JP5298519B2 - Battery learning system - Google Patents

Description

  The present invention relates to a battery learning system, and more particularly to a battery learning system that sequentially updates output characteristics of a battery operated by an electrochemical reaction by learning of actually measured characteristic values.

  For example, in order to obtain an operating point when controlling the fuel cell, current-voltage characteristics or current-power characteristics are used as output characteristics of the fuel cell. The former is a so-called IV characteristic, and the latter is a so-called IP characteristic. Since the output characteristics of the fuel cell change depending on the operating state of the fuel cell, the output characteristics are updated while learning is performed by momentary measurement.

  For example, Patent Literature 1 describes learning for correcting an IV characteristic in accordance with an operation state of a fuel cell with respect to an IV characteristic used for a power control process of the fuel cell. Here, it is determined whether or not the fuel cell system is in a steady operation. If the fuel cell system is not in a steady operation, the power control process is performed based on the currently recorded IV characteristics that are not suitable for updating the IV characteristics. When it is determined that steady operation is being performed, the process proceeds to processing for IV characteristic update. In the process for updating the IV characteristics, it is determined whether or not the current output current value of the fuel cell corresponds to a predetermined value for each predetermined step. It is disclosed that the power supply control process is performed based on the -V characteristic, and when applicable, the IV characteristic is updated based on the actual voltage value from the voltage sensor.

  Further, as Embodiment 2, it is described that a means for calculating the AC impedance of the fuel cell is provided, and a theoretical IV characteristic that excludes a voltage drop due to an internal resistance corresponding to the AC impedance is obtained. It is stated that this theoretical IV characteristic is also updated during a steady operation and at a current value for each predetermined step.

JP 2007-48628 A

  According to Patent Document 1, it is possible to learn an IV characteristic line, which is an output characteristic line, using the current output current value and voltage value of the fuel cell as characteristic values.

  As described above, in order to learn the output characteristics of the battery, it is necessary to actually measure the output value of the battery. As the output characteristic line of the battery, an output value at a predetermined characteristic step point determined in advance is used as a characteristic value to connect each characteristic value to form a discrete output characteristic line. However, it is not always performed for all characteristic step points. In this case, if the already learned output characteristic line is updated by replacing the characteristic value of the characteristic step point corresponding to the actual measured characteristic value with the actual measured characteristic value, the output characteristic line changes discontinuously before and after that. It can happen to have. The output characteristics of a battery are generally monotonically changing characteristics corresponding to the electrochemical reaction of the battery in the current-voltage characteristics or current-power characteristics. Therefore, the discontinuous change in the output characteristic line is unnatural in view of the electrochemical reaction of the battery. For example, if this output characteristic line is used as it is for power control of the battery, a control error may occur.

  An object of the present invention is to provide a battery learning system capable of obtaining a more accurate output characteristic line.

The battery learning system according to the present invention uses, as characteristic step points , a plurality of points arranged in a grid pattern in a predetermined array on a current-voltage plane having a current value on the horizontal axis and a voltage value on the vertical axis. The combination of the current value and the voltage value corresponding to the current value is a characteristic value, and the actual voltage value is detected by actually measuring the voltage value corresponding to the arbitrary current value for the battery. obtaining means for obtaining the current value of the combination of the actual voltage value as actual characteristic value, the characteristic step point having the characteristic value that matches the actual characteristic value as the actual characteristic step point, two of the actual characteristic step the characteristic step points in the region actual characteristic value is not present as a characteristic step point unlearned between points, the current value definitive the electrochemical reaction of the battery is the voltage value is constant or decreases with increasing So as to correspond to the tone reduction characteristics, a calculation means for estimating calculating the characteristic value of the characteristic step points of the unlearned as calculated characteristic values, and the actual characteristic step point, the characteristic value that matches the calculated characteristic value The characteristic step points are sequentially connected to create means for creating an initial current-voltage characteristic line, and the actual measurement obtained by the obtaining means after the creation means creates the first current-voltage characteristic line. Update means for sequentially updating current-voltage characteristic lines already obtained by learning based on characteristic values .

Further, in the battery learning system according to the present invention, the calculating means, typical current of the battery - voltage characteristic line basic current - a plurality of the as-voltage characteristic line basic current - has a voltage characteristic line in advance Preferably, the calculation characteristic value is estimated and calculated based on the basic current-voltage characteristic line closest to a partial current-voltage characteristic line connecting a plurality of the actual measurement characteristic step points .

Further, in the battery learning system according to the present invention, the calculating means, the actual characteristic closest the step point starting from the characteristic step point unlearned, towards the further characteristic step point far the unlearned, the calculated The characteristic values are preferably estimated and calculated sequentially.

  In the battery learning system according to the present invention, the battery is preferably a fuel cell.

  With the above configuration, the battery learning system uses the characteristic value of the characteristic step point of the output characteristic line corresponding to the actual characteristic value obtained by actually measuring the actual output value of the battery as the actual characteristic value for the output characteristic line that has already been obtained. Discrete output characteristic line connecting the already obtained output characteristic line, measured characteristic value, and calculated characteristic value when calculating the characteristic value of other characteristic step points as the calculated characteristic value. However, the calculated characteristic value is calculated so as to satisfy a predetermined monotonous change condition that is predetermined in accordance with the electrochemical reaction of the battery.

  The output characteristics of a battery are generally monotonically changing characteristics corresponding to the electrochemical reaction of the battery in the current-voltage characteristics or current-power characteristics. Therefore, by appropriately setting the predetermined monotonous change condition, a natural and more accurate output characteristic line can be obtained in view of the electrochemical reaction of the battery.

  The battery learning system further includes initial learning means for calculating a characteristic value of the unlearned characteristic step point as a calculated characteristic value when an unlearned characteristic step point remains. As a result, in the initial learning, the initial output characteristic line can be obtained by calculating the calculated characteristic value for the unlearned characteristic step points without actually measuring all the characteristic step points. Can finish.

  Further, in the battery learning system, when the output characteristic of the battery is the current-voltage characteristic of the battery, a condition that the voltage value monotonously decreases as the current value increases is used as the predetermined monotonous change condition. Here, the monotonic decrease is used in a broad sense that the voltage value does not increase with an increase in the current value. Therefore, the voltage value is a constant value or decreases with an increase in the current value. Thereby, it is possible to match the current-voltage characteristic of the battery having the monotonously decreasing characteristic while reflecting the actually measured characteristic value.

  In the battery learning system, the battery activation overvoltage is used to learn the battery current-activation overvoltage characteristics as the battery current-voltage characteristics. The activation overvoltage characteristic of the battery excludes the influence of the internal resistance of the battery, and is relatively less susceptible to the operating conditions of the battery. Therefore, a more accurate output characteristic line can be obtained by learning this current-activation overvoltage characteristic and updating the output characteristic such as the current-voltage characteristic based on this.

In the battery learning system, when the battery output characteristics are the battery current-power characteristics,
As the predetermined monotonous change condition, a condition is used in which the power value monotonously increases with an increase in the current value. Here, the monotonic increase is used in a narrow sense that the power value does not decrease with respect to the increase in the current value. Therefore, the power value increases with a positive gradient with respect to the current increase. Here, it is assumed that the power / current gradient between adjacent characteristic step points is larger than a predetermined threshold gradient. Thereby, it is possible to match the current-power characteristic of the battery having the monotonously increasing characteristic while reflecting the actually measured characteristic value.

  Also, in the battery learning system, the calculated characteristic values are calculated sequentially starting from the other characteristic step point closest to the measured characteristic step point and further toward the other characteristic step point. By reducing the number, a more accurate output characteristic line can be obtained.

  In the battery learning system, since the battery is a fuel cell, the output characteristic line of the fuel cell can be made more accurate.

  Embodiments according to the present invention will be described below in detail with reference to the drawings. Hereinafter, a fuel cell will be described as a battery for learning the output characteristic line. However, other batteries may be used as long as they operate by an electrochemical reaction. For example, a secondary battery such as a lithium ion secondary battery, a nickel hydride secondary battery, a lead storage battery, an alkaline battery, or a manganese battery may be used.

  In the following, a fuel cell system that includes a battery learning system and performs drive control of a power source including the fuel cell based on an output characteristic line of the fuel cell obtained by the learning will be described. Alternatively, the battery learning system alone may be configured. In this case, the battery learning system can be configured by the fuel cell and the computer that executes learning, and the learning result can be transmitted to the fuel cell drive control device by separate data transfer or the like.

  In the following, a configuration including a high-voltage power storage device, a fuel cell, a voltage converter, and a high-voltage operating inverter will be described as a power supply circuit, but other components may be included. For example, a system main relay, a low voltage battery, a DC / DC converter operated at a low voltage, and the like can be included. Although the inverter will be described as a load that is a rotating electrical machine that is a motor / generator (M / G), it is needless to say that an inverter having a load as a fuel cell auxiliary machine may be provided.

  FIG. 1 is a diagram illustrating a configuration of a fuel cell system 10 including a battery learning system. The fuel cell system 10 includes a power supply circuit 11, a rotating electrical machine 22 that is a load, a storage device 38, and a control unit 40. Here, the battery learning system includes components of the fuel cell 30, the high-frequency signal source 32, the current detection unit 34, the voltage detection unit 36, the storage device 38, and the control unit 40 that are components of the power supply circuit 11. This corresponds to a portion including the battery learning unit 42.

  The power supply circuit 11 includes a power storage device 12, a secondary battery-side smoothing capacitor 14, a voltage converter 16, a fuel cell-side smoothing capacitor 18, a fuel cell 30, and an inverter 20. Further, the power supply circuit 11 is provided with a high frequency signal source 32, a current detection means 34, and a voltage detection means 36 on the fuel cell 30 side.

  The power storage device 12 is a chargeable / dischargeable high-voltage secondary battery that performs power interchange with the fuel cell 30 via the voltage converter 16 and responds to fluctuations in the load of the rotating electrical machine 22 and the like. Have As the power storage device 12, for example, a lithium ion assembled battery or a nickel hydride assembled battery having a terminal voltage of about 200 V to about 300 V, or a capacitor can be used. In addition, since the battery learning system can be applied to learning of the output characteristic line of a battery that operates by a battery chemical reaction, it can be applied to a lithium ion assembled battery or a nickel hydride assembled battery. The battery learning system will be described as being applied only to learning of the output characteristic line of the fuel cell 30.

  The voltage converter 16 is a circuit disposed between the power storage device 12 and the fuel cell 30. For example, when the power of the fuel cell 30 is insufficient, power is supplied to the load from the power storage device 12 via the voltage converter 16, and when charging the power storage device 12, the fuel cell is supplied via the voltage converter 16. Power is supplied to the power storage device 12 from the 30 side. As such a voltage converter 16, a bidirectional converter including a reactor can be used.

  Smoothing capacitors are provided on both sides of the voltage converter 16. That is, a smoothing capacitor 14 on the secondary battery side is provided between the positive side bus and the negative side bus connecting the voltage converter 16 and the power storage device 12, and the positive side connecting the voltage converter 16 and the fuel cell 30. A smoothing capacitor 18 on the fuel cell side is provided between the bus bar and the negative electrode side bus bar.

  The fuel cell 30 is a type of assembled battery configured to extract a high-voltage generated power of about 200 V to about 300 V by combining a plurality of fuel cells, and is called a fuel cell stack. Here, each fuel cell supplies hydrogen as a fuel gas to the anode side, supplies air as an oxidizing gas to the cathode side, and takes out necessary power by a battery chemical reaction through an electrolyte membrane that is a solid polymer membrane. It has a function.

  The inverter 20 converts the high-voltage DC power into AC three-phase driving power under the control of the control unit 40 and supplies it to the rotating electrical machine 22, and conversely, the AC three-phase regenerative power from the rotating electrical machine 22 is increased. It is a circuit having a function of converting into voltage direct current charging power. The inverter 20 can be configured by a circuit including a switching element, a diode, and the like.

  The current detection means 34 has a function of detecting the output current value of the fuel cell 30 and is arranged in series with the negative electrode bus of the fuel cell 30. In some cases, the current detection means 34 may be disposed on the positive electrode bus of the fuel cell 30. As the current detection means 34, an appropriate ammeter can be used.

The voltage detection means 36 has a function of detecting the output voltage value of the fuel cell 30 and is arranged in parallel with the fuel cell 30 between the positive electrode bus and the negative electrode bus of the fuel cell 30. Such voltage detecting means 36, it is possible to use a suitable voltage meter.

  The high-frequency signal source 32 is for applying a high-frequency signal to the negative electrode side bus of the fuel cell 30 in order to measure the impedance of the fuel cell 30. The applied high frequency signal is detected as an alternating current component separately from the direct current component by the current detection means 34 and the voltage detection means 36. From the DC component, a DC current value as the output current value of the fuel cell 30 and a DC voltage value as the output voltage value can be obtained. From the AC component, the impedance of the fuel cell 30 can be calculated by a known AC impedance method.

  The rotating electrical machine 22 is a motor / generator mounted on a vehicle, for example, and is a three-phase synchronous rotating electrical machine that functions as a motor when electric power is supplied from the power supply circuit side and functions as a generator during braking. The rotating electrical machine 22 is shown here as an example of the load of the power supply circuit 11.

  The storage device 38 has a function of storing a program or the like executed by the control unit 40, and particularly has a function of storing an output characteristic line file 39 of the fuel cell 30 here. Here, the output characteristic line of the fuel cell 30 will be described with reference to FIGS. Below, it demonstrates using the code | symbol of FIG.

  The output characteristic line of the fuel cell 30 is a discrete characteristic line obtained by connecting each characteristic value with respect to the output characteristic of the fuel cell 30 using the output value at a predetermined characteristic step point as a characteristic value.

  FIG. 2 shows characteristic step points 60 arranged in a grid pattern on an IV plane in which the output characteristics are current-voltage characteristics, the horizontal axis represents current (I) and the vertical axis represents voltage (V). The situation is shown. In the example of FIG. 2, the step ΔS regarding the current (I) at the characteristic step point 60 is shown as being constant, but this step ΔS does not necessarily have to be a constant value. That is, the arrangement of the characteristic step points 60 may be determined in advance, and the increment of the arrangement need not be a constant value with respect to the current (I) and the voltage (V). can do.

  As the current value detected by the current detection unit 34 described in FIG. 1 and the voltage value detected by the voltage detection unit 36, only values that match the characteristic step point 60 in FIG. 2 are acquired. A characteristic step point indicated by a black circle in FIG. 2 is an actually measured characteristic value 62 of the fuel cell 30 obtained by measurement. Therefore, the characteristic value is indicated by a combination of (I = actual output current value, V = actual output voltage value). A broken line in FIG. 2 is a discrete characteristic line 64 obtained by connecting the actually measured characteristic value 62, and this is an output characteristic line of the fuel cell 30.

  FIG. 3 is a diagram schematically showing a current-voltage characteristic line as an output characteristic line of the fuel cell 30. Hereinafter, the current-voltage characteristic line of the fuel cell 30 is referred to as an IV characteristic line, and the current-voltage characteristic is referred to as an IV characteristic. Since the fuel cell 30 has an internal resistance, the IV characteristic line is affected by a voltage drop due to the internal resistance. Here, the magnitude of the internal resistance can be obtained by measuring the impedance and extracting the DC component.

  FIG. 3 shows how the IV characteristic line changes depending on the magnitude of the internal resistance of the fuel cell 30. At the same current value, the voltage value increases as the internal resistance decreases. That is, the IV characteristic line 66 when the internal resistance is zero is the characteristic line having the highest voltage value, and the IV characteristic line 67 when the internal resistance is present decreases as the internal resistance increases. It becomes a characteristic line.

  The IV characteristic line 66 with the internal resistance set to zero uses a voltage value obtained by eliminating a voltage drop due to the internal resistance of the fuel cell 30 from an actual voltage value that is an actually measured output voltage value of the fuel cell 30. It will be. The voltage value obtained by eliminating the voltage drop due to the internal resistance from the actual voltage value is called an activation overvoltage or a theoretical voltage. Therefore, the IV characteristic line 66 is called a current-activation overvoltage characteristic line or a theoretical IV characteristic line.

  As shown by the IV characteristic line 66 in FIG. 3, the theoretical IV characteristic line suddenly decreases in voltage value and settles to a constant value when the current increases from zero. When the current value further increases, the voltage value suddenly decreases from a constant voltage value, and the voltage value becomes zero. As described above, the theoretical IV characteristic line has a characteristic in which the voltage value has a certain range, and in other regions, the voltage always decreases as the current increases. That is, in the theoretical IV characteristic line, when the current increases, the voltage value decreases or takes a constant value and does not increase. Since the voltage value has a characteristic that does not increase with respect to the increase in the current value, it can be referred to as a monotonously decreasing characteristic. Here, it is not a monotonous decrease in a narrow sense that the voltage value always decreases as the current value increases, but a monotonic decrease in a broad sense including that the voltage value is constant. This characteristic is based on the characteristics of the electrochemical reaction of the fuel cell 30 and is also a characteristic commonly found in cells that operate by an electrochemical reaction other than the fuel cell.

  When an IV characteristic line using the actual current value actually measured by the current detection unit 34 and the actual voltage value actually measured by the voltage detection unit 36 is referred to as a real IV characteristic line, the actual IV characteristic is obtained. The line is a theoretical IV characteristic line including a voltage drop due to internal resistance. The voltage drop is given as an IR drop, where I is the current value and R is the internal resistance value. If the internal resistance value is constant, the voltage drop always increases as the current increases. That is, as can be seen from the IV characteristic line 67 of FIG. 3, the voltage value always decreases as the current value increases. That is, in the actual IV characteristic line, when the current increases, the voltage value monotonously decreases in a narrow sense, and does not take a constant value or increase.

  In the above description, the IV characteristic line has been described as the output characteristic line of the fuel cell. However, the current-power characteristic line can be used as the output characteristic line that can be used for power source drive control of the fuel cell system 10. Hereinafter, the current-power characteristic line is referred to as an IP characteristic line, and the current-power characteristic is referred to as an IP characteristic.

  The IP characteristic line is a characteristic line indicating the relationship of the output power value with respect to the output current value of the fuel cell 30. The output power value of the fuel cell 30 can be obtained by multiplying the actual current value detected by the current detector 34 by the actual voltage value detected by the voltage detector 36. As described above, the actual voltage value in the fuel cell 30 monotonously decreases with the increase in the actual current value, so what is changed by multiplying this by the actual voltage value is actually. The power value increases monotonously with the increase in the actual current value. This is shown in FIG. As shown here, the power characteristic of the IP characteristic line 68 of the fuel cell 30 increases monotonously with an increase in the current value. However, the increase rate is large when the current value is small, and the current value increases. , The increase rate decreases. However, the power value does not become a constant value or decreases as the current value increases.

  Thus, as the output characteristic line of the fuel cell 30, an actual IV characteristic line, a theoretical IV characteristic line, and an IP characteristic line can be used. And about these, by updating using an actual measurement value, learning of an output characteristic line can be performed. The storage device 38 of FIG. 1 stores learning results of these output characteristic lines as an output characteristic line file 39.

  By the way, when learning is performed, if the characteristic value of the characteristic step point corresponding to the current actual characteristic value is replaced with the actual characteristic value and updated, the output characteristic line may have a discontinuous change before and after that. . As described above, in the fuel cell 30 that operates based on the electrochemical reaction, the output characteristic is a monotonically decreasing characteristic in a broad sense in the actual IV characteristic line, and a monotonic decreasing characteristic in the narrow sense in the theoretical IV characteristic line. This is a monotonically increasing characteristic in the I-P characteristic line, and all have a monotonous change characteristic and no discontinuous characteristic.

  Therefore, in learning of the output characteristic line of the fuel cell 30, when the characteristic value of the characteristic step point corresponding to the current actual characteristic value is replaced with the actual characteristic value and updated, the output characteristic line changes discontinuously before and after that. If so, it is necessary to perform correction so that the output characteristic line of the fuel cell 30 meets the monotonic change condition that the fuel cell 30 originally has.

  Therefore, returning to FIG. 1, the control unit 40 has a function of controlling each element constituting the fuel cell system 10 to perform a unified operation as a whole, and here, the output characteristic line of the fuel cell 30 It has a function of performing learning and performing power supply drive control of the fuel cell system 10 based on the result. In particular, the learning of the output characteristic line of the fuel cell 30 has a function of performing necessary correction when the output characteristic line has a discontinuous change when the measured characteristic value is replaced and updated.

  Specifically, the control unit 40 includes a battery learning unit 42 that performs learning based on actual measurement values for the output characteristic line of the fuel cell 30, and a power supply drive control that performs power supply drive control of the fuel cell system 10 based on the result. Module 50. Then, the battery learning unit 42 updates the initial learning characteristic module 44 that creates the first output characteristic line based on the actual measurement value, and updates and necessary corrections based on the actual measurement value for the IV characteristic line for which the initial learning has been completed. An IV characteristic line learning module 46 to perform, and an IP characteristic line learning module 48 to perform updating and necessary correction on the IP characteristic line for which the initial learning has been completed based on the actual measurement value.

  Such a function can be realized by software, specifically, by executing a fuel cell learning program. Some of these functions may be realized by hardware.

  The operation of this configuration, in particular, each function of the battery learning unit 42 of the control unit 40 will be described in detail with reference to FIGS. These figures all show the output characteristic lines of the fuel cell, wherein FIGS. 5 to 10 are IV characteristic lines, and FIGS. 11 and 12 are IP characteristic lines. Below, it demonstrates using the code | symbol of FIGS. 1-4.

  The learning of the output characteristic line includes the initial learning for creating the first output characteristic line and the normal learning for updating the output characteristic line based on the actually measured characteristic value for the already obtained output characteristic line. First, the initial learning function will be described. Here, the case where the output characteristic line is an IV characteristic line will be described. FIG. 5 to FIG. 8 are diagrams illustrating how an IV characteristic line is created and necessary correction is performed in initial learning.

In the initial learning, an IV characteristic line is created using the actual current value and the actual voltage value of the fuel cell 30 at the characteristic step points described with reference to FIG. Therefore, actual measurement may be performed for each ΔS, which is the step of the characteristic step point described with reference to FIG. 2, but it is desired to obtain the first IV characteristic line as quickly as possible in the initial learning. Therefore, after measuring the small current region, which is a region where the change is rapid on the IV characteristic line, with the step ΔS, the measurement is performed by skipping the characteristic step points. In this case, since the characteristic step point that has not been actually measured has not yet been learned, it is necessary to estimate and calculate the characteristic value for the characteristic step point to obtain the calculated characteristic value.
In order to obtain a calculated characteristic value for an unlearned characteristic step point on the IV characteristic line, first, a characteristic step point that is the starting point of the unlearned area is specified. In the following, it is assumed that the current value step is not learned from the m-th characteristic step point. In other words, the measurement steps up to m-1th characteristic step points have already been measured or learned. Next, the characteristic value of the actual measurement point is used as it is. Then, the characteristic step point of the actual measurement point is set as the nth, and the calculated characteristic value is sequentially obtained starting from the unlearned characteristic step point closest to the actual measurement characteristic step point and further toward the unlearned characteristic step point. By doing so, the accuracy of the characteristic value of the characteristic step point in the vicinity of the actual measurement point is increased, the deviation from the actual measurement characteristic value is reduced, and a more accurate output characteristic line can be obtained.

  The calculation characteristic value is calculated based on the fact that the IV characteristic line has a simple decrease characteristic based on the electrochemical reaction characteristic of the fuel cell. That is, it starts from the actual measurement point so that the voltage value of the characteristic step point with the larger current value is equal to or smaller than the voltage value of the characteristic step point with the smaller current value between two adjacent characteristic step points. Then, the calculated characteristic value is calculated toward the characteristic step points which are far away. In this case, the characteristic value is the same or increases from the nth actual measurement point toward the m-1st characteristic step point. However, if the characteristic value increases too much, the entire IV characteristic line decreases monotonously. Instead, singularities such as inflection points appear. Therefore, correction is performed so as to minimize the number of singular points.

  Such calculation characteristic values can be easily calculated by referring to a standard IV characteristic. That is, the IV characteristics that have been proven so far have the characteristic that the voltage value is always the same or decreases as the current value increases. When discontinuities occur when connecting to characteristic step points that have already been learned using the reference IV characteristics, correction that minimizes discontinuities may be performed. The procedure will be described below. The following procedure is executed by the function of the initial learning module 44 of the battery learning unit 42 described in FIG.

  FIG. 5 is a diagram illustrating a state in which actual measurement is performed in a region where current is small. Here, the actually measured part is shown as a solid line as learned. In FIG. 5, m−1 indicates that the actual measurement has been performed up to the m−1th characteristic step point, and an unlearned region starts from the mth characteristic step point. A region having a large current value is an unlearned region.

  For an unlearned area after the mth characteristic step point, an IV characteristic line is estimated based on the partial IV characteristic lines 70 obtained in the m-1st learning area. For this estimation, a characteristic line closest to an extension of the partial IV characteristic line 70 among the typical IV characteristics of the fuel cell 30 obtained so far can be used. Therefore, a characteristic line that is smooth and monotonically decreasing and continuously connected to the partial IV characteristic line 70 is used. The estimated IV characteristic line obtained in this way is close to the ideal IV characteristic line of the fuel cell and is used as a reference for the calculation of the calculated characteristic value thereafter. It will be called a characteristic line 72.

  In this way, first, a basic IV characteristic line connected continuously to the actually measured partial IV characteristic line 70 is set. This is the starting point for initial learning. Next, as described above, actual measurement is performed by skipping characteristic step points in the m-th and subsequent unlearned regions. N ≧ m, where the actual measurement is performed as the n-th characteristic step point.

  When the m-th and following are unlearned regions, the n-th characteristic step point is measured, and if the actual voltage value is on the basic IV characteristic line 72, the m-th characteristic step point is From the first to the nth, the basic IV characteristic line 72 can be used, so that the nth characteristic step point can be treated as learned. When the m-th and subsequent areas are unlearned regions, the n-th characteristic step point is actually measured, and when the actual voltage value is not on the basic IV characteristic line 72, it is necessary to correct the discrepancy.

  FIG. 6 is a diagram illustrating a first correction procedure. Here, the actual voltage value of the nth characteristic step point which is the actual measurement point 74 is not on the basic IV characteristic line 72. In this case, the basic IV characteristic line 72 is translated so as to pass through the actual measurement point 74. In FIG. 6, a parallel movement IV characteristic line 76 passing through the actual measurement point 74 is shown.

  The IV characteristic line indicates a monotonically decreasing characteristic in which the voltage value is broadly defined in accordance with the increase in the current value, based on the electrochemical reaction characteristic of the fuel cell 30. Therefore, when n ≧ m, when the parallel movement IV characteristic line 76 has a voltage value larger than the actually measured actual voltage value, it is inferior in view of the electrochemical reaction characteristics of the fuel cell 30. It will be natural. In that case, correction is performed. 7 and 8 are diagrams for explaining two examples of correction.

  FIG. 7 shows a case where the actual voltage value of the nth characteristic step point that is the actual measurement point 74 is smaller than the actual voltage value of the (m−1) th characteristic step point. At this time, the parallel IV characteristic line 76 passing through the actual measurement point 74 is used in a region where the voltage value is equal to or less than the actual voltage value of the (m-1) th characteristic step point. The parallel IV characteristic line 76 is not used in a region where the voltage value of the parallel IV characteristic line 76 exceeds the actual voltage value of the (m-1) th characteristic step point. In this region, the actual voltage value of the (m-1) th characteristic step point is used as the corrected IV characteristic line 78.

  That is, in the unlearned region where n ≧ m, the parallel IV characteristics are obtained under the condition that the voltage value is smaller than the large voltage value among the voltage value of the learned characteristic step point and the voltage value of the actually measured characteristic step point. The line 76 is corrected. The plurality of IV characteristic lines 70, 78, and 76 obtained in this way are the initial learning IV characteristic lines in this case.

  FIG. 8 shows a case where the actual voltage value of the nth characteristic step point that is the actual measurement point 75 is larger than the actual voltage value of the (m−1) th characteristic step point. Also at this time, the parallel IV characteristic line 76 passing through the actual measurement point 75 is used in a region where the voltage value is equal to or less than the actual voltage value of the (m−1) th characteristic step point. In the region where the voltage value of the parallel IV characteristic line 76 exceeds the actual voltage value of the (m-1) th characteristic step point, the measured value 75 is n without using the parallel IV characteristic line 76. The actual voltage value of the second characteristic step point is used as the corrected IV characteristic line 78 with a constant value. Even if it does in this way, since the voltage value of the correction | amendment IV characteristic line 78 is larger than the actual voltage value of the m-1th characteristic step point, the applicable range of the correction | amendment IV characteristic line 78 is an unlearned area | region. Only after the m-th characteristic step point. Next, the m−1th characteristic step point and the mth characteristic step point are connected. In the IV characteristic line 79 in this range, discontinuity is observed in the voltage value, but the discontinuity is stopped in the range of the minimum step between the characteristic step points.

  Again, except for the minimum discontinuity, in the unlearned region, the voltage value is smaller than the large voltage value of the learned characteristic step point voltage value and the measured characteristic step point voltage value. The parallel IV characteristic line 76 is corrected. The plurality of IV characteristic lines 70, 79, 78, and 76 obtained in this way are the initial learning IV characteristic lines in this case.

  As described above, in the initial learning of the IV characteristic line, the parallel IV characteristic line 76 extends from the unlearned characteristic step point closest to the actual measurement points 74 and 75 toward the further unlearned characteristic step point. Are sequentially corrected. The correction is performed under the condition that the voltage value is monotonously decreased in a broad sense with respect to the increase in the current value. In this way, it is possible to obtain an initial learning IV characteristic line that satisfies the natural monotonically decreasing characteristics in view of the electrochemical reaction of the fuel cell 30 while reflecting the results of the actual measurement points 74 and 75 while minimizing exceptions. it can.

  In the above description, the IV characteristic line is described as an actual IV characteristic line, that is, a characteristic line including the internal resistance of the fuel cell 30, but the initial learning of the theoretical IV characteristic line is performed based on the same concept. be able to. In this case, the activation overvoltage is obtained at each characteristic step point by the following procedure. That is, in the fuel cell 30, the impedance of the fuel cell 30 is obtained from the AC component detected by the current detection means 34 and the voltage detection means 36 when using the high-frequency signal source 32. Then, the internal resistance is obtained from the decomposition of the real axis component and the imaginary axis component, the voltage drop due to the internal resistance is obtained from the product of the DC current value and the internal resistance, and this is subtracted from the DC voltage value to calculate the activation overvoltage value. . In this way, the activation overvoltage at each characteristic step point is calculated from the current value, voltage value, and impedance value at each characteristic step point.

  As described above, the current-activation overvoltage characteristic line has a monotonically decreasing characteristic in which the voltage value is narrowly defined with respect to an increase in the current value. Accordingly, in the initial learning of the actual IV characteristic line, the basic IV in the initial learning of the theoretical IV characteristic line is used in the same manner as when the proven actual IV characteristic line is used as the basic IV characteristic line. A proven theoretical IV characteristic line can be used as the characteristic line. Then, by calculating the calculation characteristic value in the unlearned region under the above procedure and the narrowly monotonic decreasing condition, initial learning can be performed for the theoretical IV characteristic line.

  Next, a normal learning procedure for learning and updating the IV characteristic line obtained in the above-described initial learning based on the actual measurement values will be described with reference to FIGS. The following procedure is executed by the function of the IV characteristic curve learning module 46 of the battery learning unit 42 described in FIG.

  Here, since the learned IV characteristic line has already been obtained, there is a problem if the actual voltage value of the n-th characteristic step point measured thereafter is on the learned IV characteristic line. There is no. When the actually measured voltage value is not on the learned IV characteristic line, it is necessary to perform correction so as to satisfy the monotonic decrease characteristic. 9 and 10 show two examples of correction. In these figures, the learned IV characteristic line is the IV characteristic line 80 and the actually measured characteristic step points are the actually measured points 82 and 83.

  FIG. 9 shows a case where the actual voltage value of the nth characteristic step point that is the actual measurement point 82 is larger than the voltage value on the learned IV characteristic line 80. At this time, the voltage value of the characteristic step point having a smaller current value than the actual measurement point 82 is set to be the same as the actual voltage at the actual measurement point 82. The voltage value of the characteristic step point having a current value larger than that of the actual measurement point 82 is the voltage value on the learned IV characteristic line 80. That is, the voltage value on the learned IV characteristic line 80 is taken from the (n + 1) th characteristic step point. In this way, the entire IV characteristic line including the actual measurement point 82 can have a monotonously decreasing characteristic in a broad sense. Here, the plurality of IV characteristic lines 80, 84, 85, and 80 become the corrected IV characteristic lines from the smaller current value to the larger current value. In this way, learning reflecting the characteristic value of the actual measurement point 82 is performed.

  FIG. 10 shows a case where the actual voltage value of the nth characteristic step point that is the actual measurement point 83 is smaller than the voltage value on the learned IV characteristic line 80. At this time, the voltage value of the characteristic step point having a larger current value than the actual measurement point 83 is set to be the same as the actual voltage at the actual measurement point 83. The voltage value at the characteristic step point having a current value smaller than the actual measurement point 83 is set to the voltage value on the learned IV characteristic line 80. That is, the learned voltage value on the IV characteristic line 80 is taken from the characteristic step point before the (n-1) th. In this way, the entire IV characteristic line including the actual measurement point 83 can have a monotonously decreasing characteristic in a broad sense. Here, the plurality of IV characteristic lines 80, 87, 86, 80 become corrected IV characteristic lines from the smaller current value to the larger current value. In this way, learning reflecting the characteristic value of the actual measurement point 83 is performed.

  Again, starting from the actual measurement point, between the two adjacent characteristic step points, the voltage value of the characteristic step point having the larger current value is smaller than the voltage value of the characteristic step point having the smaller current value. Corrections are made sequentially toward characteristic step points that are far away. The correction is performed under the condition that the voltage value is monotonously decreased in a broad sense with respect to the increase in the current value. That is, in the initial learning, the correction in the learning with respect to the IV characteristic line is performed under the same conditions as the correction in the calculation of the calculated characteristic value at the unlearned characteristic step point. In the above description, the learning of the actual IV characteristic line has been described. However, the same procedure can be performed for the learning of the theoretical IV characteristic line under the same conditions.

  Next, the case of the IP characteristic line as the output characteristic line will be described. Also in the IP characteristic line, there are initial learning and learning by actually measured values after the initial learning. As already described with respect to the IV characteristic line, since the same conditions are used for both initial learning and subsequent learning, the learned IP characteristic line is obtained here by the initial learning. The learning performed after being performed will be described.

  Even in the case of learning of the IP characteristic line, the characteristic value of the actual measurement point is used as it is, as in the case of the IV characteristic line. Then, the characteristic step point of the actual measurement point is set as the nth, and the estimated characteristic value is sequentially calculated starting from the unlearned characteristic step point closest to the actual measurement characteristic step point and further toward the unlearned characteristic step point. The same is true. The difference from the case of the IV characteristic line is that the actual IV characteristic line is a simple decrease characteristic in a broad sense and the theoretical IV characteristic line is a simple decrease characteristic in a narrow sense, from the characteristics of the electrochemical reaction of the fuel cell. On the other hand, the IP characteristic line has a monotonously increasing characteristic in a narrow sense.

  Therefore, in the case of the IV characteristic line, the voltage value of the characteristic step point having the larger current value is equal to or smaller than the voltage value of the characteristic step point having the smaller current value between two adjacent characteristic step points. In the case of the IP characteristic line, the power value of the characteristic step point with the larger current value between the two adjacent characteristic step points is the characteristic step with the smaller current. Correction or the like is performed so as to be larger than the power value of the point. Here, since the IP characteristic line does not take a constant value and always has a monotonically increasing characteristic, the gradient = power value / current value is estimated to be larger on the plus side than a predetermined threshold gradient set in advance. Etc. should be performed. The threshold gradient is a positive value.

  Therefore, a procedure for learning and updating the initially learned IP characteristic line based on the actual measurement values will be described. The following procedure is executed by the function of the IP characteristic curve learning module 48 of the battery learning unit 42 described in FIG.

  Here, since the learned IP characteristic line has already been obtained, there is a problem if the actual power value of the nth characteristic step point measured after that is on the learned IP characteristic line. There is no. When the actually measured power value is not on the learned IP characteristic line, it is necessary to perform correction so as to satisfy the monotonously increasing characteristic. 11 and 12 show two examples of correction. In these figures, the learned IP characteristic line is the IP characteristic line 90, and the actually measured characteristic step points are the actually measured points 92 and 93.

  FIG. 11 shows a case where the actual power value of the nth characteristic step point that is the actual measurement point 92 is smaller than the power value on the learned IP characteristic line 90. At this time, with respect to the characteristic step point having a current value smaller than the actual measurement point 92 which is the nth characteristic step point, the gradient = power value / current value toward the actual measurement point 92 starts from the actual measurement point 92 and is greater than the threshold gradient. The power value is corrected so as to have a large positive gradient. In practice, the magnitude of the gradient may be a threshold gradient. As the threshold gradient, the smallest gradient in the entire range of the proven IP characteristic line 90 or a weighted gradient obtained by multiplying the minimum gradient by an appropriate coefficient can be used. Further, the characteristic step point having a current value larger than the actual measurement point 92 that is the nth characteristic step point is set to the same power value as the learned IP characteristic line 90. That is, the power value on the learned IP characteristic line 90 is taken from the (n + 1) th characteristic step point. In this way, the entire IP characteristic line including the actual measurement point 92 can have a monotonously increasing characteristic in a narrow sense. Here, the plurality of IP characteristic lines 90, 94, 95, 90 become corrected IP characteristic lines from the smaller current value to the larger current value. In this way, learning reflecting the characteristic value of the actual measurement point 92 is performed.

  FIG. 12 shows a case where the actual power value of the nth characteristic step point that is the actual measurement point 93 is larger than the power value on the learned IP characteristic line 90. At this time, with respect to the characteristic step point having a current value larger than the actual measurement point 93 which is the nth characteristic step point, the gradient from the actual measurement point 93 = power value / current value is less than the threshold gradient while starting from the actual measurement point 93. The power value is corrected so as to have a large positive gradient. Further, the characteristic step point having a current value smaller than the actual measurement point 93 which is the nth characteristic step point is set to be the same as the power value of the learned IP characteristic line 90. That is, the power value on the learned IP characteristic line 90 is taken from the (n-1) th characteristic step point. In this way, the entire IP characteristic line including the actual measurement point 93 can have a monotonously increasing characteristic in a narrow sense. Here, the plurality of IP characteristic lines 90, 97, 96, 90 become corrected IP characteristic lines from the smaller current value to the larger current value. In this way, learning reflecting the characteristic value of the actual measurement point 93 is performed.

  In this way, starting from the actual measurement point, the distance is gradually increased toward the characteristic step point so that the gradient = power value / current value becomes a positive gradient larger than the threshold gradient between two adjacent characteristic step points. Correction is performed. For the initial learning of the IP characteristic line, the slope = power value / current value is larger than the threshold value gradient between two adjacent characteristic step points in the unlearned region under the same conditions. Thus, the correction can be performed in a similar procedure.

It is a figure which shows the structure of a fuel cell system provided with the battery learning system of embodiment which concerns on this invention. In the battery learning system of embodiment which concerns on this invention, it is a figure which shows the mode of the characteristic step point arrange | positioned in grid | lattice form, making an output characteristic into a current-voltage characteristic. In the battery learning system of embodiment which concerns on this invention, it is a figure which shows typically the current-voltage characteristic line of a fuel cell. It is a figure which shows typically the electric current-power characteristic line of a fuel cell in the battery learning system of embodiment which concerns on this invention. In the battery learning system of embodiment which concerns on this invention, it is a figure which shows the state which measured in the area | region where an electric current is small among the procedures of initial learning. In the battery learning system of embodiment which concerns on this invention, it is a figure which shows a mode that a basic IV characteristic line is translated so that it may pass along a measurement point among the procedures of initial learning. It is a figure explaining one example which correct | amends by initial learning in the battery learning system of embodiment which concerns on this invention. It is a figure explaining the other example which correct | amends by initial learning in the battery learning system of embodiment which concerns on this invention. It is a figure explaining one example which correct | amends by learning of the IV characteristic line by which the initial learning was completed in the battery learning system of embodiment which concerns on this invention. It is a figure explaining the other example which correct | amends by learning of the IV characteristic line by which the initial learning was carried out in the battery learning system of embodiment which concerns on this invention. It is a figure explaining one example which correct | amends by learning of the IP characteristic line by which the initial learning was completed in the battery learning system of embodiment which concerns on this invention. It is a figure explaining the other example which correct | amends by learning of the IP characteristic line by which the initial learning was completed in the battery learning system of embodiment which concerns on this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 Fuel cell system, 11 Power supply circuit, 12 Power storage device, 14, 18 Smoothing capacitor, 16 Voltage converter, 20 Inverter, 22 Rotating electric machine, 30 Fuel cell, 32 High frequency signal source, 34 Current detection means, 36 Voltage detection means, 38 storage device, 39 output characteristic line file, 40 control unit, 42 battery learning unit, 44 initial learning module, 46 IV characteristic line learning module, 48 IP characteristic line learning module, 50 power drive control module, 60 characteristic Step point, 62 characteristic value, 64 characteristic line, 66, 67, 70, 72, 76, 78, 79, 80, 84, 85, 86, 87 IV characteristic line, 68, 90, 94, 95, 96, 97 IP characteristic line, 74, 75, 82, 83, 92, 93 Measured points.

Claims (4)

On the current-voltage plane with the current value on the horizontal axis and the voltage value on the vertical axis, a plurality of points arranged in a grid pattern in a predetermined array step are used as characteristic step points .
A combination of the current value and the voltage value corresponding to the current value is a characteristic value ,
An acquisition means for detecting the actual voltage value by actually measuring the voltage value corresponding to the arbitrary current value for the battery, and acquiring a combination of the arbitrary current value and the actual voltage value as an actual measurement characteristic value;
The characteristic step point having the characteristic value that matches the measured characteristic value is the measured characteristic step point, and the characteristic step point in an area where the measured characteristic value does not exist between the two actually measured characteristic step points is not learned. As a characteristic step point,
As the voltage value and the current value increases to definitive the electrochemical reaction of the battery corresponds to a monotonically decreasing characteristic of constant or decreases, estimates calculate the characteristic value of the characteristic step points of the unlearned as calculated characteristic value Calculating means for
A creation means for creating an initial current-voltage characteristic line by sequentially connecting the measured characteristic step point and the characteristic step point having the characteristic value that matches the calculated characteristic value;
Updating means for sequentially updating current-voltage characteristic lines already obtained by learning based on the measured characteristic values obtained by the obtaining means after creating the first current-voltage characteristic lines in the creating means;
Battery learning system characterized in that it comprises a. The battery learning system according to claim 1,
The calculating means includes
The actual characteristic closest the step point starting from the characteristic step point unlearned, battery learning system further away the towards the characteristic step point unlearned, characterized by successively estimating calculate the calculated characteristic value. The battery learning system according to claim 1,
The calculating means includes
A partial current-voltage having a plurality of the basic current-voltage characteristic lines in advance with a representative current-voltage characteristic line of the battery as a basic current-voltage characteristic line, and connecting a plurality of the measured characteristic step points. A battery learning system that estimates and calculates the calculated characteristic value based on the basic current-voltage characteristic line closest to the characteristic line. The battery learning system according to claim 1,
The battery learning system, wherein the battery is a fuel cell.

Images