KR20160063070A - Apparatus and method for estimating battery state - Google Patents

Abstract

The present invention relates to an apparatus and a method for estimating a battery state. In particular, the present invention relates to the technology of estimating the state of charging of battery of an electric automobile by using changes in a curved type of charging voltage. The present invention comprises: a feature extraction unit which extracts the feature of a change in a curve type of voltage corresponding to the time of charging a battery and sets up a feature vector; a classification unit which classifies a category pertaining to the feature vector extracted by the feature extraction unit through an ex post facto probability; and a battery state estimation unit which calculates an estimation for the state of charging of battery in response to the output of a classification unit and estimates the state of battery by comparing the estimation with the detected estimation. The present invention is designed to accurately estimate the deterioration degree (reduction in capacity) of battery by using the configuration change of a voltage curve which compares with the time at the time of charging battery.

Description

[0001] Apparatus and method for estimating battery state [0002]

The present invention relates to an apparatus and method for estimating a battery state, and more particularly, to a battery state estimation apparatus capable of estimating a battery remaining capacity of an electric vehicle using a curve shape change of a charging voltage.

Today, due to high oil prices and carbon dioxide regulations, the development of environmentally friendly vehicles capable of replacing conventional internal combustion engine vehicles is actively under way. In recent years, pure electric vehicles that drive electric motors, hybrid electric vehicles that use an internal combustion engine and electric motors as driving sources have already been commercialized according to automobile manufacturers, or are in the process of being commercialized.

Electric vehicles and hybrid vehicles are equipped with a high-voltage battery (main battery) as a main power source for supplying power to an electric motor (traction motor) serving as a driving source. A charging device for charging the battery, an inverter for driving the electric motor, and the like.

In addition, a battery management system (BMS) for monitoring the state of the battery is mounted. This battery controller collects battery status information regarding battery temperature, voltage, charge / discharge current, battery state of charge (SOC), and the like. Then, the battery controller provides the collected battery state information to another controller in the vehicle so that it can be used for vehicle control.

Particularly, the battery controller checks the state of the battery to manage the state of the battery to a predetermined level or higher, and prevents the life span of the battery due to deterioration of battery durability. Then, the battery controller informs the vehicle controller performing the total control of the battery SOC information so that the vehicle running can be performed considering the battery state.

However, most of the conventional battery packs for vehicles use battery capacity, voltage, and temperature to estimate battery deterioration. As a result, the degree of deterioration of the battery pack can not be accurately estimated.

In addition, the conventional vehicle battery pack does not directly recognize the battery capacity decrease due to charge / discharge repetition. That is, the capacitance variation is indirectly reflected in the calculation of the polarization voltage for SOC estimation.

Thus, it is impossible to accurately grasp the power limitation, the remaining travelable distance prediction, and the like. In such a case, the reliability of the BMS function designed based on the capacity in the long-term use of the electric vehicle battery will be reduced.

The present invention is characterized in that the deterioration degree (capacity reduction) of the battery can be accurately estimated by using a change in shape of a voltage curve compared with a charging time of the battery.

The apparatus for estimating battery condition according to an embodiment of the present invention includes a feature extraction unit for extracting a feature from a change in a shape of a voltage curve corresponding to a time when a battery is charged and setting a feature vector; A classifier for classifying a class to which the feature vector extracted by the feature extractor belongs through a posterior probability; And a battery state estimation unit for calculating an estimated value for the remaining capacity of the battery in correspondence with the output of the classification unit and for comparing the estimated value with the previously measured value to estimate the state of the battery.

A method of estimating a battery state according to an embodiment of the present invention includes: detecting data of a voltage curve shape corresponding to a time in a range of a specific voltage at the time of charging the battery; Extracting a feature from a change in a voltage curve shape and setting a feature vector; Classifying a class to which a set feature vector belongs through a posterior probability; And estimating the state of the battery by calculating an estimated value of the remaining capacity of the battery corresponding to the classification result and comparing the estimated value with the previously measured value.

The present invention makes it possible to recognize in real time on the actual vehicle that the battery capacity is reduced through the SOH estimation.

In addition, the present invention provides the effect of improving the long-term reliability of the capacity-based designed battery controller (BMS) function.

It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. .

1 is a configuration diagram of a battery state estimation apparatus according to an embodiment of the present invention;
FIG. 2 is a graph for explaining a charge voltage curve in an embodiment of the present invention; FIG.
3 is a graph for explaining the battery remaining capacity in the embodiment of the present invention.
4 is a graph for explaining a voltage change curve when the battery is charged in an embodiment of the present invention.
5 is a graph for explaining voltage change curve shapes when charging a battery in an embodiment of the present invention.
FIGS. 6 to 10 are views for explaining operations relating to the classifying unit and the battery state estimating unit of FIG. 1;
11 is a view for explaining an effect according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of a battery state estimation apparatus according to an embodiment of the present invention.

The embodiment of the present invention includes a feature extraction unit 100, a classification unit 200, and a battery state estimation unit 300.

First, the feature extraction unit 100 extracts a feature from a change in the shape of a voltage curve corresponding to a time when the battery is charged.

The battery state estimating apparatus according to the embodiment of the present invention exhibits a capacity change of the battery according to repetition of charging and discharging of the battery.

For example, the feature extraction unit 100 extracts characteristics of a voltage curve in a specific voltage section, 3.6 to 4 V, during constant current charging. Then, the time period (? T /? V = 0.1) corresponding to the change of the specific voltage level (for example, the voltage 0.1V) is set to the feature vector x.

The classifying unit 200 classifies the class (ω) to which the feature vector x extracted by the feature extracting unit 100 belongs through the posterior probability. That is, the probability P (? I | x) that the feature vector x belongs to the specific class? I (1? In? N) is found.

For example, ω1 is 0 to 5%, ω2 is 5 to 10%, ω6 is 25 to 30%.

At this time, the capacity deterioration degree can be set in units of 5%, and classified into the class having the greatest probability.

Thereafter, the battery state estimating unit 300 estimates the battery remaining capacity SOC (State Of Charge) only in the form of a voltage curve corresponding to the output of the classifying unit 200.

In other words, the classifications classified in the classifying unit 200 are classified as a whole according to the estimation error requirement of the state of health (SOH). Here, the estimation error requirement of SOH can be set to 2% or less.

For example, omega 1 is 0 to 2%, omega 2 is 2 to 4%, omega 3 is 4 to 6%, ..., omega 15 is 28 to 30%. At this time, the capacity deterioration degree can be set in units of 2%.

Then, the battery condition estimating unit 300 partially subdivides the above class. That is, the estimation error requirement is differentially assigned to secure the SOH estimation accuracy in the entire region (capacity deterioration is 0 to 30%).

For example, ω1 is 0 to 2%, ω5 is 8 to 10%, ω6 is 10 to 15%, and so on.

The operation of the battery state estimation method according to an embodiment of the present invention having such a configuration will be described with reference to FIGS. 2 to 4. FIG.

As shown in the graph of FIG. 2, the curve shape of the voltage is changed in accordance with the change of the time during charging of the battery. At this time, the curve shape of the voltage may be different depending on the number of charge / discharge cycles.

Then, as shown in the graph of FIG. 3, the remaining capacity of the battery decreases in accordance with the change of the cycle. The shape of the voltage curve and the remaining capacity of the battery can be obtained by a deterioration characteristic test (Bench test).

Then, the deterioration degree estimation target battery (Test cell) is charged. As shown in the graph of FIG. 4, the voltage of the battery is charged according to the change of time, and a voltage curve shape is displayed.

Thus, according to the deterioration degree characteristic test, each value of the class ωi corresponding to the cycle change can be obtained as shown in Table 1 below.

Bracket ω1 ω2 ω3 ω4 ω5 ω6 Cycle 0 400 800 1200 1600 2000 capacity [Ah] 38.35 36.66 35.42 34.15 32.87 31.20 capacity [norm] 100 95.6 92.4 89 65.7 81.4

As shown in the graph of FIG. 5, it can be seen that the shape of the voltage curve changes in each cycle according to the cycle change of the battery during charging. The change of shape of the voltage curve according to each class ω1 to ω6 is shown in [Table 2] below.

Class / Cycle / Voltage Range (V) ω1 ω2 ω3 ω4 ω5 ω6 0 400 800 1200 1600 2000 3.6 to 3.7 230 210 190 170 150 130 3.7 ~ 3.8 290 250 240 230 220 200 3.8 to 3.9 190 180 170 160 150 160 3.9 to 4.0 190 190 180 170 170 150

As shown in [Table 2] above, the characteristics of the existing voltage curves obtained through the degradation characteristic test can be extracted in advance.

Thereafter, the feature extraction unit 100 extracts the feature from the change of the voltage curve shape corresponding to the charging time of the estimation target battery.

That is, the characteristic of the voltage curve is extracted in the 3.6V to 4.0V voltage during charging of the constant current. At this time, the time period (? T /? V = 0.1) according to the change of the specific voltage level (for example, the voltage 0.1V) is set as the feature vector (x). Can be subdivided into vectors x = (180, 230, 170, 170) in each voltage range.

Next, the classifying unit 200 classifies the features extracted by the feature extracting unit 100 as a class having the largest probability. At this time, the classifier 200 classifies the feature vector x as shown in FIG. 6 and detects the class having the highest probability.

For example, the classifying unit 200 obtains the probability P (? I | x) that the feature vector x belongs to a specific class, and obtains the two classes P (? 3 | x) and P (? 4 | . Here, the classifier 200 classifies the specific vector x by applying a Bayesian classifier.

At this time, the classifying unit 200 classifies the vector x as? 1 if the class P (? 1 | x)> P (? 2 | x). Then, the classifying section 200 classifies the vector x as? 2 if the class P (? 1 | x) <P (? 2 | x).

Then, the classifying unit 200 can calculate the posterior probability P ([omega] i | x) using the base theorem. The method of obtaining the posterior probability P (? Ix) is shown in the following Equation (1).

In the above equation (1), the likelihood p (x | omega i) represents a Likelihood (class-conditional probability) under a given condition.

Assuming that the likelihoods follow the normal distribution, it can be expressed as a one-dimensional normal distribution and a two-dimensional normal distribution as shown in FIG. 7, and the likelihood p (x|ωi) can be expressed by the following Equation 2 .

The prior probability P (? I) can be estimated as ni / N, where ni is the number belonging to? I among the samples in the training set. At this time, if N is sufficiently large, this value is very close to the actual probability value. And, in the classification problem, p (x) does not need to be computed.

And, in the classification problem, it is necessary to compare the posterior probability values, not the posterior probability P (ωi | x) values themselves. Accordingly, if the posterior probability p (x |? 1) P (? 1)> p (x |? 2) P (? 2), classify vector x as? 1. The classifying unit 200 classifies the vector x as? 2 if the posterior probability p (x |? 1) P (? 1) <p (x |? 2) P (?

This can be expressed as a minimum error Bayesian classifier as shown in FIG. The classifier 200 classifies the vectors by extending the binary classifications into M classifications. That is, the vector x is classified as? K, and the vector x is classified as k = argmax P (x |? I) P (? I) with the vector x being k = argmax P . Here, P (x | oi) P (omega i) represents a fractional function.

That is, the classifier 200 sets the class P (? I) = 0.5, and assumes that the prior probability is the same. 9, the classifying unit 200 obtains a probability density function (PDF) corresponding to the specific vector x. Here, the probability that a continuous random variable is included in a given interval is called a probability density, and the probability density function is expressed as a function form.

Thereafter, the battery state estimating unit 300 estimates the SOH of the battery according to the two class values P (? 3 | x) = 0.54 and P (? 4 | x) = 0.46 measured by the classifying unit 200.

For example, SOH can be calculated as shown in Equation (3) below.

Next, the battery state estimating unit 300 compares the estimated value calculated in the above equation (3) with the capacity measurement value and verifies it. Here, the estimated value calculated in Equation (3) is 34.83 [Ah] and the actual capacity measurement value is 34.81 [Ah].

Then, the battery condition estimating unit 300 compares the previously measured measurement value with the above calculated estimate to generate verification data. At this time, the verification data generated by the battery state estimation unit 300 may be expressed as shown in the graph of FIG.

The verification data generated from the measured values and the estimated values in the battery state estimating unit 300 are shown in Table 3 below.

Verification data Measure Estimate X12 37.39 36.2 X23 36.05 35.97 X34 (= X) 34.81 34.83 X45 33.51 33.6 X56 32.1 32.15

As described above, when the embodiment of the present invention is implemented, the battery management system (BMS) correctly estimates the SOH and the battery capacity is reduced in real time on the real vehicle, as shown in FIG. Furthermore, embodiments of the present invention provide the effect of improving the long-term reliability of the capacity-based designed battery controller (BMS) function.

It should be understood that the embodiments described above are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

Claims (11)

A feature extraction unit for extracting a feature from a change in the shape of a voltage curve corresponding to a time when the battery is charged and setting a feature vector;
A classifier for classifying a class to which the feature vector extracted by the feature extractor belongs through a posterior probability; And
And a battery state estimation unit for calculating an estimated value of the remaining capacity of the battery corresponding to the output of the classifying unit and comparing the estimated value with the previously measured value to estimate the state of the battery. The apparatus of claim 1, wherein the feature extraction unit
And sets a time period according to a change in a specific voltage level unit at the time of charging the battery to the feature vector. The apparatus of claim 1, wherein the classifier
Wherein the capacity deterioration degree of the battery is set as a specific unit, and a probability that the feature vector belongs to a specific class is obtained. The apparatus of claim 1, wherein the battery condition estimating unit
And classifies and partially subdivides the classes classified by the classifying unit according to an estimation error requirement of the SOH (State of Health). Detecting data in the form of a voltage curve corresponding to time in a range of a specific voltage upon charging of the battery;
Extracting a feature from the change of the voltage curve shape and setting a feature vector;
Classifying a class to which the set feature vector belongs through a posterior probability; And
Estimating a state of the battery by calculating an estimated value of the remaining capacity of the battery corresponding to the classification result and comparing the estimated value with the previously measured value. 6. The method of claim 5,
Wherein a time period according to a change in a specific voltage level unit at the time of charging the battery is set to the feature vector. 6. The method according to claim 5,
And determining a probability that the feature vector belongs to a specific class by setting the capacity deterioration degree of the battery as a specific unit. 6. The method of claim 5, wherein the battery condition estimating step
Subdividing the group classified by the classifying unit according to the estimation error requirement of SOH (State of Health, SOH); And
And partially subdividing the refined class on the basis of the total capacity deterioration degree. 6. The method according to claim 5,
Wherein the posterior probability is calculated using the likelihood and the base theorem based on the prior probability. 6. The method according to claim 5,
And comparing the two posterior probability values to classify the feature vector into a specific class. 6. The method of claim 5, wherein estimating the battery condition comprises:
And setting verification data corresponding to the measurement value and the estimation value.

Images