based position estimation strategy. The method accomplishes approximation over nonlinear dynamic models, making it convenient to take the magnetic saturation into account. The dynamic model is written in the virtual flux domain, ending up with affine dynamic constraints and nonlinear output constraints. The performance of the observer is tested incorporated with vector control. The result shows the estimation strategy with higher variable dimension is capable of accurate parameter and state reconstruction. Comparing to prior work, the proposed method gains better estimating precision, more flexibility and higher stability. In particular, there is significant improvement in estimation accuracy at low speed - peak noise 0.0095 rad (<0.424 rad), standard deviation 0.0020 rad (<0.038 rad) in position estimation and peak noise 1.50 rad/s (<13.17 rad/s), standard deviation 0.587 rad/s (<6.877 rad/s) in speed estimation.
2017 IEEE International Symposium on Predictive Control of Electrical Drives and Power Electronics (PRECEDE)
A novel unified position sensorless observer is proposed using homotopy continuation. A single ob... more A novel unified position sensorless observer is proposed using homotopy continuation. A single observer is used to estimate the rotor position and speed at high speed, low speed, and standstill. The position and speed are treated as independent variables such that the PMSM dynamic equation is a parametrized algebraic function f where the zeros identify the position and speed estimate. A homotopy is identified that maps the zeros of f at sampling time instant k − 1 into the zeros of f at sampling time instant k. As the algebraic equation has multiple zeros, the homotopy is designed to identify the closest zero that, assuming reasonable sampling, identifies the correct position and speed. Benefits of the proposed method are instantaneous estimation and lack of convergence transients once a rough estimate is available through initial position estimation with polarity detection. The concept is validated at low and high speed using high-fidelity simulations.
2017 IEEE International Conference on Industrial Technology (ICIT)
The advent of Silicon-Carbide and Gallium-Nitride MOSFETs offers potential to realize higher ener... more The advent of Silicon-Carbide and Gallium-Nitride MOSFETs offers potential to realize higher energy density power converters operating at increased switching frequencies. The maximum switching frequency in a power converter is limited by the ability of the switching device package to dissipate its switching and conduction losses. At a given value of drain-source voltage and current, the turn-on losses in a MOSFET are usually greater than the turn-off losses. This paper introduces a soft-switching technique for power converters using wide bandgap devices to replace the larger turn-on losses with smaller turn-off losses and thus reduce the power dissipation of the overall system. The turn-off losses are further reduced with use of additional capacitance across the MOSFET drain-source terminals. Results from an analytical model, LTSpice simulation and experimentation are shown to match closely, with significant reduction in overall system losses.
2017 IEEE Applied Power Electronics Conference and Exposition (APEC)
Power loss calculations are critical to a power converter design, helping with estimation of effi... more Power loss calculations are critical to a power converter design, helping with estimation of efficiency, switch selection and cooling system design. Moreover, power losses in a MOSFET may limit the maximum switching frequency in a power converter. Switching energy values aren't always available in MOSFET datasheets at all operating points, and calculation of voltage and current rise-time and fall-time is needed. This paper introduces a method to obtain an estimate of switching transition times and power losses, using datasheet parameters, for SiC MOSFETs with non-flat gate-plateau region. Three methods are discussed here, two existing and a proposed method. These methods are used to evaluate a certain MOSFET product, and calculated values are compared with results from PLECS simulation and double pulse test experiment. The proposed method is shown to yield improved accuracy.
2017 IEEE Applied Power Electronics Conference and Exposition (APEC)
2019 IEEE 10th International Symposium on Sensorless Control for Electrical Drives (SLED)
This paper analyzes direct position and speed estimation from a control theoretical and numerical... more This paper analyzes direct position and speed estimation from a control theoretical and numerical standpoint. The paper develops the theory of local identifiability and defines the conditions such that position and speed can be identified uniquely based on a sufficiently accurate guess. Local identifiability typically holds at nonzero machine speeds and zero speeds with any perturbation, e.g. an injected high frequency signal in pwm control or a random switching ripple in direct control (except in isotropic machines). The position and speed is identifiable in more than 98.5% of feasible operation points "instantaneously", and can be extrapolated from past estimates in the remaining cases. Finally, numerical solving strategies are studied and a combination of the Newton and conjugate gradient method is shown to provide suitable estimates within 1-5 solver iterations depending on the required accuracy.
2021 IEEE Transportation Electrification Conference & Expo (ITEC)
Demand for the grid state estimation with partial power network observation is growing rapidly wi... more Demand for the grid state estimation with partial power network observation is growing rapidly with the increasing amount of distributed energy resources (DER) connected to the grid with incomplete measured information. The grid services that could be provided by these DER, such as electrical vehicles (EVs), have the potential to affect the resilience and efficiency of the grid. This paper proposes a constrained optimization solver based on the AC power flows to recover the incomplete information of the grid. Further reactive power constraints following the specifications in IEEE 1547 are added to the solver to explore the effects of adding grid services to the steady-state microgrid.
Pervasive and Mobile Computing
2018 IEEE Transportation Electrification Conference and Expo (ITEC), 2018
A novel optimization-based sensorless technique for Induction Machines is presented in this work.... more A novel optimization-based sensorless technique for Induction Machines is presented in this work. The algorithm is able to estimate speed and position instantaneously (limited only by the Nyquist frequency and the PLL bandwidth) at standstill low and high-speed. A cost function is introduced based on the dynamic machine model. The stability analysis of speed and position estimation shows whether the system is convex by solving the Hessian matrix of the cost function. The method is unique for the entire speed range, in which high-frequency signal is injected to the machine terminals at low speeds, to increase the convex region of the cost function.
ArXiv, 2019
This paper proposes a way to augment the existing machine learning algorithm applied to state-of-... more This paper proposes a way to augment the existing machine learning algorithm applied to state-of-charge estimation by introducing a form of pulse injection to the running battery cells. It is believed that the information contained in the pulse responses can be interpreted by a machine learning algorithm whereas other techniques are difficult to decode due to the nonlinearity. The sensitivity analysis of the amplitude of the current pulse is given through simulation, allowing the researchers to select the appropriate current level with respect to the desired accuracy improvement. A multi-layer feedforward neural networks is trained to acquire the nonlinear relationship between the pulse train and the ground-truth SoC. The experimental data is trained and the results are shown to be promising with less than 2\% SoC estimation error using layer sizes in the range of 10 - 10,000 trained in 0 - 1 million epochs. The testing procedure specifically designed for the proposed technique is e...
IECON 2018 - 44th Annual Conference of the IEEE Industrial Electronics Society
Moving Horizon Estimators (MHE)solve an optimization problem to estimate unknown states or parame... more Moving Horizon Estimators (MHE)solve an optimization problem to estimate unknown states or parameters based on a sequence of measurements containing disturbances and noise in nonlinear systems subject to input and state constraints. This research applies MHE to estimating the nonlinear behavior of interior-mount permanent magnet synchronous machines (IPMSM). The MHE estimates a disturbance term that reppresents the speed-dependent nonlinear terms of the machine model and can be interpreted as extended back-EMF. The formulation is based on a cost function that matches the measurements to the model formulation and an explicit regularization penalty is added. The term is interpreted as a gain that balances estimation accuracy with noise rejection. We demonstrate that our formulation can be solved in realtime and is effective in estimating the disturbance term on an experimental test bench both in terms of noise rejection and estimation accuracy.
2019 IEEE Applied Power Electronics Conference and Exposition (APEC), 2019
To improve power density and efficiency, a critical soft switching principle with optimal-frequen... more To improve power density and efficiency, a critical soft switching principle with optimal-frequency model predictive control method is proposed in this paper for DC/DC power converter. Firstly, this paper analyzes the boundary constraints of critical soft switching that are derived with the key parameters of the interlock time and threshold current for typical SiC and GaN devices. Then, according to the derived critical soft switching constraints, two optimal-frequency control methods are proposed based on model predictive control (MPC) to eliminate the turn-on losses especially during the transient period. Compared to the traditional PI controller, the efficiency can be further improved during the reference variation period, because of the fast response of MPC. Finally, the test results verify the theoretical analysis.
2019 IEEE 10th International Symposium on Sensorless Control for Electrical Drives (SLED), 2019
This paper analyzes direct position and speed estimation from a control theoretical and numerical... more This paper analyzes direct position and speed estimation from a control theoretical and numerical standpoint. The paper develops the theory of local identifiability and defines the conditions such that position and speed can be identified uniquely based on a sufficiently accurate guess. Local identifiability typically holds at nonzero machine speeds and zero speeds with any perturbation, e.g. an injected high frequency signal in pwm control or a random switching ripple in direct control (except in isotropic machines). The position and speed is identifiable in more than 98.5% of feasible operation points “instantaneously”, and can be extrapolated from past estimates in the remaining cases. Finally, numerical solving strategies are studied and a combination of the Newton and conjugate gradient method is shown to provide suitable estimates within 1–5 solver iterations depending on the required accuracy.
2019 IEEE Energy Conversion Congress and Exposition (ECCE)
A critical soft switching technique is proposed to reduce the switching losses of a bidirectional... more A critical soft switching technique is proposed to reduce the switching losses of a bidirectional DC/DC converter. The proposed method can improve the efficiency and the value of passive components will be largely decreased. The paper derives the boundary conditions of critical soft switching with the parameters of dead time and threshold current for existing typical SiC and GaN devices. Then, a controlling strategy is developed to estimate and optimize the total power losses of a synchronous DC/DC converter with variable frequency at the given average current and duty cycle. With the combination of critical soft switching and optimal frequency control, the efficiency can be largely improved in every operating point. A power loss curve comparison between the proposed method and hard switching shows that the proposed method can reduce the losses up to 40%. The theoretical results are verified in a rigorous testing procedure.
Journal of Power Sources, 2021
A battery equivalent circuit model (ECM) is proposed using a novel physics-based diffusion compon... more A battery equivalent circuit model (ECM) is proposed using a novel physics-based diffusion component and N resistor–capacitor (RC) pairs, hence its name the ‘DNRC model’. The DNRC model characterizes ohmic, charge transfer, and diffusion overpotentials in the time domain with physically-meaningful circuit elements. Unlike the Warburg impedance, the diffusion component has no need for frequency-domain data and is formulated entirely in the time domain. Physical interpretability is validated by comparison with physics-based model (PBM) generated data. Experimental validation is performed at a wide range of state of charge (SoC) and state of health (SoH) using pulse injection and drive cycle data. The mean absolute percent error is below 0.3% using 5 circuit elements for 4 min of an arbitrary current load. The DNRC model is grounded in physical principles, suitable for real-time estimation, and may form the basis for new approaches to degradation reduction or diagnosis in battery manag...
Uploads
Papers by Matthias Preindl