CN103870892B - Method and system for achieving railway locomotive operation control from off-line mode to on-line mode - Google Patents
Method and system for achieving railway locomotive operation control from off-line mode to on-line mode Download PDFInfo
- Publication number
- CN103870892B CN103870892B CN201410117028.3A CN201410117028A CN103870892B CN 103870892 B CN103870892 B CN 103870892B CN 201410117028 A CN201410117028 A CN 201410117028A CN 103870892 B CN103870892 B CN 103870892B
- Authority
- CN
- China
- Prior art keywords
- locomotive
- sequence
- gear
- line
- railway locomotive
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Control Of Transmission Device (AREA)
- Feedback Control In General (AREA)
Abstract
The invention provides a method and system for achieving railway locomotive operation control from an off-line mode to an on-line mode. A sequence pattern excavation method is used for obtaining a control gear sequence from original operation data of a locomotive in an off-line mode, an off-line optimization algorithm is used for optimizing the time distribution proportions of specific gears in the control gear sequence, a neural network model is constructed according to the time distribution proportion sequence with the optimal energy consumption, association rules of the specific locomotive, line parameters and the control gear sequence are obtained according to the neural network model, finally, the control sequence with the optimal energy consumption is used for guiding the locomotive to operate in an on-line mode. Due to the fact that off-line calculation is not affected by time factors so that an off-line part can own better optimized space, and in the operation process of the locomotive, a good energy-saving effect can be achieved by utilizing the control gear sequence of operation of the locomotive, wherein the control gear sequence is obtained in the off-line mode. In addition, an on-line control operation result of the locomotive serves as the data input of the off-line sequence pattern excavation method and the optimization algorithm, and therefore off-line learning can be adjusted and optimized continuously.
Description
Technical field
The present invention relates to locomotive operation optimized handling technology, more particularly to one kind is from offline to canbe used on line railway locomotive fortune
The method and system of row optimized handling.
Background technology
Railway locomotive operation control is a typical multiple target, multiple constraint, nonlinear complicated real-time change process.Cause
It is a non-linear constrained optimization problems that this manipulates problem for the operation of railway locomotive.
Existing optimization railway locomotive operation method of operating mainly utilizes fuzzy control technology, genetic algorithm and their collection
Apply to locomotive operation optimization into various optimisation techniques such as algorithm to calculate to instruct the operation of locomotive.
Angle of the above-mentioned fuzzy control technology from fuzzy reasoning in combination with PREDICTIVE CONTROL, sets up fuzzy rule, and foundation
The railway locomotive that the fuzzy rule is built towards interval operation manipulates Fuzzy Forecasting Model, then manipulates mould according to the railway locomotive
Paste forecast model manipulates the operation of railway locomotive.This fuzzy control technology belongs to on-line Algorithm, and the formulation master of fuzzy rule
Artificial experience is depended on, design difficulty is big, and the time of Optimizing Search is long.
Above-mentioned genetic algorithm is optimized and search problem by generating useful solution, with reference to railway locomotive operation
Optimized handling sequence on interval, solves to problem, finally obtains suitable railway locomotive and manipulates sequence.This heredity is calculated
The time of method search is long.
Above-mentioned several on-line optimization algorithms cannot all meet at present the requirement of On-line Control, and the time of Optimizing Search is long, nothing
Method is calculated suitable for locomotive on-line optimization, is easily constrained by railway locomotive each side condition itself, and such as run time is limited
Deng, therefore the reliability of this algorithm and the optimization of result will necessarily be restricted, it is impossible in railway locomotive running
Good energy-saving effect is obtained, and artificial experience is depended on more the initial policy of said method calculating, thus for versatility
For locomotive and circuit, its optimization reasonability is difficult to support, therefore all exists from railway locomotive run time and energy consumption angle
Very big optimization space.
The content of the invention
It is an object of the invention to provide it is a kind of from the method and system for running manipulation to canbe used on line railway locomotive offline, its
It is offline obtain railway locomotive operation and manipulate the process of gear sequence do not limited by time factor, therefore from railway locomotive run when
Between and energy consumption angle have in preferably optimization space, and railway locomotive running and obtain railway locomotive operation using offline
Good energy-saving effect can be obtained when manipulating gear sequence.
The present invention is achieved through the following technical solutions：
The invention provides a kind of, from offline, to the method that canbe used on line railway locomotive runs manipulation, it includes：
Extract railway locomotive raw operational data；
Excavated for the railway locomotive raw operational data by sequential mode mining technology, obtained different type
The locomotive operation in section manipulates gear sequence；
Gear sequence is manipulated for the locomotive operation in resulting different type section, is found out often by off-line optimization algorithm
The corresponding time-sharing ratio example sequence of manipulation gear sequence of optimal energy consumption in type section；
According to the corresponding time-sharing ratio example sequence of manipulation gear sequence for obtaining, neural network model is built, and be based on
The neural network model manipulates gear sequence according to the time-sharing ratio example sequence found out by off-line optimization algorithm to locomotive operation
The characteristic parameter of row association is learnt, and the optimal locomotive operation for obtaining each bar circuit manipulates gear sequence, and by itself and railway
Locomotive and line characteristics parameter are associated；
When railway locomotive runs in certain circuit, according to the railway locomotive and line characteristics parameter, corresponding machine is matched
Car operation manipulates gear sequence；
Gear sequence is manipulated according to the locomotive operation that matching is obtained, the real time execution of railway locomotive is manipulated.
Further, it is described a kind of from the method that manipulates is run to canbe used on line railway locomotive offline, also include：
Railway locomotive real-time running data when manipulating gear sequence using locomotive operation is recorded as into railway locomotive original
Service data.
Further, gear sequence is manipulated in the locomotive operation for obtaining each bar circuit, and by itself and railway locomotive and line
After the associated process of road characteristic parameter, also include：
Different railway locomotives and the corresponding locomotive operation of line characteristics parameter are manipulated into gear sequence and is stored in policy library
In.
Further, it is described to be dug for the railway locomotive raw operational data by sequential mode mining technology
Pick, the locomotive operation for obtaining different type section manipulates the process of gear sequence, specifically includes：
The gradient that adds is obtained according to the line information of railway locomotive traveling, is divided in railway locomotive circuit according to the gradient that adds
Into different types of section；
For different type section in same circuit, different type section pair is calculated using Sequential Pattern Mining Algorithm
The locomotive operation answered manipulates gear sequence.
The present invention also provide it is a kind of from offline to canbe used on line railway locomotive run manipulate system, it includes：
Locomotive operating data extraction unit, for extracting railway locomotive raw operational data；
Sequential mode mining unit, for being directed to the railway locomotive raw operational data by sequential mode mining technology
Excavated, the locomotive operation for obtaining different type section manipulates gear sequence；
Off-line data optimizing unit, for manipulating gear sequence for the locomotive operation in resulting different type section,
The corresponding time-sharing ratio example sequence of manipulation gear sequence of optimal energy consumption in each type section is found out by off-line optimization algorithm
Row；
Neural network learning unit, for according to the corresponding time-sharing ratio example sequence of manipulation gear sequence for obtaining, structure
Neural network model is built, and based on the neural network model according to the time-sharing ratio example sequence found out by off-line optimization algorithm
Row, learn to the characteristic parameter that locomotive operation manipulates gear serial correlation, and the locomotive operation for obtaining each bar circuit manipulates shelves
Bit sequence, and it is associated with railway locomotive and line characteristics parameter；
Manipulate gear sequences match unit, for when railway locomotive runs in certain circuit, according to the railway locomotive and
Line characteristics parameter, matches corresponding locomotive operation and manipulates gear sequence；
Locomotive control running unit, the locomotive operation for being obtained according to matching manipulates gear sequence, manipulates railway locomotive
Real time execution.
Further, it is described a kind of from the system that manipulates is run to canbe used on line railway locomotive offline, also include：
Policy library, for depositing different railway locomotive and the corresponding locomotive operation of line characteristics parameter gear sequence is manipulated
Row.
The present invention can be seen that first by sequential mode mining method from a large amount of railways by the technical scheme of foregoing invention
Offline acquisition manipulates gear sequence in locomotive operation initial data, then carries out manipulating tool in gear sequence by offline optimization algorithm
The optimizing of the time-sharing ratio example sequence of body gear, neutral net mould is built by the time-sharing ratio example sequence of optimal energy consumption
Type, and obtain railway locomotive special parameter according to the neural network model study and manipulate the correlation rule of gear sequence, so as to
Finally using the manipulation gear sequence of optimal energy consumption come online direction locomotive operation.Because calculated off line is not by time factor
Limit, therefore using the offline manipulation shelves for obtaining in the preferably optimization space, and railway locomotive running that have part ownership offline
Good energy-saving effect can be obtained during bit sequence.
In addition, using the online operation and control result of railway locomotive as offline sequential mode mining method and the number of optimizing algorithm
According to input so that the data that off-line learning can be returned constantly by online partial feedback are adjusted and optimization, offline part
Combining closely and influencing each other with online part, has weighed offline optimization algorithm effect and on-line optimization calculates matter of time,
Railway locomotive can be made to manipulate gear sequence and to obtain constantly optimization improvement, so as to be finally reached more preferable energy-saving effect.
Description of the drawings
Fig. 1 is the implementing procedure figure of first embodiment of the invention；
Fig. 2 is the structure chart of second embodiment of the invention；
Fig. 3 is the structure chart of third embodiment of the invention.
In accompanying drawing：Manipulate the online manipulation device 20, locomotive operating data of the offline generating means 10, locomotive operation of gear sequence
Extraction unit 10-1, sequential mode mining unit 10-2, off-line data optimizing unit 10-3, neural network learning unit 10-4,
Policy library 10-5, manipulation gear sequences match unit 20-1, locomotive control running unit 20-2.
Specific embodiment
To make the present invention relatively sharp, the present invention is described in detail below in conjunction with the accompanying drawings.
First embodiment of the invention offer is a kind of, and from offline, to the method that canbe used on line railway locomotive runs manipulation, it is processed
Process includes offline and online two-part content：
Collection and railway locomotive is transported by Sequential Pattern Mining Algorithm that offline part includes locomotive raw operational data
Row data are learnt, and export the gear sequence of driver control railway locomotive, are calculated by optimizing again for these gear sequence sets
Method calculates and obtains each concrete gear time-sharing ratio shared in the process of moving in the manipulation gear sequence of optimal energy consumption
Example, learns the locomotive control gear being associated with railway locomotive and line characteristics parameter finally by structure neural network model
Sequence.
It is online be partly by through the study of the Sequential Pattern Mining Algorithm of offline part, optimizing algorithm and neutral net it
The locomotive control gear sequence for being obtained afterwards, in applying to railway locomotive real time management.When railway locomotive runs in certain circuit
When, first most suitable locomotive control gear sequence and whole according to its structure is searched for from policy library according to circuit and traffic information
The optimum manipulation gear sequence that bar circuit drives, then according to the operation for manipulating gear sequence solution real time management railway locomotive.
After railway locomotive end of run, it is obtained in that it in the optimum actual operating data manipulated under gear sequence.
In railway locomotive actual moving process, the change of weather and road conditions is often run into, at this time, need railway
Trainman is adjusted to the manipulation gear sequence being currently in use, and the manipulation gear sequence after adjustment is further used as offline portion
The data input divided.Therefore above-mentioned offline part and online part tight association, interact, the closed loop energy formed between them
Locomotive control gear sequence is enough continued to optimize, the optimum of railway locomotive operation can be finally obtained and be manipulated gear sequence.
As can be seen that the present invention is directed to offline part railway locomotive service data, obtained by sequential mode mining technology
Gear sequence is manipulated, then each time for manipulating the optimal energy consumption of concrete gear in gear sequence is obtained by offline optimization algorithm
Allocation proportion sequence, finally builds neural network model and learns manipulation gear sequence with railway locomotive and circuit special parameter
（The gradient, car weight, vehicle commander etc.）Between correlation rule；The locomotive operation of offline part final output is manipulated gear by online part
Sequence applies to different circuit and road conditions, and using the actual running results as offline part data input.Off-line optimization and
Sequential mode mining technology in addition to excavating to railway locomotive raw operational data, always according to application in question gear sequence
Operation result carry out further optimizing.Offline part forms closed loop with online part can constantly to locomotive control gear sequence
Row are learnt, used and are optimized, final to obtain optimum locomotive operation and manipulate gear sequence, this offline in combination with online
Mode can weigh offline optimization algorithm effect and on-line optimization and calculate matter of time, enable to railway machine using the technology
Car obtains more preferable energy-saving effect.
First embodiment of the invention be embodied as flow process as shown in figure 1, including：
Step S101, extracts railway locomotive raw operational data.
Locomotive raw operational data includes：The parameter information of railway locomotive itself, line information, the ferrum of railway locomotive operation
Road locomotive includes real time execution shelves in different circuits and the actual operating data of the different sections of highway of circuit, these actual operating datas
Position information and real time speed information etc..
The parameter information and line information of railway locomotive itself is obtained by Railway Bureau, and the actual motion number of railway locomotive
According to by the LKJ in railway locomotive（Train Detection and Identification recording equipment）Obtain.
Step S102, is excavated by sequential mode mining technology to the railway locomotive raw operational data for obtaining, and is obtained
Gear sequence is manipulated to the corresponding railway locomotive operation in different type section.
In step S102, first according to the line information of railway locomotive traveling, such as circuit actual grade, curve, tunnel etc.
Acquisition adds the gradient.
The wherein curve gradient formula that adds is as follows：
Pc=600*Lc/(Rc*Lcars) ... ... ... ... formula 1
In above-mentioned formula, PcRepresent the gradient that adds of curve, LcRepresent length of a curve, RcRepresent sweep, Lcars
Represent the total length of railway locomotive.
The gradient computing formula that adds in tunnel is as follows：
Pt=0.00013*Lt... ... ... ... ... formula 2
In above-mentioned formula, PtRepresent the gradient that adds in tunnel, LtRepresent the length in tunnel.
The final circuit gradient that adds is added gradient superposition group jointly by the add gradient, curve of circuit actual grade, tunnel
Into.
Railway locomotive circuit is divided into by different types of section according to the size of the gradient that adds, such as steep upward slope, sharp decline surpasses
Sharp decline, gentle slope.
Secondly, for different type section in same circuit, different type is calculated using Sequential Pattern Mining Algorithm
The corresponding railway locomotive operation in section manipulates gear sequence.
The Sequential Pattern Mining Algorithm adopted in the step is existing conventional Apriori algorithm.Apriori algorithm is one
The algorithm of most influential Mining Boolean Association Rules frequent item set is planted, its core is calculated based on the recursion of two benches frequency collection thought
Method.
Below by taking the section of gentle slope as an example, illustrate to calculate manipulation of the railway locomotive in certain section operation using Apriori algorithm
Gear sequence implements process：
The gear sequence of 30 gentle slope section travelings on same circuit of railway locomotive is extracted, T is marked as1,
T2,…,T30；For each gear sequence, g is used1,g2,…,gnRepresent the gear item with sequencing.
For g1,g2,…,gnThe gear item of expression carries out first round iteration C1, obtain { g1},{g2}…{gnGear item
Collection, the radix of the gear item collection is 1.Calculate { g1},{g2}…{gnGear item collection support（Support represents corresponding shelves
The percentage ratio that position item occupies in all of gear sequence）.In these gear items, filter out less than support threshold（Support
Degree threshold value rule of thumb sets）Gear item, obtain the first round calculating gear item.
The gear item calculated for the first round carries out the second wheel iteration C2, by C1The gear item of middle acquisition passes through multiplication cross
Mode combine obtain radix be 2 gear item collection { g1,g2},{g1,g3}…{gn,gm, calculating basis are 2 gear
The support of item collection, filters out the gear item collection less than support threshold, obtains the gear item of the second wheel calculating.
Third round iteration C is carried out for the calculated gear item of the second wheel3, by C2The gear item of middle acquisition is by intersecting
The mode of multiplication combines the gear item collection { g for obtaining that radix is 31,g2,g3},{g2,g3,g5}…{gn,gm,gk, according to
All subsets of the frequent gear item collection in Apriori algorithm also must frequently this characteristic, filter out gear infrequently
Item collection；Then calculating basis are the support of 3 gear item collection, filter out the gear item collection less than support threshold.
The rest may be inferred, carries out kth wheel iteration Ck, by Ck-1The gear item collection multiplication cross of middle acquisition obtains shelves of the radix for k
Position item collection, carries out subtracting branch by the same mode of former wheels, and what iteration terminated is masked as again finding any frequent k+1's
Gear item collection.Gear item collection { the g ' for now obtaining1,g′2,…,g′nIt is manipulation shelves of the final railway locomotive in the section operation
Bit sequence.
Produce with order pass from the railway locomotive manipulation gear item collection for obtaining finally according to the formula of following confidence level
The manipulation gear sequence of connection rule, such as gm,gn→gk, represent that this is closed by what the correlation rule shown in formula 3 was produced with order
The manipulation sequence of connection is gm,gn,gk。
Confidence(A->B)=P(B|A)=support_count(AB)/support_count(A)
... ... ... formula 3
In formula 3, Confidence (A->B) confidence level is represented；A represents manipulation gear item gm,gn；B represents manipulation gear
Item gk；P (B | A) represents the probability of the B under the conditions of A；Support support_count (AB) represents that A, B occur simultaneously when；
Support_count (A) represents the support that A occurs.
Sequential mode mining is carried out to the section of different gradient type by algorithm above respectively, each type is finally obtained
The railway locomotive operation in section manipulates gear sequence.
Step S103, the manipulation gear sequence in the different type section for obtaining in step S102, by off-line optimization
Algorithm finds out the railway locomotive of optimal energy consumption in each type section and manipulates the corresponding time-sharing ratio example sequence of gear sequence.
The off-line optimization algorithm adopted in the step is existing conventional genetic algorithm.Genetic algorithm is search optimal solution
One of common method, railway locomotive operation is calculated using off-line optimization algorithm the corresponding time-sharing ratio example of gear sequence is manipulated
In sequence process, the whole mathematical model of genetic algorithm is as follows：
... ... ... ... formula 4
Wherein:vi<vi_lim
While TiMeet：And
X ∈ G ... ... ... ... ... ... ... formula 5
G ∈ U ... ... ... ... ... ... ... formula 6
Wherein E is locomotive overall operation energy consumption；I is step-length counting；H is step-length sum；giFor i step-lengths when gear；viFor
Locomotive speed during i step-lengths；TiFor i step-lengths when run time；△Ei(gi,vi) for i step-lengths when specific energy consumption；vi_limFor i
Operation speed limit during step-length；Δ T is total run time error；T is plan total run time；TmaxPermit for total run time error
Perhaps it is worth；X is decision variable, and it represents the corresponding time-sharing ratio example sequence of gear rule middle gear sequence formulated；G represents institute
The set being made up of the time-sharing ratio example sequence for meeting time-constrain and speed limiting constraint condition；U represents all time-sharing ratios
The fundamental space that example sequence is constituted.
The corresponding time-sharing ratio example sequence of manipulation gear sequence for obtaining optimum is found by traversal, according to this most afterwards
The excellent corresponding time-sharing ratio example sequence of gear sequence that manipulates obtains optimal operation energy consumption.
Still illustrate how by taking the section of gentle slope as an example below to be grasped from the locomotive that step S102 is obtained using off-line optimization algorithm
The corresponding time-sharing ratio example sequence of gear sequence of optimum is found in vertical gear sequence.Implement process as follows:
Maximum evolutionary generation T is set；
If it is { g that railway locomotive manipulates gear sequence1,g2,g3, wherein g1,g2,g3Correspond to respectively different gear items according to
Railway locomotive manipulates the self-defined linear three-dimensional vector of radix of gear sequence, generates several at random from solution space and manipulates gear
The corresponding time-sharing ratio example sequence { a of sequence1,b1,c1},{a2,b2,c2}…{ak,bk,ck, these time-sharing ratio example sequences
Set Q is constituted, wherein set Q is initial population P (0)；akRepresent operation g1The ratio of whole section of running time shared by gear item,
bkRepresent operation g2The ratio of whole section of running time, c shared by gear itemkRepresent operation g3Shared by gear item during whole section of traveling
Between ratio, wherein ak+bk+ck=100%。
Each time-sharing ratio example sequence in traversal set Q, calculates each individual（That is each time-sharing ratio example sequence
Row）Fitness, that is, correspond to each time-sharing ratio example sequence under it is corresponding manipulate gear sequence total energy consumption；By choosing
Select computing and retain the less time-sharing ratio example sequence of energy consumption, and the time-sharing ratio example sequence is arrived as advantage individual inheritance
It is of future generation.
The time-sharing ratio example sequence individual to advantage carries out intersection and averages computing, obtains new time-sharing ratio example
Sequence is individual.
Time assigned sequence for obtaining is individual, and by mutation operator the corresponding time scale ginseng of each gear item is obtained
Number, is suitably adjusted according to driving experience to the time scale parameter, obtains next generation colony P (t+1), and wherein t is represented and worked as
Front colony's algebraically is t for colony.During final t=T, then the corresponding time-sharing ratio of gear sequence with optimal energy consumption is obtained
Example sequence.
Can be seen that by intersection, mutation operator, from random time-sharing ratio by the genetic algorithm of previous step S103
Gradually approach to optimal or near optimal energy consumption in example sequence, it is possible to obtain the gear sequence in each type section is whole
The time-sharing ratio example sequence of the optimal energy consumption taken during the traveling of individual circuit.
Step S104, according to the railway locomotive of optimal energy consumption in each type section for obtaining gear sequence time point is manipulated
With ratio sequence, neural network model is built；And the time point found out according to off-line optimization algorithm based on the neural network model
With ratio sequence, the characteristic parameter that manipulation gear serial correlation is run to railway locomotive learns, and obtains the ferrum of different circuits
Road locomotive operation manipulates gear sequence, and it is associated with railway locomotive operation characteristic parameter.
The corresponding time-sharing ratio example sequence of manipulation gear sequence obtained in step S103 is due to being subject to several factors
Impact, rule of thumb, same type of section, the time scale of railway locomotive actual motion can be very different, not
More times adjust in time the railway locomotive of the optimal energy consumption of offline optimization and manipulate the corresponding time-sharing ratio example of gear sequence
Sequence, therefore the corresponding time scale of locomotive control gear sequence of the optimal energy consumption that offline optimization is learnt opens as one kind
Send out, build neural network model, analyze and the related characteristic parameter of conversion between concrete gear, to railway locomotive and circuit
Characteristic parameter（Such as grade information, locomotive overall length, locomotive load-carrying parameter）Learnt, it is final obtain special parameter certain
The optimal manipulation gear sequence adopted in concrete scope.
The structure of above-mentioned neural network model is techniques known, is not described in detail here
The all operations of above-mentioned offline part are performed in railway locomotive running, therefore offline part
Exploitation method for digging needs not be under the constraint of the factors such as time, and is obtained in that preferable optimum results.Through offline portion
The locomotive operation for separately winning manipulates gear sequence application on site in railway locomotive running, and concrete implementation status is as follows：
Step S105, manipulates the different corresponding locomotive operations of railway locomotive operation characteristic parameter gear sequence and is stored in
In policy library.
Offline part, to step S104, is obtained in that locomotive operation behaviour corresponding with each type section by step S101
Vertical gear sequence, the locomotive operation manipulates matching in gear sequence some railway locomotives and line characteristics parameter, such as gradient letter
Breath, railway locomotive load-carrying, railway locomotive length, if for empty wagons etc., the final offline locomotive operation manipulation gear sequence for obtaining
Preserved in the application in the form of dynamic link policy library, wherein every kind of locomotive operation manipulates gear sequence pair answers railway machine
Car and line characteristics parameter.
Step S106, when railway locomotive runs in certain circuit, according to the circuit and the traffic information of circuit, from plan
Corresponding locomotive operation is slightly matched in storehouse and manipulates gear sequence.
After line information is obtained, circuit is segmented according to the grade information of circuit（Steep to go up a slope, sharp decline surpasses
Sharp decline, gentle slope）, because the locomotive operation for generating in offline part optimization before manipulates gear sequence and different types of section
Matching（A kind of corresponding locomotive operation in each type of section manipulates gear sequence）, therefore may finally be according to grade information, ferrum
The features such as road locomotive load-carrying, railway locomotive length remove dynamic link policy library, and it is most suitable therefrom to search each type section
Locomotive operation manipulates gear sequence, and the locomotive operation for matching suitable railway locomotive operation manipulates gear sequence.The locomotive operation
Manipulate comprising the gear information run under railway locomotive different time in gear sequence, it is an orderly sequence results.
Step S107, manipulates the locomotive operation in each section for being exported gear sequence and constitutes whole piece by operation sequential
Route correspondence locomotive operation manipulates gear sequence.
If only having a type of section in whole piece route, this step S107 can also be skipped, directly performed next
Step.
Step S108, according to the locomotive operation obtained in step S107 the real-time fortune that gear sequence manipulates railway locomotive is manipulated
OK.
Step S109, actual operating data when railway locomotive application optimum locomotive operation is manipulated into gear sequence is recorded as
Railway locomotive raw operational data, and proceed to step S101.
In above-mentioned first embodiment, it is also possible to only including the implementation process of step S101 to step S108, such case
The manipulation gear sequence of ripe optimization is obtained suitable for having been directed towards specific railway locomotive.
Second embodiment of the invention provides a kind of from offline to the system that canbe used on line railway locomotive runs manipulation, the system
Structure chart as shown in Fig. 2 including the locomotive fortune for manipulating the offline generating means 10 of gear sequence and online part of offline part
The online manipulation device 20 of row.Wherein partly include offline：Locomotive operating data extraction unit 10-1, sequential mode mining unit
10-2, off-line data optimizing unit 10-3, neural network learning unit 10-4.Partly include online：Manipulate gear sequences match
Unit 20-1, locomotive control running unit 20-2.
Locomotive operating data extraction unit 10-1, for extracting railway locomotive raw operational data；
Sequential mode mining unit 10-2, for being directed to the locomotive raw operational data by sequential mode mining technology
Excavated, the locomotive operation for obtaining different type section manipulates gear sequence；
Off-line data optimizing unit 10-3, for manipulating gear sequence for the locomotive operation in resulting different type section
Row, by off-line optimization algorithm the corresponding time-sharing ratio example of locomotive control sequence of optimal energy consumption in each type section is found out
Sequence；
Neural network learning unit 10-4, for according to the locomotive control shelves of optimal energy consumption in each type section for obtaining
The corresponding time-sharing ratio example sequence of bit sequence, builds neural network model, and based on the neural network model according to seeking offline
The time-sharing ratio example sequence that excellent algorithm is found out, learns to the characteristic parameter that locomotive operation manipulates gear serial correlation, obtains
The locomotive operation for obtaining each bar circuit manipulates gear sequence, and it is associated with railway locomotive and line characteristics parameter；
Gear sequences match unit 20-1 is manipulated, for when railway locomotive runs in certain circuit, according to the railway machine
The characteristic parameter of car operation, matches corresponding locomotive operation and manipulates gear sequence；
Locomotive control running unit 20-2, the locomotive operation for being obtained according to matching manipulates gear sequence, manipulates railway
The real time execution of locomotive.
Third embodiment of the invention provides another kind of from offline to the system that canbe used on line railway locomotive runs manipulation, the reality
The structure of example is applied as shown in figure 3, in addition to all units in second embodiment are included, also including：Policy library 10-5.The policy library
10-5 is used to deposit different railway locomotive and the corresponding locomotive operation of line characteristics parameter manipulates gear sequence.
The present invention can be seen that first by sequential mode mining method from a large amount of railways by the technical scheme of foregoing invention
Offline acquisition manipulates gear sequence in locomotive operation initial data, then carries out manipulating tool in gear sequence by offline optimization algorithm
The optimizing of the time-sharing ratio example sequence of body gear, neutral net mould is built by the time-sharing ratio example sequence of optimal energy consumption
Type, and obtain railway locomotive special parameter according to the neural network model study and manipulate the correlation rule of gear sequence, so as to
Finally using the manipulation gear sequence of optimal energy consumption come online direction locomotive operation.Because calculated off line is not by time factor
Limit, therefore using the offline manipulation shelves for obtaining in the preferably optimization space, and railway locomotive running that have part ownership offline
Good energy-saving effect can be obtained during bit sequence.
In addition, using the online operation and control result of railway locomotive as offline sequential mode mining method and the number of optimizing algorithm
According to input so that the data that off-line learning can be returned constantly by online partial feedback are adjusted and optimization, offline part
Combining closely and influencing each other with online part, has weighed offline optimization algorithm effect and on-line optimization calculates matter of time,
Railway locomotive can be made to manipulate gear sequence and to obtain constantly optimization improvement, so as to be finally reached more preferable energy-saving effect.
Although the present invention is disclosed as above with preferred embodiment, embodiment is not for limiting the present invention's.Not
Depart from the spirit and scope of the present invention, any equivalence changes done or retouching also belong to the protection domain of the present invention.Cause
The content that this protection scope of the present invention should be defined with claims hereof is as standard.
Claims (6)
1. it is a kind of from offline to the method that canbe used on line railway locomotive runs manipulation, it is characterised in that described method includes：
Extract railway locomotive raw operational data；
Dug for the railway locomotive raw operational data by the sequential mode mining technology based on Apriori algorithm
Pick, according to the locomotive operation that the formula of following confidence level obtains different type section gear sequence is manipulated：
Confidence(A->B)=P (B | A)=support_count (AB)/support_count (A)
... ... ... formula 3
In formula 3, Confidence (A->B) confidence level is represented；A represents manipulation gear item gm,gn；B represents manipulation gear item gk；
P (B | A) represents the probability of the B under the conditions of A；Support support_count (AB) represents that A, B occur simultaneously when；
Support_count (A) represents the support that A occurs；
Gear sequence is manipulated for the locomotive operation in resulting different type section, by the off-line optimization based on genetic algorithm
Algorithm sets up following mathematical model, and finds out the manipulation gear sequence of optimal energy consumption in each type section according to the mathematical model
Corresponding time-sharing ratio example sequence：
Wherein:vi＜ vi_lim
While TiMeet：And Δ T≤Tmax
X ∈ G ... ... ... ... ... ... ... formula 5
G ∈ U ... ... ... ... ... ... ... formula 6
Wherein E is locomotive overall operation energy consumption；I is step-length counting；H is step-length sum；giFor i step-lengths when gear；viFor i steps
Locomotive speed when long；TiFor i step-lengths when run time；ΔEi(gi,vi) for i step-lengths when specific energy consumption；vi_limFor i steps
Operation speed limit when long；Δ T is total run time error；T is plan total run time；TmaxFor the permission of total run time error
Value；X is decision variable, and it represents the corresponding time-sharing ratio example sequence of gear rule middle gear sequence formulated；G represents all
Meet the set of the time-sharing ratio example sequence composition of time-constrain and speed limiting constraint condition；U represents all time-sharing ratio examples
The fundamental space that sequence is constituted；
According to the corresponding time-sharing ratio example sequence of manipulation gear sequence for obtaining, neural network model is built, and based on the god
Jing network modeies according to the time-sharing ratio example sequence found out by off-line optimization algorithm, pair with the time-sharing ratio example sequence
Corresponding locomotive operation manipulates the characteristic parameter of gear serial correlation and is learnt, and the locomotive operation for obtaining each bar circuit manipulates shelves
Bit sequence, and it is associated with railway locomotive and line characteristics parameter；
When railway locomotive runs in certain circuit, according to the railway locomotive and line characteristics parameter, corresponding locomotive fortune is matched
Row manipulates gear sequence；
Gear sequence is manipulated according to the locomotive operation that matching is obtained, the real time execution of railway locomotive is manipulated.
2. according to claim 1 a kind of from the method that manipulates is run to canbe used on line railway locomotive offline, its feature exists
In methods described also includes：
Railway locomotive real-time running data when manipulating gear sequence using locomotive operation is recorded as into the original operation of railway locomotive
Data.
3. according to claim 1 and 2 a kind of from offline to the method that canbe used on line railway locomotive runs manipulation, its feature
It is to manipulate gear sequence in the locomotive operation for obtaining each bar circuit, and it is related to railway locomotive and line characteristics parameter
After the process of connection, also include：
Different railway locomotives and the corresponding locomotive operation of line characteristics parameter are manipulated into gear sequence to be stored in policy library.
4. according to claim 3 a kind of from the method that manipulates is run to canbe used on line railway locomotive offline, its feature exists
In, it is described to be excavated for the railway locomotive raw operational data by sequential mode mining technology, obtain different type
The locomotive operation in section manipulates the process of gear sequence, specifically includes：
The gradient that adds is obtained according to the line information of railway locomotive traveling, is divided into railway locomotive circuit not according to the gradient that adds
The section of same type；
For different type section in same circuit, different type section is calculated using Sequential Pattern Mining Algorithm corresponding
Locomotive operation manipulates gear sequence.
5. it is a kind of from offline to the system that canbe used on line railway locomotive runs manipulation, it is characterised in that described system includes：
Locomotive operating data extraction unit, for extracting railway locomotive raw operational data；
Sequential mode mining unit, for being directed to the railway machine by the sequential mode mining technology based on Apriori algorithm
Car raw operational data is excavated, and according to the locomotive operation that the formula of following confidence level obtains different type section gear is manipulated
Sequence：
Confidence(A->B)=P (B | A)=support_count (AB)/support_count (A)
... ... .... formula 3
In formula 3, Confidence (A->B) confidence level is represented；A represents manipulation gear item gm,gn；B represents manipulation gear item gk；
P (B | A) represents the probability of the B under the conditions of A；Support support_count (AB) represents that A, B occur simultaneously when；
Support_count (A) represents the support that A occurs；
Off-line data optimizing unit, for manipulating gear sequence for the locomotive operation in resulting different type section, passes through
Following mathematical model is set up based on the off-line optimization algorithm of genetic algorithm, and is found out in each type section according to the mathematical model
The corresponding time-sharing ratio example sequence of manipulation gear sequence of optimal energy consumption：
Wherein:vi＜ vi_lim
While TiMeet：And Δ T≤Tmax
X ∈ G ... ... ... ... ... ... ... formula 5
G ∈ U ... ... ... ... ... ... ... formula 6
Wherein E is locomotive overall operation energy consumption；I is step-length counting；H is step-length sum；giFor i step-lengths when gear；viFor i steps
Locomotive speed when long；TiFor i step-lengths when run time；ΔEi(gi,vi) for i step-lengths when specific energy consumption；vi_limFor i steps
Operation speed limit when long；Δ T is total run time error；T is plan total run time；TmaxFor the permission of total run time error
Value；X is decision variable, and it represents the corresponding time-sharing ratio example sequence of gear rule middle gear sequence formulated；G represents all
Meet the set of the time-sharing ratio example sequence composition of time-constrain and speed limiting constraint condition；U represents all time-sharing ratio examples
The fundamental space that sequence is constituted；
Neural network learning unit, for according to the corresponding time-sharing ratio example sequence of manipulation gear sequence for obtaining, building god
Jing network modeies, and based on the neural network model according to the time-sharing ratio example sequence found out by off-line optimization algorithm, it is right
Locomotive operation manipulates the characteristic parameter of gear serial correlation and is learnt, and the locomotive operation for obtaining each bar circuit manipulates gear sequence
Row, and it is associated with railway locomotive and line characteristics parameter；
Gear sequences match unit is manipulated, for when railway locomotive runs in certain circuit, according to the railway locomotive and circuit
Characteristic parameter, matches corresponding locomotive operation and manipulates gear sequence；
Locomotive control running unit, the locomotive operation for being obtained according to matching manipulates gear sequence, manipulates the reality of railway locomotive
Shi Yunhang.
6. according to claim 5 a kind of from the system that manipulates is run to canbe used on line railway locomotive offline, its feature exists
In described system also includes：
Policy library, for depositing different railway locomotive and the corresponding locomotive operation of line characteristics parameter gear sequence is manipulated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410117028.3A CN103870892B (en) | 2014-03-26 | 2014-03-26 | Method and system for achieving railway locomotive operation control from off-line mode to on-line mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410117028.3A CN103870892B (en) | 2014-03-26 | 2014-03-26 | Method and system for achieving railway locomotive operation control from off-line mode to on-line mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103870892A CN103870892A (en) | 2014-06-18 |
| CN103870892B true CN103870892B (en) | 2017-05-10 |
Family
ID=50909405
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410117028.3A Active CN103870892B (en) | 2014-03-26 | 2014-03-26 | Method and system for achieving railway locomotive operation control from off-line mode to on-line mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103870892B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106647279B (en) * | 2017-01-13 | 2019-11-12 | 清华大学 | A kind of locomotive smart steering optimized calculation method based on fuzzy rule |
| CN106802553B (en) * | 2017-01-13 | 2019-10-11 | 清华大学 | A kind of railway locomotive operation control system hybrid tasks scheduling method based on intensified learning |
| CN106844621B (en) * | 2017-01-18 | 2020-06-26 | 清华大学 | Method for constructing energy-saving operation real-time optimization control strategy library of rail locomotive |
| CN106842925B (en) * | 2017-01-20 | 2019-10-11 | 清华大学 | A kind of locomotive smart steering method and system based on deeply study |
| CN107515537A (en) * | 2017-08-21 | 2017-12-26 | 南京理工大学 | A kind of train bicycle energy-saving operating method based on modified force search method |
| CN111664927B (en) * | 2020-05-28 | 2022-04-26 | 首钢京唐钢铁联合有限责任公司 | Method and device for judging metering state of rail weigher |
| CN116756808B (en) * | 2023-05-23 | 2024-04-23 | 济青高速铁路有限公司 | Railway line shape optimization method and device, electronic equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19950186A1 (en) * | 1999-10-19 | 2001-04-26 | Abb Patent Gmbh | Arrangement for optimized operation of journeys in track connected traffic, with which vehicles belonging to traffic undertakings or vehicle groups are controlled to determine |
| WO2002049900A1 (en) * | 2000-12-20 | 2002-06-27 | Central Queensland University | Vehicle dynamics prediction system and method |
| CN101214819A (en) * | 2008-01-22 | 2008-07-09 | 北京交通大学 | Method and guiding device for diesel traction engine optimizing operation |
| CN102275601A (en) * | 2011-04-29 | 2011-12-14 | 北京全路通信信号研究设计院有限公司 | Method and device for intelligently controlling train |
| CN102509170A (en) * | 2011-10-10 | 2012-06-20 | 浙江鸿程计算机系统有限公司 | Location prediction system and method based on historical track data mining |
| CN102566416A (en) * | 2011-12-20 | 2012-07-11 | 北京交通大学 | Intelligent identification and evaluation method for operation condition of hybrid shunting locomotive |
| CN102981408A (en) * | 2012-12-10 | 2013-03-20 | 华东交通大学 | Running process modeling and adaptive control method for motor train unit |
| US8412393B2 (en) * | 2008-07-01 | 2013-04-02 | General Electric Company | Apparatus and method for monitoring of infrastructure condition |
-
2014
- 2014-03-26 CN CN201410117028.3A patent/CN103870892B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19950186A1 (en) * | 1999-10-19 | 2001-04-26 | Abb Patent Gmbh | Arrangement for optimized operation of journeys in track connected traffic, with which vehicles belonging to traffic undertakings or vehicle groups are controlled to determine |
| WO2002049900A1 (en) * | 2000-12-20 | 2002-06-27 | Central Queensland University | Vehicle dynamics prediction system and method |
| CN101214819A (en) * | 2008-01-22 | 2008-07-09 | 北京交通大学 | Method and guiding device for diesel traction engine optimizing operation |
| US8412393B2 (en) * | 2008-07-01 | 2013-04-02 | General Electric Company | Apparatus and method for monitoring of infrastructure condition |
| CN102275601A (en) * | 2011-04-29 | 2011-12-14 | 北京全路通信信号研究设计院有限公司 | Method and device for intelligently controlling train |
| CN102509170A (en) * | 2011-10-10 | 2012-06-20 | 浙江鸿程计算机系统有限公司 | Location prediction system and method based on historical track data mining |
| CN102566416A (en) * | 2011-12-20 | 2012-07-11 | 北京交通大学 | Intelligent identification and evaluation method for operation condition of hybrid shunting locomotive |
| CN102981408A (en) * | 2012-12-10 | 2013-03-20 | 华东交通大学 | Running process modeling and adaptive control method for motor train unit |
Non-Patent Citations (2)
| Title |
|---|
| "列车节能操纵优化方法研究";金炜东等;《铁道学报》;19971231;第19卷(第6期);第58页第16-20行、第59页第8-14、26-30行、第60页第1-16行 * |
| "我国铁路运输系统节能问题的研究分析";胡辉;《华东交通大学学报》;20111231;第28卷(第6期);第3.1节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103870892A (en) | 2014-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103870892B (en) | Method and system for achieving railway locomotive operation control from off-line mode to on-line mode | |
| CN106842925A (en) | A kind of locomotive smart steering method and system based on deeply study | |
| CN103847749A (en) | Method for generating double-deck multi-objective locomotive optimized manipulating sequence | |
| Larki et al. | Solving the multiple traveling salesman problem by a novel meta-heuristic algorithm | |
| CN103218660B (en) | A kind of airway selection method based on extensive fuzzy competition nerve net | |
| CN107368920A (en) | A kind of off-peak period multi-train movement energy conservation optimizing method | |
| CN113536509B (en) | Micro-grid topology identification method based on graph convolution network | |
| CN104616062A (en) | Nonlinear system recognizing method based on multi-target genetic programming | |
| CN106570562A (en) | Adaptive-DE-algorithm-based fuzzy modeling method for bridge crane | |
| CN111271071A (en) | Shield tunneling machine attitude control method based on fuzzy adaptive neural network | |
| CN112862065A (en) | Power grid branch parameter identification method based on multitask graph neural network | |
| CN112183288B (en) | Multi-agent reinforcement learning method based on model | |
| CN105512755A (en) | Decomposition-based multi-objective distribution estimation optimization method | |
| CN106647269B (en) | A kind of locomotive smart steering optimized calculation method | |
| Harandi et al. | A new feature selection method based on task environments for controlling robots | |
| Cheng et al. | Intelligent driving methods based on sparse LSSVM and ensemble CART algorithms for high-speed trains | |
| CN106647279B (en) | A kind of locomotive smart steering optimized calculation method based on fuzzy rule | |
| Zhang et al. | A network traffic prediction model based on quantum inspired PSO and neural network | |
| CN108181816A (en) | A kind of synchronization policy update method for optimally controlling based on online data | |
| Amrani et al. | Train speed profiles optimization using a genetic algorithm based on a random-forest model to estimate energy consumption | |
| CN103679267A (en) | Method and device for constructing RBF neural network based on unmarked samples | |
| CN104598741B (en) | A kind of construction method of track saturation degree forecast model | |
| Shi et al. | A discrete particle swarm optimization algorithm for travelling salesman problem | |
| CN107239599A (en) | Based on Ground surface settlement method caused by the shield-tunneling construction of neural fuzzy inference system | |
| CN107139777A (en) | A kind of vehicle energy management method and its system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 100084 Beijing city Tsinghua University Haidian District Huaye building area four three Applicant after: BEIJING THSOFT INFORMATION TECHNOLOGY CO., LTD. Address before: 100084 Beijing city Tsinghua University Haidian District Huaye building area four three Applicant before: T H Soft Information Technology Co., Ltd. |
|
| COR | Change of bibliographic data | ||
| CB03 | Change of inventor or designer information |
Inventor after: Huang Jin Inventor after: Yang Ying Inventor after: Chen Xinjie Inventor after: Du Fangyu Inventor after: Chen Cuanyue Inventor after: Cang Dacuan Inventor before: Huang Jin Inventor before: Chen Xinjie Inventor before: Du Fangyu Inventor before: Chen Cuanyue Inventor before: Cang Dacuan |
|
| COR | Change of bibliographic data | ||
| C41 | Transfer of patent application or patent right or utility model | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20161207 Address after: 100084 Beijing city Tsinghua University Haidian District Huaye building area four three Applicant after: BEIJING THSOFT INFORMATION TECHNOLOGY CO., LTD. Applicant after: CRRC DALIAN INSTITUTE CO., LTD. Address before: 100084 Beijing city Tsinghua University Haidian District Huaye building area four three Applicant before: BEIJING THSOFT INFORMATION TECHNOLOGY CO., LTD. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |