Simulation device, operation management device, simulation method, and operation management method

Abstract

Provided are a method for simulating a storage battery operation plan and a device for achieving the same with which higher speed can be achieved than before while ensuring accuracy. The device performs simulation of an operation based on an operation plan for a storage battery which includes a plurality of module batteries each composed of a plurality of unit cells, the device comprising: a learning processing unit for learning a relationship between a feature amount, which characterizes an operation state of the storage battery, and an index value change amount, which is a change amount of at least one state index value of the storage battery for each predetermined time interval, in a predetermined learning target period determined for a past operation period of the storage battery; and a simulation execution unit for simulating at least one state index value in an execution period of the operation plan on the basis of a result of the learning in the learning processing unit.

Description

Simulation device, operation management device, simulation method, and operation management method

[0001] The present invention relates to the simulation of a storage battery including a plurality of module batteries. [[IF=5]]

[0002] There is already a known device (see, for example, Patent Document 1) that acquires an operation plan (charge / discharge plan) of a sodium-sulfur battery (NaS battery), which is a type of storage battery, and determines whether the state of the NaS battery is within an allowable range when the operation plan is executed, based on a simulation of battery temperature and remaining capacity, and outputs guidance regarding the operation of the NaS battery. The operation plan is created, for example, by a user of a power storage system including the NaS battery and its associated devices (control devices and auxiliary machines such as heaters and fans). There is also already a known device that can calculate the charge / discharge conditions that can be adopted in a high-temperature operating secondary battery with a small amount of calculation (see, for example, Patent Document 2).

[0003] As disclosed in Patent Documents 1 and 2, a NaS battery is usually configured by connecting a plurality of module batteries in series and in parallel. Each module battery is configured by incorporating a plurality of single cells into one housing container.

[0004] In a NaS battery, when over-discharge progresses, sodium polysulfide is generated on the anode side and sodium in the cathode is depleted, making subsequent charge / discharge impossible. Also, when over-charging progresses, the solid electrolyte is damaged, making subsequent charge / discharge impossible. Furthermore, when the temperature of a single cell becomes too high, phenomena such as the internal pressure of each single cell becoming abnormally high, or the negative electrode active material directly contacting the positive electrode active material due to damage of the solid electrolyte and the temperature of each single cell rising abnormally may occur, and there is a possibility that the housing container will be damaged.

[0005] In order to be able to avoid such over-discharge, over-charging, and other problems, as described above, the guidance device disclosed in Patent Document 1 determines the feasibility of an operation plan created by a user of the power storage system prior to the execution of the operation plan.

[0006] There is a need to frequently and quickly change operational plans in order to use energy storage systems for bidding on the electricity market. On the other hand, obtaining the optimal operational plan that maximizes profits requires multiple simulations.

[0007] To achieve both of these goals, it is crucial to be able to perform battery simulations quickly and make prompt decisions on their feasibility based on the results.

[0008] In Patent Document 1, the internal resistance of the battery, the amount of heat dissipated by the module battery, and the thermal capacity of the module battery are used as parameters when simulating battery temperature and remaining capacity. However, when applying the method disclosed in Patent Document 1 to a storage battery, it is necessary to perform charging and discharging with an operation plan suitable for setting the parameters to acquire data, set simulation parameters for each module battery from the acquired data, and then perform the simulation using the set simulation parameters.

[0009] Furthermore, an example of an operational plan suitable for calculating simulation parameters is one consisting of a simple operational pattern in which discharge and charge occur only once within the planned period. However, such an operational plan does not necessarily coincide with the optimal operational plan that maximizes profits.

[0010] Therefore, the method described in Patent Document 1 has the problem that it requires a lot of time for preparatory processing before the simulation is performed, due to reasons such as the enormous amount of computation required to set the simulation parameters and the need for operation based on an operation plan suitable for calculating the simulation parameters prior to operation for profitability.

[0011] On the other hand, the method disclosed in Patent Document 2 is intended to reduce the amount of computation, but it does not take into account the influence of each module battery when the storage battery is composed of multiple module batteries. When adopting the method disclosed in Patent Document 2, it is necessary to perform charging and discharging with an operation plan suitable for creating a lookup table to acquire data, create a lookup table from the acquired data, and then perform a simulation using the created lookup table. An operation plan suitable for creating a lookup table is the same as an operation plan suitable for calculating simulation parameters, but it is desirable to operate with varying SOC at the start of discharge and acquire data. Therefore, even when adopting the method in Patent Document 2, there is a problem that preparation processing such as operating with an operation plan suitable for creating a lookup table and setting up the lookup table takes time prior to performing the simulation.

[0012] Japanese Patent Publication No. 2008-210586, International Publication No. 2016 / 136445

[0013] The present invention has been made in view of the above problems, and aims to provide a method for simulating battery operation plans that does not require time-consuming preparation processes like those in the prior art, and that is faster than conventional methods while ensuring accuracy, as well as an apparatus for realizing this method.

[0014] To solve the above problems, a first aspect of the present invention is a device for performing an operation simulation based on an operation plan for a storage battery comprising a plurality of module batteries, each composed of a plurality of single cells, comprising: a learning processing unit that learns the relationship between a feature quantity that characterizes the operating state of the storage battery during a predetermined learning target period set up to cover the past operating period of the storage battery, and an index value change amount which is the amount of change of at least one state index value of the storage battery at predetermined time intervals; and a simulation execution unit that performs a simulation of the at least one state index value during the execution period of the operation plan based on the learning results of the learning processing unit.

[0015] A second aspect of the present invention is a simulation apparatus according to the first aspect, characterized in that the learning processing unit acquires the actual temperature, actual voltage, and actual current of each of the plurality of module batteries for at least the predetermined learning target period, calculates the feature quantities based on the actual temperature, actual voltage, and actual current at predetermined time intervals, creates a decision tree for determining the assumed values ​​of the change in the index value during the execution period of the operation plan according to the conditions of the feature quantities based on the learning results that associate the change in the index value with the feature quantities at predetermined time intervals, and performs a simulation of the at least one state index value during the execution period of the operation plan by sequentially identifying the assumed values ​​of the change in the index value during the execution period of the operation plan at predetermined time intervals based on the decision tree, and sequentially adding the identified assumed values.

[0016] A third aspect of the present invention is a simulation apparatus according to the first or second aspect, characterized in that the at least one state index value includes the temperature of the storage battery.

[0017] A fourth aspect of the present invention is a device for performing an operation simulation based on an operation plan for a storage battery comprising a plurality of module batteries, each composed of a plurality of single cells, comprising: a learning processing unit that learns the relationship between feature quantities that characterize the operating state of the storage battery and at least one state index value of the storage battery during a predetermined learning target period set up to cover the past operating period of the storage battery; and a simulation execution unit that performs a simulation of the at least one state index value during the execution period of the operation plan based on the learning results of the learning processing unit.

[0018] A fifth aspect of the present invention is a simulation apparatus according to the fourth aspect, characterized in that the learning processing unit acquires the actual temperature, actual voltage, and actual current of each of the plurality of module batteries for at least the predetermined learning target period, calculates the feature quantities at predetermined time intervals based on the actual temperature, actual voltage, and actual current, creates a decision tree for determining the assumed value of the at least one indicator value during the execution period of the operation plan according to the conditions of the feature quantities based on the learning results that associate the at least one indicator value with the feature quantities at predetermined time intervals, and the simulation execution unit performs a simulation of the at least one state indicator value during the execution period of the operation plan by sequentially identifying the assumed value of the at least one state indicator value during the execution period of the operation plan at predetermined time intervals based on the decision tree.

[0019] A sixth aspect of the present invention is a simulation apparatus according to the fourth or fifth aspect, characterized in that the at least one state index value includes the internal resistance of the storage battery.

[0020] A seventh aspect of the present invention is a simulation apparatus according to the fourth or fifth aspect, characterized in that the at least one state index value includes at least one of the discharge depth at the end of discharge, which is the discharge depth when the storage battery reaches the discharge stop voltage, and the charge depth at the end of discharge, which is the discharge depth when the storage battery reaches the charge stop voltage.

[0021] An eighth aspect of the present invention is a device for managing the operation of a storage battery comprising a plurality of module batteries, each composed of a plurality of single cells, comprising: an operation feasibility determination unit that determines whether or not to operate the storage battery according to the operation plan based on the results of a simulation in a simulation device according to the first or second aspect, and a charge / discharge management unit that sends the operation plan, which has been determined to be operational by the operation feasibility determination unit, to a control device of the storage battery, and causes the control device to operate the storage battery according to the operation plan.

[0022] A ninth aspect of the present invention is a method for simulating the operation of a storage battery, which comprises a plurality of module batteries, each composed of a plurality of single cells, based on an operation plan, comprising: a learning process step of learning the relationship between a feature quantity that characterizes the operating state of the storage battery during a predetermined learning target period set up to cover the past operating period of the storage battery, and an index value change amount which is the amount of change of at least one state index value of the storage battery at predetermined time intervals; and a simulation step of simulating the at least one state index value during the execution period of the operation plan based on the learning results in the learning process step.

[0023] A tenth aspect of the present invention is a simulation method according to the ninth aspect, characterized in that, in the learning process, the actual temperature, actual voltage, and actual current of each of the plurality of module batteries are acquired for at least the predetermined learning target period, the feature quantities are calculated based on the actual temperature, the actual voltage, and the actual current at predetermined time intervals, a decision tree is created for determining the assumed values ​​of the change in the indicator value during the execution period of the operation plan according to the conditions of the feature quantities based on the learning results that associate the change in the indicator value with the feature quantities at predetermined time intervals, and in the simulation process, the assumed values ​​of the change in the indicator value during the execution period of the operation plan are sequentially identified at predetermined time intervals based on the decision tree, and the identified assumed values ​​are sequentially added together to simulate the at least one state indicator value during the execution period of the operation plan.

[0024] An eleventh aspect of the present invention is a simulation method according to the ninth or tenth aspect, characterized in that the at least one state index value includes the temperature of the storage battery.

[0025] A twelfth aspect of the present invention is a method for simulating the operation of a storage battery, which comprises a plurality of module batteries, each composed of a plurality of single cells, based on an operation plan, comprising: a learning process step of learning the relationship between feature quantities that characterize the operating state of the storage battery and at least one state index value of the storage battery during a predetermined learning target period set up to cover the past operating period of the storage battery; and a simulation step of simulating the at least one state index value during the execution period of the operation plan based on the learning results in the learning process step.

[0026] A thirteenth aspect of the present invention is a simulation method according to the twelfth aspect, characterized in that, in the learning process, the actual temperature, actual voltage, and actual current of each of the plurality of module batteries are acquired for at least the predetermined learning target period, the feature quantities are calculated based on the actual temperature, the actual voltage, and the actual current at predetermined time intervals, a decision tree is created for determining the assumed value of the at least one indicator value during the execution period of the operation plan according to the conditions of the feature quantities based on the learning results that associate the at least one indicator value with the feature quantities at predetermined time intervals, and in the simulation process, the assumed value of the at least one state indicator value during the execution period of the operation plan is sequentially identified at predetermined time intervals based on the decision tree, thereby simulating the at least one state indicator value during the execution period of the operation plan.

[0027] A fourteenth aspect of the present invention is a simulation method according to the twelfth or thirteenth aspect, characterized in that the at least one state index value includes the internal resistance of the storage battery.

[0028] A fifteenth aspect of the present invention is a simulation method according to the twelfth or thirteenth aspect, characterized in that the at least one state index value includes at least one of the discharge depth at the end of discharge, which is the discharge depth when the storage battery reaches the discharge stop voltage, and the charge depth at the end of discharge, which is the discharge depth when the storage battery reaches the charge stop voltage.

[0029] A sixteenth aspect of the present invention is a method for managing the operation of a battery storage system comprising a plurality of module batteries, each composed of a plurality of single cells, comprising: an operation feasibility determination step of determining whether or not to operate the battery storage system according to the operation plan based on the results of a simulation method according to the ninth or tenth aspect; and a charge / discharge management step of sending the operation plan, which was determined to be operational in the operation feasibility determination step, to a control device of the battery storage system and causing the control device to operate the battery storage system according to the operation plan.

[0030] According to the first to sixteenth aspects of the present invention, simulation results for typical temperature changes in a storage battery can be obtained more quickly and with reasonable accuracy than when using conventional methods, without having to perform preparatory processes such as calculating simulation parameters, creating lookup tables, and setting simulation parameters and lookup tables, which were necessary with conventional methods.

[0031] This is a block diagram showing the relationships between various devices related to the simulation of charge and discharge operations. This is a schematic diagram showing the electrical connection configuration of multiple single cells 13c within the containment container 11a. This is a diagram showing the procedure for the learning process. This is a diagram illustrating an example of a decision tree obtained through the learning process. This is a diagram that schematically shows the flow of the simulation process, along with the preceding and succeeding processes. This is a diagram showing the procedure for identifying the value of the assumed temperature change ΔT(i) at the simulation target time i. This is a diagram illustrating an example of a decision tree consisting of multiple layers created using SOC as a feature. This is a diagram illustrating a decision tree in which layers based on non-continuous features are interposed between layers based on continuous features. This is a diagram showing the procedure for the learning process when internal resistance is the target of the simulation. This is a schematic diagram showing the flow of the simulation process targeting internal resistance. This is a graph schematically showing the relationship between discharge depth and voltage in a NaS battery.

[0032] <First Embodiment> <Overview> Figure 1 is a block diagram showing the relationship between an energy storage system 1 that performs charging and discharging operations according to a pre-set operating pattern (also referred to as an operating plan or charge / discharge plan) and various devices related to the simulation of charging and discharging operations based on the created operating pattern.

[0033] The energy storage system 1 comprises a storage battery 11 and a control device 12 that controls the charging and discharging operations of the storage battery 11.

[0034] The storage battery 11 comprises a plurality of module batteries 11(1) to 11(n) (where n is a natural number). Each module battery 11(1) to 11(n) is generally composed of a battery assembly 13 formed by connecting a plurality of single cells 13c in series and parallel, a heater 14 for heating and maintaining the temperature of the battery assembly 13, a fan 15 for dissipating heat from the battery assembly 13, and a temperature sensor 16 for measuring the temperature of the module batteries 11(1) to 11(n), all housed in a single container 11a. The storage battery 11 is then constructed by connecting the battery assemblies 13 of each module battery 11(1) to 11(n) in series.

[0035] The single cell 13c is, for example, a sodium-sulfur cell (NaS cell) with sulfur as the positive electrode active material and metallic sodium as the negative electrode active material. Alternatively, the single cell 13c may be a lithium-ion cell. In the following explanation, we will mainly use the case where the single cell 13c is an NaS cell as an example.

[0036] Figure 2 schematically shows the electrical connection configuration of multiple single cells 13c within the housing container 11a. As described above, in each of the module batteries 11(1) to 11(n), multiple single cells 13c are housed in a single housing container 11a to constitute a battery assembly 13. Note that the heater 14, fan 15, and temperature sensor 16 are not shown in Figure 2.

[0037] The battery assembly 13 is composed of one or more blocks 13b, each consisting of several individual cells 13c. The housing 11a is also provided with a positive external terminal 60 and a negative external terminal 62 on its exterior, and one or more blocks 13b are connected in series from the positive external terminal 60 to the negative external terminal 62. Each block 13b is formed by connecting two or more strings 13s, each consisting of two or more individual cells 13c connected in series, in parallel.

[0038] The battery 11 is used by the owner of the energy storage system 1 (energy storage business operator), or by power generators, transmission and distribution companies, retail electricity companies, etc., for adjusting the supply and demand of electricity with the external commercial power grid, or further for generating revenue through the buying and selling of electricity in the electricity market (e.g., wholesale electricity market, capacity market, supply and demand adjustment market).

[0039] The operation pattern creation device 2 is a device responsible for creating operation patterns for performing charge and discharge operations in the battery 11. The operation pattern creation device 2 generates operation pattern data that describes the contents of the created operation pattern and sends this to the battery simulation device 3. The operation pattern data describes the target operation period, the charge and discharge periods and output during the operation period, etc. Furthermore, the timing of the ON / OFF operation of the heater 14 and fan 15 may also be described.

[0040] The battery simulation device 3 is a device that simulates the charging and discharging operation when the operating pattern created in the operating pattern creation device 2 is applied to the battery 11. More specifically, the battery simulation device 3 simulates at least one state index value (more precisely, its change over time) that indicates the state of the battery 11 when the charging and discharging operation is performed, over the target time of the operating pattern (also referred to as the operating output time), and outputs the time-series data.

[0041] The state indicators included in the simulation will include, at a minimum, the temperature of the storage battery (battery temperature).

[0042] The data output by the battery simulation device 3 (simulation result data) is used to pre-determine, prior to the execution of the operation pattern, whether it is possible to actually execute the charge and discharge operation based on the created operation pattern in the battery 11.

[0043] The battery simulation device 3 includes a simulation execution unit 31 and a learning processing unit 32.

[0044] The simulation execution unit 31 performs a simulation on the time change of at least one of the above-described state index values when the battery 11 is assumed to be operated based on a certain operation pattern. The simulation execution unit 31 uses the result of the learning process in the learning processing unit 32 during such a simulation.

[0045] Prior to the execution of the simulation in the simulation execution unit 31, the learning processing unit 32 learns the relationship between various feature amounts characterizing the operation state of the battery 11 (for each of the module batteries 11(1) to 11(n)) in a predetermined learning target period determined for the past operation period of the battery 11, and the time change of the measured values of the state index values to be simulated in each of the module batteries 11(1) to 11(n).

[0046] Note that the feature amount does not necessarily have to be a quantitative value indicated by continuous values, and may be determined alternatively or qualitatively. Examples of such feature amounts include the operation state (discharge, charge, stop) of the battery 11, charge and discharge power, remaining capacity (SOC) (or depth of discharge (DOD)), actual temperature, operation states (on / off) of the heater 14 and the fan 15, charge and discharge cycle numbers, number of cells with lost function, and charge and discharge history index values. By representing the charge and discharge power as a positive value during discharge, a negative value during charge, and 0 during charge and discharge stop, the charge and discharge power reflecting the operation state may be used as a feature amount. Alternatively, decision trees may be created individually for each of the discharge, charge, and charge and discharge stop cases.

[0047] Including the operation state and the charge and discharge power as feature amounts takes into account that the heat generation amounts in the module batteries 11(1) to 11(n) differ depending on these.

[0048] The reason for including the SOC (or DOD) as a characteristic quantity is that the internal resistance of the module batteries 11(1) to 11(n) and the chemical reactions occurring in the single cell 13c differ depending on the value of the SOC (or DOD), and thus the amount of heat generated is different. In the range of DOD where the amount of heat generated increases, the temperature rise amplitude increases, and the difference between the simulation value and the measured value also tends to increase.

[0049] The reason for including the operating states of the heater 14 and the fan 15 as characteristic quantities is that the way the temperature of the module batteries 11(1) to 11(n) rises is different when they are operating and when they are not operating.

[0050] Also, the number of cells with loss of function refers to the number of single cells 13c that have stopped operating due to failures or abnormalities after the start of use of the storage battery 11.

[0051] Examples of the charge / discharge history index value include the integrated value of the charge / discharge power (≈ SOC change amount) over a past predetermined time when the charge power is made negative and the discharge power is made positive, and the elapsed time after the start of discharge (or charge).

[0052] Details of the simulation in the storage battery simulation device 3 will be described later.

[0053] The storage battery operation management device 4 is a device for managing the operation of the storage battery 11. The storage battery operation management device 4 includes an operation permission determination unit 41 and a charge / discharge management unit 42.

[0054] The operation permission determination unit 41 acquires, from the storage battery simulation device 3, simulation result data that describes the result of the simulation in the storage battery simulation device 3 and the operation pattern targeted by such simulation. Then, based on such simulation result data, it determines whether it is possible to execute the charge / discharge operation based on the operation pattern targeted by the simulation in the storage battery 11.

[0055] The results of the determination are notified to the operation pattern creation device 2 as appropriate. If it is determined that execution of a certain operation pattern is "not" and this result is notified to the operation pattern creation device 2, the operation pattern creation device 2 will appropriately modify the operation pattern in question or recreate a new operation pattern.

[0056] The charge / discharge management unit 42 sends data of the operation pattern that the operation feasibility determination unit 41 has determined to be "permissible" (operation feasible pattern data) to the control device 12 of the energy storage system 1.

[0057] The operation pattern creation device 2, the battery simulation device 3, and the battery operation management device 4 are all realized by executing a predetermined program on a general-purpose or dedicated computer that has a control unit equipped with a CPU, ROM, RAM, etc., an input operation unit such as a mouse, keyboard, or touch panel, a display or other display unit, a storage unit such as a hard disk or SSD, etc. Therefore, the components of each device, such as the simulation execution unit 31 and the learning processing unit 32 of the battery simulation device 3, and the operation feasibility determination unit 41 and the charge / discharge management unit 42 of the battery operation management device 4, are functional components that are virtually realized by the computer.

[0058] Furthermore, the battery simulation device 3 and the battery operation management device 4 may be implemented by a single computer, or the operation pattern creation device 2 and the battery simulation device 3 may be implemented by a single computer. Alternatively, the operation pattern creation device 2, the battery simulation device 3, and the battery operation management device 4 may all be implemented by a single computer. In addition, the operation feasibility determination unit 41 of the battery operation management device 4 may be provided in the battery simulation device 3 instead of the battery operation management device 4.

[0059] When the control device 12 of the energy storage system 1 obtains operational pattern data from the charge / discharge management unit 42 of the battery operation management device 4, it controls the charging and discharging of the battery 11 based on the contents of that data. At the same time, the control device 12 also controls the ON / OFF status of the heaters 14 and fans 15 of each of the module batteries 11(1) to 11(n).

[0060] Furthermore, in the battery 11, the temperature sensor 16 measures the temperature (actual temperature) of each module battery 11(1) to 11(n) regardless of whether it is operating (charging / discharging) or in standby mode. The measurement results from the temperature sensor 16 are sequentially sent to the learning processing unit 32 of the battery simulation device 3. In addition, the real-time voltage and current values ​​(actual voltage and actual current) of each module battery 11(1) to 11(n) are also sequentially sent to the learning processing unit 32 of the battery simulation device 3. These actual temperature, actual voltage, and actual current values ​​are used in the learning process in the learning processing unit 32.

[0061] <Simulation Details> Next, the content of the simulation performed by the battery simulation device 3 will be explained. In this embodiment, the battery temperature will be assumed to be the state index value targeted for simulation. In this case, the battery simulation device 3 generally has a learning processing unit 32 that learns in advance the relationship between the temperature change and feature quantities in each of the module batteries 11(1) to 11(n) at predetermined time intervals Δt, and the simulation execution unit 31 then performs a simulation of the temperature change of the battery 11 at time intervals Δt, assuming that the operation pattern targeted for simulation is executed, based on the learning results.

[0062] (Learning Process) Figure 3 shows the procedure for the learning process in the learning processing unit 32. Figure 4 is an example of a decision tree obtained through the learning process.

[0063] In general terms, the learning process involves learning the combination of the temperature change ΔTm(j) (where m is a natural number from 1 to n) from the previous time j-1 for each module battery 11(1) to 11(n) at every time j at each time interval Δt of a predetermined learning period, and the feature quantities at that time j. Based on the learned content, a decision tree consisting of multiple layers with progressively set sorting conditions is created. The learned content is stored in a memory unit (not shown) of the battery simulation device 3. The learning result data, which describes the contents of the decision tree, is then passed to the simulation execution unit 31 and used in the simulation in the simulation execution unit 31.

[0064] Examples of learning periods include one day, ten days, one month, and one year. Alternatively, the learning period may be set to the time it takes for the battery to reach 1000 charge-discharge cycles or the time it takes for the discharged energy to reach 100 MWh. The learning period may also be extended according to the elapsed time since the initial operation of the battery 11.

[0065] To enable such learning processing, the learning processing unit 32 acquires actual temperature, actual voltage, and actual current from each of the module batteries 11(1) to 11(n) when the storage battery 11 is in operation. Then, for each time interval Δt, these measured values ​​and the values ​​of feature quantities calculated based on these measured values ​​are stored in a storage unit (not shown) of the storage battery simulation device 3. The data (learning data) stored in this manner will be used for simulation in the future. The techniques disclosed in Patent Documents 1 and 2 may be used as appropriate for calculating the values ​​of the feature quantities. In addition, some feature quantities may be identified based on the description content of the operational pattern data, or acquired from the control device 12 of the energy storage system 1.

[0066] The learning process can be started at any appropriate time after the battery 11 first starts operating, and the simulation can be run as long as the learning process is performed. However, immediately after the battery 11 starts operating, there is not enough time for the learning target period, and therefore the amount of learning target data accumulated is small. The longer the total operating time of the battery 11, and the more time is available for the learning target period, the better the accuracy of the simulation will be.

[0067] As shown in Figure 3, in the learning process using a decision tree, the learning processing unit 32 first reads the value of a feature at a past time j included in the learning period as the j-th data to be learned from the storage unit (not shown) of the battery simulation device 3 (step S1).

[0068] Then, the learning processing unit 32 uses the value described in the j-th data to be learned that has been read out to calculate the temperature change amount ΔTm(j) based on equation (2) (step S2).

[0069] The learning processing unit 32 also identifies the decision tree node corresponding to the feature value of the j-th data to be learned, and records the value of the temperature change amount ΔTm(j) calculated based on equation (2) in the storage unit, associating it with each node (step S3). As a result, the value of the temperature change amount ΔTm(j) corresponding to the j-th data to be learned is distributed to one of the termination nodes.

[0070] In Figure 4, for simplicity, an example is shown where a decision tree is defined using the operating state, SOC, and fan 15 operating state from the features described above. In the decision tree shown in Figure 4, the temperature change ΔTm(j) is distributed based on whether the operating state is discharge or not at the top level (root node), whether the SOC is less than 50% or not at the next level, and whether the fan 15 is on or not at the final level. As a result, eight termination nodes are set.

[0071] For example, if the j-th data point to be learned is from a time when the battery 11 was discharged, the SOC was 50% or more, and the fan 15 was not operating, then the temperature change ΔTm(j) will be allocated to the fourth and final node from the left.

[0072] If there is further training data that has not been sorted (recorded) based on this decision tree (NO in step S4), steps S1 to S3 are repeated.

[0073] Once all the training data has been sorted (recorded) (YES in step S4), the average of the associated (sorted) temperature change amounts ΔTm(j) is calculated as the average temperature change amount ΔTd(Ne) for each termination node Ne, and this is recorded in association with each termination node Ne (step S5). Figure 4 also shows the calculation results of the average temperature change amount ΔTd(Ne) for each termination node Ne.

[0074] (Simulation Process) Figure 5 is a schematic diagram showing the flow of the simulation process in this embodiment, along with the preceding and succeeding processes.

[0075] First, an operation pattern for the battery 11 is created in advance in the operation pattern creation device 2, and the simulation execution unit 31 of the battery simulation device 3 acquires operation pattern data from the operation pattern creation device 2, which describes the contents of the operation pattern (step S11). To clarify, the operation pattern to be simulated is assumed to be adopted in the actual operation of the battery 11, and can be an optimal operation plan that maximizes profits when bidding in the electricity market, for example. It is not necessary to use an operation plan consisting of a simple operation pattern, as was used in conventional simulation methods to calculate simulation parameters.

[0076] An operational pattern is created for a specific operational output time. Typically, this operational output time is divided into multiple "frames," and the operational pattern is set for each frame. One frame is, for example, 30 minutes long.

[0077] The simulation execution unit 31, having acquired the operational pattern data, sequentially performs simulations of the battery temperature for all simulation target time points i at each time interval Δt in the operational output time, which is the target of the operational pattern setting, based on the contents of the description. If the operational output time is divided into multiple frames as described above, the time interval Δt may be set to be the same as the time of one frame (for example). In addition, the end time of each frame may be set as the simulation target time point i.

[0078] Specifically, first, for a given simulation target time i, the value of the assumed temperature change amount ΔT(i) from the previous simulation target time i-1 is identified (step S12).

[0079] Figure 6 shows the procedure for determining the value of the assumed temperature change ΔT(i) at the simulation target time i.

[0080] The simulation execution unit 31 first reads out the feature values ​​(expected values) used for distribution based on the decision tree at the simulation target time i as feature values ​​at the simulation target time i (step S21). These expected values ​​are set based on the description of the operation pattern data. When using the decision tree shown in Figure 4, the operating state of the battery 11, the SOC value, and the operating state of the fan 15 that are expected at the simulation target time i are read out as feature values ​​at the simulation target time i.

[0081] Then, the simulation execution unit 31 identifies the node corresponding to the i-th feature that was read out, according to the decision tree created by the learning processing unit 32, reads out the value of the average temperature change amount ΔTd(Ne) associated with the corresponding termination node Ne (step S22), and identifies (outputs) that value as the assumed temperature change amount ΔT(i) from time i-1 (step S23).

[0082] For example, if, at a certain simulation target time i, the battery is in a charged state, the SOC value is 45%, and the fan 15 is operating, then these feature quantities will be associated with the fifth termination node Ne from the left in Figure 4. Therefore, the simulation execution unit 31 sets the average temperature change amount ΔTd(Ne) of 0.4, which is associated with that termination node, as the value of the assumed temperature change amount ΔT(i) from time i-1 at the simulation target time i.

[0083] The simulation execution unit 31, having identified the assumed temperature change ΔT(i), calculates the battery temperature T(i) (more specifically, its assumed target value) of the storage battery 11 (each of the module batteries 11(1) to 11(n)) at the simulation target time i based on the following equation (1) (step S13).

[0084] T(i) = T(i-1) + ΔT(i) ... (1) In equation (1), the initial value T(0) of the battery temperature T(i) is the temperature of the battery 11 at the start of the operation pattern to be simulated. This value may be set based on the characteristic quantities of the battery 11 at the start, or, if the start of the operation pattern is the same as the end of another operation pattern that has been simulated in advance, the value of the battery temperature T(i) at the end of the simulation of the other operation pattern may be adopted.

[0085] (1) If the simulation target time i that is the subject of the calculation based on equation is not the last time of the operating period (NO in step S14), the simulation execution unit 31 repeats the processes of steps S12 and S13 for a new simulation target time i that is advanced by a time interval Δt from the simulation target time i.

[0086] (1) If the simulation target time i that was the subject of the calculation based on equation is the last time of the operating period (YES in step S14), then the battery temperature T(i) has been determined for all simulation target time i included in the operating period. In other words, the simulation of the battery temperature for the operating period assumed in the operating pattern has been completed.

[0087] It should be noted that the simulation result for battery temperature obtained in this embodiment is unique, and individual simulation results are not obtained for each of the module batteries 11(1) to 11(n). This is because it is assumed that the temperature changes of each module battery 11(1) to 11(n) are the same, but the obtained simulation result is based on the results of a learning process targeting the temperature changes in each of the multiple module batteries. Therefore, the simulation result has a reasonable degree of accuracy that approximates the expected temperature changes for each of the multiple module batteries. In other words, the simulation result for battery temperature obtained in this embodiment can be said to represent the typical temperature changes in the module batteries 11(1) to 11(n) provided in the storage battery 11.

[0088] The simulation execution unit 31 generates simulation result data that describes the simulation results along with the operation pattern that was the subject of the simulation, and passes this data to the battery operation management device 4.

[0089] In the battery operation management device 4, which has acquired the simulation result data, the operation feasibility determination unit 41 makes a determination as to whether the battery 11 is operational (step S15).

[0090] The operation feasibility determination unit 41 refers to the description of the simulation result data and determines whether or not there are modules that cannot be operated, thereby determining whether or not the battery 11 can be operated (charged / discharged) based on the operation pattern that was the subject of the simulation.

[0091] Specifically, the system determines, based on the description in the simulation result data, whether the battery temperature deviates from the predetermined operating condition range (tolerance range) at any point during the operation output time (at any of the simulation target time points i).

[0092] If there is a simulation target time i that deviates from the operating condition range (NG in step S15), the operation feasibility determination unit 41 determines that operation of the storage battery 11 based on the operation pattern data targeted for simulation is "not possible".

[0093] If there is no simulation target time i that deviates from the operating condition range (OK in step S15), the operation feasibility determination unit 41 determines that operation of the storage battery 11 based on the operation pattern data targeted for simulation is "possible".

[0094] The judgment result (operation feasibility judgment result) is notified to the operation pattern creation device 2. If there are modules that cannot be operated and operation is determined to be "not" possible, the operation pattern creation device 2 will modify or recreate the operation pattern data.

[0095] On the other hand, if operation is determined to be "possible," the operation pattern data (operation-possible pattern data) is passed from the charge / discharge management unit 42 to the control device 12 of the energy storage system 1. The control device 12, having acquired the operation-possible pattern data, controls the battery 11 according to its description, thereby putting the battery 11 into operation (actual operation) (step S16).

[0096] Furthermore, during actual operation, the learning processing unit 32 acquires actual temperature, actual current, and actual voltage data (actual data) for each of the module batteries 11(1) to 11(n) and records them in a storage unit (not shown) of the battery simulation device 3 (step S17). This actual data will be used for new learning processing in the learning processing unit 32.

[0097] As described above, in the simulation of operating patterns targeting the battery temperature of the storage battery performed in this embodiment, simulation results for the temperature change of the storage battery are obtained based on the results of the learning process. In such a simulation, it is not necessary to set the simulation parameters that were previously required, and it is not necessary to prepare an operating plan for such settings and to operate based on it. In other words, since it is not necessary to operate the storage battery in order to set the simulation parameters, the operational efficiency of the storage battery is improved. Moreover, the simulation results have a reasonable degree of accuracy that approximates the temperature change expected for each of the multiple battery modules.

[0098] Furthermore, since the determination of the temperature change is performed based on a pre-created decision tree, and the calculation of the battery temperature based on equation (1) is a fixed-value operation targeting the determined temperature change, simulation results can be obtained more quickly compared to conventional methods.

[0099] In other words, according to this embodiment, simulation results for typical temperature changes in a storage battery can be obtained more quickly and with reasonable accuracy than when using conventional methods. As a result, the determination of whether or not operation is possible based on these simulation results can also be made more quickly than with conventional methods.

[0100] Therefore, the simulation method according to this embodiment is suitable for energy storage systems used in bidding on the electricity market, where rapid and frequent simulations are required.

[0101] <Creating a Decision Tree with Continuous Values ​​as Thresholds> When creating a decision tree to determine the value of the assumed temperature change ΔT(i), features that change continuously (take continuous values) are sometimes used. Figure 7 is an example of a multi-layered decision tree created using SOC, one such feature.

[0102] In the decision tree shown in Figure 7, the temperature change ΔTm(j) is allocated at the top level (root node) based on whether the State of Temperature (SOC) is less than 62% or not. If the SOC is less than 62%, the temperature change ΔTm(j) is allocated at the next level based on whether the SOC is less than 37% or not. On the other hand, if the SOC is not less than 62% (i.e., 62% or more), the temperature change ΔTm(j) is allocated at the next level based on whether the SOC is (62% or more) less than 92% or not. As a result, four termination nodes Ne are set.

[0103] When creating a decision tree using features that take continuous values ​​in this way, it is preferable to set the threshold so that the impurity (variability) of the temperature change ΔTm(j) values ​​applicable to each node is small. This impurity can be expressed, for example, by the variance (or standard deviation) of the corresponding temperature change ΔTm(j) values. In this case, the threshold is set using all the temperature change ΔTm(j) values ​​accumulated during the training period, so along with setting the threshold, all the temperature change ΔTm(j) values ​​that were targeted for training will be assigned to one of the nodes.

[0104] Even when the features take continuous values, it is possible to set a threshold in advance and then sequentially perform learning processing on the individual temperature change values ​​ΔTm(j) in the same manner as shown in Figure 3. Depending on the situation, either method may be adopted as appropriate, or both may be used in combination.

[0105] Figure 8 illustrates a decision tree in which layers based on non-continuous features are interposed between layers based on continuous features. In the decision tree shown in Figure 8, a layer based on the operating state of fan 15 is interposed after the decision tree shown in Figure 7 when the SOC is less than 62%. The layers after this layer are the same as those in the decision tree shown in Figure 7. As a result, six termination nodes Ne are set.

[0106] When setting a threshold for a hierarchy based on features that take continuous values ​​to minimize impurity, depending on the size of the set threshold, it may be possible to perform more appropriate sorting by interposing a hierarchy based on features that do not take continuous values, as shown here.

[0107] Figure 8 illustrates a case where the same feature (specifically, SOC) is used as a branching condition at different levels, and the same threshold is used at lower levels regardless of the branching result at higher levels. However, when constructing a regression tree to minimize the aforementioned impurity, the branching conditions at each level usually differ from node to node.

[0108] <Addressing Degradation Over Time> When the battery 11 is operated continuously, the characteristics of individual module batteries 11(1) to 11(n) change due to degradation over time. Therefore, if a decision tree created with the initial operation of the battery 11 or a period close to it set as the learning target period is used continuously, the risk of the difference between the simulated values ​​and the measured values ​​becoming larger will gradually increase. It is preferable to perform the learning process for creating the decision tree while taking this degradation over time into consideration.

[0109] (Example 1) For example, the learning period may be reset at successive intervals and the decision tree may be updated. By performing simulations based on a new decision tree that reflects the deterioration over time, the difference between the simulated values ​​and the measured values ​​can be kept small.

[0110] (Example 2) For example, multiple (new and old) learning periods of the same length, such as every month, may be set sequentially, and a decision tree may be created for each period using the same features. When setting the correction amount Te_m(i), the node corresponding to the i-th feature may be identified in each decision tree, the value of the mean deviation ΔTd(Ne) associated with the corresponding termination node Ne may be read, and the average of the mean deviations ΔTd(Ne) obtained from each decision tree may be used as the value of the correction amount Te_m(i). In this case, the maximum number of decision trees to use may be kept constant, and if the number of decision trees exceeds the maximum number when a newly created decision tree is added, the oldest decision tree may be excluded.

[0111] (Example 3) When using multiple decision trees, both old and new, as in Example 2, the old decision tree may be corrected before use. Specifically, when the average value of the output of the root node of the old decision tree (the average value of all temperature change ΔTm(j) values ​​used to create the decision tree) is set as Ave_old, and the average value of the output of the root node of the latest decision tree (the average value of all temperature change ΔTm(j) values ​​used to create the decision tree) is set as Ave_new, the average temperature change ΔTd(Ne) value associated with each termination node Ne of the old decision tree may be corrected using equation (3) below.

[0112] Corrected ΔTd(Ne) = Uncorrected ΔTd(Ne) + (Ave_new - Ave_old) ... (3) (Example 4) In the case of using multiple decision trees, both old and new, as in Example 2, the weighting coefficient of the average temperature change ΔTd(Ne) obtained from the new decision tree is increased, and the weighting coefficient of the average temperature change ΔTd(Ne) obtained from the old decision tree is decreased. The weighted average of the average temperature change ΔTd(Ne) obtained from the old decision tree may be used as the value of the assumed temperature change ΔT(i). In this case, since the number of data to be learned is secured, the change in the state of the battery 11 can be reflected in the value of the assumed temperature change ΔT(i) while ensuring the stability of the value applied as the assumed temperature change ΔT(i).

[0113] <Second Embodiment> <Expansion of Simulation Scope> In the simulation in the first embodiment, the temperature changes in each of the module batteries 11(1) to 11(n) and the feature quantities of the storage battery 11 are learned, and a decision tree created based on the results of this learning is used to obtain a simulation result showing a representative temperature change in the storage battery 11. This enables a simulation that takes into account the change in battery temperature in each of the module batteries 11(1) to 11(n) more quickly than conventional methods.

[0114] In this embodiment, the simulation concept described above is extended to several state indicator values ​​other than battery temperature, which were the subject of the simulation in the first embodiment. That is, a simulation of the state indicator values ​​during the operating period of the operation pattern that is the subject of the simulation is performed based on the results of a pre-trained learning process, using a procedure similar to the procedure shown in Figures 3 and 5.

[0115] This allows for the rapid and reasonably accurate acquisition of representative simulation results for state indicators other than battery temperature. Furthermore, by including state indicators other than battery temperature in the simulation, the state of the storage battery 11 can be simulated in more detail.

[0116] To enable such responses, in this embodiment as in the first embodiment, the learning processing unit 32 acquires actual temperature, actual voltage, and actual current from each of the module batteries 11(1) to 11(n). Then, for each time interval Δt, these measured values ​​and the values ​​of feature quantities calculated based on these measured values ​​are stored in a storage unit (not shown) of the battery simulation device 3. This data will be used for future simulations. In addition, some feature quantities are identified based on the description content of the operational pattern data, and may also be acquired from the control device 12 of the energy storage system 1.

[0117] Furthermore, in this embodiment as well, the simulation results are described in the simulation result data, similar to the first embodiment. The simulation results related to state index values ​​other than battery temperature, described in the data, are used according to each state index value. Similar to the first embodiment, this may also be used to determine whether the storage battery 11 is operational or not.

[0118] Examples of state indicator values ​​that can be simulated in this way include internal resistance, discharge end-stop depth, and charge end-stop depth. These will be explained in detail below. Note that the types of features to be learned and the learning period may be the same as or different from those used for learning battery temperature. Furthermore, all or part of battery temperature, internal resistance, discharge end-stop depth, and charge end-stop depth may be simulated in parallel.

[0119] (Internal Resistance) Figure 9 shows the procedure for the learning process in the learning processing unit 32 when internal resistance is the target of the simulation. Figure 10 is a schematic diagram showing the flow of the simulation process targeting internal resistance.

[0120] When internal resistance is the target of the simulation, the learning processing unit 32 first learns the combinations of the internal resistance R m(j) of each module battery 11(1) to 11(n) and the feature quantities at each time j for every time interval Δt of a predetermined learning period, and creates a decision tree consisting of multiple layers in which sorting conditions are set in stages based on the learned content.

[0121] Specifically, the learning processing unit 32 first reads the value of a feature at a past time j included in the learning period as the j-th learning target data from the storage unit (not shown) of the battery simulation device 3, similar to the case of battery temperature simulation (step S101).

[0122] Next, the learning processing unit 32 uses the value described in the j-th data to be learned that was read out to calculate the internal resistance Rm(j) at time j for each of the module batteries 11(1) to 11(n) (step S102).

[0123] The learning processing unit 32 also identifies the decision tree node corresponding to the feature value of the j-th data to be learned, and records the calculated internal resistance Rm(j) value in the storage unit, associating it with each node (step S103). As a result, the internal resistance Rm(j) value corresponding to the j-th data to be learned is distributed to one of the termination nodes.

[0124] If there is further training data that has not been sorted (recorded) based on this decision tree (NO in step S104), steps S101 to S103 are repeated.

[0125] When all the training data has been sorted (recorded) (YES in step S104), the average value of the internal resistance Rm(j) associated with each termination node Ne is calculated as the average internal resistance Rd(Ne) and recorded in association with each termination node Ne (step S105).

[0126] Then, when the simulation is executed, the simulation execution unit 31 first acquires the operation pattern data to be simulated from the operation pattern creation device 2, similar to the case of the battery temperature simulation (step S111).

[0127] Next, the simulation execution unit 31 identifies (outputs) the value of the assumed internal resistance (assumed internal resistance) R(i) at the simulation target time i, based on the description of the operation pattern data. Specifically, it first reads out the value (assumed value) of the feature at the simulation target time i. Then, it identifies the node corresponding to the i-th feature that was read out, according to the decision tree created by the learning processing unit 32, and identifies the value of the average internal resistance Rd(Ne) associated with the corresponding termination node Ne as the assumed internal resistance R(i).

[0128] If the simulation target time i for which the internal resistance is specifically targeted is not the last time of the operating period (NO in step S113), the simulation execution unit 31 repeats the process in step S112 for a new simulation target time i that is advanced by a time interval Δt from the said simulation target time i.

[0129] If the specific simulation time i targeted for internal resistance is the last time of the operating period (YES in step S113), then the internal resistance R(i) has been determined for all simulation time i included in the operating period. In other words, the simulation of internal resistance for the operating period assumed in the operating pattern is complete.

[0130] The simulation execution unit 31 generates simulation result data that describes the simulation results along with the operation pattern that was the subject of the simulation. The simulation results for internal resistance are used for calculations such as the battery temperature (for example, the calculation of the initial value T(0)).

[0131] Then, when the battery 11 is operated (actually operated) based on the operational pattern data that was the subject of the simulation, the learning processing unit 32 acquires actual temperature, actual current, and actual voltage data (actual data) for each of the module batteries 11(1) to 11(n) (step S114) and records it in a storage unit (not shown) of the battery simulation device 3 (step S115). This actual data will be used for new learning processing in the learning processing unit 32.

[0132] It should be noted that the internal resistance values ​​of each module battery 11(1) to 11(n) may normally differ. On the other hand, the simulation results for internal resistance obtained in this embodiment only show typical changes in the module batteries 11(1) to 11(n) provided in the storage battery 11.

[0133] From the viewpoint of more reliably maintaining the internal resistance of each module battery 11(1) to 11(n) within an acceptable range during the operation of the storage battery 11, it is ideally desirable to individually simulate the internal resistance of each module battery 11(1) to 11(n). However, simulating the internal resistance in the manner of this embodiment is preferable in that it ensures a reasonable level of accuracy while performing the simulation more quickly than conventional methods, similar to the case of the first embodiment.

[0134] (End-of-discharge and end-of-charge depth) First, the end-of-discharge depth and end-of-charge depth will be explained based on Figure 11. Figure 11 is a schematic graph showing the relationship between the end-of-discharge depth and voltage in a NaS battery.

[0135] As shown in Figure 11, the voltage (charge / discharge voltage) of the NaS battery is kept at a generally constant value V0 except near the end of discharge where the depth of discharge is maximum and near the end of charging where the depth of discharge is minimum. However, it decreases near the end of discharge and increases near the end of charging. Based on these characteristics, the charge / discharge voltages of each module battery 11(1) to 11(n) of the storage battery 11 are predetermined to be the voltage at which the end of discharge is considered to have been reached and further discharge is stopped (discharge stop voltage) VL, and the voltage at which the end of charging is considered to have been reached and further charging is stopped (charge stop voltage) VH.

[0136] The discharge depth at which the voltage of each module battery 11(1) to 11(n) reaches the discharge stop voltage VL, and the discharge depth at which it reaches the charge stop voltage VH, are referred to as the discharge end stop depth and the charge end stop depth, respectively.

[0137] The values ​​of the discharge end-stop depth and the charge end-stop depth serve as indicators for operating the battery 11 without causing a halt in charging or discharging in each module battery 11(1) to 11(n), and can therefore be treated as state indicator values ​​for simulation.

[0138] When these discharge end-stop depths and charge end-stop depths are the targets of the simulation, the learning processing unit 32 learns in advance the combinations of the discharge end-stop depth DLm(j) and charge end-stop depth DHm(j) of module batteries 11(1) to 11(n) and the feature quantities at each time j for every time interval Δt of a predetermined learning period, and creates a decision tree consisting of multiple layers with sorting conditions set in stages based on the learned content.

[0139] Then, when the simulation is executed, the simulation execution unit 31 first reads out the value (expected value) of the feature quantity at the simulation target time i, which is set based on the description of the operation pattern data, similar to the case of the battery temperature simulation. Next, the simulation execution unit 31 identifies the node corresponding to the i-th feature quantity that was read out, according to the decision tree created by the learning processing unit 32, and identifies (outputs) the value of the average discharge end-side stop depth DLd(Ne) and the value of the average charge end-side stop depth DHd(Ne) associated with the corresponding end node Ne, as the discharge end-side stop depth DL(i) and charge end-side stop depth DH(i), respectively.

[0140] The acquisition and recording of actual data during subsequent operational use is the same as in the case of the internal resistance simulation.

[0141] As shown in Figure 9, the voltage behavior near the end of discharge and near the end of charge in each module battery 11(1) to 11(n) is not necessarily uniform. Therefore, the values ​​of the end-of-discharge stop depth and the end-of-charge stop depth in each module battery 11(1) to 11(n) can usually differ. On the other hand, the simulation results for the end-of-discharge stop depth and the end-of-charge stop depth obtained in this embodiment represent only typical changes in the module batteries 11(1) to 11(n) provided in the storage battery 11.

[0142] From the viewpoint of operating the battery 11 to its maximum capacity without causing any cessation of charging or discharging in each of the module batteries 11(1) to 11(n), it is ideally desirable to individually simulate the discharge-end cessation depth and charge-end cessation depth of each module battery 11(1) to 11(n). However, simulating the discharge-end cessation depth and charge-end cessation depth in the manner of this embodiment is preferable in that it ensures a reasonable level of accuracy while performing the simulation more quickly than conventional methods, similar to the case of the first embodiment.

[0143] In the case of lithium-ion batteries, the battery voltage tends to increase with increasing charge level, but the dependence between charging and discharging differs, and this dependence also changes depending on the battery's degradation state. Therefore, even when a battery is composed of multiple lithium-ion batteries, it is meaningful to simulate the discharge end-stop depth and charge end-stop depth in the manner described above, as is the case with NaS batteries.

[0144] <Modification> In the above-described embodiment, a single simulation result is obtained for the storage battery 11, assuming that the temperature changes of each module battery 11(1) to 11(n) are the same. However, the simulation in the above-described embodiment may also be performed individually for each module battery 11(1) to 11(n).

Claims

1. A simulation device for performing an operation simulation based on an operational plan for a storage battery comprising multiple module batteries, each composed of multiple single cells, comprising: a learning processing unit that learns the relationship between a feature quantity that characterizes the operating state of the storage battery during a predetermined learning target period defined for the past operating period of the storage battery, and an index value change amount which is the amount of change of at least one state index value of the storage battery at predetermined time intervals; and a simulation execution unit that performs a simulation of the at least one state index value during the execution period of the operational plan based on the learning results of the learning processing unit.

2. A simulation device according to claim 1, wherein the learning processing unit acquires the actual temperature, actual voltage, and actual current of each of the plurality of module batteries for at least the predetermined learning target period, calculates the feature quantities based on the actual temperature, the actual voltage, and the actual current at predetermined time intervals, creates a decision tree for determining the assumed values ​​of the change in the index value during the execution period of the operation plan according to the conditions of the feature quantities based on the learning results that associate the change in the index value with the feature quantities at predetermined time intervals, and performs a simulation of the at least one state index value during the execution period of the operation plan by sequentially identifying the assumed values ​​of the change in the index value during the execution period of the operation plan based on the decision tree at predetermined time intervals, and sequentially adding the identified assumed values.

3. A simulation apparatus according to claim 1 or claim 2, characterized in that the at least one state index value includes the temperature of the storage battery.

4. A simulation device for performing an operation simulation based on an operational plan for a storage battery comprising multiple module batteries, each composed of multiple single cells, comprising: a learning processing unit that learns the relationship between feature quantities that characterize the operating state of the storage battery and at least one state index value of the storage battery during a predetermined learning target period set up to cover the past operating period of the storage battery; and a simulation execution unit that performs a simulation of the at least one state index value during the execution period of the operational plan based on the learning results of the learning processing unit.

5. A simulation apparatus according to claim 4, wherein the learning processing unit acquires the actual temperature, actual voltage, and actual current of each of the plurality of module batteries for at least the predetermined learning target period, calculates the feature quantities at predetermined time intervals based on the actual temperature, the actual voltage, and the actual current, creates a decision tree for determining the assumed value of the at least one indicator value during the execution period of the operation plan according to the conditions of the feature quantities based on the learning results that associate the at least one indicator value with the feature quantities at predetermined time intervals, and performs a simulation of the at least one state indicator value during the execution period of the operation plan by sequentially identifying the assumed value of the at least one state indicator value during the execution period of the operation plan based on the decision tree at predetermined time intervals.

6. A simulation apparatus according to claim 4 or claim 5, characterized in that the at least one state index value includes the internal resistance of the storage battery.

7. A simulation apparatus according to claim 4 or claim 5, characterized in that the at least one state index value includes at least one of the discharge end-stop depth, which is the discharge depth when the storage battery reaches the discharge stop voltage, and the charge end-stop depth, which is the discharge depth when the storage battery reaches the charge stop voltage.

8. An operation management device for managing the operation of a battery, each comprising a plurality of module batteries, each composed of a plurality of single cells, comprising: an operation feasibility determination unit that determines whether or not to operate the battery according to the operation plan based on the results of a simulation in the simulation device described in claim 1 or claim 2; and a charge / discharge management unit that sends the operation plan, which has been determined to be operational by the operation feasibility determination unit to be operational, to a control device of the battery, and causes the control device to operate the battery according to the operation plan.

9. A method for simulating the operation of a battery storage system, each comprising multiple module batteries, each composed of multiple single cells, based on an operational plan, comprising: a learning process step of learning the relationship between a feature quantity that characterizes the operating state of the battery storage system during a predetermined learning target period defined for the past operating period of the battery storage system, and an index value change amount which is the amount of change of at least one state index value of the battery storage system at predetermined time intervals; and a simulation step of simulating the at least one state index value during the execution period of the operational plan based on the learning results in the learning process step.

10. A simulation method according to claim 9, characterized in that, in the learning process step, for at least the predetermined learning target period, the actual temperature, actual voltage, and actual current of each of the plurality of module batteries are acquired; the feature quantities are calculated based on the actual temperature, actual voltage, and actual current at predetermined time intervals; a decision tree is created at each predetermined time interval based on the learning result of associating the change in the index value with the feature quantities, to determine the assumed value of the change in the index value during the execution period of the operation plan according to the conditions of the feature quantities; and in the simulation step, the assumed value of the change in the index value during the execution period of the operation plan is sequentially identified at predetermined time intervals based on the decision tree, and the identified assumed values ​​are sequentially added together to perform a simulation of at least one state index value during the execution period of the operation plan.

11. A simulation method according to claim 9 or claim 10, characterized in that the at least one state index value includes the temperature of the storage battery.

12. A method for simulating the operation of a battery storage system, each comprising multiple module batteries, each composed of multiple single cells, based on an operational plan, comprising: a learning process step of learning the relationship between feature quantities that characterize the operating state of the battery storage system during a predetermined learning target period defined for the past operating period of the battery storage system, and at least one state index value of the battery storage system; and a simulation step of simulating the at least one state index value during the execution period of the operational plan based on the learning results in the learning process step.

13. A simulation method according to claim 12, characterized in that, in the learning process step, the actual temperature, actual voltage, and actual current of each of the plurality of module batteries are acquired for at least the predetermined learning target period; the feature quantities are calculated at predetermined time intervals based on the actual temperature, the actual voltage, and the actual current; a decision tree is created at predetermined time intervals based on the learning results that associate the at least one index value with the feature quantities, to determine the assumed value of the at least one index value during the execution period of the operation plan according to the conditions of the feature quantities; and in the simulation step, the assumed value of the at least one state index value during the execution period of the operation plan is sequentially identified at predetermined time intervals based on the decision tree, thereby simulating the at least one state index value during the execution period of the operation plan.

14. A simulation method according to claim 12 or claim 13, characterized in that the at least one state index value includes the internal resistance of the storage battery.

15. A simulation method according to claim 12 or claim 13, characterized in that the at least one state index value includes at least one of the discharge depth at the end of discharge, which is the discharge depth when the storage battery reaches the discharge stop voltage, and the charge depth at the end of discharge, which is the discharge depth when the storage battery reaches the charge stop voltage.

16. A method for managing the operation of a battery storage system comprising a plurality of module batteries, each composed of a plurality of single cells, comprising: an operation feasibility determination step of determining whether or not to operate the battery storage system according to the operation plan based on the results of a simulation in the simulation method described in claim 9 or claim 10; and a charge / discharge management step of sending the operation plan, which was determined to be operational in the operation feasibility determination step, to a control device of the battery storage system, and causing the control device to operate the battery storage system according to the operation plan.

Images