JP4689756B1 - Battery internal state estimation device and battery internal state estimation method - Google Patents

Abstract

An SOH is calculated based on a simulation model of a battery.
A storage means (RAM10c) for storing a plurality of parameters included in a simulation model, a measurement means (CPU10a) for measuring a terminal voltage and a discharge current of a battery at a predetermined period, and a measurement result by the measurement means. An adaptive learning means (CPU 10a) for performing adaptive learning on the parameters, an actual measurement means (I / F 10d) for actually measuring the internal resistance of the battery, an actual measurement value R _meas of the internal resistance obtained by the actual measurement means, And estimation means (CPU 10a) for estimating SOH indicating the deterioration state of the battery based on the parameter value and / or the parameter correction value obtained by the adaptive learning means.
[Selection] Figure 2

Description

  The present invention relates to a battery internal state estimation device and a battery internal state estimation method.

  Patent Document 1 discloses a technique for estimating SOH (State of Health) of a lead storage battery based on impedance. Patent Document 2 discloses a method for estimating a state of charge (SOC) of a lead storage battery using a neural network.

JP 2007-78661 A Special table 2007-518873

  Incidentally, the technique disclosed in Patent Document 1 has a problem that an error is large because SOH is estimated only from impedance. In addition, the technique disclosed in Patent Document 2 does not disclose a specific mathematical formula for estimating SOH, and therefore has a problem that SOH cannot be estimated using the technique.

  Therefore, an object of the present invention is to provide a battery internal state estimation device and a battery internal state estimation method capable of calculating SOH based on a battery simulation model.

In order to solve the above-described problems, a battery internal state estimation device according to the present invention is a battery internal state estimation device that estimates the internal state of a battery based on a battery simulation model, and includes a plurality of parameters included in the simulation model. Storing means for storing; measuring means for measuring the terminal voltage and discharge current of the battery in a predetermined cycle; and adaptive learning means for performing adaptive learning on the parameter based on a measurement result by the measuring means; The actual measurement means for actually measuring the internal resistance of the battery, the actual measurement value R _{meas of the} internal resistance obtained by the actual measurement means, the value of the parameter and / or the correction value of the parameter obtained by the adaptive learning means. An estimation means for estimating SOH (State of Health) indicating the deterioration state of the battery based on Features.
According to such a configuration, it is possible to calculate SOH based on a battery simulation model.

According to another invention, in addition to the above invention, the estimation means estimates a deterioration state based on a value of a parameter η indicating an aging change in internal resistance of the simulation model as the value of the parameter. And
According to such a configuration, SOH can be obtained with higher accuracy than when only the measured internal resistance value is used.

According to another invention, in addition to the above invention, the estimating means multiplies a parameter R0 indicating the internal resistance of the simulation model by a parameter η indicating the secular change of the internal resistance as a correction value of the parameter. The deterioration state is estimated based on a value obtained by correcting the obtained value based on the average current _Iavrg flowing through the load and the stable open circuit voltage.
According to such a configuration, SOH can be obtained with higher accuracy than when only the parameter η is used.

According to another invention, in addition to the above invention, the estimating means multiplies a parameter R0 indicating the internal resistance of the simulation model by a parameter η indicating the secular change of the internal resistance as a correction value of the parameter. Then, with respect to the value obtained by correcting the obtained value based on the average current _Iavrg flowing through the load and the stable open circuit voltage, the SOC (State of Charge) indicating the state of charge of the battery, the current flows through the load The deterioration state is estimated based on a value further corrected based on an average voltage V _{avrg of the} battery when the battery is running and a voltage V _start of the battery in a state before a current flows through the load.
According to such a configuration, the SOH can be obtained with higher accuracy than in the case of correction using the average current _Iavrg and the stable open circuit voltage.

In another aspect of the invention, in addition to the above-described invention, the estimation unit multiplies the measured value R _{meas of the} internal resistance by each of the parameter value and the parameter correction value by a predetermined constant, The deterioration state of the battery is estimated based on a value obtained by adding the obtained results.
According to such a configuration, the SOH can be accurately estimated for any type of battery by adjusting the constant value.

According to another aspect of the present invention, there is provided a battery internal state estimation method for estimating an internal state of a battery based on a battery simulation model, a storage step of storing a plurality of parameters included in the simulation model in a memory, and a terminal of the battery A measurement step for measuring the voltage and the discharge current at a predetermined period; an adaptive learning step for performing adaptive learning on the parameter based on a measurement result in the measurement step; and an actual measurement step for actually measuring the internal resistance of the battery SOH indicating the deterioration state of the battery based on the measured value R _{meas of the} internal resistance obtained in the measured step and the value of the parameter and / or the corrected value of the parameter obtained in the adaptive learning step An estimation step of estimating (State of Health) Features.
According to such a method, it becomes possible to calculate SOH based on the simulation model of the battery.

  ADVANTAGE OF THE INVENTION According to this invention, it becomes possible to provide the battery internal state estimation apparatus and battery internal state estimation method which can calculate SOH based on the simulation model of a battery.

It is a figure which shows the structural example of the battery internal state estimation apparatus which concerns on embodiment of this invention. It is a figure which shows the structural example of the control part shown in FIG. It is a figure for demonstrating the processing algorithm performed in this embodiment. It is an example of the simulation model of a lead acid battery. It is a figure which shows the impedance characteristic of the simulation model shown in FIG. It is a figure which shows the equivalent circuit of CPE shown in FIG. It is a flowchart for demonstrating the process performed in 1st Embodiment shown in FIG. It is a figure which shows distribution of estimated SOH and measured SOH by 1st Embodiment. It is a figure which shows the relationship between SOH estimation error and frequency by 1st Embodiment. (A) shows the relationship between the estimation error and the relative estimation error according to the first embodiment, and (B) shows the corrected R-square value. It is a figure which shows distribution of the estimation SOH by a prior art example, and measurement SOH. It is a figure which shows the relationship between the SOH estimation error by a prior art example, and frequency. (A) shows the relationship between the estimation error and the relative estimation error according to the conventional example, and (B) shows the corrected R-square value. It is a figure which shows distribution of estimated SOH and measured SOH by 2nd Embodiment. It is a figure which shows the relationship between SOH estimation error by 2nd Embodiment, and frequency. (A) shows the relationship between the estimation error and the relative estimation error according to the second embodiment, and (B) shows the corrected R-square value. It is a figure which shows distribution of estimated SOH and measured SOH by 3rd Embodiment. It is a figure which shows the relationship between the SOH estimation error by 3rd Embodiment, and frequency. (A) shows the relationship between the estimation error and the relative estimation error according to the third embodiment, and (B) shows the corrected R-square value. It is a figure which shows distribution of estimated SOH and measured SOH by 4th Embodiment. It is a figure which shows the relationship between SOH estimation error by 4th Embodiment, and frequency. (A) shows the relationship between the estimation error and the relative estimation error according to the fourth embodiment, and (B) shows the corrected R-square value.

  Next, an embodiment of the present invention will be described.

(A) Description of Configuration of First Embodiment FIG. 1 is a diagram illustrating a configuration example of a battery internal state estimation device according to an embodiment of the present invention. As shown in FIG. 1, the battery internal state estimation device 1 of the present embodiment corresponds to a control unit 10 and a voltage detection unit 11 (part of “measurement means” and part of “observation means” in the claims). ), Current detector 12 (corresponding to part of “measurement means” and part of “observation means” in the claims), and discharge circuit 14 as main components, and lead storage battery 13 (in claims) Corresponding to “battery”). In this example, an alternator 15, a cell motor 16, and a load 17 are connected to the lead storage battery 13 via the current detection unit 12. In the present embodiment, the case where the battery internal state estimation device 1 is mounted on a vehicle such as an automobile will be described as an example. Needless to say, it may be used for other purposes.

  As shown in FIG. 2, the control unit 10 includes a CPU (Central Processing Unit) 10a (corresponding to “estimating means” and “adaptive learning means” in the claims), a ROM (Read Only Memory) 10b, a RAM (Random Access). Memory) 10c (corresponding to “storage means” in claims) and I / F (Interface) 10d (corresponding to part of “measurement means” and part of “observation means” in claims) As a component. The CPU 10a controls each part of the apparatus based on a program 10ba stored in the ROM 10b. The ROM 10b is constituted by a semiconductor memory, and stores the program 10ba and other information. The RAM 10c is constituted by a semiconductor memory, and stores the parameter 10ca and other information in a rewritable manner. The I / F 10d converts detection signals from the voltage detection unit 11 and the current detection unit 12 into digital signals and inputs them, and supplies a control signal to control the discharge circuit 14.

  The voltage detection unit 11 detects the terminal voltage of the lead storage battery 13 and notifies the control unit 10 of it. The current detection unit 12 detects a current flowing through the lead storage battery 13 and notifies the control unit 10 of the current. The discharge circuit 14 incorporates a semiconductor switch, for example, and discharges the lead storage battery 13 by turning the semiconductor switch on or off according to the control of the control unit 10. The alternator 15 is driven by a prime mover (not shown) such as a reciprocating engine, for example, and generates DC power to charge the lead storage battery 13. The cell motor 16 is constituted by, for example, a DC motor, and is rotated by DC power supplied from the lead storage battery 13 to start the prime mover. The load 17 is composed of, for example, an automobile headlight, a wiper, a direction indication light, a navigation device, and other devices.

FIG. 3 is a diagram for explaining an outline of a processing algorithm realized by executing the program 10ba. As shown in this figure, in this embodiment, a simulation model 30 of a lead storage battery 13 having a plurality of parameters is set. Then, an observation value is obtained by observing the target lead storage battery 13 and a calculated value corresponding to the observation value is obtained based on the simulation model 30. Deviations between these observed values and calculated values are calculated, and optimal parameters are estimated by adaptive learning using the extended Kalman filter 31. Then, the simulation model 30 can be optimized by updating the simulation model 30 with the estimated parameters. The SOH calculation module 32 determines the parameters (η, R0) optimally learned in this way, the observed values (I _avrg , V _avrg , V _start , R _meas ), and values (SOC ( The SOH is calculated by substituting the state of charge and the stable open circuit voltage (stable OCV (Open Circuit Voltage)) into a predetermined mathematical formula described later, thereby calculating the SOH with high accuracy. The “stable open circuit voltage (stable OCV)” is an open circuit voltage in an electrochemically balanced state, and will be simply referred to as “OCV” below. However, it is not necessary to be in an electrochemically completely equilibrium state, and a state close to equilibrium may be included.

  In this specification, “adaptive learning” refers to a method in which a flexible and general model having parameters is prepared and the parameters are optimized statistically and adaptively by learning. In the following embodiment, an extended Kalman filter is used as an example of adaptive learning. However, the present invention is not limited to such a case. For example, adaptive learning using a neural network model or a genetic algorithm is used. It is also possible to use adaptive learning using a model. That is, any method can be used as long as it creates a learning target model and optimizes the parameters constituting the model based on the results obtained by observation.

  FIG. 4 is a diagram illustrating an example of the simulation model 30 (in this example, an electrical equivalent circuit) of the lead storage battery 13. In this example, the simulation model 30 includes a resistance R0, an inductance L, impedances Z1 and Z2, and a voltage source V0 as main components.

Here, the resistance R0 is an internal resistance whose main elements are the conductive resistance of the electrode of the lead storage battery 13 and the liquid resistance of the electrolytic solution. The inductance L is an inductive component caused by an electric field generated by a current flowing through the electrode of the lead storage battery 13 or the like. The inductance L is very small compared to the inductance value of the cable connected to the lead storage battery 13 and can be ignored as necessary. The impedance Z1 is an equivalent circuit corresponding to the anode of the lead storage battery 13 and the electrolytic solution in contact with the anode. The impedance Z1 basically has a characteristic based on the Butler-Volmer formula, and the constant phase element CPE1. It can be expressed as a parallel connection circuit with the resistor R1. Details of the impedance Z1 will be described later. The impedance Z2 is an equivalent circuit corresponding to the cathode of the lead storage battery 13 and the electrolytic solution in contact therewith, has a characteristic based on the aforementioned Butler-Volmer equation, and a parallel connection circuit of the constant phase element CPE2 and the resistor R2 Can be expressed as Details of the impedance Z2 will also be described later. Voltage source V0, the internal impedance is the ideal voltage source 0, a voltage source whose voltage is represented the electrolytic solution concentration C _N and the electrolyte concentration C _P in the vicinity of the anode in the vicinity of the cathode as a parameter.

  FIG. 5 is a diagram showing impedance characteristics of the equivalent circuit shown in FIG. In FIG. 5, the vertical axis represents the imaginary component of the impedance (Im (Z)), and the horizontal axis represents the real component of the impedance (Re (Z)). The thick line in the figure indicates the impedance characteristics of the equivalent circuit. In this example, as the frequency increases, the impedance of the equivalent circuit moves from the right to the left on the thick line, first moves in a locus so as to approach the semicircle indicated by Z1, and then Z2 Draw a trajectory and move asymptotically to the semicircle indicated by. Then, at a high frequency, the real number component is asymptotic to a straight line of R0, and due to the inductance L, the impedance increases as the frequency increases.

  FIG. 6 shows an equivalent circuit of impedances Z1 and Z2 used in the present embodiment. In this example, the impedance Z1 is formed by connecting a plurality of RC parallel units in which resistors Ra1, Ra2 and Ra3 and capacitors Ca1, Ca2 and Ca3 are connected in parallel, in series. Specifically, the resistor Ra1 and the capacitor Ca1 are connected in parallel to form one RC parallel unit, and similarly, the resistor Ra2 and the capacitor Ca2 and the resistor Ra3 and capacitor Ca3 are connected in parallel to form an RC parallel unit. . The impedance Z2 is formed by connecting a plurality of RC parallel units in which resistors Rb1, Rb2, and Rb3 and capacitors Cb1, Cb2, and Cb3 are connected in parallel.

(B) Description of Operation of First Embodiment Next, the operation of the present embodiment will be described. FIG. 7 is a flowchart for explaining the flow of processing executed in the embodiment shown in FIG. When the processing of this flowchart is started, the following steps are executed.

In step S <b> 10, the CPU 10 a measures the internal resistance actual measurement value R _meas in a state where the discharge current is not flowing from the lead storage battery 13. Specifically, for example, when entering the vehicle, the CPU 10a drives the discharge circuit 14 with the door being unlocked as a trigger, and the lead storage battery 13 with a predetermined cycle (for example, several tens of Hz cycle). And the internal resistance R _meas of the lead storage battery 13 is actually measured based on the voltage value and current value at that time. The internal resistance R _meas may be corrected to the reference state based on the SOC and the lead storage battery temperature T. That is, since the internal resistance R _meas is dependent on the SOC and the lead storage battery temperature T, for example, the actually measured internal resistance R _meas is corrected based on the formula “R _meas ' = α · R _meas + β · SOC + γ · T + δ”. You may make it do. Α, β, γ, and δ are constants obtained by actual measurement or the like. Of course, it goes without saying that the correction may be made based on expressions other than those shown above.

  In step S11, the CPU 10a determines whether or not the cell motor 16 has been activated. If the cell motor 16 has been activated (step S11: YES), the process proceeds to step S12, and otherwise (step S11: NO). The same process is repeated. For example, when the ignition key is operated to start the prime mover, the process proceeds to step S12.

  In step S12, the CPU 10a acquires the observation value of the lead storage battery 13 while the cell motor 16 is rotating (while the largest load current is flowing). Specifically, the CPU 10a acquires information indicating the terminal voltage of the lead storage battery 13 output from the voltage detection unit 11 and information indicating the discharge current output from the current detection unit 12, and sequentially stores them in the RAM 10c as observation values. Store. The observation acquisition in step S12 is executed at a predetermined cycle (for example, 10 ms cycle), and the observation value acquired in this way is held in the RAM 10c until the calculation of SOH is completed.

In step S <b> 13, the CPU 10 a performs adaptive learning on the parameters constituting the simulation model 30. In the present embodiment, adaptive learning is executed by the extended Kalman filter. Specifically, a state vector having at least a parameter as a resistance R0 shown in FIG. 4 and η indicating a change with time of the resistance R0 is set. Then, the current state vector (= n) is predicted based on the previous state vector (= n−1) (n is an arbitrary natural number). Furthermore, a more accurate state is estimated by updating the state vector based on the current observation value. As the state vector, other than this, for example, element values and the resistors and capacitors constituting the equivalent circuit of Z1, Z2 shown in FIG. 6, selecting the voltage source V0 and C _N, the C _P as a parameter Can do. Of course, other parameters may be used.

  In step S14, the CPU 10a determines whether or not to end the adaptive learning process. If it ends (step S14: YES), the process proceeds to step S15. Otherwise (step S14: NO), the step proceeds to step S15. Returning to S12, the same processing as described above is repeated. For example, when the rotation of the cell motor 16 is stopped (or when the prime mover is started), the process proceeds to step S15.

  In step S15, the CPU 10a acquires the parameter R0 and the parameter η obtained by the adaptive learning in step S12.

  In step S16, the CPU 10a calculates OCV and SOC. Specifically, the terminal voltage V0 of the lead storage battery 13 measured immediately before the start of the lead storage battery 13 or the stable open circuit voltage of the lead storage battery 13 estimated from the charge / discharge state of the lead storage battery 13 is OCV. In addition, there are a method for obtaining the SOC by combining the OCV and the integrated current value, and a method for obtaining the SOC using the current-voltage characteristic (I / V) in the operating environment, and any method may be used. Since the voltage detector 11 has a very high internal resistance and little current flows during measurement, the voltage measured immediately before the start of the lead storage battery 13 is hardly affected by the resistor R0 and the impedances Z1 and Z2. The voltage V0 of the voltage source can be measured.

In step S17, CPU 10a _calculates the _{I avrg, V avrg, V start} . _{Here, I avrg} and _{V avrg} at the time of engine starting by the cell motor 16, the average voltage of the average current and a lead-acid battery 13 flows from the lead-acid battery 13 to the load (primarily cell motor _{16), V start} prime mover starting by the cell motor 16 This is the voltage of the lead storage battery 13 immediately before. It _may be measured by the voltage detection unit 11 a voltage immediately before rotating the starter motor 16 for _{V start.} As for I _avrg and V _avrg , the average values of the current value and the voltage value acquired during the rotation of the _cell motor 16 and stored in the RAM 10c may be calculated.

In step S18, CPU 10a _is configured to calculate the _{R corr,} to calculate the _{R corr} 'based on the calculated _{R corr.} Here, R _corr is defined by the following equation (2) using resistors R0, η, and g () shown in FIG. As described above, η is a parameter indicating the change over time of the resistance R0, and has a value close to 1 when the lead storage battery 13 is new, and a value greater than 1 when the lead storage battery 13 deteriorates over time. Have Further, g () represents a predetermined function with OCV and _Iavrg as variables. Note that g () can be expressed by, for example, the following formula (1).
g (OCV, _Iavrg ) = 1 / (A1 + A2 · OCV) / (A3 + A4 · _Iavrg ) (1)

Here, A1 to A4 are predetermined constants (for example, constants determined according to the type of the lead storage battery 13 and obtained by actual measurement), for example, reading and using those stored in the ROM 10b in advance. can do. R _corr is expressed by the following formula (2).
R _corr = η · R0 · g (OCV, _Iavrg ) (2)

Then, CPU 10a _is calculated based on _{R corr} 'is a correction value of _{R corr} in equation (3) below.
R _corr ′ = R _corr + f (SOC, OCV, I _avrg , V _avg , V _start ) (3)

_{Here, f (SOC, OCV, I} avrg, V avrg, V start) , for example, each parameter _{SOC, OCV, I avrg, V} avrg, a function represented by a linear equation of the _{V start} and a predetermined constant is there. Specifically, for example, it can be represented by the following formula (4). B1 to B5 are predetermined constants, and for example, those stored in advance in the ROM 10b can be read and used. As a specific example, when the 20-hour rate (0.05C) rated capacity of the lead storage battery 13 is 60 Ah, B1 to B5 are -2.599388 × 10 ⁻⁶ , −8.027024 × 10 ⁻⁴ , 1. 388216 × 10 ⁻⁵ , −4.6602935 × 10 ⁻⁴ , and −4.872664 × 10 ⁻⁴ can be used, respectively.
_{f (SOC, OCV, I avrg} , V avrg, V start) = B1 · SOC + B2 · OCV + B3 · I avrg + B4 · V avrg + B5 · V start ··· (4)

In step S19, the CPU 10a calculates SOH based on the following formula, and ends the process.
SOH = C1 · η + C2 · R _corr '+ C3 · R _meas + C4 (5)
C1, C2, C3, and C4 are constants obtained in advance (for example, constants determined according to the type of the lead storage battery 13 and obtained by actual measurement), for example, those stored in advance in the ROM 10b. It can be read and used. As a specific example, when the 20-hour rate (0.05C) rated capacity of the lead storage battery 13 is 60 Ah, as C1 to C4, 92.71322, -285252.8, -4879.45, -596.149, respectively. Can be used.

  According to the above processing, when the cell motor 16 is started, adaptive simulation is performed on the simulation model 30 to optimize the parameters, and the SOH is calculated based on the parameters and the actually measured values based on the above-described equation (5). Can be calculated. The fact that the SOH calculated in this way is very accurate will be described below based on experimental results.

8 to 10 show actual measurement results according to the first embodiment described above. On the other hand, FIGS. 11 to 13 show actual measurement results when SOH is estimated by the prior art. Here, as conventional technology, SOH is estimated by the following equation.
SOH = C1 · R _meas + C2 (6)
Here, R _meas is an actually measured value of the internal resistance of the lead storage battery, and C1 and C2 are constants obtained in advance. As a specific example, when the 20 hour rate (0.05 C) rated capacity of the lead storage battery 13 is 60 Ah, −7026.74106, 117.2042 can be used as C1 and C2.

  FIG. 8 is a diagram showing the variation between the estimated SOH value and the measured SOH in the first embodiment, and FIG. 11 is a diagram showing the same variation in the conventional example. From the comparison of these figures, compared to the conventional example, in the first embodiment, each sample is concentrated and distributed on a straight line (measured value = estimated 45-degree straight line starting from the origin). That there is little variation.

  FIG. 9 shows a distribution of SOH estimation errors in the first embodiment, and FIG. 12 shows a similar distribution in the conventional example. In these figures, the horizontal axis indicates the SOH error range, and the vertical axis indicates the frequency of samples belonging to each error range. From the comparison of these figures, there is a sample group belonging to the range of 100 to 110% in the conventional example, and the variation is large as a whole, but such a sample group does not exist in the first embodiment. The variation is small.

  FIG. 10A is a table showing respective items of estimation error and relative estimation error for all samples according to the first embodiment, and FIG. 13A is a table showing similar measurement results according to the prior art. Here, the “estimation error” indicates a difference value between the measured value and the estimated value Ah. The “relative estimation error” is the difference between the measured value and the estimated value Ah as a percentage of the measured value. As is clear from the comparison between FIG. 10A and FIG. 13A, in the conventional technique, the “maximum error”, “minimum error”, “average error”, and “standard deviation” in the estimation error of all 81 samples. Are "19.1", "-14.8", "0.2", and "7.8" in the first embodiment as "12.1", "-12.9". ”,“ −0.3 ”, and“ 5.0 ”. Further, the “maximum error”, “minimum error”, “average error”, and “standard deviation” in the relative estimation error are “103.7”, “−39.4”, “5.9”, and What was “30.6” is improved to “41.6”, “−34.2”, “1.5”, and “14.3”.

  FIG. 10B and FIG. 13B show “corrected R-square values” called degree-of-freedom-determined determination coefficients. In general, sufficient estimation accuracy can be obtained with about 0.8. It is an index value considered to be. In the case of the first embodiment, as shown in FIG. 10B, the corrected R-square value is 0.8063, which indicates that the estimation accuracy is high. On the other hand, in the case of the conventional technique, as shown in FIG. 13B, the corrected R-square value is 0.59972, indicating that the estimation accuracy is lower than that of the present embodiment. From the above comparison, it can be understood that the estimation accuracy of the first embodiment is high.

(C) Description of Second Embodiment In the second embodiment, the formula for calculating SOH is different from that in the first embodiment. Other configurations are the same as those in the first embodiment. More specifically, in the first embodiment, the SOH is calculated by the equation (5), but in the second embodiment, the SOH is calculated based on the following equation (7).
SOH = C1 · η + C2 · R _meas + C3 (7)
Here, C1, C2, and C3 are predetermined constants (for example, constants determined according to the type of the lead storage battery 13 and obtained by actual measurement), and η is a parameter indicating a change with time of the resistance R0. , R _meas is an actually measured value of the internal resistance. Note that η and R _meas are the same parameters as described above, and C1, C2, and C3 are calculated separately from the above constants. As a specific example, when the 20 hour rate (0.05 C) rated capacity of the lead storage battery 13 is 60 Ah, −8.44404, −657.13129, 124.95939 can be used as C1, C2, and C3. .

  14-16 has shown the actual measurement result by 2nd Embodiment. FIGS. 14 and 15 correspond to FIGS. 11 and 12, respectively, and show the correspondence between the estimated SOH and the measured SOH and the distribution of the SOH estimation error. FIG. 16A shows an estimation error and a relative estimation error, and FIG. 16B shows a corrected R-square value. From comparison between FIGS. 14 and 15 and FIGS. 11 and 12, it can be seen that the characteristics of the second embodiment are improved as compared with the conventional example. Also, from the comparison between FIG. 16 and FIG. 13, the second embodiment is improved over the related art for each of the estimation error, the relative estimation error, and the corrected R-square value. From the above, it can be understood that the estimation accuracy of the second embodiment is high.

(D) Description of Third Embodiment In the third embodiment, the formula for calculating SOH is different from that in the first embodiment. Other configurations are the same as those in the first embodiment. More specifically, in the first embodiment, the SOH is calculated by the equation (5). However, in the third embodiment, the SOH is calculated based on the following equation (8).
SOH = C1 · R _corr + C2 · R _meas + C3 (8)
Here, C1, C2, and C3 are predetermined constants (for example, constants determined according to the type of the lead storage battery 13 and obtained by actual measurement), and R _corr is expressed by the equation (2). _meas is an actually measured value of the internal resistance. Note that R _corr and R _meas are the same parameters as described above, and C1, C2, and C3 are calculated separately from the above constants. As a specific example, when the 20-hour rate (0.05C) rated capacity of the lead storage battery 13 is 60 Ah, -11150.18291, -3748.003121, 142.47629 can be used as C1, C2, and C3. .

  17 to 19 show actual measurement results according to the third embodiment. FIGS. 17 and 18 correspond to FIGS. 11 and 12, respectively, and show the correspondence between the estimated SOH and the measured SOH and the distribution of the SOH estimation error. FIG. 19A shows an estimation error and a relative estimation error, and FIG. 19B shows a corrected R-square value. From comparison between FIGS. 17 and 18 and FIGS. 11 and 12, it can be seen that the characteristics of the third embodiment are improved as compared with the conventional example. Further, from the comparison between FIG. 19 and FIG. 13, the third embodiment is improved over the related art for each of the estimation error, the relative estimation error, and the corrected R-square value. The corrected R-square value is 0.77955, which is an improvement over the second embodiment. From the above, it can be understood that the estimation accuracy of the third embodiment is high.

(E) Explanation of 4th Embodiment In 4th Embodiment, compared with 1st Embodiment, the formula for calculating SOH differs. Other configurations are the same as those in the first embodiment. More specifically, in the first embodiment, the SOH is calculated by the equation (5). However, in the fourth embodiment, the SOH is calculated based on the following equation (9).
SOH = C1 · η + C2 · R _corr + C3 · R _meas + C4 (9)
Here, C1, C2, C3, and C4 are predetermined constants (for example, constants determined according to the type of the lead storage battery 13 and obtained by actual measurement), and η is a parameter indicating a change with time of the resistance R0. R _corr is obtained by equation (2), and R _meas is an actually measured value of the internal resistance. Note that η, R _corr , and R _meas are the same parameters as described above, and C1, C2, C3, and C4 are calculated separately from the above constants. As a specific example, when the 20-hour rate (0.05 C) rated capacity of the lead storage battery 13 is 60 Ah, C1, C2, C3, and C4 are 6.89056, -12602.7054, -3688.8798, 139. 4371 can be used.

  20 to 22 show actual measurement results according to the fourth embodiment. FIGS. 20 and 21 correspond to FIGS. 11 and 12, respectively, and show the correspondence between the estimated SOH and the measured SOH and the distribution of the SOH estimation error. FIG. 22A shows an estimation error and a relative estimation error, and FIG. 22B shows a corrected R-square value. From the comparison between FIGS. 20 and 21 and FIGS. 11 and 12, it can be seen that the characteristics of the fourth embodiment are improved as compared with the conventional example. Further, from the comparison between FIG. 22 and FIG. 13, the fourth embodiment is improved over the related art for each of the estimation error, the relative estimation error, and the corrected R-square value. The corrected R-square value is 0.78603, which is an improvement over the third embodiment. From the above, it can be understood that the estimation accuracy of the fourth embodiment is high.

As described in the above second to fourth embodiments, the equation for calculating SOH can be calculated using the measured value of the internal resistance R _meas and / or η and R _corr. The SOH thus obtained is more accurate than the conventional SOH calculated based only on the internal resistance R _meas .

(F) Modified Embodiment Each of the above embodiments is an example, and there are various modified embodiments other than this. For example, in the above embodiment, adaptive learning is performed using an extended Kalman filter, but other methods may be used. Specifically, adaptive learning may be performed using a neural network model, or adaptive learning may be performed using a genetic algorithm model.

  In the above embodiment, two impedances Z1 and Z2 are assumed. However, for example, it may be assumed that only one of them exists. Moreover, although impedance Z1, Z2 is set as the structure comprised by three RC parallel units, respectively, the number of RC parallel units may differ in each of impedance Z1, Z2. Moreover, instead of three, it may be one or two, or four or more.

  In the above embodiment, adaptive learning is executed based on the current flowing through the cell motor 16, but adaptive learning may be executed based on the current flowing through other loads. For example, when there is a motor for driving the vehicle (for example, a motor of a hybrid vehicle), adaptive learning may be performed based on the current flowing through the motor. In addition, for example, in the case of a photovoltaic power storage battery used in a general household, it is possible to use a current flowing in various loads used in the household (for example, an air conditioner in which a large current flows at startup). .

  In each of the above embodiments, the constants C1 to C4, the constants A1 to A4, and the constants B1 to B5 are constants determined according to, for example, the type and usage environment of the lead storage battery 13 and can be obtained by actual measurement. it can. Specifically, it can obtain | require by using the least squares method about the sample of the some lead acid battery in which SOH is known.

  In the above embodiments, specific values are shown for the constants C1 to C4, the constants A1 to A4, and the constants B1 to B5. However, these values are only examples and are limited to such cases. Is not to be done. In addition, the constants C1 to C4, the constants A1 to A4, and the constants B1 to B5 have been described on the assumption that they are fixed values. It may be possible to rewrite according to

  In each of the above embodiments, the number of adaptive learning parameters used is different. However, the more parameters used, the higher the accuracy of SOH estimation, but the more complicated the calculation. For this reason, a desired form is selected and used from the first to fourth embodiments according to required calculation accuracy and available resource processing capacity (processing capacity of the control unit 10). That's fine.

  Moreover, although the above embodiment demonstrated lead acid battery as an example, it is possible to apply this invention also about batteries other than this (for example, nickel cadmium battery etc.). In this case, the simulation model 30 may be changed according to the type of battery.

  In the above embodiment, the control unit 10 is configured by a CPU, a ROM, a RAM, and the like, but may be configured by, for example, a DSP (Digital Signal Processor).

DESCRIPTION OF SYMBOLS 1 Battery internal state estimation apparatus 10 Control part 10a CPU (estimation means, adaptive learning means)
10b ROM
10c RAM (storage means)
10d I / F (part of actual measurement means, part of observation means)
11 Voltage detector (part of actual measurement means, part of observation means)
12 Current detector (part of actual measurement means, part of observation means)
13 Lead acid battery (battery)
14 Discharge circuit 15 Alternator 16 Cell motor (electric motor)
17 Load 30 Simulation model 31 Extended Kalman filter R0 Internal resistance Z1, Z2 Impedance

Claims (6)

In the battery internal state estimation device that estimates the internal state of the battery based on the battery simulation model,
Storage means for storing a plurality of parameters included in the simulation model;
Measuring means for measuring the terminal voltage and discharge current of the battery at a predetermined period;
Adaptive learning means for performing adaptive learning on the parameter based on a measurement result by the measurement means;
Actual measurement means for actually measuring the internal resistance of the battery,
Based on the measured value R _{meas of the} internal resistance obtained by the measured means and the value of the parameter and / or the correction value of the parameter obtained by the adaptive learning means, SOH (State of estimation),
A battery internal state estimation device comprising:   2. The battery internal state estimation device according to claim 1, wherein the estimation unit estimates a deterioration state based on a value of a parameter η indicating an aging of an internal resistance of the simulation model as the value of the parameter. . The estimation means multiplies the parameter R0 indicating the internal resistance of the simulation model by a parameter η indicating the secular change of the internal resistance as a correction value of the parameter, and uses the obtained value as an average current I flowing through the load. The battery internal state estimation device according to claim 1 or 2, wherein the deterioration state is estimated based on a value corrected based on _avrg and a stable open circuit voltage. The estimation means multiplies the parameter R0 indicating the internal resistance of the simulation model by a parameter η indicating the secular change of the internal resistance as a correction value of the parameter, and uses the obtained value as an average current I flowing through the load. _An SOC (State of Charge) indicating a state of charge of the battery with respect to a value corrected based on _avrg and a stable open circuit voltage, an average voltage V _{avrg of the} battery when current flows through the load, and 3. The battery internal state estimation device according to claim 1, wherein a deterioration state is estimated based on a value further corrected based on a voltage V _start of the battery in a state before a current flows through the load. The estimation means multiplies the measured value R _{meas of the} internal resistance by a predetermined constant for each of the parameter and the correction value of the parameter value, and adds the obtained results to each other. The battery internal state estimation device according to any one of claims 1 to 4, wherein the deterioration state of the battery is estimated. In the battery internal state estimation method for estimating the internal state of the battery based on the battery simulation model,
Storing a plurality of parameters included in the simulation model in a memory;
A measurement step of measuring the terminal voltage and discharge current of the battery at a predetermined period;
An adaptive learning step for performing adaptive learning on the parameter based on a measurement result in the measurement step;
An actual measurement step for actually measuring the internal resistance of the battery,
Based on the measured value R _{meas of the} internal resistance obtained in the measured step and the value of the parameter and / or the correction value of the parameter obtained in the adaptive learning step, SOH (State of Health), and an estimation step
A battery internal state estimation method comprising:

Images