Map learning method and map learning device

Abstract

To accelerate control by reducing the time for retrieving indices that satisfy request values, respectively.SOLUTION: In a power control system, each of one or two or more charge / discharge elements each receiving each of index values transmitted by broadcasting, autonomously controls its own charge / discharge power based on each index value, and a controller controls charge / discharge power of an entire group, using a map indicating an estimated value of charge / discharge power of the entire group, which is estimated as a response of the entire group including one or two or more charge / discharge elements to each index value. A map learning method includes: calculating first index values corresponding to request values PTi requested to the entire group from the map (S01); transmitting first index values Si+1 by broadcasting (S02); measuring first output values Pi+1 of the charge / discharge power of the entire group with respect to the first index values Si+1 (S03); and leaning the map using the request values PTi and the first output values Pi+1 (S07).SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] This invention relates to a map learning method and a map learning device. [Background technology]

[0002] Patent Document 1 discloses a power system comprising multiple power conditioners and a centralized control device for managing these multiple power conditioners. The centralized control device detects the power to be adjusted and transmits indicators to the multiple power conditioners for controlling the individual output power of each power conditioner so that the power to be adjusted becomes the target power. Based on the indicators transmitted from the centralized control device, the multiple power conditioners calculate their individual target power by solving a constrained optimization problem and control their individual output power. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] International Publication No. 2017 / 150376 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] The centralized management device described in Patent Document 1 updates the index by adding a value obtained by multiplying the difference in power between the target power and the power to be adjusted by a constant to the current index, and repeats updating and transmitting the index until the difference in power becomes zero. Therefore, if the difference between the current index and the index that satisfies the target power is large, the time required to search for the index becomes longer, and the control responsiveness deteriorates.

[0005] The objective of this invention is to speed up control by reducing the time required to search for indicators that satisfy the control objective. [Means for solving the problem]

[0006] To solve the above-mentioned problems, one aspect of the present invention is a map learning method for a power control system in which each of one or more charge / discharge elements that receives a broadcasted index value autonomously controls its own charge / discharge power based on the index value, and a controller controls the charge / discharge power of the entire group using a map that shows an assumed value of the charge / discharge power of the entire group, which is assumed to be the response of the entire group including one or more charge / discharge elements to the index value. The map learning method calculates a first index value corresponding to the requested value of the charge / discharge power required for the entire group from the map, broadcasts the first index value, measures a first output value of the charge / discharge power of the entire group in relation to the first index value, and learns the map using the requested value and the first output value. [Effects of the Invention]

[0007] According to one aspect of the present invention, control can be sped up by reducing the time required to search for an indicator that satisfies the required value. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 is a block diagram showing the configuration of the controller 1 and electric vehicles (EV1, EV2, EV3, ..., hereafter collectively referred to as EVs) that constitute the power control system according to the first to third embodiments. [Figure 2] Figure 2 shows a map learning method according to the first to third embodiments, where (a) is a flowchart showing the map learning method performed by the controller 1 in Figure 1, and (b) is a flowchart showing the charge / discharge control method performed by each electric vehicle (EV). [Figure 3] Figure 3 is a flowchart (part 1) illustrating an example of the specific steps of the map learning step shown in step S07 of Figure 2, which is a map learning method according to the second and third embodiments. [Figure 4] Figure 4 is a flowchart (part 2) illustrating an example of the specific steps of the map learning step shown in step S07 of Figure 2, which is a map learning method according to the second and third embodiments. [Figure 5A]Figure 5A is a line graph (an example of a map) to aid in explaining the map learning steps shown in Figures 3 and 4. [Figure 5B] Figure 5B is a line graph (an example of a map) to aid in explaining the map learning steps shown in Figures 3 and 4, and is a diagram showing a learning example where the output power (Pi+1) is smaller than the required power (PTi) (Part 1). [Figure 5C] Figure 5C is a line graph (an example of a map) to aid in explaining the map learning steps shown in Figures 3 and 4, and is a diagram showing a learning example where the output power (Pi+1) is smaller than the required power (PTi) (part 2). [Figure 6A] Figure 6A is a line graph (an example of a map) to aid in explaining the map learning steps shown in Figures 3 and 4, and is a diagram showing a learning example where the output power (Pi+1) is greater than the required power (PTi) (Part 1). [Figure 6B] Figure 6B is a line graph (an example of a map) to aid in explaining the map learning steps shown in Figures 3 and 4, and is a diagram showing a learning example where the output power (Pi+1) is greater than the required power (PTi) (part 2). [Modes for carrying out the invention]

[0009] (First Embodiment) Embodiments and their modified forms will be described below with reference to the drawings. In the drawings, identical parts are denoted by the same reference numerals and their descriptions are omitted.

[0010] Referring to Figure 1, the configuration of the controller 1 (an example of a map learning device) and electric vehicles (EV1, EV2, EV3, ..., hereafter collectively referred to as EVs) that constitute the power control system according to the first embodiment will be described. The power control system is an autonomous distributed control system in which the controller 1 broadcasts a signal indicating an index value to one or more electric vehicles EV, and each of the one or more electric vehicles EV that receives the signal indicating the index value autonomously controls its own charging and discharging power based on the index value. One or more electric vehicles EV form one group 2.

[0011] Each electric vehicle (EV) includes, for example, a battery 22 for storing electrical energy and a power adjustment device 21 that controls the charging and discharging power of the battery 22 based on an indicator value. The power adjustment device 21 includes, for example, a state acquisition unit 23 that acquires information indicating its own state, including the current SOC (State of Charge) of the battery 22, a target SOC after charging and discharging, and the planned start and end times of charging and discharging; an indicator value receiving unit 27 that receives an indicator value broadcast by the controller 1; a charging and discharging power calculation unit 26 that calculates a target value for charging and discharging power based on the received indicator value and its own state; and an output control unit 25 that controls the charging and discharging power of the battery 22 in accordance with the calculated target value. The above-described configuration of the electric vehicle (EV) is an example, and each electric vehicle (EV) may have other configurations that autonomously and distributedly control its own charging and discharging power based on an indicator value. Furthermore, the electric vehicle (EV) is an example of a charging and discharging element, and charging and discharging elements include any equipment and devices other than electric vehicles (EVs) that are capable of charging or discharging at least one of the two.

[0012] Controller 1 may have a memory 15 that stores a map showing the expected value of the charge / discharge power of the entire group 2 (hereinafter sometimes referred to as "expected power"), which is assumed as the group 2's response to the index value. Specifically, "expected power" is the power assumed to be the sum of the power that each electric vehicle EV in group 2 charges or discharges, i.e., the power that it outputs, when the index value is transmitted to the electric vehicles EVs included in group 2. The map is data that shows the relative relationship between the index value and the expected power. Instead of memory 15, the map may be stored in another storage device, such as a cloud server.

[0013] Controller 1 further includes an index value calculation unit 12 that calculates an index value corresponding to the requested charge / discharge power (hereinafter sometimes referred to as "requested power") required for the entire group 2 from the map, a power measurement unit 11 that measures the requested power, and a broadcast transmission unit 13 that broadcasts the calculated index value. The index value broadcast is the index value obtained by applying the requested power to the assumed power on the map. The requested power is, for example, the charge / discharge power for the entire group 2 required from the power system 3. In addition to the requested power, the power measurement unit 11 also measures the output value of the charge / discharge power for the entire group 2 (hereinafter sometimes referred to as "output power"), which is the response of the entire group 2 to the broadcasted index value. Specifically, "output power" is the sum of the power actually charged or discharged by each electric vehicle EV in group 2 that received the index value, i.e., the power output. The output power is output to the power system 3 via the power measurement unit 11. In other words, when the entire group 2 discharges, the total amount of discharged power for the entire group 2 is supplied from group 2 to the power system 3. On the other hand, when Group 2 is charged as a whole, the total amount of power for Group 2 is supplied to Group 2 from power grid 3.

[0014] Power System 3 includes smart grids, smart communities, and microgrids that manage energy supply from the source to end consumption via a communication network within a limited area such as a business or factory, as well as energy management systems such as xEMS, EnMS, and EMS. xEMS includes BEMS (Building Energy Management System), HEMS (Home Energy Management System), MEMS (Condominium Energy Management System), FEMS (Factory Energy Management System), and CEMS (Community Energy Management System).

[0015] Controller 1 controls the charging and discharging power of the entire group 2 using a map. Specifically, Controller 1 searches the map for an index value that satisfies the required power. For example, Controller 1 updates the index value by adding a value obtained by multiplying the difference in power between the required power and the output power by a constant to the index value previously transmitted, and then transmits the updated index value. Controller 1 searches the map for an index value that satisfies the required power by repeatedly updating the index value and transmitting the index value alternately until the difference in power becomes zero, that is, until the required power and the output power match. Further details can be found in the disclosure in International Publication No. 2017 / 150376.

[0016] Controller 1 further includes a map learning unit 14 that learns a map using the requested power and output power. In the first embodiment, the specific learning method by the map learning unit 14 is not particularly limited, and various methods can be used to improve the responsiveness of power control for the entire group 2. Specifically, the map learning unit 14 learns by correcting the map so that the assumed power (=requested power) of the map approaches the actual output power. For example, multiple different index values ​​may be transmitted broadcastly, the output power for each index value may be measured, and an approximation curve of these measured values ​​may be obtained. On the other hand, embodiments specifying a particular learning method will be described in the second and third embodiments.

[0017] Referring to Figure 2, the map learning method performed by the controller 1 in Figure 1 and the charge / discharge control method performed by the power adjustment device 21 of each electric vehicle (EV) will be explained. The flowchart shown in Figure 2 may be executed repeatedly at a predetermined learning cycle. In step S01, the power measurement unit 11 measures the current output power (P i ), and required power (P Ti ) measures. Controller 1 measures the current index value (S i ) obtain information indicating the current index value (S i ) corresponds to the index value calculated in step S02 of the previous control cycle (i) and transmitted in step S03. Current output power (P i This corresponds to the output power measured in step S06 of the previous control cycle (i).

[0018] Proceed to step S02. The index value calculation unit 12 calculates an index value (S i ) corresponding to the required power (P Ti ) from the map (M i+1 ). Proceed to step S03. The broadcast transmission unit 13 broadcasts a signal indicating the calculated index value (S i+1 ) to all electric vehicles EV belonging to group 2.

[0019] Proceed to step S04. The index value reception unit 27 of each electric vehicle EV receives a signal indicating the broadcast transmitted index value (S i+1 ). Proceed to step S05. The power adjustment device 21 of each electric vehicle EV autonomously controls its own charge / discharge power based on the index value (S i+1 ). Specifically, the state acquisition unit 23 acquires information indicating its own state, the charge / discharge power calculation unit 26 calculates a target value of its own charge / discharge power based on an optimization problem using the index value and its own state. Then, the output control unit 25 controls the charge / discharge power of the battery 22 in accordance with the calculated target value. Details of the charge / discharge control of each electric vehicle EV can refer to the control of individual output power in International Publication No. 2017 / 150376.

[0020] Proceed to step S06. The power measurement unit 11 of the controller 1 measures the output power (P i+1 ) which is the response of the entire group 2 to the index value (S i+1 ) broadcast in step S03. Proceed to step S07. The map learning unit 14 learns the map (M Ti ) using the required power (P i+1 ) and the output power (P i ), and stores the learned map (M i+1 ) in the memory 15.

[0021] The map learning unit 14 learns the map using the requested power and output power, thereby improving the accuracy of the map. In other words, the error between the assumed power in the map (M) and the actual output power (P) can be reduced. Therefore, by the controller 1 using the learned map to control the charging and discharging power of the entire group 2, the time required to search for an index value that satisfies the requested power is shortened, and the controller 1 can control the charging and discharging power of the entire group 2 with good responsiveness. As a result, the power control of the entire group 2 in a power control system in which an index value is broadcast to one or more electric vehicles (EVs), and each EV that receives the index value autonomously controls its own charging and discharging power is accelerated.

[0022] (Variation 1) The number of electric vehicles (EVs) belonging to Group 2 and their status, such as State of Charge (SOC), vary depending on fluctuating factors including region, season, month, weekday / holiday, day of the week, and time of day. However, if these fluctuating factors are consistent, the number and status of EVs may show the same trend. For example, if power grid 3 is a microgrid that manages energy supply from the source to the end consumption point via a communication network within a limited area such as a workplace or shopping mall, it is expected that the number of EVs belonging to Group 2 will differ significantly depending on whether it is a weekday or holiday. On the other hand, it is also expected that the same trend will be observed even if the day of the week is different, as long as it is a weekday. Thus, even if the indicator values ​​are the same, the total output power of Group 2 will also differ depending on the region, season, month, weekday / holiday, day of the week, or time of day.

[0023] Therefore, the map learning unit 14 may learn different maps depending on the region, season, month, weekday / holiday, day of the week, or time zone. Furthermore, it may learn different maps by combining two or more variable elements (region, season, month, weekday / holiday, day of the week, or time zone). This reduces the time required to search for an index value corresponding to the requested power compared to using a single common map for power control, allowing the controller 1 to control the charging and discharging power of the entire group 2 with good responsiveness. For example, maps for Japan / North America / Europe, spring / summer / autumn / winter, weekday / holiday, commuting time / business time, etc., can be prepared and learned separately in advance. In addition, if there are specific circumstances of particular vehicles belonging to group 2, such as electric vehicles (EVs) belonging to group 2 at a fixed time, electric vehicles (EVs) moving at a fixed time, or multiple electric vehicles (EVs) with fixed priorities, maps may be created individually according to those circumstances.

[0024] (Second Embodiment) The second and third embodiments describe embodiments that specify the detailed procedure of step S07 (map learning step) in Figure 2. Figures 3 and 4 are flowcharts illustrating the map learning method according to the second and third embodiments, and are examples of specific procedures for the map learning step shown in step S07 in Figure 2. In the second and third embodiments, the configuration of the controller 1 and the configuration of each electric vehicle EV are the same as those shown in Figure 1, and the overall configuration of the map learning method and the charge / discharge control method performed by each electric vehicle EV are the same as those shown in Figure 2.

[0025] In the second embodiment, with reference to Figures 3, 4, 5A, and 5B, a map learning method is shown for when the search is conducted in the direction of increasing the index value because the output power is lower than expected. Figure 5A shows the map (M i A line graph is shown as an example of this. The horizontal axis shows the index value (S), and the vertical axis shows the assumed power. The assumed power on the map is shown along with the required power (P). Ti By applying ), the corresponding index value (S i+1 ) can be calculated. Note that the map also shows the minimum value of the index (Smin ) and maximum value (S max ) is set, and the minimum and maximum values ​​of the assumed power (P max ) correspond to each other. In Figure 5A, the map (M i As an example, a line graph is shown connecting the minimum and maximum values ​​of the indicator value and assumed power with a line segment consisting of a curve or a straight line, but the map may take other forms.

[0026] The map learning unit 14 executes the steps shown in Figures 3 and 4. First, in step S701, the required power (P) measured in step S01 is used. Ti ) is the current output power (P) measured in the same step S01. i Determine whether it is greater than (P). Figure 5A shows the case where step S701 is YES. If step S701 is YES, proceed to step S702 and determine whether the output power (P) measured in step S06 is greater than or equal to. i+1 ) is the required power (P) measured in step S01. Ti Determine whether it is greater than ). Figure 5B shows the output power (P i+1 ) is the required power (P Ti This indicates the case where the output power (P) is smaller than the map assumption. If NO in step S702, the output power (P) is smaller than the map assumption. i+1 Because the value is small, controller 1 will search in the direction of increasing the index value. In this case, proceed to step S703, and map learning unit 14 will search the map (M i ) Index value (S i+1 Reduce the power increase rate in the first range which is smaller than ) and map (M i ) Index value (S i+1 By increasing the power growth rate of the second range which is greater than ) the map (M i ) is learned. Through this map learning, the required power (P Ti The corresponding index value can be increased. Therefore, the time spent searching in the direction of increasing the index value is shortened, and control becomes faster.

[0027] More specifically, the map learning unit 14 calculates the index value (S) in step S03. i+1 Output power (P) before broadcasting )i ) The corresponding index value on the map (S i ) from the index value (S i+1 Map to ) (M i The power increase rate (derivative value) of ) is reduced, i.e., the slope of the line graph is reduced, and the index value (S i+1 ) from the maximum value of the index (S max Increase the rate of power increase up to ). At this time, the index value (S i ) from the index value (S i+1 The rate of increase in power up to ) may be uniformly reduced. Index value (S i+1 ) from the maximum value of the index (S max The power increase rate up to ) may be uniformly increased. Map (M i ) The index value above (S i+1 The assumed power corresponding to ) is the output power (P i+1 ) may be matched with the map (M i ) The index value above (S i ) and the maximum value of the index (S max The assumed power values ​​corresponding to ) do not need to be changed. Index value (S i ) and index value (S i+1 The power increase rate may be made continuous in ). Step S703 is performed to obtain the learned map (M shown in Figure 5B) i+1 ) can be obtained.

[0028] The map after training (M) is shown in Figure 5B. i+1 According to the required power (P Ti As the index value corresponding to ), the index value (S i+1 ) a larger index value (S i+2 This allows us to calculate the index value. Therefore, the time spent searching in the direction of increasing the index value is reduced, and control becomes faster.

[0029] Proceeding to step S704, the map learning unit 14 calculates the index value (S) as shown in Figure 5B. i+1 ) from the maximum value of the index (S max Determine whether the average power increase rate up to 32a is greater than the threshold. The threshold is determined by map (M iIt is set based on the maximum value β of the power growth rate (slope) in ). A larger power growth rate (slope) indicates the presence of a higher density of electric vehicles (EVs). For example, by setting a threshold value of 1.2 times the maximum value β, the index value (S i+1 ) from the maximum value of the index (S max During this time, it is possible to determine whether or not the number of electric vehicles (EVs) is expected to be more than 1.2 times the maximum density of electric vehicles (EVs). If the answer is YES in step S704, the maximum value of the index (S max The maximum value of the assumed power (P') for ) max Reduce the power by the equivalent of one electric vehicle (EV), return to step S703, and again, map (M i+1 ) is learned. As shown in Figure 5C, the map learning unit 14 learns the index value (S i+1 ) from the maximum value of the index (S max Steps S703 to S705 are repeated until the average value of the power increase rate up to 32b falls below the threshold. According to steps S704 and S705, the power increase rate (slope) of the map can be kept within the expected range for the maximum density of electric vehicles (EVs). Therefore, the search time when searching in the direction of increasing the index value is further shortened, and the power control of the entire group 2 is accelerated. If the answer in step S704 is NO, proceed to step S709.

[0030] Although an example was shown in which steps S704 and S705 are performed after step S703, it is also acceptable to proceed to step S709 without performing steps S704 and S705.

[0031] In step S709, the map learning unit 14 multiplies the pre-learned map (M) by the learning rate k (where k is a number between 0 and 1). i The trained map (M) is obtained by multiplying the value obtained by subtracting the learning rate k from 1. i+1 By adding ) the learned map (M i+1 ) corrects the map (M) before training in the learning cycle shown in Figure 2. i ) The entire map after training (M i+1If it is replaced with i ), there is a risk that the control after a special day when an event or the like is held and more power is consumed than usual may slow down. Therefore, by using the learning rate (k) to combine the map before learning (M i+1 ) and the map after learning (M

[0032] On the other hand, if the answer is NO in step S701, the process proceeds to S710 in FIG. 4. Similarly to step S702, it is determined whether the output power (P i+1 ) measured in step S06 is greater than the required power (P Ti ) measured in step S01. If the answer is NO in step S710, since the output power (P i+1 ) is smaller than the assumption of the map, the controller 1 will search in the direction of increasing the index value. In this case, the process proceeds to step S711. Similarly to step S703, the map learning unit 14 reduces the power increase rate in the first range smaller than the index value (S i ) in the map (M i+1 ), and increases the power increase rate in the second range larger than the index value (S i ) in the map (M i+1 ) to learn the map (M i ). By this map learning, the index value corresponding to the required power (P Ti ) can be increased. Therefore, the time for searching in the direction of increasing the index value is shortened, and the control is speeded up.

[0033] More specifically, the map learning unit 14 reduces the power increase rate of the map (M<0,000088>) from the minimum value (S i+1 ) of the index value to the index value (S i ), and from the index value (S i+1 ) to the index value (S i+1 ) corresponding to the output power (P i ) before broadcasting the index value (S i ) on the map corresponding to the output power (P i ) before notifying the index value (S i) Increase the power increase rate up to this point. At this time, the power increase rate from the minimum value (S min ) to the index value (S i+1 ) may be uniformly decreased. The power increase rate from the index value (S i+1 ) to the index value (S i ) may be uniformly increased. The assumed power corresponding to the index value (S i+1 ) on the map may be made to coincide with the output power (P i+1 ). The assumed powers corresponding to the minimum value (S min ) and the index value (S i ) on the map may not be changed. The power increase rate may be made continuous at the index value (S i ) and the index value (S i+1 ). By step S711, the learned map (M i+1 ) is obtained. After step S711, the process proceeds to step S709 in FIG. 3.

[0034] According to the map (M i+1 ) after learning in step S711, as the index value corresponding to the required power (P Ti ), an index value larger than the index value (S i+1 ) can be calculated. Therefore, the time for searching in the direction of increasing the index value is shortened, and the control is speeded up.

[0035] (Third Embodiment) In the third embodiment, referring to FIGS. 3, 4, 6A, and 6B, a map learning method in the case of searching in the direction of decreasing the index value because the output power is larger than expected is shown. When the output power (P i+1 ) is larger than the required power (P Ti ), over-discharge or over-charge to the power system 3 occurs, causing various effects on the power system 3 such as a decrease in the frequency of the commercial power supply, reverse power flow, system malfunction, etc. Therefore, the controller 1 needs to immediately eliminate the over-discharge or over-charge to the power system 3.

[0036] FIG. 6A shows the case where YES in step S702 of FIG. 3, that is, the output power (P i+1 ) is the required power (P TiThis indicates the case where it is greater than ). In this case, the map (M i The output power (P) was higher than expected. i+1 Because the value is large, controller 1 will search in the direction of reducing the index value. Proceeding to step S706, the map learning unit 14 will search the map (M i ) Index value (S i+1 Increase the power increase rate in the third range smaller than ) and map (M i ) Index value (S i+1 By reducing the power increase rate in the fourth range greater than ) the map (M i ) is learned. Through this map learning, the required power (P Ti The corresponding index value can be reduced. Therefore, the time spent searching in the direction of reducing the index value is shortened, and control becomes faster. Over-discharge or overcharge of power system 3 can be immediately resolved.

[0037] More specifically, the map learning unit 14 calculates the index value (S) in step S03. i+1 Output power (P) before broadcasting ) i ) corresponding map (M i ) The index value above (S i ) from the index value (S i+1 Map to ) (M i ) Increase the power growth rate and the index value (S i+1 ) from the maximum value of the index (S max Reduce the rate of power increase up to ). At this time, the index value (S i ) from the index value (S i+1 The rate of increase in power up to ) may be uniformly increased. Index value (S i+1 ) from the maximum value of the index (S max The power increase rate up to ) can be uniformly reduced. Map (M i ) The index value above (S i+1 The assumed power corresponding to ) is the output power (P i+1 ) may be matched with the map (M i ) The index value above (S i ) and the maximum value of the index (S max The assumed power values ​​corresponding to ) do not need to be changed. Index value (S i) and index value (S i+1 The power increase rate may be made continuous in ). Step S706 is performed to obtain the learned map (M i+1 ) can be obtained.

[0038] Map after training (M) in Figure 6A i+1 According to the required power (P Ti As the index value corresponding to ), the index value (S i+1 ) A smaller index value (S i+2 This makes it possible to calculate the index value. Therefore, the time spent searching for a smaller index value is reduced, and control becomes faster. Over-discharge or overcharge of power system 3 can be immediately resolved.

[0039] Proceed to step S707, output power (P i+1 ) is the required power (P Ti It is determined whether a loss occurs if the output power (P) is greater than ). i+1 ) is the required power (P Ti If the value is greater than ), specifically, it means that over-discharge or over-charge is occurring in power system 3, and it is necessary to immediately lower the index value in order to resolve the over-discharge or overcharge. Therefore, in step S707, it is determined whether or not a loss has occurred due to over-discharge or over-charge. If a loss has occurred (YES in step S707), proceed to step S708; if no loss has occurred (NO in step S707), skip step S708 and proceed to step S709.

[0040] In step S708, as shown in Figure 6B, the index value (S i+1 Output power (P) before broadcasting ) i ) corresponding map (M i+1 ) The index value above (S i ) from the index value (S i+1 Map to ) (M i The power increase rate of ) is set to a constant. In step S708, the learned map (M' shown in Figure 6B) is obtained. i+1 ) is obtained. Figure 6B shows the trained map (M' i+1According to the map (M' i+1 Compared to the index value (S i+2 This allows the value to be made even smaller. Therefore, the time spent searching for a smaller index value is reduced, and control becomes faster. Over-discharge or overcharge conditions in power system 3 can be immediately resolved.

[0041] Furthermore, after step S706, it is acceptable to proceed directly to step S709 without performing step S707 or steps S707 and S708. Alternatively, after step S706, it is acceptable to proceed directly to step S708 without determining whether there is a loss in S707.

[0042] On the other hand, if the case in step S710 in Figure 4 is YES, that is, the output power (P) measured in step S06 i+1 ) is the required power (P) measured in step S01. Ti If it is greater than ), controller 1 will search in the direction of decreasing the index value. In this case, proceed to step S712, and map learning unit 14 will search the map (M i ) Index value (S i+1 Increase the power increase rate in the third range smaller than ) and map (M i ) Index value (S i+1 By reducing the power increase rate in the fourth range greater than ) the map (M i ) is learned. Through this map learning, the required power (P Ti The corresponding index value can be reduced. Therefore, the time spent searching in the direction of reducing the index value is shortened, and control becomes faster. Over-discharge or overcharge of power system 3 can be immediately resolved.

[0043] More specifically, the map learning unit 14 determines the minimum value of the index value (S min ) from the index value (S i+1 Map to ) (M i ) Increase the power growth rate and the index value (S i+1 ) From step S03, the index value (S i+1 Output power (P) before broadcasting ) i ) corresponding map (Mi ) The index value above (S i Reduce the rate of power increase up to ). At this time, the minimum value of the index (S min ) from the index value (S i+1 The rate of increase in power up to ) may be uniformly increased. Index value (S i+1 ) from the index value (S i The power increase rate up to ) can be uniformly reduced. Map (M i ) The index value above (S i+1 The assumed power corresponding to ) is the output power (P i+1 ) may be made to match. Index value (S i ) and index value (S i+1 The power increase rate may be made continuous in step S712. i+1 ) can be obtained.

[0044] Map after learning in step S712 (M i+1 According to the required power (P Ti As the index value corresponding to ), the index value (S i+1 This allows for the calculation of an index value smaller than the given value. Therefore, the time spent searching for a smaller index value is reduced, and control becomes faster. Over-discharge or over-charge of power system 3 can be immediately resolved.

[0045] Proceed to step S713, and in the same manner as in step S707, output power (P i+1 ) is the required power (P Ti If the value is greater than ), it is determined whether a loss will occur. If a loss has occurred (YES in step S713), proceed to step S714. If no loss has occurred (NO in step S713), skip step S714 and proceed to step S709 in Figure 3.

[0046] In step S714, the minimum value of the index (S min ) from the index value (S i+1 Map to ) (M i+1 The power increase rate of ) is set to a constant. Step S714 sets the learned map (M' i+1 ) is obtained. Map by step S708 (M'i+1 According to the map (M) in step S712 i+1 Compared to the index value (S i+2 This allows the value to be made even smaller. Therefore, the time spent searching for a smaller index value is reduced, and control becomes faster. Over-discharge or overcharge conditions in power system 3 can be immediately resolved.

[0047] Controller 1 can be implemented using a microcomputer equipped with a CPU (Central Processing Unit), memory 15, and an input / output unit. A computer program (map learning program) for making the microcomputer function as Controller 1 is installed on the microcomputer and executed. As a result, the microcomputer functions as one of the multiple information processing units (11-14) of Controller 1. Instead of implementing Controller 1 with software, dedicated hardware may be provided to perform each information processing. Dedicated hardware may include devices such as application-specific integrated circuits (ASICs) or conventional circuit components arranged to perform the functions described in the above embodiment. Alternatively, the multiple information processing units (11-14) may be configured with individual hardware.

[0048] The embodiments described above are merely examples of the present invention. Therefore, the present invention is not limited to the embodiments described above, and various modifications are possible in forms other than those described above, as long as they do not depart from the technical spirit of the present invention, depending on the design and other factors. [Explanation of Symbols]

[0049] 1. Controller (Map Learning Device) Group 2 3 Power system 11 Power Measurement Unit 12. Indicator Value Calculation Section 13. Broadcast transmission unit 14. Map Learning Section EV1, EV2, EV3, ..., EV Electric Vehicle (Charging and Discharging Elements) PTi Required power (required value) P i Output power (second output value) P i+1 Output power (first output value) S i+1 Indicator value (first indicator value) S max Maximum value of the indicator S min Minimum value of the index

Claims

1. A map learning method for learning the map in a power control system in which each of one or more charge / discharge elements that receives a transmitted index value autonomously controls its own charge / discharge power based on the index value, and a controller controls the charge / discharge power of the entire group using a map that shows an assumed value of the charge / discharge power of the entire group, which is assumed to be the response of the entire group including the one or more charge / discharge elements to the index value, From the aforementioned map, a first index value corresponding to the required charge / discharge power for the entire group is calculated. The first indicator value is transmitted via broadcast, The first output value of the charge / discharge power of the entire group is measured relative to the first index value. The map is learned using the aforementioned request value and the first output value. Map learning methods.

2. The map learning method according to claim 1, wherein if the first output value is smaller than the required value, the map is learned by reducing the power increase rate in a first range smaller than the first index value in the map and increasing the power increase rate in a second range larger than the first index value in the map.

3. The map learning method according to claim 2, wherein if the first output value is smaller than the requested value, and the second output value of the total charge / discharge power of the group before the first index value is broadcast is smaller than the requested value, the map is learned by reducing the power increase rate of the map from the second index value on the map corresponding to the second output value to the first index value, and by increasing the power increase rate from the first index value to the maximum value of the index value.

4. The map learning method according to claim 3, wherein if the average value of the power increase rate from the first index value to the maximum value of the index value is greater than a threshold, the assumed value for the maximum value of the index value is reduced and the map is learned again.

5. The map learning method according to claim 2, wherein if the first output value is smaller than the requested value and the second output value of the total charge / discharge power of the group before the first index value is broadcast is larger than the requested value, the map is learned by reducing the power increase rate of the map from the minimum value of the index value to the first index value and increasing the power increase rate of the map from the first index value to the second index value on the map corresponding to the second output value.

6. The map learning method according to claim 1, wherein if the first output value is greater than the required value, the map is learned by increasing the power increase rate in a third range smaller than the first index value in the map and decreasing the power increase rate in a fourth range greater than the first index value in the map.

7. The map learning method according to claim 1, wherein if the first output value is greater than the requested value and the second output value of the total charge / discharge power of the group before the first index value is broadcast is less than the requested value, the map is learned by increasing the power increase rate of the map from the second index value on the map corresponding to the second output value to the first index value, and decreasing the power increase rate from the first index value to the maximum value of the index value.

8. The map learning method according to claim 7, wherein if the first output value is greater than the requested value, and the second output value of the total charge / discharge power of the group before the first index value is broadcast is less than the requested value, the power increase rate of the map from the second index value to the first index value is set to a constant.

9. The map learning method according to claim 6, wherein if the first output value is greater than the requested value, and the second output value of the total charge / discharge power of the group before the first index value is broadcast is greater than the requested value, the power increase rate of the map from the minimum value of the index value to the first index value is increased, and the power increase rate of the map from the first index value to the second index value on the map corresponding to the second output value is decreased, thereby learning the map.

10. The map learning method according to claim 9, wherein if the first output value is greater than the requested value, and the second output value of the total charge / discharge power of the group before the first index value is broadcast is greater than the requested value, the power increase rate of the map from the minimum value of the index value to the first index value is set to a constant.

11. The map learning method according to claim 1, wherein the map after learning is modified by adding the map after learning, which is obtained by multiplying the map before learning by a learning rate k (where k is a number from 0 to 1), by a value obtained by subtracting the learning rate k from 1.

12. The map learning method according to claim 1, which learns different maps according to region, season, month, weekday / holiday, day of the week, or time of day.

13. In a power control system in which one or more charge / discharge elements that receive a transmitted index value autonomously control their own charge / discharge power based on the index value, and a controller controls the charge / discharge power of the entire group using a map that shows an assumed value of the charge / discharge power of the entire group, which is assumed to be the response of the entire group including the one or more charge / discharge elements to the index value, a map learning device for learning the map, An index value calculation unit calculates a first index value corresponding to the required charge / discharge power for the entire group from the map, A broadcast transmission unit that broadcasts the first indicator value, A power measurement unit that measures a first output value of the total charge / discharge power of the group relative to the first index value, A map learning unit that learns the map using the requested value and the first output value, A map learning device having the following features.

Images