DC Feeder Voltage Calculation Device, DC Feeder Voltage Control System, DC Feeder Voltage Calculation Program, and DC Feeder Voltage Calculation Method

Abstract

In a DC feeder voltage control system (7) that controls the operation of substations (4-1 to 4-m), a DC feeder voltage calculation device (1) includes: a communication unit (30) that acquires substation information including substation voltages applied from the substations (4-1 to 4-m) to feeders in DC electrified sections of an electric railway and directional currents flowing from the substations (4-1 to 4-m) to feeders in each direction in the DC electrified section, as well as running history information that indicates the position of a train and the state of power; and a processing unit (20) that calculates and plans control content for the substations (4-1 to 4-m) based on an inference model created by learning using learning data created from a simulation using the running history information and the substation information.

Description

[Technical Field]

[0001] The present disclosure relates to a DC feeder voltage calculation device, a DC feeder voltage control system, a DC feeder voltage calculation program, and a DC feeder voltage calculation method for calculating a substation voltage, which is a voltage applied from a substation to a feeder line in a DC electrified section of an electric railway. [Background technology]

[0002] Conventionally, in DC electrified sections of electric railways, the regenerative power generated by regenerative brakes installed on trains has been effectively utilized for the purpose of energy conservation, etc. For example, Patent Document 1 discloses a technology for reducing the amount of power consumption in DC electrified sections by repeatedly performing power simulations using model information, train running history information, condition information, etc., and calculating the optimal substation voltage setting value for each condition. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2018 / 225194 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the conventional technology requires that an optimal voltage setting value be calculated in advance, which poses a problem in that if the pre-calculated voltage setting value differs from the optimal voltage setting value for the current operating conditions, appropriate voltage control cannot be performed, potentially resulting in energy loss in the feeder lines in DC electrified sections.

[0005] The present disclosure has been made in view of the above, and has an object to provide a DC feeder voltage calculation device that can reduce energy loss in a feeder line in a DC electrified section. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the objectives, the DC feeder voltage calculation device disclosed herein is characterized by comprising a communication unit that acquires substation information including the substation voltage applied from the substation to a feeder in a DC electrified section of an electric railway and the directional current flowing from the substation to each direction feeder in the DC electrified section, as well as running history information indicating the train's position and power status, and a processing unit that calculates control content for the substation based on an inference model and substation information created by learning using learning data created from a simulation using the running history information. [Effects of the Invention]

[0007] The DC feeder voltage calculation device of the present disclosure has the effect of reducing energy loss in a feeder line in a DC electrified section. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram showing an overview of processing by a DC feeder voltage calculation device according to a first embodiment. [Figure 2] FIG. 1 is a diagram showing a configuration example of a DC feeding voltage control system according to a first embodiment; [Figure 3] FIG. 1 is a diagram showing an example of running history information stored in a running history information storage unit of the DC feeding voltage calculation device according to the first embodiment; [Figure 4] FIG. 1 is a first diagram showing an example of learning data stored in a learning data storage unit of a DC feeder voltage calculation device according to Embodiment 1. [Figure 5] FIG. 1 is a diagram showing an example of a line to which a DC feeding voltage control system according to a first embodiment is applied. [Figure 6] FIG. 2 is a second diagram showing an example of learning data stored in the learning data storage unit of the DC feeder voltage calculation device according to the first embodiment; [Figure 7] FIG. 1 is a diagram showing an example of substation information stored in a substation information storage unit of the DC feeder voltage calculation device according to the first embodiment. [Figure 8] FIG. 10 is a diagram showing an example of control content stored in a control content storage unit of the DC feeder voltage calculation device according to the first embodiment. [Figure 9] A flowchart showing a process for creating an inference model by the DC feeder voltage calculation device according to the first embodiment. [Figure 10] 1 is a flowchart showing a process for calculating control details performed by a DC feeder voltage calculation device according to a first embodiment. [Figure 11] 1 is a flowchart showing a process in which the DC feeding voltage calculation device according to the first embodiment outputs control content to each substation. [Figure 12] 1 is a flowchart showing a process for creating learning data by a DC feeder voltage calculation device according to a first embodiment. [Figure 13] FIG. 1 is a diagram showing an example in which a processing circuit realizing a DC feeder voltage calculation device according to a first embodiment is configured with a processor and a memory. [Figure 14] FIG. 1 is a diagram showing an example in which a processing circuit realizing a DC feeder voltage calculation device according to a first embodiment is configured with dedicated hardware. [Figure 15] FIG. 10 is a diagram showing a configuration example of a DC feeding voltage control system according to a second embodiment. [Figure 16] FIG. 10 is a diagram showing an example of learning data stored in a learning data storage unit of the DC feeder voltage calculation device according to the second embodiment. [Figure 17] FIG. 10 is a diagram showing an example of substation statistical information stored in a substation statistical information storage unit of the DC feeder voltage calculation device according to the second embodiment. [Figure 18] FIG. 10 is a diagram showing an example of control content stored in a control content storage unit of a DC feeder voltage calculation device according to a second embodiment. [Figure 19] FIG. 10 is a diagram showing a configuration example of a DC feeding voltage control system according to a third embodiment. [Figure 20] FIG. 10 is a diagram showing an example of conditions stored in a condition information storage unit of a DC feeding voltage calculation device according to a third embodiment. [Figure 21] FIG. 10 is a diagram showing an example of learning data stored in a learning data storage unit of the DC feeder voltage calculation device according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A DC feeder voltage calculation device, a DC feeder voltage control system, a DC feeder voltage calculation program, and a DC feeder voltage calculation method according to embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0010] Embodiment 1 FIG. 1 is a diagram illustrating an overview of processing by a DC feeder voltage calculation device 1 according to a first embodiment. The DC feeder voltage calculation device 1 acquires substation information from a plurality of substations 4-1 to 4-m, including substation voltages applied from each substation to a feeder 5 in a DC electrified section of an electric railway and directional currents flowing from each substation to the feeder 5 in each direction in the DC electrified section. In the following description, when there is no need to distinguish between the substations 4-1 to 4-m, they may be referred to as substations 4. Note that m is an integer of 2 or greater. The substation voltages applied from each substation 4 to the feeder 5 are measured by a voltage sensor 53 for each substation 4, and the directional currents flowing from each substation 4 to the feeder 5 in each direction are measured by current sensors 54-1 and 54-2 for each substation 4. In the following description, when there is no need to distinguish between the current sensors 54-1 and 54-2, they may be referred to as current sensors 54. In the example of Figure 1, two current sensors 54 are connected to each substation 4, but the number of current sensors 54 connected to each substation 4, i.e., the number of directions of the feeder line 5 as seen from the substation 4, is not limited to two.

[0011] The DC feeder voltage calculation device 1 also acquires running history information indicating the position and power status of each train from multiple trains 6-1 to 6-n. In the following description, when the trains 6-1 to 6-n are not distinguished from one another, they may be referred to as train 6. Note that n is an integer equal to or greater than 2. When the train 6 is performing ground-to-car communication with a train traffic control system (not shown) that manages the operation of the train 6, the DC feeder voltage calculation device 1 can acquire the running history information via the train traffic control system. The power status refers to, for example, the power demand of the train 6. During normal acceleration, the train 6 receives power supply from the substation 4 via the feeder line 5 to run, but during deceleration using regenerative braking, the train 6 supplies regenerative power generated by the regenerative braking to the feeder line 5. Therefore, while a value indicating the power status of the train 6 is a positive value when power is supplied from the substation 4 via the feeder line 5, it may become a negative value during regenerative braking. Trains 6-1 to 6-n are assumed to be equipped with regenerative brakes, i.e., capable of supplying regenerative power to feeder line 5, but trains 6 may also be included that are not equipped with regenerative brakes, i.e., cannot supply regenerative power to feeder line 5.

[0012] The DC feeder voltage calculation device 1 acquires running history information for offline pre-processing and substation information for real-time processing, and uses this information to calculate and plan the optimal control content for each substation 4 so that feeder voltage control can be performed that reduces energy loss in the feeder lines 5 in DC electrified sections.

[0013] Specifically, as offline pre-processing, the DC feeder voltage calculation device 1 performs a simulation using running history information acquired the day before or earlier and simulation model information described below, and creates learning data in which the substation voltage applied to the feeder 5 from each substation 4 and the directional current flowing from the substation 4 to the feeder 5 in each direction in the DC electrified section are used as feature data, i.e., explanatory variables, and the optimal no-load voltage for each substation 4 is used as the control content, i.e., target variable. The DC feeder voltage calculation device 1 also creates an inference model to be used in real-time processing by learning using the learning data. As real-time processing, the DC feeder voltage calculation device 1 calculates the control content for each substation 4 using the substation information acquired at a specified cycle and the inference model, and issues instructions to each substation 4.

[0014] FIG. 2 is a diagram showing an example of the configuration of a DC feeding voltage control system 7 according to the first embodiment. The DC feeding voltage control system 7 includes a DC feeding voltage calculation device 1 and a DC feeding voltage control device 2. The DC feeding voltage control system 7 controls the operation of a substation 4. The DC feeding voltage control device 2 is connected to substations 4-1 to 4-m via a network 3. The DC feeding voltage control device 2 includes a substation voltage control unit 41 and a substation information acquisition unit 42. The substation 4 includes a voltage control unit 51 and a voltage / current acquisition unit 52.

[0015] Although not shown in FIG. 2 , the voltage and current acquisition unit 52 of the substation 4 is connected to a voltage sensor 53 and a current sensor 54, and acquires the substation voltage applied from the substation 4 to the feeder 5 at a specified period from the voltage sensor 53, and acquires the direction-specific currents flowing from the substation 4 to the feeder 5 in each direction from the current sensor 54. The voltage and current acquisition unit 52 of the substation 4 outputs the acquired substation voltages and direction-specific currents as substation information to the DC feeding voltage control device 2 via the network 3. The substation information acquisition unit 42 of the DC feeding voltage control device 2 acquires substation information from the substations 4-1 to 4-m, and outputs the acquired substation information of the substations 4-1 to 4-m to the DC feeding voltage calculation device 1.

[0016] When the substation voltage control unit 41 of the DC feeding voltage control device 2 acquires from the DC feeding voltage calculation device 1 the control content for each substation 4 calculated by processing of the DC feeding voltage calculation device 1, which will be described later, the substation voltage control unit 4 outputs the control content corresponding to each substation 4 to each substation 4 via the network 3. The voltage control unit 51 of the substation 4 controls the substation voltage to be applied to the feeder 5 based on the acquired control content. In this way, the DC feeding voltage control device 2 controls the substation voltage to be applied from the substation 4 to the feeder 5 in the DC electrified section of the electric railway based on the control content calculated by the DC feeding voltage calculation device 1. The DC feeding voltage control system 7 is a system that controls the substation voltage from each substation 4 to be applied to the feeder 5 in the DC electrified section of the electric railway.

[0017] The configuration and operation of the DC feeding voltage calculation device 1 will be described in detail. As shown in Fig. 2, the DC feeding voltage calculation device 1 includes a storage unit 10, a processing unit 20, and a communication unit 30. The storage unit 10 includes a simulation model information storage unit 11, a running history information storage unit 12, a learning data storage unit 13, an inference model storage unit 14, a substation information storage unit 15, and a control content storage unit 16. The processing unit 20 includes a learning data creation unit 21, an inference model creation unit 22, a control content calculation unit 23, and an information acquisition unit 24.

[0018] In the storage unit 10, the simulation model information storage unit 11 stores simulation model information used in a simulation performed when the learning data creation unit 21 creates learning data. The simulation model information includes, for example, a train model, a substation model, a power grid model, and the like.

[0019] The train model is information for controlling the amount of regenerative power throttling for each train 6 located in a DC electrified section, and includes information on the regenerative throttling start voltage and regenerative throttling end voltage. The substation model is information for controlling the substation voltage of each substation 4, and includes information such as the internal resistance of each substation 4, the controllable range of substation voltage, and the number of multiple power converters that can be operated in parallel that each substation 4 is equipped with. The controllable range of substation voltage is the voltage range over which the substation voltage can be controlled at the substation 4, and is the no-load voltage range that the DC feeder voltage calculation device 1 can set as the control content for the substation 4. The feeder network model includes information such as the location information of each substation 4, the connection relationship between each substation 4 and the feeder 5, the connection relationship between the feeders 5, the length of the feeder 5, the resistivity of the feeder 5, the connection relationship between the rails in the DC electrified section, the length of the rails, and the resistivity of the rails.

[0020] The simulation model information may further include a station auxiliary power supply model including control information for a station auxiliary power supply installed in a station where a specified type of train 6 stops, and a power storage device model including control information for a power storage device that stores regenerative power generated by the train 6. The station auxiliary power supply model includes information about the station auxiliary power supply, such as its absorption start voltage, rated voltage, and conversion efficiency. The power storage device model includes information about the power storage device, such as its charge start voltage, charge conversion efficiency, discharge start voltage, discharge conversion efficiency, minimum SoC (State of Charge), and SoC initial value. The simulation model information may include at least one of the station auxiliary power supply model and the power storage device model. In this case, the power feeder network model includes location information for the station auxiliary power supply if the simulation model information includes the station auxiliary power supply model, and includes location information for the power storage device if the simulation model information includes the power storage device model. The DC feeder voltage calculation device 1 can perform simulations with higher accuracy by using a large amount of simulation model information.

[0021] The running history information storage unit 12 stores running history information acquired from each train 6, the information indicating the position and power status of the train 6. The running history information includes information indicating the position and power status of trains 6 that have run on the previous day or earlier. FIG. 3 is a diagram illustrating an example of running history information stored in the running history information storage unit 12 of the DC feeder voltage calculation device 1 according to the first embodiment. The running history information includes information such as the date and time when the running history information was acquired from each train 6, a train ID (identifier) ​​for identifying the train 6, a direction indicating the direction of travel of the train 6, the position of the train 6, and the power status of the train 6, i.e., the power demand. The running history information illustrated in FIG. 3 is an example, and other information may be included in the running history information. Regarding the power demand, a positive value represents the power used for powering the train 6, and a negative value represents the power that the train 6 can regenerate. The running history information is assumed to be actually measured, but may also be generated by a running simulation of the train 6.

[0022] The learning data storage unit 13 stores the learning data that the learning data creation unit 21 creates by performing a simulation.

[0023] In the DC feeder voltage calculation device 1, the learning data creation unit 21 randomly extracts a no-load voltage from candidate voltage values, which are candidate values ​​for the no-load voltage of each substation 4 indicated by the controllable range of the substation voltage included in the substation model stored in the simulation model information storage unit 11. The learning data creation unit 21 also extracts time-section running history information from the running history information stored in the running history information storage unit 12. The learning data creation unit 21 performs a simulation using the extracted information, calculates the substation voltage and direction-specific current of each substation 4, and sets them as feature data, i.e., explanatory variables. The learning data creation unit 21 searches for an optimal no-load voltage for each substation 4 from the candidate voltage values ​​and sets it as a response variable. The learning data creation unit 21 stores the pair of the explanatory variable and the response variable as one sample of learning data in the learning data storage unit 13. FIG. 4 is a first diagram illustrating an example of learning data stored in the learning data storage unit 13 of the DC feeder voltage calculation device 1 according to the first embodiment. In the first embodiment, the learning data includes explanatory variables indicating substation voltages, which are the voltages at each substation 4, and direction-specific currents, and a target variable indicating the optimum no-load voltage to be applied from each substation 4 to the feeder 5. The data for one sample described above corresponds to one row of data shown in FIG. 4. In the example of FIG. 4, substation 4-1 is denoted as "SS1" and substation 4-2 is denoted as "SS2." This also applies to the following.

[0024] Here, the explanatory variables of the learning data shown in FIG. 4 are, for example, explanatory variables calculated at a time cross section at 00:00:00. The explanatory variables at the time cross section at 00:00:00 are calculated by the learning data creation unit 21 through simulation using the driving history information at the time cross section at 00:00:00 and a random no-load voltage. The objective variables of the learning data shown in FIG. 4 are, for example, objective variables calculated at a time cross section at 00:00:01. This assumes a control time delay of 1 second. When the control time delay is 1 second, the learning data creation unit 21 repeatedly performs simulations at a time cross section 1 second later to search for the optimal no-load voltage. The difference between the time cross sections of the explanatory variables and the time cross section of the objective variable may be set appropriately depending on the control time delay. The reason for using a random no-load voltage in the simulation by the learning data creation unit 21 is to enable the learning data creation unit 21 to determine the optimal no-load voltage at 00:00:01, regardless of the no-load voltage at the time cross section at 00:00:00. Here, as an example, the interval of the time slice, that is, the unit of the time slice, is 1 second as shown in FIG.

[0025] An example of a route on which the train 6 runs corresponding to the learning data shown in Fig. 4 will be described. Fig. 5 is a diagram showing an example of a route to which the DC feeding voltage control system 7 according to the first embodiment is applied. The example of Fig. 5 shows a case where the route on which the train 6 runs is a double track on which the train 6 can run independently in the uphill and downhill directions.

[0026] Train 6-1 is traveling in the downhill direction to the right of the figure, and can receive power from substations 4-1 and 4-2 via feeder 5-1. Train 6-2 is traveling in the uphill direction to the left of the figure, and can receive power from substations 4-1 and 4-2 via feeder 5-2. Substation 4-1 can also supply power to train 6 traveling to the left of substation 4-1, which is not shown in Figure 5, via feeders 5-1 and 5-2. Substation 4-2 can also supply power to train 6 traveling to the right of substation 4-2, which is not shown in Figure 5, via feeders 5-1 and 5-2.

[0027] In the example of Figure 5, the current flowing from substation 4 to downlink feeder 5-1 in the up direction is referred to as directional current 1, the current flowing from substation 4 to downlink feeder 5-1 in the down direction is referred to as directional current 2, the current flowing from substation 4 to uplink feeder 5-2 in the up direction is referred to as directional current 3, and the current flowing from substation 4 to downlink feeder 5-2 in the up direction is referred to as directional current 4. directional current 1 is measured by current sensor 54-1, directional current 2 is measured by current sensor 54-2, directional current 3 is measured by current sensor 54-3, and directional current 4 is measured by current sensor 54-4. In Figure 5, directional currents are indicated by arrows or the like only for substation 4-1, but the same applies to substation 4-2. The direction-specific currents 1 to 4 of the substation 4 shown in FIG. 5 correspond to the direction-specific currents 1 to 4 of the substation 4 indicated by the explanatory variables of the learning data in FIG.

[0028] Focusing on train 6-1, power is supplied to train 6-1 from substation 4-1, and power is also supplied to train 6-1 from substation 4-2. In the state shown in FIG. 5, train 6-1 is closer to substation 4-1 than to substation 4-2. Therefore, it is expected that train 6-1 will receive more power from substation 4-1 than from substation 4-2. In this case, it is expected that directional current 2 for train 6-1, contained in the substation information for substation 4-1, will be greater than directional current 1 for train 6-1, contained in the substation information for substation 4-2. Furthermore, as train 6-1 continues traveling to the right of FIG. 5, i.e., in the downhill direction, train 6-1 will gradually approach substation 4-2 and gradually move away from substation 4-1. In this case, it is thought that the direction-specific current 1 for the direction of train 6-1 included in the substation information of substation 4-2 will gradually increase, and the direction-specific current 2 for the direction of train 6-1 included in the substation information of substation 4-1 will gradually decrease. Taking the above into consideration, the DC feeder voltage calculation device 1 can estimate the approximate operating status of train 6 running between adjacent substations 4 by using the substation information of adjacent substations 4.

[0029] FIG. 6 is a second diagram showing an example of the learning data stored in the learning data storage unit 13 of the DC feeder voltage calculation device 1 according to the first embodiment. Although the description in FIG. 6 is simplified, in reality, the learning data creation unit 21 repeatedly performs simulations to determine specific values ​​even for parts where specific values ​​are not described in the explanatory variables and the objective variable. Since the optimal no-load voltage is changed and simulations are performed even at the same time cross section, the feature data, i.e., the explanatory variables, change even at the same time cross section. On the other hand, since the same running history information is used at the same time cross section, the optimal no-load voltage, i.e., the objective variable, does not change at the same time cross section.

[0030] The inference model storage unit 14 stores the inference model created by the inference model creation unit 22 through learning using the training data. The learning in the inference model creation unit 22 is supervised machine learning, and existing methods such as neural networks and gradient boosting may be used. The inference model creation unit 22 may tune the hyperparameters of the inference model by dividing the training data into training data and validation data and performing cross-validation as necessary, while observing the accuracy of predictions made using the validation data.

[0031] The substation information storage unit 15 stores substation information acquired from substations 4. The substation information is acquired by the DC feeding voltage calculation device 1 from each substation 4 via the network 3 and the DC feeding voltage control device 2. In the DC feeding voltage calculation device 1, the information acquisition unit 24 of the processing unit 20 stores the substation information acquired from the DC feeding voltage control device 2 via the communication unit 30 in the substation information storage unit 15. FIG. 7 is a diagram showing an example of substation information stored in the substation information storage unit 15 of the DC feeding voltage calculation device 1 according to the first embodiment. The substation information storage unit 15 stores the substation information acquired from each substation 4 at one-second intervals in the example of FIG. 7 for each time the substation information is acquired. Note that if the acquisition times of the substation information acquired from each substation 4 are different, a process for aligning the times is required. Therefore, when the acquisition times of the substation information acquired from each substation 4 are different, for example, the information acquisition unit 24 of the processing unit 20 performs processing to align the acquisition times of the substation information from each substation 4. The processing to align the acquisition times of the substation information includes, but is not limited to, interpolation processing using linear interpolation or the like.

[0032] The control content storage unit 16 stores the control content for each substation 4 calculated by the control content calculation unit 23. FIG. 8 is a diagram showing an example of the control content stored in the control content storage unit 16 of the DC feeding voltage calculation device 1 according to the first embodiment. In the example of FIG. 8, the control content includes the date and time when the control content is calculated by the control content calculation unit 23, a control time indicating the date and time when control will start using the control content, and a no-load voltage to be instructed to each substation 4 as the specific control content. That is, in the first embodiment, the control content is an optimum no-load voltage to be applied from the substation 4 to the feeder line 5. For example, the information acquisition unit 24 of the processing unit 20 reads out the control content from the control content storage unit 16 in accordance with the control time and outputs the read control content to the DC feeding voltage control device 2 via the communication unit 30.

[0033] The processing unit 20 calculates and plans control details for the substation 4 based on the substation information and an inference model created by learning using learning data created from a simulation using the travel history information. The processing unit 20 performs a specified number of simulations by changing the combination of candidate voltage values ​​for each substation 4. The candidate voltage values ​​for the substation 4 are one or more specific voltage values ​​or voltage values ​​selected from a voltage range that can be set as candidate voltage values.

[0034] Specifically, in the processing unit 20, the learning data creation unit 21 performs a simulation using random candidate voltage values ​​for the substation 4, using the simulation model stored in the simulation model information storage unit 11 and running history information stored in the running history information storage unit 12, which indicates the positions and power states of multiple trains 6 that have run on the DC electrified section on or before the previous day. The learning data creation unit 21 creates learning data through the simulation, which is a combination of explanatory variables including the substation voltage of each substation 4 at each time cross section and directional current, which is the current value of the feeder 5 for each direction of the train 6, and a target variable corresponding to the control content for each substation 4 to reduce the power used in running the train 6. The content of the simulation performed by the learning data creation unit 21 is as described in the section about the learning data storage unit 13. The learning data creation unit 21 stores the created learning data in the learning data storage unit 13.

[0035] Here, if there are a large number of candidate voltage values ​​for each substation 4, the number of combinations of candidate voltage values ​​for each substation 4 when the candidate voltage values ​​for each substation 4 are changed also becomes enormous. In such a case, if the learning data creation unit 21 performs simulations for all combinations of candidate voltage values, the time required for the simulations will also become enormous. For this reason, it is assumed that an upper limit on the number of simulations is set in the learning data creation unit 21. If the number of combinations of all candidate voltage values ​​is equal to or less than the upper limit, the learning data creation unit 21 performs simulations for all combinations of candidate voltage values, and if the number of combinations of all candidate voltage values ​​is greater than the upper limit, the learning data creation unit 21 performs simulations for the number of combinations of candidate voltage values ​​indicated by the upper limit.

[0036] The number of candidate voltage values ​​for each substation 4 may be the same for all substations 4, or may be different for each substation 4. For example, if there are three candidate voltage values ​​for substation 4-1, there may be four candidate voltage values ​​for substation 4-2. Note that the number of candidate voltage values ​​for some substations 4 may be one.

[0037] The inference model creation unit 22 creates an inference model by learning using the learning data stored in the learning data storage unit 13. The content of the learning performed by the inference model creation unit 22 is as explained in the inference model storage unit 14. The inference model creation unit 22 stores the created inference model in the inference model storage unit 14.

[0038] The control content calculation unit 23 calculates the control content using current substation information including the substation voltage detected by the voltage sensor 53 of the substation 4 and the directional current detected by the directional current sensor 54 of the train 6 of the substation 4, and an inference model. Specifically, the control content calculation unit 23 creates feature data required for inference using the inference model, using the substation information at a desired time stored in the substation information storage unit 15. The control content calculation unit 23 calculates the control content by inputting the created feature data into the inference model stored in the inference model storage unit 14 and performing inference. The control content calculation unit 23 stores the calculated control content in the control content storage unit 16.

[0039] The information acquisition unit 24 acquires the driving history information acquired by the communication unit 30 from the communication unit 30 and stores it in the driving history information storage unit 12. The information acquisition unit 24 also acquires the substation information acquired by the communication unit 30 from the communication unit 30 and stores it in the substation information storage unit 15. The information acquisition unit 24 also reads out the control contents stored in the control content storage unit 16 and outputs it to the communication unit 30.

[0040] The communication unit 30 acquires substation information output from the substation 4 via the network 3 and the DC feeding voltage control device 2. The communication unit 30 also acquires running history information output from the train 6. The communication unit 30 also outputs the control content acquired from the information acquisition unit 24 to each substation 4 via the DC feeding voltage control device 2 and the network 3.

[0041] 9 is a flowchart showing a process in which the DC feeder voltage calculation device 1 according to the first embodiment creates an inference model. In the DC feeder voltage calculation device 1, the learning data creation unit 21 acquires simulation model information from the simulation model information storage unit 11 (step S101), and acquires running history information from the running history information storage unit 12 (step S102). The learning data creation unit 21 performs a simulation using the acquired simulation model information and running history information to create learning data (step S103). The learning data creation unit 21 stores the created learning data in the learning data storage unit 13 (step S104). The inference model creation unit 22 acquires learning data from the learning data storage unit 13 (step S105). The inference model creation unit 22 creates an inference model by learning using the learning data (step S106). The inference model creation unit 22 stores the created inference model in the inference model storage unit 14 (step S107).

[0042] 10 is a flowchart showing a process of calculating a control content by the DC feeder voltage calculation device 1 according to the first embodiment. In the DC feeder voltage calculation device 1, the control content calculation unit 23 checks whether or not the train 6 has started operating (step S201). The control content calculation unit 23 may check whether or not the train 6 has started operating by making an inquiry to a train operation management system (not shown) that manages the operation of the train 6, or may check whether or not running history information from the train 6 has been acquired by the information acquisition unit 24 or the communication unit 30. If the train 6 has not started operating (step S201: No), the control content calculation unit 23 ends the process. If the train 6 has started operating (step S201: Yes), the control content calculation unit 23 acquires an inference model from the inference model storage unit 14 (step S202).

[0043] The information acquisition unit 24 checks whether the operation of the train 6 has ended (step S203). The information acquisition unit 24 can check whether the operation of the train 6 has ended in the same manner as in step S201 by the control content calculation unit 23 described above. If the operation of the train 6 has ended (step S203: Yes), the information acquisition unit 24 ends the processing. If the operation of the train 6 has not ended (step S203: No), the information acquisition unit 24 acquires substation information from each substation 4 via the communication unit 30 (step S204). The information acquisition unit 24 stores the acquired substation information in the substation information storage unit 15 (step S205).

[0044] The control content calculation unit 23 acquires substation information from the substation information storage unit 15, and creates feature data for an inference model from the acquired substation information (step S206). The control content calculation unit 23 acquires an inference model from the inference model storage unit 14, and calculates control content for each substation 4 using the feature data for the created inference model and the acquired inference model (step S207). The control content calculation unit 23 stores the calculated control content in the control content storage unit 16 (step S208). The DC feeder voltage calculation device 1 returns to step S203 and performs the same processing as described above.

[0045] 11 is a flowchart showing a process in which the DC feeding voltage calculation device 1 according to the first embodiment outputs control content to each substation 4. In the DC feeding voltage calculation device 1, the information acquisition unit 24 checks whether or not the train 6 has started operating (step S301). The information acquisition unit 24 can check whether or not the train 6 has started operating in the same manner as in step S201 by the control content calculation unit 23 described above. If the train 6 has not started operating (step S301: No), the information acquisition unit 24 ends the process. If the train 6 has started operating (step S301: Yes), the information acquisition unit 24 checks whether or not the train 6 has ended operating (step S302).

[0046] The information acquisition unit 24 can check whether the operation of the train 6 has ended in the same manner as in step S203 by the control content calculation unit 23 described above. If the operation of the train 6 has ended (step S302: Yes), the information acquisition unit 24 ends the process. If the operation of the train 6 has not ended (step S302: No), the information acquisition unit 24 acquires the control content from the control content storage unit 16 (step S303). The information acquisition unit 24 outputs the acquired control content to each substation 4 via the communication unit 30 (step S304). Note that the processing unit 20 of the DC feeding voltage calculation device 1 can asynchronously perform the process shown in the flowchart of FIG. 10 and the process shown in the flowchart of FIG. 11.

[0047] FIG. 12 is a flowchart showing a process in which the DC feeder voltage calculation device 1 according to the first embodiment creates learning data. In the DC feeder voltage calculation device 1, the learning data creation unit 21 checks whether the data of time slice 1 and the data of time slice 2 are stored in the running history information storage unit 12 (step S401). Time slice 1 corresponds to, for example, the "time slice at 00:00:00" for the learning data shown in the upper part of FIG. 6, and time slice 2 corresponds to, for example, the "time slice at 00:00:01" for the learning data shown in the upper part of FIG. 6. If the data of time slice 1 and the data of time slice 2 are stored in the running history information storage unit 12 (step S401: Yes), the learning data creation unit 21 acquires the data of time slice 1 and the data of time slice 2 from the running history information storage unit 12 (step S402). The learning data creation unit 21 searches for optimal control content for each substation 4 in the time slice 2 (step S403). In the first embodiment, the optimal control content for each substation 4 refers to the optimal no-load voltage for each substation 4, as shown in Fig. 6 etc. The optimal control content can be searched for, for example, by using a method such as Bayesian optimization, but is not limited to this.

[0048] The learning data creation unit 21 determines whether the simulation has been performed a specified number of times (step S404). The specified number of times is the upper limit of the number of simulations described above. If the simulation has not been performed the specified number of times (step S404: No), the learning data creation unit 21 randomly determines control content for each substation 4 in the time slice 1 (step S405). The learning data creation unit 21 executes the simulation in the time slice 1 (step S406) and creates feature data (step S407). The learning data creation unit 21 stores, as learning data, pairs of the feature data, which are explanatory variables, and the optimal control content, which are objective variables, in the learning data storage unit 13 (step S408). After step S408, the learning data creation unit 21 returns to the processing of step S404. If the learning data creation unit 21 has not performed the simulation the specified number of times (step S404: No), in step S405 it changes the control content for each substation 4 in the time slice 1 and repeats the above-mentioned process. Note that, as described above, if the number of combinations of all candidate voltage values ​​is equal to or less than the upper limit, the learning data creation unit 21 may determine step S404: Yes when simulations have been performed for all candidate voltage value combinations.

[0049] When the learning data creation unit 21 has performed the simulation the specified number of times (step S404: Yes), it updates the time slice 1 and the time slice 2 to the next time (step S409). Updating the time slice 1 and the time slice 2 to the next time means, for example, updating the combination of the time slice 1 and the time slice 2 from the "00:00:00 time slice" and the "00:00:01 time slice" shown in the upper part of FIG. 6 to the "00:00:01 time slice" and the "00:00:02 time slice" shown in the middle part of FIG. 6. The number of learning data corresponding to each combination of the time slices shown in FIG. 6 corresponds to the specified number of times in step S404. After step S409, the learning data creation unit 21 returns to the processing of step S401. If the driving history information storage unit 12 has data on time slice 1 and data on time slice 2 (step S401: Yes), the learning data creation unit 21 changes the combination of data on time slice 1 and data on time slice 2 and repeats the above-mentioned process. If the driving history information storage unit 12 does not have data on time slice 1 and data on time slice 2 (step S401: No), the learning data creation unit 21 ends the process.

[0050] Next, a description will be given of the hardware configuration of the DC feeder voltage calculation device 1. In the DC feeder voltage calculation device 1, the storage unit 10 is a memory. The processing unit 20 and the communication unit 30 are realized by a processing circuit. The processing circuit may be a processor and memory that executes a program stored in the memory, or may be dedicated hardware.

[0051] FIG. 13 is a diagram illustrating an example in which a processing circuit 90 that realizes the DC feeder voltage calculation device 1 according to the first embodiment is configured with a processor 91 and a memory 92. When the processing circuit 90 is configured with the processor 91 and the memory 92, each function of the processing circuit 90 of the DC feeder voltage calculation device 1 is realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in the memory 92. In the processing circuit 90, each function is realized by the processor 91 reading and executing the program stored in the memory 92. In other words, the processing circuit 90 includes the memory 92 for storing a program that results in the processing of the DC feeder voltage calculation device 1 being executed. It can also be said that these programs cause a computer to execute the procedures and methods of the DC feeder voltage calculation device 1.

[0052] The above program can also be said to be a program that causes the DC feeder voltage calculation device 1 to execute the following steps: a communication step of acquiring substation information including the substation voltage applied from the substation 4 to the feeder 5 in the DC electrified section of the electric railway and the directional current flowing from the substation 4 to the feeder 5 in each direction in the DC electrified section, as well as running history information indicating the position and power status of the train 6; and a processing step of calculating the control content for the substation 4 based on the substation information and an inference model created by learning using learning data created from a simulation using the running history information.

[0053] Here, the processor 91 may be a CPU (Central Processing Unit), a processing device, an arithmetic device, a microprocessor, a microcomputer, or a DSP (Digital Signal Processor), etc. Furthermore, the memory 92 may be, for example, a non-volatile or volatile semiconductor memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable ROM), or an EEPROM (registered trademark) (Electrically EPROM), a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, or a DVD (Digital Versatile Disc).

[0054] Fig. 14 is a diagram showing an example in which the processing circuit 93 that realizes the DC feed voltage calculation device 1 according to the first embodiment is configured with dedicated hardware. When the processing circuit 93 is configured with dedicated hardware, the processing circuit 93 shown in Fig. 14 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination of these. Each function of the DC feed voltage calculation device 1 may be realized by the processing circuit 93 separately for each function, or all functions may be realized by the processing circuit 93 together.

[0055] It should be noted that some of the functions of the DC feeder voltage calculation device 1 may be realized by dedicated hardware and some by software or firmware. In this way, the processing circuit can realize each of the above-mentioned functions by dedicated hardware, software, firmware, or a combination of these.

[0056] As described above, according to this embodiment, the DC feeding voltage calculation device 1 performs offline pre-processing by performing a simulation using the running history information of the train 6 up to the previous day and the simulation model information to create learning data, and then creates an inference model that learns the relationship between the substation information and the optimal substation voltage through learning using the learning data. Next, as real-time processing, the DC feeding voltage calculation device 1 uses the current substation information acquired from each substation 4 and the inference model to calculate the control details for each substation 4, or in the first embodiment, the optimal no-load voltage for each substation 4. This allows the DC feeding voltage calculation device 1 to reduce energy loss in the feeder 5 in the DC electrified section.

[0057] Furthermore, even when the control period is short, the DC feeding voltage calculation device 1 can calculate the optimal control content for each substation 4 in a short calculation time, and can calculate the optimal control content for each substation 4 that reflects the current operation status of the train 6. Furthermore, by using current substation information acquired from each substation 4 as real-time processing, the DC feeding voltage calculation device 1 can control the voltage of the substation 4 so that there is a large interchange of regenerative power and loss in the feeder line 5 is reduced, thereby realizing energy-saving train operation.

[0058] Embodiment 2 In the first embodiment, the DC feeding voltage calculation device 1 calculates the control content for each time slice individually. In the second embodiment, a case will be described in which the DC feeding voltage calculation device 1a calculates the control content for a specified time period.

[0059] FIG. 15 is a diagram showing an example of the configuration of a DC feeding voltage control system 7a according to the second embodiment. The DC feeding voltage control system 7a includes a DC feeding voltage calculation device 1a and a DC feeding voltage control device 2. In FIG. 15, the DC feeding voltage control device 2, network 3, and substations 4-1 to 4-m are the same as the DC feeding voltage control device 2, network 3, and substations 4-1 to 4-m shown in FIG. 2. The DC feeding voltage calculation device 1a includes a storage unit 10a, a processing unit 20a, and a communication unit 30. The storage unit 10a is configured by replacing the learning data storage unit 13 and the control content storage unit 16 with a learning data storage unit 13a and a control content storage unit 16a in the storage unit 10 of the first embodiment shown in FIG. 2, and further adding a substation statistical information storage unit 17. The processing unit 20a is obtained by replacing the learning data creation unit 21 and the control content calculation unit 23 with a learning data creation unit 21a and a control content calculation unit 23a in the processing unit 20 of the first embodiment shown in FIG. 2, and further adding a substation statistical information calculation unit 25.

[0060] Hereinafter, a specific example will be described in which the DC feeder voltage calculation device 1a performs control in 10-minute control units; however, the control unit of the DC feeder voltage calculation device 1a is not limited to 10 minutes. The control unit of the DC feeder voltage calculation device 1a may be 1 minute or 1 hour. For example, in the first embodiment, the control unit is 1 second, so the running history information required for one control is 1 second of running history information in a 1-second time section. However, if the control unit is 1 minute, the running history information required for one control is 60 seconds of running history information in a 1-second time section. Similarly, if the control unit is 10 minutes, the running history information required for one control is 600 seconds of running history information in a 1-second time section. If the control unit is 1 hour, the running history information required for one control is 3600 seconds of running history information in a 1-second time section. Note that the DC feeder voltage calculation device 1a can also change the control unit as appropriate in cases such as when an accident occurs with the train 6 or when the weather is bad.

[0061] Similar to the learning data creation unit 21 in the first embodiment, the learning data creation unit 21a performs a simulation using simulation model information and running history information to create learning data in which the average value of the substation voltage, which is the voltage at each substation 4 in a specified period, and the average value of the direction-specific current in the specified period are explanatory variables. The learning data creation unit 21a stores the created learning data in the learning data storage unit 13a. FIG. 16 is a diagram showing an example of learning data stored in the learning data storage unit 13a of the DC feeder voltage calculation device 1a according to the second embodiment. In the second embodiment, the learning data includes explanatory variables indicating the average value of the substation voltage, which is the voltage at each substation 4 in a specified period, and the average value of the direction-specific current in the specified period, and a target variable indicating an optimal no-load voltage to be applied from each substation 4 to the feeder 5. If the explanatory variables of the learning data shown in Fig. 16 are, for example, explanatory variables calculated over a time section from 00:00:00 to 00:09:59, which is statistical information, then the objective variables of the learning data shown in Fig. 16 will be objective variables calculated over a time section from 00:10:00 to 00:19:59, which is statistical information. Note that, in addition to the average value, the maximum value, minimum value, etc. over the control unit period may also be used as the statistical information.

[0062] The DC feeder voltage calculation device 1a of the second embodiment can perform the same processing as that of the DC feeder voltage calculation device 1 of the first embodiment shown in the flowchart of Fig. 12 by replacing time cross section 1 with time cross section group 1 and time cross section 2 with time cross section group 2. For example, in the first embodiment, if time cross section 1 is set to 00:00:00 and time cross section 2 is set to 00:00:01, in the second embodiment, time cross section group 1 is set to 00:00:00 to 00:09:59 and time cross section group 2 is set to 00:10:00 to 00:19:59.

[0063] The substation statistical information calculation unit 25 acquires substation information for a specified period, here a period of 10 minutes which is the control unit, from the substation information storage unit 15, and calculates an average value of the substation information as statistical information of the substation information. FIG. 17 is a diagram showing an example of substation statistical information stored in the substation statistical information storage unit 17 of the DC feeder voltage calculation device 1a according to the second embodiment. Specifically, the substation statistical information calculation unit 25 calculates an average value of the substation voltage for each substation 4, and calculates an average value of the direction-specific current for each direction. The substation statistical information calculation unit 25 stores the calculated statistical information of the substation information as shown in FIG. 17 in the substation statistical information storage unit 17.

[0064] In the DC feeder voltage calculation device 1a of the second embodiment, for example, in the process of creating learning data by the DC feeder voltage calculation device 1 of the first embodiment shown in the flowchart of Fig. 10, after the information acquisition unit 24 stores the substation information in the substation information storage unit 15 in step S205, the substation statistical information calculation unit 25 calculates statistical information of the substation information and stores the calculated statistical information of the substation information in the substation statistical information storage unit 17. In the second embodiment, in step S206, the control content calculation unit 23a acquires the statistical information of the substation information from the substation statistical information storage unit 17 and creates feature amount data for an inference model from the acquired statistical information of the substation information.

[0065] The control content calculation unit 23a stores the control content calculated for a specified period, here a 10-minute control unit, in the control content storage unit 16a. FIG. 18 is a diagram illustrating an example of control content stored in the control content storage unit 16a of the DC feeder voltage calculation device 1a according to the second embodiment. The control content calculation unit 23a calculates the control content to be used for the next control unit period just before the start of the next control unit, or one second before the control start time in the example of FIG. 18. In the second embodiment, the control content is the optimum no-load voltage to be applied from the substation 4 to the feeder line 5, as in the first embodiment. In the example of FIG. 18, the control end time of the control content in the first row is 2024-10-01 00:09:59, and the control start time of the control content in the next second row is 2024-10-01 00:10:00. For convenience of description, it appears as if there is no one-second period without control when transitioning from control according to the control content in the first row to control according to the control content in the second row. However, in reality, each substation 4 continues control according to the previous control content until the next control content is instructed, so there is no period when control content is switched and no control is performed.

[0066] In the DC feeding voltage calculation device 1a of the second embodiment, for example, in the process of outputting control content to each substation 4 by the DC feeding voltage calculation device 1 of the first embodiment shown in the flowchart of Fig. 11, when the information acquisition unit 24 acquires the control content from the control content storage unit 16a in step S303, the information acquisition unit 24 checks the control start time shown in Fig. 18. The information acquisition unit 24 acquires the control content at a timing when the control content of each substation 4 can be changed at the control start time.

[0067] As described above, according to this embodiment, the DC feeding voltage calculation device 1a uses substation statistical information such as the average values ​​of substation voltages and direction-specific currents for a specified period to calculate control details for each substation 4, and in the second embodiment, the optimum no-load voltage for each substation 4. Even in this case, the DC feeding voltage calculation device 1a can obtain the same effects as in the first embodiment. Furthermore, compared to the DC feeding voltage calculation device 1 of the first embodiment, the DC feeding voltage calculation device 1a requires more processing to calculate substation statistical information, but can reduce the number of times the control details are calculated and the number of times the control details are output to each substation 4, thereby reducing the overall processing load.

[0068] Embodiment 3 In the third embodiment, a case will be described in which the DC feeder voltage calculation device 1b creates learning data by performing a simulation using running history information according to various conditions. Although the third embodiment can be applied to the first and second embodiments, the following description will be made taking as an example a case in which it is applied to the second embodiment.

[0069] FIG. 19 is a diagram illustrating an example of the configuration of a DC feeding voltage control system 7b according to a third embodiment. The DC feeding voltage control system 7b includes a DC feeding voltage calculation device 1b and a DC feeding voltage control device 2. In FIG. 19, the DC feeding voltage control device 2, network 3, and substations 4-1 to 4-m are the same as the DC feeding voltage control device 2, network 3, and substations 4-1 to 4-m shown in FIG. 15. The DC feeding voltage calculation device 1b includes a storage unit 10b, a processing unit 20b, and a communication unit 30. The storage unit 10b is configured by replacing the learning data storage unit 13a with a learning data storage unit 13b in the storage unit 10a of the second embodiment shown in FIG. 15, and further adding a timetable information storage unit 18 and a condition information storage unit 19. The processing unit 20b is configured by replacing the learning data creating unit 21a with a learning data creating unit 21b in the processing unit 20a of the second embodiment shown in FIG.

[0070] The timetable information storage unit 18 stores timetable information indicating the operation plan of the train 6. Although not shown in the figure, the timetable information storage unit 18 includes information such as the timetable revision date, the timetable type for weekdays, holidays, or days of the week, the train number, the departure station, the departure time, the arrival station, and the arrival time.

[0071] The condition information storage unit 19 stores various conditions used when the learning data creation unit 21b extracts running history information from the running history information storage unit 12 to create learning data by performing a simulation. FIG. 20 is a diagram illustrating an example of conditions stored in the condition information storage unit 19 of the DC feeder voltage calculation device 1b according to the third embodiment. The conditions include, for example, a day type (e.g., weekday or holiday), a day type (e.g., Monday through Sunday), a season type (e.g., spring, summer, autumn, or winter), a weather type (e.g., sunny), a time period during which the train 6 operates, and a delay type indicating a delay status of the train 6. In FIG. 20, each row contains a different combination of various conditions. Therefore, when creating learning data based on a certain combination of conditions, the learning data creation unit 21b only needs to extract one row of condition information from the condition information storage unit 19. Note that the learning data creation unit 21b does not need to use all of the conditions stored in the condition information storage unit 19. It is possible to use a combination of several conditions, or it is also possible to use at least one condition. For example, the learning data storage unit 13b can use only the day type condition.

[0072] The learning data creation unit 21b extracts driving history information that meets specified conditions from the driving history information storage unit 12, and creates learning data by performing a simulation using the simulation model information and the extracted driving history information. The learning data creation unit 21b stores the created learning data in the learning data storage unit 13b. FIG. 21 is a diagram showing an example of learning data stored in the learning data storage unit 13b of the DC feeder voltage calculation device 1b according to the third embodiment. The learning data in the third embodiment further includes, as explanatory variables, at least one condition from among day type, day of the week type, season type, weather type, time period type, and delay type, in addition to the learning data described in the second embodiment. FIG. 21 shows an example in which day type and delay type are added as explanatory variables to the learning data of the second embodiment shown in FIG.

[0073] For example, in the process of creating learning data by the DC feeder voltage calculation device 1 of the first embodiment shown in the flowchart of Fig. 12, the learning data creation unit 21b of the third embodiment acquires timetable information from the timetable information storage unit 18 and acquires information on the conditions to be used from the condition information storage unit 19 before step S401. In steps S401 and S402 shown in the flowchart of Fig. 20, the learning data creation unit 21b extracts driving history information that satisfies the condition information from the driving history information storage unit 12, performs learning, creates learning data, and stores the learning data in the learning data storage unit 13b. The learning data creation unit 21b performs the process shown in the flowchart of Fig. 12 for each piece of condition information, i.e., for each piece of condition information shown in each row of Fig. 20, creates learning data, and stores the learning data in the learning data storage unit 13b.

[0074] In the DC feeder voltage calculation device 1b, the inference model creation unit 22 may create only one inference model corresponding to all the condition information, or may create an inference model for each piece of condition information, since the learning data stored in the learning data storage unit 13b also includes conditions.

[0075] Furthermore, the control content calculation unit 23a of the third embodiment may add current condition information to the feature amount data when creating the feature amount data in step S206 in the process of calculating the control content by the DC feeding voltage calculation device 1 of the first embodiment shown in the flowchart of Fig. 10. The control content calculation unit 23a may acquire information about the current delay state of the train 6 from, but is not limited to, the aforementioned train traffic control system (not shown). The control content calculation unit 23a may acquire information about the current delay state of the train 6 from a source other than the train traffic control system as long as it can acquire the information. Note that in the third embodiment, the control content is the optimum no-load voltage to be applied from the substation 4 to the feeder 5, as in the first and second embodiments.

[0076] The types of condition information are not limited to those shown in Fig. 20 etc. The DC feeder voltage calculation device 1b can also use, as conditions, conditions such as passenger flow information on the train 6, congestion information at stations where the train 6 stops, the number of cars indicating the number of cars that make up the train 6, and train type indicating the operating mode of the train 6, such as a local train or an express train.

[0077] As described above, according to this embodiment, the DC feeding voltage calculation device 1b adds conditions such as the type of day and the type of day of the week, compared to the first and second embodiments, creates learning data using only information that matches the conditions, creates an inference model, and calculates control details, and in the third embodiment, the optimal no-load voltage for each substation 4. As a result, when the current operating status of the train 6 matches the above-mentioned conditions, the DC feeding voltage calculation device 1b can calculate control details with higher accuracy than the first and second embodiments, thereby making it possible to further reduce energy loss.

[0078] Embodiment 4 In the first to third embodiments, the control content of the DC feeding voltage calculation devices 1 to 1b for each substation 4 is to instruct the no-load voltage for each substation 4. In the fourth embodiment, a case will be described in which the DC feeding voltage calculation devices 1 to 1b further instruct the substation 4 on the number of parallel-operated power converters to be operated in parallel. Note that the fourth embodiment can be applied to the first to third embodiments, but hereinafter, a case in which it is applied to the first embodiment will be described as an example.

[0079] In the fourth embodiment, the configuration of the DC feeding voltage calculation device 1 is the same as the configuration of the DC feeding voltage calculation device 1 of the first embodiment shown in FIG. 2. For example, in the DC feeding voltage calculation device 1, the control content calculation unit 23 acquires in advance information on the number of power converters provided in each substation 4. The control content calculation unit 23 calculates the number of power converters to be operated in parallel in each substation 4, along with the no-load voltage for each substation 4, and stores the calculated information in the control content storage unit 16 as control content. Although not shown, the control content calculated by the control content calculation unit 23 in the fourth embodiment is the same as the control content stored in the control content storage unit 16 of the first embodiment shown in FIG. 8, with information on the number of power converters to be operated in parallel in each substation 4 added to the right of the column for the no-load voltage for each substation 4. That is, the control content in the fourth embodiment further includes the number of power converters to be operated in parallel among the multiple power converters provided in the substation 4, in addition to the control content described in the first to third embodiments.

[0080] Similarly, although not shown, the learning data created by learning data creation unit 21 in embodiment 4 is obtained by adding information on the number of power converters to be operated in parallel at each substation 4 to the right of the column for the no-load voltage for each substation 4 to the learning data stored in learning data storage unit 13 of embodiment 1 shown in Fig. 4. The learning data in embodiment 4 further includes, as a target variable, the number of power converters to be operated in parallel among the multiple power converters included in substation 4, in addition to the learning data described in embodiments 1 to 3.

[0081] As described above, according to this embodiment, the DC feeding voltage calculation device 1 adds, as control content, information on the number of power converters to be operated in parallel at each substation 4. This allows the DC feeding voltage calculation device 1 to perform more detailed control over each substation 4 compared to the first to third embodiments.

[0082] The configurations shown in the above embodiments are merely examples, and may be combined with other known technologies, or different embodiments may be combined with each other. It is also possible to omit or modify parts of the configurations as long as they do not deviate from the gist of the invention. [Explanation of symbols]

[0083] 1, 1a, 1b DC feeding voltage calculation device, 2 DC feeding voltage control device, 3 Network, 4-1 to 4-m Substation, 5, 5-1, 5-2 Feeder, 6-1 to 6-n Train, 7, 7a, 7b DC feeding voltage control system, 10, 10a, 10b Memory unit, 11 Simulation model information memory unit, 12 Running history information memory unit, 13, 13a, 13b Learning data memory unit, 14 Inference model memory unit, 15 Substation information memory unit, 16, 16a Control content memory unit, 17 Substation statistical information memory unit, 18 Diagram information memory unit, 19 Condition information memory unit, 20, 20a, 20b Processing unit, 21, 21a, 21b Learning data creation unit, 22 Inference model creation unit, 23, 23a Control content calculation unit, 24 Information acquisition unit, 25 Substation statistical information calculation unit, 30 communication unit, 41 substation voltage control unit, 42 substation information acquisition unit, 51 voltage control unit, 52 voltage / current acquisition unit, 53 voltage sensor, 54-1 to 54-4 current sensors, 90, 93 processing circuit, 91 processor, 92 memory.

Claims

1. a communication unit that acquires substation information including substation voltages applied from substations to feeders in DC electrified sections of electric railways and directional currents flowing from the substations to feeders in each direction in the DC electrified sections, as well as running history information indicating train positions and power conditions; a processing unit that calculates control details for the substation based on an inference model created by learning using learning data created from a simulation using the driving history information and the substation information; A DC feeding voltage calculation device comprising:

2. The processing unit a simulation using random candidate voltage values ​​for the substation, using simulation model information including a train model including information for controlling the amount of regenerative power throttling in the train located in the DC electrified section, a substation model including control information for the substation voltage, and a power grid model including location information of the substation, and the running history information indicating the locations and power states of a plurality of trains that have run in the DC electrified section on or before the previous day; 2. The DC feeding voltage calculation device according to claim 1 .

3. The processing unit creating the learning data that is a combination of explanatory variables including the substation voltage of each substation at each time cross section and the directional current, which is a current value of a feeder line for each direction of the train, and a target variable corresponding to the control content for each substation to suppress the power used for operating the train; 3. The DC feeding voltage calculation device according to claim 2.

4. The processing unit creating the inference model by learning using the learning data; Calculating the control content using the current substation information including the substation voltage detected by a voltage sensor in the substation and the direction-specific current detected by a direction-specific current sensor in the substation for the train, and the inference model.

3. The DC feeding voltage calculation device according to claim 2.

5. The explanatory variables are the substation voltage and the direction-specific current, The objective variable and the control content are an optimal no-load voltage applied from the substation to the feeder line.

4. The DC feeding voltage calculation device according to claim 3.

6. the explanatory variables are an average value of the substation voltage in a specified period and an average value of the direction-specific current in the specified period; The objective variable and the control content are an optimal no-load voltage applied from the substation to the feeder line.

4. The DC feeding voltage calculation device according to claim 3.

7. The explanatory variables further include at least one condition of a day type, a day of the week type, a season type, a weather type, a time period type, and a delay type.

6. The DC feeding voltage calculation device according to claim 5.

8. The objective variable and the control content further include the number of parallel-operated power converters to be operated in parallel among the plurality of power converters included in the substation.

6. The DC feeding voltage calculation device according to claim 5.

9. The simulation model information further includes at least one of a station auxiliary power supply model including control information for a station auxiliary power supply installed in a station at which the specified type of train stops, and a power storage device model including control information for a power storage device that stores regenerative power generated by the train.

3. The DC feeding voltage calculation device according to claim 2.

10. the processing unit performs a specified number of simulations by changing the combination of the candidate voltage values ​​for each substation.

3. The DC feeding voltage calculation device according to claim 2.

11. the candidate voltage value for the substation is one or more specific voltage values ​​or a voltage value selected from a voltage range that can be set as the candidate voltage value; The DC feeding voltage calculation device according to claim 10.

12. The DC feeder voltage calculation device according to any one of claims 1 to 11, a DC feeding voltage control device that controls a substation voltage applied from a substation to a feeder line in a DC electrified section of an electric railway based on the control content calculated by the DC feeding voltage calculation device; A DC feeding voltage control system comprising:

13. a communication step of acquiring substation information including a substation voltage applied from a substation to a feeder in a DC electrified section of an electric railway and a directional current flowing from the substation to a feeder in each direction in the DC electrified section, and running history information indicating the position and power status of a train; a processing step of calculating a control content for the substation based on an inference model created by learning using learning data created from a simulation using the driving history information and the substation information; A DC feeder voltage calculation program characterized by causing a computer to execute the above.

14. A DC feeder voltage calculation method executed by a computer, comprising: a communication step of acquiring substation information including a substation voltage applied from a substation to a feeder in a DC electrified section of an electric railway and a directional current flowing from the substation to a feeder in each direction in the DC electrified section, and running history information indicating the position and power status of a train; a processing step of calculating a control content for the substation based on an inference model created by learning using learning data created from a simulation using the driving history information and the substation information; A DC feeder voltage calculation method comprising:

Images