A Multiobjective Linear Time-varying Model Predictive Control Strategy for a Battery/Supercapacitor Hybrid Energy Storage System

Chao Jia Author

03/29/2021 Added

84 Plays

Description

This work proposes a multiobjective, linear, time-varying model predictive control strategy is proposed to optimize the current split between battery and supercapacitor of a hybrid energy storage system used in electric vehicles.

Searchable Transcript

Search:

[00:00:00.660]Hello. Welcome to my presentation.
[00:00:02.700]I'm Chao, a PhD student of electrical engineering.
[00:00:06.780]Here, I'm going to present my research that I conducted with Dr.
[00:00:10.350]Wei Qiao on multiobjective linear time-varying model
[00:00:14.370]predictive control strategy for a battery supercapacitor hybrid energy storage
[00:00:19.320]system.
[00:00:21.090]Battery and supercapacitor are two types of energy storage devices.
[00:00:25.080]Since they have complimentary power properties,
[00:00:28.980]they can be integrated as a battery supercapacitor hybrid energy storage
[00:00:33.480]system, which is referred to as the HESS hereafter.
[00:00:37.890]HESS has been widely used in electric vehicles,
[00:00:40.740]storage-enabled microgrid, and uninterruptible power supplies.
[00:00:45.630]With a suitable power management strategy, hybridization of battery
[00:00:49.740]and supercapacitor can take advantage of both energy storage devices
[00:00:54.810]to improve the energy efficiency of the HESS while prolonging battery lifetime.
[00:00:59.070]Based on the literature study,
[00:01:02.490]the HESS prediction models in prior literatures do not consider parameter
[00:01:07.050]variations with respect to the batteries state-of-charge and,
[00:01:11.880]therefore, cannot guarantee the satisfactory model accuracy
[00:01:16.770]over the entire SOC range. To bridge this research gap,
[00:01:22.290]this work proposes a new, linear time-varying (LTV for short) prediction model.
[00:01:27.150]for the HESS. Furthermore, according to the proposed model,
[00:01:31.830]a multiobjective LTV-MPC strategy is proposed to
[00:01:36.540]optimally split the current between the battery and the supercapacitor.
[00:01:41.250]At last, a scaled-down experimental
[00:01:44.250]setup is developed to validate the proposed strategy.
[00:01:50.100]Figure 1 shows you the block diagram of the electric vehicle powertrain
[00:01:53.790]configuration studied in this work.
[00:01:56.700]The major components include the battery, supercapacitor.
[00:01:59.790]DC/DC converter, inverter,
[00:02:01.920]electric motor, and the power management strategy. The battery equivalent circuit
[00:02:06.750]model is shown in Figure 2,
[00:02:09.710]which is composed of three pairs of resistor-capacitor branches,
[00:02:15.360]an open circuit voltage, and a series resistance.
[00:02:19.620]The values of these eight parameters depend on the batteries state-of-charge.
[00:02:24.900]Figure 3 shows the equivalent model of the supercapacitor
[00:02:29.870]The three resistor-capacitor branches represent different time constants.
[00:02:35.220]These two equivalent circuit models
[00:02:39.750]can be expressed as a group of differential equations,
[00:02:43.740]which include unknown parameters that
[00:02:46.710]are required to be identified by the testing data.
[00:02:51.480]Figure 4 and Figure 5 show the battery and supercapacitor voltage curves
[00:02:55.140]of the test and simulation results, respectively,
[00:03:00.490]The simulation results agree well with the experimental results,
[00:03:05.590]so the equivalent circuit models proposed in this work
[00:03:09.460]offer a good modeling accuracy.
[00:03:13.270]Next, I will introduce the proposed
[00:03:17.590]LTV-MPC strategy.
[00:03:19.780]The model developed in the last section can be stacked into a discrete-time
[00:03:24.220]state-space model of the HESS,
[00:03:28.180]given by these two equations.
[00:03:29.680]where x is the state vector, u is the control variable, D is the
[00:03:34.120]measured disturbance,
[00:03:36.430]y is the output vector, and A, B, C,
[00:03:40.210]and D are matrices of the state-space model.
[00:03:44.860]There cost functions J1, J2, and J3 are considered in the MPC
[00:03:49.450]strategy of the HESS. J1 is the power losses of HESS.
[00:03:54.010]J2 represents the battery
[00:03:56.320]current variations. J3 is a penalty
[00:03:59.320]term on the state-of-charge of the supercapacitor.
[00:04:02.890]Then, the following combined objective function J is used to evaluate
[00:04:07.870]the overall performance.
[00:04:10.660]where omega1 and omega2 are weighting factors. By minimizing the
[00:04:15.280]combined objective function J, the energy efficiency of the HESS
[00:04:19.480]can be improved to offer a longer driving range
[00:04:22.900]via J1 and the battery lefttime can be prolonged via J2
[00:04:29.680]According to the LTV prediction model, the control action at time k
[00:04:33.700]is obtained by minimizing the following multiobjective function
[00:04:38.440]over a prediction horizon Np steps. We need to solve this
[00:04:43.390]online constrainted optimization problem. where Nc is the
[00:04:47.860]control horizon. U(k) is the sequence or the supercapacitor currrent
[00:04:52.690]to be optimized. At the time step k,
[00:04:56.830]the MPC receives the new measurements or estimations of the current states and
[00:05:01.390]then solves the constrained
[00:05:04.740]optimization problem to obtain the optimal value of U(k).
[00:05:09.700]Then, the controller only applies the first optimal action to the plant.
[00:05:16.560]The proposed LTV-MPC-based
[00:05:17.790]current split strategy is compared
[00:05:22.710]with two state-of-the-art power management strategies,
[00:05:27.310]a rule-based strategy and a frequency-decouping
[00:05:29.560]strategy for the HESS in EV application.
[00:05:34.590]Figure 6 shows a scaled-down experimental setup for the HESS,
[00:05:38.760]including two battery module, two supercapacitor modules, a bidirectional DC
[00:05:43.230]power supply, a bidirectional DC/DC converter,
[00:05:46.500]and a real-time controller.
[00:05:51.420]Figure 7 compares accumulated energy losses and Figure 8 compares
[00:05:56.340]accumulated current variations of three strategies for a vehicle driving cycle.
[00:06:01.910]The proposal strategy has the least total energy losses.
[00:06:06.560]The accumulated
[00:06:07.730]current variation of the proposed strategy is less than 6.5%
[00:06:12.560]of those of the other two strategies.
[00:06:17.380]Finally, the conclusions are summarized here.
[00:06:21.400]A multiobjective
[00:06:22.570]LTV-MPC strategy was proposed properly distribute the
[00:06:28.030]load current between battery and the supercapacitor to reduce the
[00:06:32.770]energy losses and battery current variations.
[00:06:36.640]The simulation and experimental results validated superiority of the
[00:06:41.440]proposed LTV-MPC
[00:06:42.670]strategy over a rule-based strategy and a frequency
[00:06:47.590]-decoupling strategy.
[00:06:50.200]This work has been accepted on this conference regarding the power electronics
[00:06:55.630]for distributed generation systems.
[00:06:58.720]I would like to thank the Nebraska Center for Energy Sciences Research
[00:07:03.490]and the Nebraska Public Power District for supporting this work.

The screen size you are trying to search captions on is too small!

You can always jump over to MediaHub and check it out there.

Comments

0 Comments

A Multiobjective Linear Time-varying Model Predictive Control Strategy for a Battery/Supercapacitor Hybrid Energy Storage System

Description

Searchable Transcript

Comments icon comment

Related Channels

Comments