DESCRIPTION STATE-OF-CHARGE ESTIMATION APPARATUS Field [0001] The present invention relates to an apparatus that estimates a state of charge (SOC) of a secondary battery. Background [0002] Generally, when batteries are used in a railway system in order to supply energy during running or to recover regenerative energy at the time of deceleration and halt, many cell batteries are used with being connected to each other in series-parallel combination, because of a large amount of electric energy to be consumed. At this point, it is necessary to obtain an SOC for the running of a vehicle and control of charge/discharge of batteries. [0003] There has been conventionally known a method of estimating an SOC of a battery by a combination of a system using integration of charge and discharge currents and a system using a relation between an open-circuit voltage and the SOC of a secondary battery held in equilibrium (for example, Patent Literature 1 mentioned below). The method disclosed in Patent Literature 1 is to estimate an SOC from current integration at the time of charge or discharge, and to estimate an SOC by estimating the open-circuit voltage during a standby time. [0004] In relation to the open-circuit voltage estimation method, there has been known an example of estimating an open-circuit voltage using a voltage change after charge or discharge (for example, Patent Literature 2 mentioned below). Citation List Patent Literatures [0005] Patent Literature 1: Japanese Patent Application Parenr Literature 2: Japanese Patent Application Laid-open No. 2007-333474 Summary Technical Problem [0006] However, the method of Patent Literature 1 mentioned above has the following problems. That is, because an equivalent circuit of a battery is used to estimate an open-circuit voltage, an internal resistance dependent on a temperature or an SOC is required, and it is also required to take battery degradation into consideration. [0007] Meanwhile, when an SOC of a power storage device composed of many batteries is estimated, it is desirable to rsaszre a battery voltage and a battery current for the entire batteries that constitute the power storage device. However, sensors for measuring the battery voltage and the battery current are not high in resolution and accuracy, so that the measurement values often include ripples. According to Patent Literature 2 mentioned above, the open-circuit volrage is estimated by measuring a total voltage of the batteries and making addition of a predetermined value when the change rate of the total voltage is equal to or lever th=n a predetermined value after charge or discharge. Therefore, it is possible to estimate the open-c±rcuiz voltage only by measuring the total voltage without using any equivalent circuit constants. However, there are problems in that it is required to ensure a high voltage-measuring accuracy to obtain a stable value because a significant difference is caused in the SOC when the open-circuit voltage of cells differs only by 1 mV. [0008] The present invention has been achieved in view of the above-mentioned circumstances, and an object of the present invention. is zz prcvica =. state-of-charge estimation apparams "tha- can suppress change in an SOC estimated value. Solution to Problem [0009] In order to solve the above-mentioned problems and achieve the object, the present invention provides a state-of-charge estimation apparatus that is connected to a power storage device constituted by connecting a plurality of cell batteries, and estimates a state of charge of the power storage device, the state-of-charge estimation apparatus comprising: a first computation unit that computes a current integration value based on a current from a detector than detects a current flowing to or from the power storage device from or to a power conversion unit while a previous valne of a state-of-charge estimated value is set as an initial value, and that outputs the computed current integration value as a first state-of-charge estimated value; a second computation unit that estimates, after suspension of charge or discharge in the power storage device, an open-circuit voltage based on a voltage detected in a connection point between the power conversion unit and the power storage device, a previous value of the voltage detected in the connection point, and a gain that decreases with passage of time after the suspension of the charge or discharge in nne power storage device, and that outputs the estimated open—circuit voltage as a second state-of-charge estimated value; and a selection unit that selects any one of the first state-of-charge estimated value and the second state-of-charge estimated value. Advantageous Effects of Invention [0010] According to the present invention, the state-of-charge estimation apparatus includes a first SOC -computation unit that .esrima.tfiSr.a first state-of-charge estimated value, a second SOC computani" estimates a second state-of-charge esrinatea valTie, and an SOC sudden-change prevention unit that suppress sudden change in the second state-of-charge estimated value based on a variable gain and a previous SOC value. Therefore, it is possible to suppress the change in the SOC estimated value even when the battery-voltage measurement accuracy is low or the measured voltage has ripples-Brief Description of Drawings [0011] FIG. 1 is a diagram showing a configuration of an electric rolling stock to which a state-of-charge estimation apparatus according to the present invention is applied. FIG. 2 is a diagram showing configuration of a power storage device shown in FIG. 1. FIG. 3 is a diagram showing a configuration of a state-of-charge estimation apparatus according to a first embodiment of the present invention. FIG. 4 is a diagram showing a configuration of a first SOC computation unit shown in FIG. 3. FIGS. 5 is a diagram showing a configuration of a first flag generation unit shown in FIG. 3. FIGS. 6 is a diagram showing a configuration of a second flag generation unit shown in FIG. 3. FIG. 7 is a diagram showing a conf±gnr=t:ion cf a second SOC computation unit shown in FIG-. 3. FIG. 8 is a chart showing a relation between a cell open-circuit voltage and an SOC. FIG. 9 is a chart showing one example of a gain K3 shown in FIG. 7. FIG. 10 is a diagram showing a configuration of an SOC sudden-change prevention unit shown in FIG. 3. FIG. 11is.-a. diagram, snowing a configuration of an SOC selection unit shown in FIG. 3. FIG. 12 is a diagram showing one example of a circz-.-that sets an initial value of a unit-delay unit shown in FIG. 3. FIG. 13 is a diagram showing a configuration of a state-of-charge estimation apparatus according to a second embodiment of the present invention. FIG. 14 is a diagram showing a configuration of a. second SOC computation unit according to a third embodiment of the present invention. Description of Embodiments [0012] Embodiments of a state-of-charge estimation apparatus according to the present invention will be described below in detail with reference to the drawings. The present invention is not limited to the embodiments. [0013] First embodiment. FIG. 1 is a diagram showing a configuration of an electric rolling stock to which a state-of-charge estimation apparatus 10 according to the present invention. is applied, FIG. 2 is a diagram showing a configuration of a power storage device 11 shown in FIG. 1, and FIG. 3 is a diagram showing a configuration of the state-of-charge estimation apparatus 10 according to the first embodiment of the present invention. [0014] A power conversion device (power conversion T~-~) 12 shown in FIG. 1 is connected to a power collector 14 and wheels 15 that serve as a return circuit of a return current, and receives power from an overhead wire 13 connected to an electric power substation (not shown) that serves as a direct-current (DC) power supply. A voltage detector 18 shown in FIG. 1 detects a voltage in a connection point between the power conversion device 12 =^H the power storage.de.vice.il... A current detector 17 detects ,. a current that flows to or from the power storage device 11. The state-of-charge estimation apparatus 10 computes and outputs an SOC estimated value based on a current command Iall* from the power conversion device 12, a total voltage Vail detected by the voltage detector 18, and a total current Iall detected by the current detector 17. [0015] With reference to FIG. 2, the power storage device 11 is configured to include battery modules 21-1 to 2n-m. The battery module 21-1 is constituted by connecting cell batteries 6-1 to 6-k in series. Similarly, the battery module 2n-l is constituted by connecting cell batteries 7-1 to 7-k in series. The battery module 21-m is constituted by connecting cell batteries 4-1 to 4-k in series. The battery module 2n-m is constituted by connecting cell batteries 5-1 to 5-k in series. Each cell battery is a repeatedly dischargeable/rechargeable secondary battery such as a lithium ion secondary battery, and an SOC of each cell battery can be observed by measuring an open-circuit voltage. A breaker, a battery monitoring device and the like are not shown here although these are often provided in the power storage device 11. [0016] A voltage between terminals of the power storage device 11 is assumed as the total voltage Vail. It-is assumed that a discharge/charge current is the total current Iall and that the value of the total current Iall is positive in a charging state. It is also assumed that the number of parallel-connected cells (modules) is m and the number of series-connected cells (modules) is n in the. power storage device 11, where n=(the number of modules)x(the number of series-connected cells in each module). Because of the large number of cells that constitute the power storage device 11, a voltage resulting from a resxstarrcer-component of -conductors or cables used for the connecting the terminals is added to the total voltage Vail during charge/discharge. [0017] In FIG. 1, the voltage detector 18 and the current detector 17 provided on an intermediate between the power conversion device 12 and the power storage device 11 are used to measure the total voltage Vail and the total current Iall. However, the total voltage Vail and the total current Iall are not limited to those measured in this manner, and a total voltage Vail and a total current Iall measured by a charge/discharge circuit side (not shown) may be used. In this case, the resistance component of the cable led to the state-of-charge estimation apparatus 10 becomes larger, and therefore it is desirable to consider an attenuation amount caused by this resistance component. [0018] (State-of-charge estimation apparatus 10) FIG. 3 is a diagram showing a configuration of the state-of-charge estimation apparatus 10 according to the first embodiment of the invention. The state-of-charge estimation apparatus 10 is configured to include, as main constituent elements, a first flag generation unit 21, a second flag generation unit 22, a second SOC computation unit 24, a first SOC computation unit 20, an SOC selection unit 23, an SOC sudden-change prevention unit (change-amount restriction unit) 25, and a unit-delay unit 26. [0019] The first SOC computation unit 20 computes and outputs a first state-of-charge estimated value .(hereinafter, simply referred to as "first estimated value") SOC1 based on the total current Iall, a first flag (FLG1) generated by the first flag generation unit 21, and an SOCz that is a previous value of SOC from the unit-delay unit 2 6 (hereinafter, referred to as "previous value SOCz"). [00 20] The-second SOC computation unit 24 computes and a second state-of-charge estimated value SOC20 based on the total voltage Vail and a second flag (FLG2) and a third flag (FLG3) from the second flag generation unit 22. [0021] The SOC sudden-change prevention unit (hereinafter, simply referred to as "sudden-change prevention unit") 25 outputs a third state-of-charge estimated value (hereinafter, simply referred to as "third estimated value") SOC2 based on the second state-of-charge estimated value (hereinafter, simply referred to as "second estimated value) SOC20 from the second SOC computation unit 24 and the previous value SOCz from the unit delay-unit 26. [0022] The SOC selection unit 23 selects, as the SOC, any one of the first estimated value SOC1 from the first SOC computation unit 20 and the third estimated value S0C2 from the sudden-change prevention unit 25 according to the value of the FLG2 from the second flag generation unit 22. [0023] An overall computation cycle T1 [sec] of the state-of-charge estimation apparatus 10 is 0.1 [sec], for example. As described above, the power storage device 11 is composed of many cells. Generally, series-connected batteries are provided with a battery-voltage equalization unit (not shown), and the parallel-connected series cells have a substantially equal voltage value. Therefore, it is considered that SOC variations among the cell batteries are not remarkable. For this reason, the SOC is handled as a single value. [0024] A configuration of the state-of-charge estimation apparatus 10 is explained below in detail. [0025] (First SOC computation unit 20) FIG. 4 is a diagram showing a configuration of the first SOC computation unit shown in FIG. 3, FIG. 5 is a diagram showing a- configuration of the first flag gener shrxz: in FIG. 3, and FIG. 6 is a diagram showing a configuration of the second flag generation unit shown in FIG. 3. [0026] In FIG. 4, the first SOC computation unit 20 is configured to include a gain unit 41, a multiplier 42, an integrator 43, a gain unit 44, and a gain unit 45. [0027] The gain unit 41 obtains a cell-current average value Icell by multiplying the total current Iall by a gain 1/m. That is, the gain unit 41 obtains an average value of cell currents by dividing the total current Iall by the number m of parallel-connected cell batteries. [0028] The multiplier 42 multiplies the cell-current average value Icell from the gain unit 41 by the FLG1 and outputs a multiplication result. This multiplier 42 stops integration when the FLG1 is 0 and executes integration when the FLG1 is 1. The FLG1 is generated by the first flag generation unit 21 described later, and the FLG1=1 indicates execution of integration and the FLG1=0 indicates stop of integration. [0029] The integrator 43 integrates an output from the multiplier 42, and the integrator 43 is reset every time in a case where FLG2 is 1. [0030] The gain unit 45 multiplies the previous value SOCz by a gain K2, thereby converting the value SOCz into an initial c~irrexit-integration value (an initial value) . In this way, the initial current-integration value is obtained from the previous value SOCz when the integrator 43 is reset by the FLG2. The gain K2 of the gain unit 45 is K2 = battery capacity [Ah] x 3600 [sec] / 100 [%]. The FLG2 is generated by the second flag generation unit 22 described later. [0031] The gain unit 44 multiplies a current integration value frois the integrator 43 by a gain Kl, so as to convert the integration value in~ zhe first estimated value S0C1. Note that Kl = 1 / battery capacity IAhj / 360 [sec] x 100 [0032] The first SOC computation unit 20 is configured to stop integration to prevent an error in the SOC value from increasing due to the influence of a current detection error when the FLG1 is 0. This configuration is described. A case of satisfying a relation of |Iall| present value)- from the. gain unit ..-51a. The gain unit 52 multiplies the difference from the subtracter 58 by the gain K3 and outputs the multiplication result. The adder 59 adds the output (the previous value) from the unit-delay unit 55 to the output from the gain unit 52, thereby obtaining an open-circuit voltage estimated value (an open-circuit voltage) Vocell. [0048] After passage of a sufficient time from the charge or discharge, Vocell=Vcell. However, right after the charge, VocelKVcell, and a difference between the Vocell and the Vcell then gradually decreases. Furthermore, right after the discharge, Vocell>Vcell, and the difference therebetween then gradually decreases. Therefore, the second SOC computation unit 24 obtains the open-circuit voltage estimated value Vocell at a present time (k) using the following Equation. Vocell(k)=Vcell(k-1)+K3x(Vcell(k) -Vcell(k-1) ) (1) [0049] That is, multiplication of a change in the cell voltage average value Vcell from k-1 to k by K3 is added to the cell voltage average value Vcell(k-1) t3 before then. Accordingly, when the gain K3=l, the open-circuit voltage estimated value Vocell is equal to the present cell voltage average value Vcell, and when the gain K3<1, the influence of the change in the cell voltage average value Vcell on the open-circuit voltage estimated value Vocell decreases. Herein, the gain K3 is a variable gain. [0050] The limiter 53 limits a value of the open-circuit voltage estimated value Vocell from the adder 59 to a practical value for reference to a table 54. [0051] In the table 54a, the open-circuit voltage estimated value Vocell is caused to correspond to the SOC as shown in FIG. 8. The correspondence between the open-circuit voltage- estimated value Vocell and the SOG can be by a charge/discharge test on the battery cell. The second SOC computation unit 24 outputs, as the second estimated value SOC20, the value of the SOC corresponding to the open-circuit voltage estimated value Vocell limited by the limiter 53 while referring to the table 54a. An approximate expression may be used in place of the table 54a. [0052] The value of the gain K3 set for the gain unit 52 is set to be smaller with the passage of time and to become smaller than 1 as a final value as shown in FIG. 9. FIG. 9 shows the value of the gain K3 in the case where t2 is 20 [sec] and t3=5 [sec], as one example. The time when the second SOC computation unit 24 that has detected FLG3=1 starts operating is set as 0 [sec]. When the second SOC computation unit 24 performs computation at an interval of t3, for example, the value of the gain K3 during the period t2 (0 to 20 [sec]) is about 6. [0053] The reason for setting the value of the gain K3 as described above is as follows. A state where the power storage device 11 is caused to suspend its discharge when being discharged is described by way of example. [0054] Around the period t2, which is just after the end of the discharge, the difference between Vcell(k-l) and Vcell(k)-Vcell(k-1) is large. When the value of the gain 13 is small around the period t2, a change in the open-rr. r-cn- - voltage estimated value Vocell is large and the second estimated value SOC20 greaT1y changes accordingly. Until just before the end of the discharge, the SOC is estimated by the current integration. However, around the period t2 just after the end of the discharge, the total voltage Vail rapidly recovers (rises). With a view to estimate the stable SOC while suppressing the change in the second estimated -value, SQC20, .the, second SOC computation unit 24 according tc -the present embodiment sets the value of the gain K3 to be large around the period t2 just after the end of the discharge, thereby preventing the excess of the SOC. The excess of the SOC means that the SOC estimated value by the estimation of the open-circuit voltage right after the end of the discharge falls below the SOC estimated value by the current integration just after the end cf the discharge. [0055] The value cf the gain K3 is set to gradually decrease after the passage of the period t2 and to be equal to 0.5 with the passage of about 50 [sec]. The reason for setting the value of the gain K3 as described above is as follows. The change in the cell voltage average value Vcell is originally supposed to be small because of the passage of a certain time after the end of the discharge. However, even when the discharge is suspended, the cell voltage average value Vcell fluctuates because the power conversion device 12 continues switching and the current detector 17, therefore, detects some ripple current or the voltage detector 18 has the voltage detection error. When the open-circuit volrage estimated value Vocell changes along with this change, the second estimated value SOC20 changes accordingly. To suppress such a change in the second estimated vaJLr-e SOC20, the second SOC computation unit 24 according tc the present embodiment gradually decreases the value of the gain K3 after the passage of the period t2 and sets the gain K3 to K3=0.5 after the passage of about 50 [sec], thereby suppressing the change in the open-circuit voltage estimated value Vocell. [0056] The gain K3 during the charge is also set to a value that decreases with the passage of time similarly to that during the discharge. For example, the value of the gain K3-is set to beargea-r-ound the period t2 and to gradually decrease after the passars cf niie p=crod t2. This is because an absolute value cf a change rate of the voltage average value Vcell is large to small both after the discharge and after the charge although the magnitude relation between the open-circuit voltage estimated value Vocell and the voltage average value Vcell after the discharge is opposite to that after the charge. [0057] (SOC sudden-change prevention unit 25) FIG. 10 is a diagram showing a configuration of the sudden-change prevention unit 25 shown in FIG. 3. The sudden-change prevention unit 25 is configured to include, as main constituent elements, a subtracter 63, a gain unit 61, a limiter 62, and a subtracter 64. [0058] The subtracter 63 outputs a difference between the previous value SOCz and the seccnd estimated value SOC20. The gain unit 61 multiplies the difference from the subtracter 63 by K4, and the limiter 62 limits an output from the gain unit 61 to ±0.02 [%] per 0.1 [sec] and outputs the limited result. The subtracter 64 adds an output from the limiter 62 to the previous value SOCz and outputs an addition result as the rhird estimated value SOC2. [0059] The above processing is specifically explained. First, the third estimated value SOC2 [%] is computed by the following Equation. Third estimated value SOC2 [%] = SOCz - Limit(K4(SOCz- SOC20)) (2) where the value of K4 is 0.8, for example, and for the limiter 62, ±0.02 [%] is set, for example. [0060] Because the computation cycle of the sudden-change prevention unit 25 is 0.1 [sec], the third estimated value SOC2 changes at a maximum by 0.02 [%] for 0.1 [sec]. Furthermore, the computa'tion>cycle-t3 of the second SQC .computation unit 24 shown in FIG. 7 is 5 [sec]. Tie—=fore, the second estimated value SOC20 from the second SOC computation unit 24 is updated once every 50 times of executions of computations of the sudden-change prevention unit 25. That is, the second estimated value SOC20 changes at a maximum by 0.02x50=1 [%] per 5 [sec]. As a specific example, when the second estimated value SOC20 inputted to the sudden-change prevention unit 25 decreases frccn. 50%, for example, the second estimated value SOC20 becc-ves 49.98 [%] 0.1 [sec] later and becomes 49.96 [%] further 0.1 Isec] later. That is, the second estimated value SOC20 5 [sec] after being inputted to the sudden-change prevention unit 25 decreases by 1 [%] to 49.0 [%] . Similarly, when a. value of the second estimated value SOC20 increases from 5Q%, the second estimated value SOC20 becomes 51.0 [%] 5 [sec] lazer. [0061] In this way, the sudden-change prevention unit 25 prevents the third estimated value SOC2 from suddenly changing by limiting the change in the second estimated value SOC20 from the second SOC computation unit 24 re 1% for 5 seconds. When the sudden-change prevention unit 25 does not have the limiter (Limit) 62, the sudden-change prevention unit 25 outputs a K4 multiple of the difference between the second estimated value SOC20 and the previous value SOCz as the third estimated value SOC2. For ex=3=?Ie, when K4=l, the difference between the second estimated value SOC20 and the previous value SOCz is direcT1y reflected in the third estimated value S02. Furthermore, the above-mentioned value of K4 and a set value of the limiter 62 are given as an example and the invention is not limited to these values. [0062] FIG. 11 is a diagram showing a configuration of the SOC selection unit shown in FIG. 3. The SOC selection unit'23 selects' any"one of the -first estimated value- SOC1 estimated by the first SOC computation unit 2 0 and the third estimated value SOC2 that is estimated by the second SOC computation unit 24 and has the sudden change limited by the sudden-change prevention unit 25 by a switch 71 that operates according to the value of the FLG2, and outputs the selected value as SOC that is the final SOC estimated value. For example, the first estimated value SOC1 is selected when the FLG2=0, and the third estimated value SOC2 is selected when the FLG2=1. [0063] Specifically, the FLG2=0 is held during the charge and discharge and for the period (t2) of the passage of about 20 seconds from the charge or discharge is suspended. Therefore, the switch 71 selects the first estimated value SOC1 from the first SOC computation unit 20 and outputs the first estimated value SOC1. That is, when IIall|in. the situation the_ number of series . connected cell batteries exceeds 160. On the other hand, according to the table 54a shown in FIG. 8, the relation between the open-circuit voltage and the SOC is that the value of SOC/open-circuit voltage increases as the open-circuit voltage and the SOC rise. Therefore, when the SOC is larger, the influence of a voltage change increases. When the open-circuit voltage changes by 10 mV (the total voltage is 1.6 V for the number of series-connected cell batteries is 160), for example, the SOC is considered to change by 2 to 3%. Therefore, it is difficult to keep the change in the SOC below a 1% level in consideration of resolution and measurement accuracy of the voltage detector 18. Furthermore, a DC-DC converter (not shown) that is included in the power conversion device 12 and executes the charge and discharge for the power storage device 11 is caused to operate with the current command being 0, it is thought that a voltage ripple resulting from a ripple current occurs. Under these conditions, it is necessary to provide an estimation method that allows the SOC estimated value to include a few errors and prevents the SOC estimated value from remarkably deviating from an actual SOC and from excessively changing under these conditions. Moreover, in the railway system, it is assumed to frequenT1y use not only constant-current operation but also constant-voltage operation during the charge or discharge so as to improve battery utilization efficiency and to protect the batteries. [0070] The state-of-charge estimation apparatus 10 according to the present embodiment can ensure a high estimation accuracy both during the charge/discharge and during suspension thereof by using a combination of SOC computation based on the current integration (the first SOC computation, unit 2Hi and .SOC computation, based on the; open-circuit voltage (the second SOC ccmpu"cax:ion unit 24) . [0071] Moreover, it is unnecessary to use a battery equivalent circuit model for the estimation of the open-circuit voltage, and it is possible to obtain a stable estimated value by the use of only the measured value of the total voltage Vail without any large deviation from the actual SOC and the excessive change in the estimated value. Furthermore, the state-of-charge estimation apparatus 10 can be less influenced by noise. [0072] Further, even when the charge or discharge continues at a low current at which the influence of the error is larger in the case of estimation based on the current integration, the state-of-charge estimation apparatus 10 according to the present embodiment does not erroneously estimate the open-circuit voltage although the estimated value can not be updated but the charge or discharge continues. Therefore, the state-of-charge estimation apparatus 10 according to the present embodiment can handle not only the constant-current charge and discharge but also the constant-voltage charge and discharge. [0073] As described above, the state-of-charge estimation apparatus 10 according to the present embodiment includes: the first SOC computation unit 20 that computes the current integration value of the total current Iall detected by the current detector 17 and computes the first estimated value SOC1 based on the current integration value while the previous value SOCz is set as the initial value; the second SOC computation unit 24 that computes, after the charge or discharge in the power storage device 11 is suspended, the open-circuit voltage estimated value Vocell based on the cell..voltage-average value Vcell of the total - voltage Vail detected zy ~zz.z. -rclnara detector 18, the previous value of the cell volrage average value Vcell, and the gain K3 that changes with the passage of time after the charge or discharge is suspended, and outputs the open-circuit voltage estimated value Vocell as the second estimated value SOC20; and the sudden-change prevention unit 25 that limits the change amount in the second estimated value S0C2D in the computation cycle t3 of the second SOC computation "unit 24 to be equal to or smaller than the predetermined value (1%) based on the previous value SOCz and the second estimated value SOC20. Therefore, it is possible to improve the accuracy of estimating the SOC during the charge cr discharge and during the suspension of the charge and discharge. [0074] The second 5CC computation unit 24 according to the present embodiment can compute the open-circuit voltage estimated value Vocell only based on the total voltage Vail. Therefore, it is possible to estimate the SOC from the open-circuit voltage estimated value Vocell without using any equivalent circuit model as in the conventional techniques. [0075] The second SOC computation unit 24 according to the present embodiment computes the open-circuit voltage estimated value Vocell bv -the use of the gain unit 52. Therefore, it is possible to suppress the change in the open-circuit voltage estimated value Vocell after the end of the charge or discharge, and to obtain the stable SOC estimated value (the second estimated value SOC20). [007 6] The first SOC computation unit 20 according to the present embodiment controls the first estimated value SOC1 to be constant when FLG1=0. Therefore, it is possible to estimate the SOC even during the period (t2) during "Which the constant-srol-tage .charge or discharge continues and the open-circuit voltage estimated Tilie Tocell not be estimated. [0077] In the above descriptions, the cell batteries included in the power storage device 11 are arranged in n series-connected cell batteries by m parallel-connected cell batteries; however, needless to mention, the values of n and m can be also each set to an arbitrary value equal to or larger than 1. Furthermore, values of the computation cycles T1 to t3 described above are given only as examples and the computation cycles T1 to t3 are net limited to these values. Further, as one example, the value of the gain K3 shown in FIG. 9 is set to be constant during the period t2 (0 to 20 seconds) and then gradually decrease. Alternatively, similar effects can be attained even when the state-of-charge estimation apparatus 13 is configured to, for example, gradually change the value of the gain K3 from the time (0 second) when the charge or discharge ends. [0078] Second embodiment. FIG. 13 is a diagram showing a configuration of the state-of-charge estimation apparatus 10 according to a second embodiment of the present invention. Elements identical to those in the first embodiment are denoted by the same reference signs and description thereof is omitted, only parts different from the first embodiment are described here. The state-of-charge estimation apparatus 10 according to the second embodiment difrers from that shown in FIG. 3 in that a table 9 is provided. The secondary batteries applied to the power storage device 11 degrade with the passage of service time and decrease in capacity. It is significant to accurately ascertain the SOC of a secondary battery. For example, when the secondary batteries are overcharged because of inability to accurately asce