JP3350229B2 - Railway load prediction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an electric railway load predicting apparatus used as a power load managing apparatus for a railway substation and for predicting the electric power consumption of a substation after a predetermined time.

[0002]

2. Description of the Related Art In conventional railways, the load of a substation exceeds the contracted power by an operator, judging from the number of trains on a feeder section, for the purpose of efficient use of electric power for train operation and reduction of electric charges. When it is predicted that the train operation is notched, an instruction is issued to limit the train operation notch, thereby reducing the power for train operation.

[0003] However, when the operator predicts the load power of the substation from only the number of trains on the feeder section in this way, there is a problem in terms of prediction accuracy, the burden on the operator is large, and the prediction is incorrect. If the power load exceeds the contracted power with the power company, the surplus contracted power will be penalized.

[0004] Therefore, there has been proposed a railway load predicting apparatus for automatically predicting a railway load performed by such an operator. This is a mileage power prediction method for predicting substation load power based on the mileage of a train traveling in a target feeding section and the number of vehicles.

[0005]

However, this traveling distance power prediction system has the following problems, and FIGS. 8 to 10 are diagrams for explaining this problem. FIGS. 8 to 10 show a 70 km section including three stations.
FIG. 8 is a diagram showing the relationship between the train position and the train speed, FIG. 9 is a diagram showing the relationship between the train position and the traction force of the train, and FIG. 10 is a diagram showing the train position and usage. It is a figure which shows the relationship between electric power and a train position, and the integrated electric energy.

Here, in FIG. 8, the ATC signal is
It means a speed limit signal for a train in the case of a railway employing the ATC (Automatic Train Control) signal system as a security system. In this ATC signal system, deceleration control is automatically performed when the running speed of the train exceeds the ATC signal.

The following can be understood from these simulation results. (1) As shown in FIG. 10, the power used by the train fluctuates greatly. (2) When the train accelerates from the station A to the station B, the tractive force and the electric power used are maximized as shown in FIGS. The same applies when the train accelerates from station B to station C. Therefore, the traction force and the electric power used during acceleration traveling are about 1.10 when traveling at a constant speed.
2 to 2.0 times.

Therefore, it can be said that even if the train travels the same travel distance, the power used greatly varies depending on the acceleration travel distance. For example, immediately after the accident condition is released and train operation is restored or immediately after the notch restriction is released, if the trains accelerate at the same time from low speed and the accelerating mileage of each train overlaps, the substation in the feeder section The amount of load power at a location becomes very large. In such a case, in the above-described traveling distance power prediction method, since the prediction does not reflect the overlap of the acceleration traveling distances, the prediction cannot be correctly performed, and the contract power amount may be exceeded. Therefore, the above-described prediction method has a problem in prediction accuracy.

Therefore, an object of the present invention is to provide a more accurate
It is an object of the present invention to provide a railway load prediction device capable of predicting electric power consumption, accurately performing train operation notch restriction and canceling the restriction based on the prediction result, and maintaining electric power consumption equal to or less than contract electric power.

[0010]

In order to achieve the above-mentioned object, the invention according to claim 1 is an invention according to claim 1, wherein a plan corresponding to a current train position and a schedule prepared for each train type is prepared for each train. Based on the traveling pattern, a traveling pattern predicting means for predicting a predicted traveling pattern of the target feeding section, and a traveling distance and an acceleration of the target feeding section for each train from the predicted traveling pattern predicted by the traveling pattern predicting means. A travel distance / acceleration distance calculation means for calculating the distance, a number of trains in the train, and a travel distance / acceleration distance calculated by the travel distance / acceleration distance calculation means; And a power prediction unit that predicts a load power amount of a substation in a target feeding section using a model.

In order to achieve the above object, the invention according to claim 2 is based on the present train position and a planned traveling pattern suitable for each train type prepared in advance for each train. A traveling pattern predicting means for predicting a predicted traveling pattern of the target feeding section; and a traveling distance for calculating a traveling distance and an acceleration distance of the target feeding section for each train from the predicted traveling pattern predicted by the traveling pattern predicting means. Acceleration distance calculation means, the number of train cars,
A power prediction for predicting the load power amount of the substation in the target feeding section from the mileage and the acceleration distance calculated by the mileage / acceleration distance calculation means, using a regression model created in advance from the past actual power consumption data. Means, the load power amount of the substation predicted by the power prediction means and the contract power amount with the electric power company, and when the load power amount exceeds the contract power amount, the operation restriction command for each train is issued. And an operation restriction command determining means for determining whether or not to output the electric load.

[0012] In order to achieve the above object, the invention according to claim 3 is based on the present train position and a planned traveling pattern suitable for each train type prepared in advance for each train. From the traveling pattern predicting means for predicting the predicted traveling pattern of the target feeding section, and from the predicted driving pattern predicted by the traveling pattern predicting means, for each train, the traveling distance and the acceleration distance for obtaining the traveling distance and the acceleration distance of the target feeding section Acceleration distance calculating means, actual running pattern creating means for creating an actual running pattern from the track circuit boundary passing time for each train, and the target feeder of each train from the actual running pattern created by the actual running pattern creating means. Actual mileage / acceleration mileage calculating means for calculating the mileage and acceleration distance of the section, the number of trains, and the actual mileage・ Based on the mileage and acceleration distance calculated by the acceleration mileage calculation means and the actual value of the load power at the substation at this time, regression of the mileage, acceleration distance, number of vehicles and the substation load power of each train Regression model learning means for learning model parameters, and the number of vehicles of the train, and the travel distance and acceleration distance of the target feeding section obtained by the travel distance / acceleration distance calculation means, obtained by the regression model learning means. And a power prediction unit that predicts the load power amount of the substation in the target feeding section from the obtained regression model.

[0013] In order to achieve the above object, the invention according to claim 4 is based on the present train position for each train and a planned traveling pattern suitable for a train prepared for each train type prepared in advance. From the traveling pattern prediction means for predicting the predicted traveling pattern of the target feeding section for the next T minutes, and the predicted traveling pattern predicted by the traveling pattern prediction means, the traveling distance and the acceleration distance of the target feeding section for each train are calculated. Calculated traveling distance / acceleration distance calculating means, actual running pattern creating means for creating an actual running pattern from track circuit boundary passing times for each train, and each train from the actual running pattern created by the actual running pattern creating means Actual mileage / acceleration mileage calculating means for calculating the mileage and acceleration distance of the vehicle, Based on the mileage and acceleration distance calculated by the fast mileage calculation means and the actual value of the substation load power at this time, a regression model of the mileage, acceleration distance, number of vehicles and the substation load power of each train Regression model learning means for learning parameters; and, when a temporary speed limit section is set in the target feeding section, the travel predicted by the travel pattern prediction means for the section affected by the speed limit and the subsequent sections. And an overspeed setting traveling pattern correction unit for correcting a pattern.

[0014] In order to achieve the above object, the invention according to claim 5 is based on the present train position and the planned traveling pattern suitable for each train type prepared in advance for each train. From the traveling pattern prediction means for predicting the predicted traveling pattern of the target feeding section for the next T minutes, and the predicted traveling pattern predicted by the traveling pattern prediction means, the traveling distance and the acceleration distance of the target feeding section for each train are calculated. Calculated traveling distance / acceleration distance calculating means, actual running pattern creating means for creating an actual running pattern from track circuit boundary passing times for each train, and each train from the actual running pattern created by the actual running pattern creating means Actual mileage / acceleration mileage calculation means for calculating the mileage and acceleration distance of the target feeding section of the target, Based on the mileage and acceleration distance calculated by the mileage / acceleration mileage calculation means and the actual value of the substation's load power at this time, the mileage, acceleration distance, number of vehicles and the substation load power of each train Regression model learning means for learning the amount of regression model parameters, and if a train is disturbed due to an accident or the like, the accident or the like that causes the train disturbance is eliminated and the train returns to the planned driving pattern. And an abnormal running pattern correction unit that corrects the running pattern predicted by the running pattern prediction unit assuming that the vehicle is running.

[0015]

According to the invention corresponding to claim 1, in addition to the travel distance of the train traveling in the feeder section,
Since the acceleration distance with the largest power consumption is used, it is possible to accurately predict the power consumption according to the acceleration distance that differs depending on the running pattern of each train, thereby keeping the substation load power below the contracted power. Will be possible. According to the second aspect of the invention, the operation restriction command determining means is added to the first aspect of the invention.
As compared with the invention corresponding to (1), the burden on the operator is reduced, and the operation restriction is more accurately and accurately performed.

According to the third aspect of the present invention, the actual running pattern creating means, the actual running distance / acceleration running distance calculating means, and the regression model learning means are added to the invention according to the first aspect. More accurate power prediction can be performed from past performance data, and the timing of operation restriction cancellation becomes more accurate.

According to the fourth aspect of the present invention, since the temporary speed setting traveling pattern correction means is added to the third aspect of the present invention, high-precision electric power when the temporary speed limit section is set is obtained. Prediction becomes possible, and train operation notch control is performed more accurately and accurately.

According to the fifth aspect of the present invention, the abnormal state running pattern correction means is added to the third aspect of the present invention, so that the accuracy of the power prediction at the time of abnormality is further improved. The use of the railway load prediction device of the invention corresponding to claims 1 to 5 described above has the effect of greatly reducing the contracted power surcharge that is taken as a penalty when the used power exceeds the contracted power with the power company. Having.

[0019]

Embodiments of the present invention will be described below with reference to the drawings. First, the principle of the present invention will be described.
From the simulation examples of FIGS. 8 to 10 described above, the power consumption when the train is accelerating is much larger than the power consumption when the train is traveling at a constant speed. In addition, the power consumption is predicted in consideration of the acceleration traveling distance.

For this reason, the following modeling is performed in the present invention. FIG. 2A is a diagram illustrating a relationship between a train position and a speed, and FIG. 2B is a diagram illustrating power consumption of the train. When a certain train travels along the travel pattern distance L1 shown in FIG. 2A and accelerates and runs only within the distance L2, when the train power used at this time becomes as shown in FIG. 2B. Approximated. More specifically, the electric power Pi used by a certain train i is determined by considering the number Mi of trains of the train i and the traveling distance L1i of the train i and the acceleration distance L2i of the train i (of the distance L1i,
Approximation was made as shown in equation (1) using the distance the train accelerated.

Pi = (K1 × L1i + K2 × L2i) × Mi (1) where K1 and K2 are coefficients and the load power of a substation in a feeder section (a range in which a single substation supplies electric power). Since the amount P is considered to be the sum of the electric power used by each train traveling in the feeder section, if the total number of trains traveling in the feeder section is N,
Equation (2) is obtained.

[0022]

(Equation 1) And the load power amount P of the substation in a feeder section is represented by the number Mi of vehicles of each train running in the feeder section, the running distance L1i, and the acceleration distance L2i.

Here, the parameters of the coefficients K1 and K2 are as follows:
Although different for each vehicle performance, they are approximated as the same value. (3) Coefficients K1 and K2 of the model equation shown in equation (3)
If it is known, the load of the future substation can be predicted from the number M of trains running on the feeder section, the running distance L1i, and the acceleration distance L2i therein.

According to the present invention, the load power of the railway substation is predicted after a predetermined time (for example, T minutes) by using the model formula shown in the formula (3), and the past actual data is statistically processed. Further, the method is characterized in that the regression model parameters of the equation (3) are learned.

Hereinafter, embodiments of the present invention will be specifically described. FIG. 1 is a functional block diagram showing a first embodiment according to the present invention.
Traveling speed correcting means 3 for abnormal speed setting, traveling pattern correcting means 5 for abnormal time, predicted distance / acceleration distance calculating means 7
Power prediction means 8, operation restriction command determination means, for example, notch restriction command determination means 9, actual traveling pattern creating means 10, actual traveling distance / acceleration distance calculating means 11, and regression model learning means 12. I have.

The running pattern predicting means 1 calculates a current running position and a notch limit on / off command for each train, and a planned running pattern database 2 prepared in advance for each train type schedule. The predicted traveling pattern of the target feeding section for the next T minutes is predicted.

The temporary speed setting traveling pattern correction means 3 determines whether or not a temporary speed limit section is set in the target power feeding section based on the temporary speed setting information. Then, a correction pattern for the predicted traveling pattern predicted by the traveling pattern prediction means 1 is created for the section affected by the above and the subsequent sections based on the traveling pattern database 4 prepared in advance.

More specifically, a correction pattern for a predicted pattern is created for the speed drop in the section affected by the speed limit and the resulting delay in travel time. The abnormal-time running pattern correction means 5 is configured to, when the train is disturbed due to an abnormal state such as an accident, eliminate the accident or the like causing the train disorder and return the train to the planned driving pattern database 2. Assuming that the vehicle is traveling, a correction pattern for the predicted traveling pattern predicted by the traveling pattern prediction means 1 is created based on the abnormally accelerated traveling pattern database 6 prepared in advance.

The traveling pattern predicting means inputs each correction pattern created by the overspeed setting traveling pattern correcting means 3 and the abnormal time traveling pattern correcting means 5 and outputs the corrected predicted traveling pattern.

The running distance / acceleration distance calculating means 7 calculates the running distance L1i and the acceleration distance L2i of the target power section for the next T minutes for each train from the predicted running pattern output from the running pattern predicting means 1. .

The power predicting means 8 calculates the traveling distance L1i and the acceleration distance L2i calculated by the traveling distance / acceleration distance calculating means 7.
And the number Mi of vehicles, and the load power amount of the substation in the target feeding section is predicted by the regression model parameter database 13 created in advance by statistical processing from the past power consumption data.

The notch limit command judging means 9 compares the predicted power amount P of the substation predicted by the power predicting means 8 with the contracted power amount with the power company, and the substation power amount P does not exceed the contracted power amount. Thus, it is determined whether or not to output a notch restriction command for each train. The result of the determination by the notch limit command determining means 9 is displayed on a limit command determination display means (not shown).

The actual running pattern creating means 10 calculates a predetermined past time (for example, T
), An actual running pattern for a predetermined time in the past (for example, T minutes) is created from the passage time of the track circuit boundary for each train. It is created by recording the time and calculating the average speed of each track circuit.

The track circuit uses a rail in a certain section as a part of an electric circuit, and can detect the presence of a train by short-circuiting between rails on the axle of a train or vehicle in the section. A circuit that can be made.

Actual running distance / accelerated running distance calculating means 11
Calculates the travel distance L1i and the acceleration distance L2i of the target feeding section for the last T minutes of each train from the actual travel pattern created by the actual travel pattern creating means 10.

The regression model learning means 12 calculates the mileage and acceleration distance for the past T minutes calculated by the actual mileage / acceleration mileage calculation means 11, and also the number of vehicles and the actual load power amount of the substation for the last T minutes. To learn the regression model parameters of the running distance, acceleration distance, number of vehicles, and substation load power of each train.

In each configuration described above, means for predicting the load power amount from the current operation status, that is, learning the regression model parameters of equation (3) from the portion above the broken line in FIG. 1 and the past actual data. It consists of the means, ie, the part below the dashed line in FIG.

With respect to the operation of the first embodiment of the railway load prediction device having the above configuration, the operation of the means for estimating the load electric energy will be described as (S10) and the operation of the means for learning the regression model will be described as (S20). I do. First, the procedure of (S10) will be described. These series of operations are started every certain time Δt.

(S10) -1 The traveling pattern prediction means 1 predicts the future traveling pattern of each train. The temporary speed setting travel pattern correcting means 3 is configured to, when a temporary speed limit section is set in the target feeding section, determine the travel pattern predicted by the travel pattern prediction means 1 for the section affected by the speed limit and the subsequent sections. Is corrected.

The abnormal-time running pattern correcting means 5 runs the train so that when the train is disturbed due to an accident or the like, the accident or the like causing the train disruption is eliminated and the train returns to the planned running pattern. As a result, the traveling pattern predicted by the traveling pattern prediction means 1 is corrected.

(S10) -2 The travel distance / acceleration distance calculation means 7 calculates the travel distance L1i and the acceleration distance L2i for the train traveling in the target feeding section based on the travel pattern from the travel pattern prediction means 1. .

(S10) -3 The power predicting means 8 predicts the load power P from the running distance L1i and the acceleration distance L2i calculated by the running distance / acceleration distance calculating means 7 and the number of vehicles Mi by the equation (3).

(S10) -4 Notch limit command judging means 9 compares the result of (S10) -3 with the contracted electric energy to determine whether or not to limit the notch.

Here, the specific contents of each procedure will be described in detail. (S10) -1 In order to predict the load power amount P of the substation in the future by the model formula shown in the formula (3), it is necessary to predict the future mileage and acceleration distance of each train. . For this reason, here, the predicted driving pattern (position,
Speed, time). First, a planned traveling pattern is created for each train in advance, and this is stored in the planned traveling pattern database 2. In this case, the planned traveling pattern is data representing the traveling state (position, speed, time) between stations that satisfies the train schedule for which the arrival / departure time of each station is planned. As a way of holding this data, for example, point sequence data for each certain speed section is used.

Using this planned traveling pattern, the traveling pattern of each train for the next T minutes is predicted as follows. First, it is determined whether or not the current train operation is normal. If it is determined that the operation is normal,
Assume that normal operation continues for minutes. In other words, it is assumed that traveling for the next T minutes will be performed according to a planned traveling pattern prepared in advance. Specifically, a planned travel pattern is applied to each train based on the current train position, and this is used as a predicted travel pattern for the next T minutes. If it is determined that the operation of the field is normal, a planned driving pattern is applied to each train based on the current position, and it is assumed that the train will travel along the planned driving pattern for the next T minutes. A running pattern for the next T minutes will be used.

This will be described with reference to FIG. 3. A broken line curve in the figure represents a planned traveling pattern (position, speed, time) of a certain train by time and position. For example, if the current train is in the state of (A) and is delayed by Δt from the planned driving pattern, if the delay of Δt is within a range that can be regarded as normal operation, the planned driving pattern is parallel by Δt in the time direction. The moved pattern is used as the future running pattern of this train. Therefore, the predicted traveling pattern of this train for the next T minutes is predicted as a solid line (A) to (B) in the figure.

On the other hand, the prediction when the operation of the train at the present time is determined to be an abnormal operation is performed as follows. It is described below that (a), (b),
This is the case of (c).

(A) In the case where a temporary speed limit section is set in the feeder section In a section not affected by the temporary speed limit, the vehicle travels along the planned travel pattern.
Is partially corrected for the predicted traveling pattern predicted by the above.

In general, the setting of the temporary speed limit section is limited to a number of patterns in the set section and the speed limit. Therefore, a correction travel pattern (position) in a range affected by the speed limit is set in advance for each set pattern. , Speed, relative time)
Is stored in the corrected travel pattern database 4 and the predicted travel pattern is corrected by applying this.

For example, as shown in FIG. 4, the corrected traveling pattern is applied to a range cc to dd in which the temporary speed limit section is received, and after that point, the vehicle is moved in parallel in the direction of the delay time Δt1 caused by the speed limit. bp is the predicted driving pattern before correction, ap
Is a predicted traveling pattern after correction.

(B) When notch restriction is imposed at the present time When notch restriction is imposed, it is a state in which the timing for releasing the notch restriction next is planned.
As described above, in the related art, immediately after the notch restriction is released, the trains start accelerating all at once, and the used power often exceeds the contracted power. Therefore, here, assuming that the notch limit was lifted at the next moment, in order to surely keep it below the contract power,
A running pattern assuming a driving in which each train accelerates to return to the planned running pattern is predicted.

For this, the following method is used. First,
A traveling pattern (relative position, speed, relative time) in which the train outputs the maximum traction force and accelerates (hereinafter referred to as an abnormal acceleration traveling pattern) is stored in the abnormal acceleration traveling pattern 6 in advance. It is assumed that each train accelerates from the present time to return to the planned traveling pattern. For each train, the abnormal acceleration running pattern is applied based on the current train speed. That is, the relative position and relative time at which the speed value of the pattern coincides with the current train speed are defined as the current train position and the current time, and the abnormal acceleration acceleration pattern thereafter is defined as the future travel pattern of this train. After the position where the speed of the abnormal-time accelerated traveling pattern exceeds the speed of the planned traveling pattern with respect to the position, the train travels along the planned traveling pattern in exactly the same manner as the prediction performed in (a). As a matter of course, a traveling pattern is predicted.

(C) When the current running state (position, speed, time) of the train is compared with the planned running pattern at the present time, ie, when there is a large deviation from the planned running pattern, Or, in the case where the time lag is equal to or greater than a certain threshold value, as in (b), a traveling pattern that assumes a case where an accident or the like causing such a state is eliminated is predicted. That is, a traveling pattern in which the accident or the like is resolved and the train accelerates to return to the planned traveling pattern is predicted. This prediction is performed in the same manner as in (b).

(S10) -2 Calculation of running distance and acceleration distance from running pattern Here, the running distance of each train in the target feeding section and the acceleration in the running pattern are calculated from the running pattern predicted in (S10) -1. Calculate the distance.

This will be described with reference to FIGS. 5A and 5B. FIG. 5A is a characteristic diagram showing the relationship between time and position for the same traveling pattern, and FIG. 5B is a characteristic diagram showing the relationship between speed and position. Assuming that the predicted traveling pattern of a certain train for the next T minutes is between aa and bb, since the target feeding section includes between a1a1 and bb, the traveling distance L1 of the target feeding section is a1a1 It is the difference between the position and the position of bb. Since the acceleration distance L2 is the sum of the portions where the speed change rate is positive in the entire travel distance L1, for example, the travel distance L1 is divided into a certain section ΔL, and the average speed change amount is positive. Take the sum of the sections.

(S10) -3 Prediction of substation load electric energy Here, the model formula shown in equation (3) is used to calculate the number Mi of vehicles of each train, the travel distance L1i and the acceleration distance L2i of the target power section.
To predict the load power P of the substation in the target feeding section for the next T minutes.

(S10) -4 Judgment of ON / OFF of notch limit Here, the predicted value of the load electric energy performed in (S10) -3 is compared with the contract electric energy with the electric power company, and the notch restrict command ON / OFF is determined. Make a decision to turn off. At the present time, when the notch limit command is off, if the load power amount predicted value exceeds the contracted power amount, the notch limit command is turned off → on and output.

On the other hand, when the notch limit command is ON at the present time, in order to prevent the contract power from exceeding the contracted power again in the near future after the notch limit is turned ON → OFF, the notch limit command is turned OFF. Is set lower than the contracted power amount, and if the predicted value is lower than the threshold value, a notch restriction release command is output.

(S20) Expression (3) based on past performance data
The means and operation for learning the model parameters shown in FIG. The following (S20) -1, (S20) -2,
(S20) -3 and (S20) -4 will be sequentially described in detail with reference to FIG.

(S20) -1 For each train, an actual running pattern for the last T minutes is calculated. In FIG. 6 (a), a traveling pattern is obtained from the passing times T1, T2,... Of the track circuit boundaries P1, P2,.

For example, when the passage times of the track circuit boundaries Pn, Pn + 1 are Tn, Tn + 1,...
The average speed Vn in the section of Pn + 1 is calculated by equation (4). Vn = (Pn + 1-Pn) / (Tn + 1-Tn) (4) (S20) -2 From the actual running pattern of (S20) -1, the travel distance and the acceleration distance of the train that ran in the target feeding area were calculated. calculate. Specifically, the travel distance L1 is calculated as the distance between the current train position and the train position T minutes ago, which is the target feeding section. This corresponds to L1 in FIG.

As for the acceleration distance, first, (S20) -1
The average speed change rate An of each track circuit is calculated by the equation (5) using the average speed Vn of each track circuit obtained in the above. An = (Vn + 1−Vn) / ΔT (5) where ΔT = (Tn + 2 + Tn + 1) / 2− (Tn + 1 + T)
n) / 2 (ΔT is n after passing through the center point of the n-th track circuit.
(Time required to pass the center point of the +1 track circuit)) Then, the sum of the distances of the sections (between the centers of the track circuits) where the average speed change rate An is positive is calculated, and the acceleration distance of the target feeding section is calculated. And This corresponds to L2 in FIG.

(S20) -3 The value of the substation load electric energy at this time is fetched. Here, the substation load electric energy measured value for the past T minutes is taken. (S20) -4 Accumulated data in which a set of the traveling distance, the acceleration distance, the number of vehicles, and the actual substation load power value of (S20) -3 in (S20) -2 is accumulated for a certain period, and the formula ( The learning of the parameters K1 and K2 of the model equation shown in 3) is performed.

Here, a method of learning a regression model will be described. (S30) -1 A data set (P, Lm1, Lm2) obtained every time t minutes as an actual value is stored, and when some data set is accumulated (for example, for one month), the data set is accumulated at this time. Based on the data set, (S30) -3 described later
The coefficients K1 and K2 (referred to as regression coefficients) are calculated by the method shown in (2). (S30) -2 The method of (S30) -1 is more desirable as the number of data sets is larger. The new parameters K1 and K2 are calculated by performing the same operations as (S30) -1 (every t minutes), and replaced with the old parameters K1 and K2. This is the parameter K
1, K2 learning.

(S30) -3 Actual data set (P, L
The method for calculating the regression model coefficients K1, K2 from (m1, Lm2) will be described. The data set of results is M as shown in FIG.
Assume that there are pairs. From Lm1 (j) and Lm2 (j), the coefficient K1
, K2 is calculated as follows: Pr (j) = K1 × Lm1 (j) + K2 × Lm2 (j)
ε (j) = ε (j) = P (j) −Pr (j) = P (j) − [K1 × Lm1 (j) +
K2 × Lm2 (j)] ε (j) ² = jP (j) − [K1 × Lm1 (j) + K2
× Lm2 (j)]｝ ^{2 For} these, j is determined for 1 to M. Then the sum of squares is

[0066]

(Equation 2)

Then, K1 and K2 are solved such that the right side of the equation (6) is minimized. More specifically, the equation (7) is obtained by partially differentiating the right side of the equation (6) with respect to each of K1 and K2 and setting it to 0, and solving this for K1 and K2.

[0068]

(Equation 3)

By transforming equation (7), K as in equation (8)
It becomes a simultaneous equation of 1 and K2. aK1 + bK2 = Z1 cK1 + dK2 = Z2 (8) From this, K1 and K2 are obtained.

According to the embodiment described above, the traveling along the diagram, the traveling at the time of setting the temporary speed, and the like are previously patterned and made into a database, and the future traveling pattern is predicted using this. As a result, the processing time required for the prediction can be shortened, and the frequency of the prediction can be increased. Therefore, the notch limit command can be output at an appropriate timing, and the output timing of the notch limit command can be delayed. Is reduced.

[0071]

According to the present invention, the power consumption is more accurately predicted, and based on the prediction result, the train operation notch is restricted and the restriction is released accurately, and the power consumption is maintained at or below the contract power. It is possible to provide an electric railway load prediction device that can perform the operation.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a first embodiment of a railway load prediction device according to the present invention.

2A is a characteristic diagram illustrating a relationship between speed and position for explaining the operation in FIG. 1, and FIG. 2B is a diagram illustrating a relationship between power consumption and position for explaining the operation in FIG. Characteristic diagram.

FIG. 3 is a view for explaining a planned traveling pattern of FIG. 1;

FIG. 4 is a view for explaining a planned traveling pattern when a temporary speed limit section in FIG. 1 is set;

FIG. 5 is a view for explaining a method of calculating a traveling distance and an acceleration distance from the traveling pattern of FIG. 1;

6A is a diagram for explaining a method of learning model parameters from past performance data of FIG. 1, and FIG. 6B is a diagram illustrating a method of calculating a traveling distance and an acceleration distance from an actual traveling pattern. Figure to do.

FIG. 7 is a view for explaining a method of calculating a regression model coefficient from the actual data of FIG. 1;

FIG. 8 is a simulation diagram illustrating a relationship between a train speed and a position for explaining a problem to be solved by the present invention.

FIG. 9 is a simulation diagram showing a relationship between tractive force and position for explaining a problem to be solved by the present invention.

FIG. 10 is a simulation diagram illustrating a relationship between an integrated power amount and a position for describing a problem to be solved by the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... running pattern prediction means, 2 ... planned running pattern database, 3 ... temporary speed setting running pattern correction means, 4 ... corrected running pattern database, 5 ... abnormal running pattern correcting means, 6 ... abnormal running pattern database
7: travel distance / acceleration distance calculation means, 8: electric power prediction means,
9: Notch limit command determining means, 10: Actual running pattern creating means, 11: Actual running distance / acceleration distance calculating means, 1
2 ... regression model learning means, 13 ... regression model parameter database.

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Tadashi Okada 2-4-2, Shibata, Kita-ku, Osaka-shi, Osaka Inside West Japan Railway Company (72) Inventor Hiroyuki Hirayama 2-chome, Shibata, Kita-ku, Osaka, Osaka No. 4-24 Inside West Japan Railway Company (72) Inventor Tetsuro Sakamoto 2-4-2 Shibata Kita-ku, Osaka City, Osaka Prefecture Inside West Japan Railway Company (72) Inventor Mieko Hayashi 1 Toshiba-cho, Fuchu-shi, Tokyo Address Toshiba Corporation Fuchu Plant (56) References JP-A-5-252563 (JP, A) JP-A-7-304353 (JP, A) JP-A-5-176457 (JP, A) JP-A-5 -16808 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) B60M 3/00 B61L 27/00 H02J 3/00

Claims (5)

(57) [Claims] 1. A running pattern prediction for predicting a predicted running pattern of a target feeding section for each train based on a current train position and a planned running pattern corresponding to a schedule prepared for each train type prepared in advance. Means, a traveling distance / acceleration distance calculating means for calculating a traveling distance and an acceleration distance of a target feeding section for each train from a predicted traveling pattern predicted by the traveling pattern prediction means, A power predicting means for predicting the load power amount of a substation in a target feeding section from a running distance and an acceleration distance calculated by the running distance / acceleration distance calculating means, based on a regression model created in advance from past power consumption data. And a railway load prediction device comprising: 2. A traveling pattern prediction for predicting a predicted traveling pattern of a target feeding section for each train based on a current train position and a planned traveling pattern prepared for each train type prepared in advance. Means, a traveling distance / acceleration distance calculating means for calculating a traveling distance and an acceleration distance of a target feeding section for each train from the predicted traveling pattern predicted by the traveling pattern prediction means, A power predicting means for predicting the load power amount of a substation in a target feeding section from a running distance and an acceleration distance calculated by the running distance / acceleration distance calculating means, based on a regression model created in advance from past power consumption data. Comparing the load power amount of the substation predicted by the power prediction means with the contracted power amount with the power company, and when the load power amount exceeds the contracted power amount, And a driving restriction command determining means for determining whether or not to output a driving restriction command for each of the trains. 3. A traveling pattern prediction for predicting a predicted traveling pattern of a target feeding section for each train based on a current train position and a prepared traveling pattern suitable for a diagram for each train type prepared in advance. Means, a travel distance / acceleration distance calculation means for calculating a travel distance and an acceleration distance of a target feeding section for each train from the predicted travel pattern predicted by the travel pattern prediction means, and a track circuit boundary passing for each train. An actual traveling pattern creating means for creating an actual traveling pattern from the time; and an actual traveling distance / acceleration distance for calculating a traveling distance and an acceleration distance of a target feeding section of each train from the actual traveling pattern created by the actual traveling pattern creating means. Acceleration mileage calculation means, the number of trains, the mileage calculated by the actual mileage / acceleration mileage calculation means, Fast distance and a regression model learning means for learning regression model parameters of a running distance, an acceleration distance, the number of vehicles and a substation load power amount of each train based on the actual value of the load power amount of the substation at this time; From the regression model obtained by the regression model learning means, based on the number of vehicles and the running distance and acceleration distance of the target feeding section obtained by the running distance / acceleration distance calculating means, And a power prediction means for predicting a load power amount at a station. 4. For each train, a predicted travel pattern of a target power section for the next T minutes is predicted based on a current train position and a planned travel pattern prepared for each train type prepared in advance. Traveling pattern prediction means, and a traveling distance / acceleration distance calculating means for calculating a traveling distance and an acceleration distance of a target feeding section for each train from the predicted traveling pattern predicted by the traveling pattern prediction means; Actual running pattern creating means for creating an actual running pattern from the track circuit boundary passing time, and actual running distance / acceleration for calculating the running distance and acceleration distance of each train from the actual running pattern created by the actual running pattern creating means Mileage calculating means, the number of trains, mileage calculated by the actual mileage / acceleration mileage calculating means, acceleration Regression model learning means for learning regression model parameters of the distance of each train, the acceleration distance, the number of vehicles, and the substation load power based on the separation and the actual value of the substation load power at this time; When a temporary speed limit section is set in the electric section, a temporary speed setting travel pattern correction means for correcting the travel pattern predicted by the travel pattern prediction means for the section affected by the speed limit and the subsequent sections; A railway load prediction device comprising: 5. For each train, a predicted travel pattern of a target power section for the next T minutes is predicted based on a current train position and a planned travel pattern prepared for each train type prepared in advance. Traveling pattern prediction means, and a traveling distance / acceleration distance calculating means for calculating a traveling distance and an acceleration distance of a target feeding section for each train from the predicted traveling pattern predicted by the traveling pattern prediction means; An actual traveling pattern creating means for creating an actual traveling pattern from the track circuit boundary passing time; and a traveling distance and an acceleration distance of each train in a target feeding section are calculated from the actual traveling pattern created by the actual traveling pattern creating means. Actual mileage / acceleration mileage calculation means, and the number of trains and the actual mileage / acceleration mileage calculation means A regression model learning means for learning regression model parameters of a running distance, an acceleration distance, the number of vehicles and a substation load power amount of each train based on a running distance, an acceleration distance, and a load power amount actual value of the substation at this time; When the train is disturbed due to an accident or the like, the traveling pattern predicting means predicts that the accident or the like causing the disruption of the train is resolved and the train travels to return to the planned traveling pattern. An electric train load predicting device, comprising: an abnormal running pattern correcting means for correcting the corrected running pattern.