CN107516107A - A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power - Google Patents

A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power Download PDF

Info

Publication number
CN107516107A
CN107516107A CN201710645911.3A CN201710645911A CN107516107A CN 107516107 A CN107516107 A CN 107516107A CN 201710645911 A CN201710645911 A CN 201710645911A CN 107516107 A CN107516107 A CN 107516107A
Authority
CN
China
Prior art keywords
mrow
msub
operating mode
speed
forecasting methodology
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.)
Pending
Application number
CN201710645911.3A
Other languages
Chinese (zh)
Inventor
王伟达
韩立金
丁峰
项昌乐
马越
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Institute of Technology BIT
Original Assignee
Beijing Institute of Technology BIT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Institute of Technology BIT filed Critical Beijing Institute of Technology BIT
Priority to CN201710645911.3A priority Critical patent/CN107516107A/en
Publication of CN107516107A publication Critical patent/CN107516107A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/29—Graphical models, e.g. Bayesian networks
    • G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G07—CHECKING-DEVICES
    • G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
    • G07C5/00—Registering or indicating the working of vehicles
    • G07C5/08—Registering or indicating performance data other than driving, working, idle, or waiting time, with or without registering driving, working, idle or waiting time
    • G07C5/0808—Diagnosing performance data

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention discloses a kind of producing condition classification Forecasting Methodology of vehicle driving-cycle, this method includes：Vehicle driving-cycle is categorized as steady operating mode and the fast class of variable working condition two under off-line state using K mean cluster algorithm, and operating mode classification is presently in on-line stage real-time judge vehicle；For steady operating mode, using based on markovian speed prediction method, and be directed to fast variable working condition, using the speed prediction method based on radial base neural net, the advantages of two methods are comprehensively utilized with this prediction effect to be optimal.

Description

A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power
Technical field
The present invention relates to a kind of Forecasting Methodology of vehicle driving-cycle, the traveling work of especially a kind of motor vehicle driven by mixed power Condition classification Forecasting Methodology.
Background technology
Motor vehicle driven by mixed power is one of current effective way for solving vehicle energy and overrunning with air pollution quality. Wherein double mode transmission system can better meet heavy non-rice habitats car compared to the hybrid power transmission scheme of other forms The specific demands such as speed adjustable range is wide and driving power is big, but program structure is complex, and energy control method is wanted Ask higher, the optimal energy control method that design to use in real time, which will turn into, ensures that double mode motor vehicle driven by mixed power can The core content normally and efficiently run.
At present, most commonly used in industrial quarters is rule-based energy control method, and regular design mostly derives from Heuristic finding and engineer experience, although its design is simple, it is easy to accomplish, it is poor to different adaptability for working condition, it can not obtain To optimal control effect.In order to pursue more preferable control effect, academia has done substantial amounts of scientific research and explored based on optimization Energy control method, its main thought is to establish aims of systems cost function and constraints, is solved by optimized algorithm To optimum control amount.Wherein dynamic programming algorithm is most widely used, but it needs to know global operating mode in advance, so only It can be used to emulate.Equivalent fuel consumption strategy can be used with real-time online, but be had for different operating mode Reliability equivalence factors hardly possible The drawbacks of to set.And predictive control algorithm (Model Predictive Control, MPC) developed in recent years is adopted Tested with multistep, the thinking of rolling optimization and feedback compensation, obtained good real-time control effect.This method is largely Upper depend on effectively is predicted to following speed, assumes that following speed keeps constant in the prior art；Or assume speed Exponentially change, these methods are simply but inaccurate；Or obtain vehicle future travel car by onboard navigation system Speed；Or following speed is predicted by identifying the repetition operating mode of Special Work vehicle, these methods are needed by GPS system Or priori work information, it is not particularly suited for no alignment system and perceives the off roader of radar.
The content of the invention
For problem above, the present invention proposes a kind of driving cycle classification Forecasting Methodology, using K averages (Kmeans) Clustering algorithm is by producing condition classification, and the side for different types of operating mode using Markov Chain or radial base neural net Method is predicted to following speed, realizes the improvement of non-rice habitats double mode motor vehicle driven by mixed power performance.
The main object of the present invention is to provide a kind of operating mode Forecasting Methodology of vehicle driving-cycle, especially a kind of mixing The producing condition classification Forecasting Methodology of power car driving cycle.
The purpose of the present invention can be realized by following approach:
A kind of producing condition classification Forecasting Methodology of vehicle driving-cycle, this method include：
Vehicle driving-cycle is categorized as steady operating mode and fast variable working condition under off-line state using K mean cluster algorithm Two classes, and it is presently in operating mode classification in on-line stage real-time judge vehicle；
For steady operating mode, using based on markovian speed prediction method, and fast variable working condition is directed to, using base In the speed prediction method of radial base neural net, the advantages of comprehensively utilizing two methods with this, is imitated with the prediction being optimal Fruit.
The producing condition classification Forecasting Methodology of the present invention, further, wherein according to following characteristics parameter：Peak acceleration (m/ S2), maximum deceleration (m/s2), average acceleration (m/s2), vehicle speed standard variance (km/h), max. speed and minimum speed Difference (km/h), acceleration standard variance (m/s2) classifies vehicle driving-cycle.
The producing condition classification Forecasting Methodology of the present invention, further, wherein using K mean cluster algorithm, by calculating sample Between close and distant degree carry out data classification, finally realize that the data in same class have larger characteristic similarity, inhomogeneity Between then differ greatly, specific operating mode judgment step is as follows：
Off-line phase：(1) multiple standard cycle operating modes are combined and forms sample；
(2) each sampling instant calculates the operating mode feature parameter of 10 seconds in the past in state of cyclic operation, obtains characteristic parameter sample Notebook data [x11,x12,...,x1m], [x21,x22,...,x2m] ... ..., [xn1,xn2,...,xnm], wherein m is characterized parameter sequence Number, n is state of cyclic operation length；
(3) K mean cluster algorithm is applied, randomly selects cluster centre c1=[c11,c12,...,c1m], c2=[c21, c22,...,c2m], the distance of all samples and cluster centre is calculated, and sample is grouped according to Nearest Neighbor Method, ownership is different θm(k) Clustering Domain, wherein k are iterations, then adjust cluster centre as the following formula：
If cm(k+1)≠cm(k), then continue to adjust cluster centre, until cluster centre no longer changes, then it is assumed that classification It is stable, obtain the cluster centre c1 of steady operating mode and the cluster centre c2 of fast variable working condition；
On-line stage：(1) during vehicle actual travel, current sample time calculate over 10 seconds operating mode it is special Levy parameter value [x1,x1,...,xm]；
(2) characteristic ginseng value [x is calculated according to following formula1,x1,...,xm] to two cluster centres c1 and c2 distance d：
In formula：J=1,2 correspond to two class operating modes；
(3) if d1≤d2, then current time is judged for steady operating mode, if d1＞ d2, then current time is judged to exchange work soon Condition.
The producing condition classification Forecasting Methodology of the present invention, further, wherein assuming vehicle at each moment under steady operating mode Acceleration is unrelated with historical information, is only determined by current information, it is thus regarded that the acceleration change of vehicle is a kind of Markov Process, the changing rule of speed and acceleration is simulated using Markov chain model, and to following speed under steady operating mode It is predicted.
The producing condition classification Forecasting Methodology of the present invention, further, wherein for fast variable working condition, with radial direction base nerve net Network is theoretical, and the prediction of following speed is carried out by the study online to driver's driving behavior.
The producing condition classification Forecasting Methodology of the present invention, further, wherein radial basis neural network is according to history speed With current driver's pedal information, following speed is predicted, meanwhile, current caused information of vehicles is to believe as new history Breath, chosen by self-organizing center and pseudoinverse technique determines that the method for weights realizes the adaptive on-line study of neutral net.
The invention also provides a kind of producing condition classification Forecasting Methodology of hybrid car travel operating mode, this method uses Any of the above-described limited operating mode Forecasting Methodology.
Brief description of the drawings
Fig. 1 is a kind of double mode hybrid vehicle system structural representation.
Fig. 2 is the energy control method flow chart of the present invention.
Fig. 3 is the state of cyclic operation schematic diagram that the present invention combines.
Fig. 4 is the Markov chain model transition probability matrix figure of the present invention.
Fig. 5 is the radial base neural net speed prediction structure chart of the present invention.
Fig. 6 is the online operating mode judged result figure of the present invention.
Fig. 7 is the speed prediction result figure of the present invention.
Fig. 8 is part state of cyclic operation prediction error comparison diagram.
Fig. 9 be the present invention speed, EVT Mode, battery SOC simulation result figure.
Figure 10 is the engine of the present invention, motor A and motor B rotational speed and torque simulation result figure.
Figure 11 is the engine working point distributed simulation result figure of the present invention.
Figure 12 is the engine working point simulation result contrast of the present invention：(a) PREDICTIVE CONTROL, (b) keep control, and (c) is dynamic State plans that (d) is regular.
Figure 13 is the battery SOC simulation result contrast of the present invention：(a) PREDICTIVE CONTROL, (b) keep control, and (c) dynamic is advised Draw, (d) rule.
Embodiment
Illustrate the embodiment of the present invention by taking certain type double mode motor vehicle driven by mixed power as an example.System architecture such as Fig. 1 institutes Show, major parameter is as shown in table 1.Two kinds of hybrid electric working pattern EVT1 (Electronically Controlled of system Continuously Variable Transmission, EVT) switching of pattern and EVT2 patterns can be by manipulating clutch Device and brake realize that work as clutch separation, brake is EVT1 patterns when engaging；When clutch engages, brake separates When be EVT2 patterns.
The double mode motor vehicle driven by mixed power major parameter of table 1
In the modeling of Control-oriented, ignore planetary gear inertia and each element between friction, and assume connection be all It is rigid, actuation system models can be obtained,
EVT1 patterns：
EVT2 patterns：
In formula：k1, k2, k3The intrinsic parameter of respectively three planet rows, ωA, ωBFor motor A, B rotating speed, TA, TBFor Motor A, B torque, ωi, ωoFor coupling mechanism input and output end rotating speed, Ti, ToRespectively coupling mechanism input and Output end torque.
Simultaneously because the mechanical connection between part, system also meet relationship below：
ωe=iqωi (5)
V=rwωo/if (6)
Te=Ti/iq (7)
Tw=ifTo (8)
In formula：ωeFor engine speed, TeFor motor torque, V is speed, TwFor output torque on wheel, iqTo be preceding Transmission gear ratio, rwFor radius of wheel, ifFor rear transmission gear ratio.
Engine mockup uses the MAP model constructed by test data, it is assumed that engine preheats completely, its fuel oil Consumption rate is the static function of rotational speed and torque：
In formula：mfConsumed for engine fuel, feFor MAP.
The state-of-charge SOC of battery is a significant variable in motor vehicle driven by mixed power energy control method, is used here Internal resistance model modeling, and ignore the influence of temperature, obtain following SOC expression formulas：
In formula：VocFor battery open circuit voltage, RbattFor the internal resistance of cell, CbattFor battery capacity, ηA, ηBRespectively motor A, B Efficiency, index kA, kBWhen motor charges the battery, equal to 1, when motor is battery discharge, equal to -1.
Lateral direction of car and vertical motion are not considered, ignores the gradient, then can obtain vehicle according to vehicle traveling equilibrium equation moves Mechanical model：
In formula：M is complete vehicle quality, and ρ is atmospheric density, CdFor coefficient of air resistance, AfFor vehicle front face area, μ is wheel Coefficient of rolling resistance, g are acceleration of gravity.
The energy control method main purpose of double mode motor vehicle driven by mixed power is to meet traveling demand and system restriction Under the conditions of, online reasonable distribution demand power, engine working point is adjusted, to obtain more preferably fuel economy, and maintain electricity Pond SOC.The present invention carries out the real-time optimization of power distribution based on predictive control algorithm online, chooses engine speed and torque For system control amount u, definition system state amount is x, and systematic observation input quantity is v, and system output quantity is y, then can will be towards Control system model is expressed as：
In formula：X=[SOC], u=[Teωe]T, v=[V Tw]T, y=[SOC m＆f]T。
In each sampling instant k, predict that the optimization object function in time domain is：
In formula：wsAnd wmThe respectively weight coefficient of respective items, SOCrFor battery SOC reference value, P is prediction time domain.Together When, following physical constraint needs to be satisfied：
In formula：*_maxWith *_minThe respectively bound of respective items.
By system model discretization during solving-optimizing problem, because prediction time domain is shorter, and battery SOC each moment can Row domain scope is small, so dynamic programming algorithm can be used to online Real-time solution optimization problem, it is assumed that U*(k)=[u* (k),...,u*(k+P-1) it is] the optimum control amount sequence in prediction time domain, then controlled quentity controlled variable used by current time system For
U (x (k))=u*(k) (15)
The core concept of PREDICTIVE CONTROL is exactly to be asked in each sampling instant solving an optimization in limited prediction time domain Topic, the optimal control sequence in prediction time domain is calculated, but only implement the optimum control of the sampling instant and give up other controls Amount, then repeat this process in next sampling instant.PREDICTIVE CONTROL is used for double mode motor vehicle driven by mixed power energy control method In, you can according to current driver's pedal information, and the information of vehicles such as speed, battery SOC, engine speed torque are combined, led to Cross real-time optimization and carry out power distribution, improve VE Vehicle Economy.Due to whole state of cyclic operation can not be predicted, so the plan Summary can not obtain globally optimal solution, but be their ability to on-line implement, and global near-optimization is obtained by way of rolling optimization Solution, while can take into account uncertain caused by the factors such as model mismatch, interference so that control keeps the optimal of reality. In each sampling instant k, energy control method flow chart is as shown in Fig. 2 specifically, will follow the steps below：
(1) current system conditions, including vehicle speed of operation, driver pedal information, battery SOC etc. are observed.
(2) judge current EVT Mode state with driver pedal information according to vehicle speed of operation, and update current system Model and system restriction.And assume that EVT Mode state keeps constant in prediction time domain.
(3) the following speed in prediction time domain is predicted, obtains predicting systematic observation input quantity v in time domain, in detail Method will be introduced below.
(4) the structure forecast control optimization problem in prediction time domain, and numerical value is carried out by dynamic programming algorithm online and asked Solution.
(5) optimal control sequence in prediction time domain is calculated.
(6) only with first group of optimum control amount, system is acted in the sampling instant, gives up remaining controlled quentity controlled variable.
(7) this process is repeated in subsequent time.
It is how rationally accurate using vehicle history and current data in the case of no any driving cycle prior information Prediction vehicle future speed, will largely influence energy control method effect of optimization.The present invention is gathered using K averages Producing condition classification is steady operating mode and the fast class of variable working condition two under off-line state by class algorithm, and in on-line stage real-time judge car It is presently in operating mode classification.For steady operating mode, using based on markovian speed prediction method, and become for fast Operating mode, using the speed prediction method based on radial base neural net, the advantages of comprehensively utilizing two methods with this, is to reach most Excellent prediction effect.Meanwhile it will predict that speed substitutes into the demand torque that formula (11) can be calculated in prediction time domain.
Steady operating mode and fast variable working condition main difference is that the size of the fluctuation of speed and acceleration in operating mode, in order to Two kinds of operating mode types are distinguished, it is necessary to according to the characteristic parameter in operating mode by two kinds of producing condition classifications, table 2 is the operating mode feature chosen Parameter.
The operating mode feature parameter of table 2
Using K mean cluster algorithm, data classification is carried out by calculating the close and distant degree between sample, is finally realized same Data in class have larger characteristic similarity, are then differed greatly between inhomogeneity, and specific operating mode judgment step is as follows：
Off-line phase：(1) multiple standard cycle operating modes are combined and forms sample, as shown in Figure 3.
(2) each sampling instant calculates the operating mode feature parameter of 10 seconds in the past in state of cyclic operation, obtains characteristic parameter sample Notebook data [x11,x12,...,x1m], [x21,x22,...,x2m] ... ..., [xn1,xn2,...,xnm], wherein m is characterized parameter sequence Number, n is state of cyclic operation length.
(3) K mean cluster algorithm is applied, randomly selects cluster centre c1=[c11,c12,...,c1m], c2=[c21, c22,...,c2m], the distance of all samples and cluster centre is calculated, and sample is grouped according to Nearest Neighbor Method, ownership is different θm(k) Clustering Domain, wherein k are iterations, then adjust cluster centre as the following formula：
If cm(k+1)≠cm(k), then continue to adjust cluster centre, until cluster centre no longer changes, then it is assumed that classification It is stable, obtain the cluster centre c1 of steady operating mode and the cluster centre c2 of fast variable working condition.
On-line stage：(1) during vehicle actual travel, the operating mode feature of 10 seconds is calculated in current sample time Parameter value [x1,x1,...,xm]。
(2) characteristic ginseng value [x is calculated according to following formula1,x1,...,xm] to two cluster centres c1 and c2 distance d：
In formula：J=1,2 correspond to two class operating modes.
(3) if d1≤d2, then current time is judged for steady operating mode, if d1＞ d2, then current time is judged to exchange work soon Condition.
Assuming that vehicle is unrelated with historical information in the acceleration at each moment, only determined by current information, it is thus regarded that car Acceleration change be a kind of markoff process, here can using Markov chain model come simulate speed with plus The changing rule of speed, and following speed is predicted under steady operating mode[18,19]。
According to different driver pedal aperture α≤0,0 ＜ α≤0.2,0.2 ＜ α≤0.4,0.4 ＜ α≤0.6,0.6 ＜ α ≤ 0.8,0.8 ＜ α≤1, establish six groups of corresponding single order Markov chain models.Each group of Markov chain model is by car Fast V (0 to 30m/s) and acceleration a (- 1.5 to 1.5m/s2) discrete mesh space is formed, definition car speed is current shape State amount, p section is divided into, is indexed by i ∈ { 1 ..., p }, definition vehicle acceleration is subsequent time output quantity, will It is divided into q section, is indexed by j ∈ { 1 ..., q }.Then the transition probability matrix T of each group of Markov chain model can be with It is expressed as：
In formula：n∈{1,...,NpIt is the required object time for predicting speed, T in prediction time domainijFor at current time Vehicle velocity Vk+n=ViIn the case of, vehicle acceleration is developed to a in subsequent timejProbability.
In an initial condition, the steady operating mode of selection typical case, Markov chain model transition probability is calculated according to following formula Matrix,
In formula：NijIt is that i subsequent times are the number that j occurs for current time.It is in driver pedal aperture shown in Fig. 4 During 0 ＜ α≤0.2, Markov chain model transition probability matrix.
In real time execution, Markov chain model needs online self-recision with the change of adaptation condition, when current K is carved, if previous moment vehicle velocity Vk-1=Vi, this moment generation ak=aj, then Markov chain tra nsfer probability square under this event Battle array adaptive correction be：
T(k)ij=T (k-1)ij+λ (20)
In formula：S ∈ { 1 ..., q }, s ≠ j, λ are adaptation coefficient.Formula (20) observes this thing occurred at current time Part, and by the probability amendment of this event in Markov Chain transition probability matrix, formula (21) is corrected should when this event occurs The probability of other output valves under state.It is noted that during actual adaptive correction, current time transition probability matrix In only a row probability data be updated, other probability keep constant.
According to above Markov chain model, you can predict subsequent time vehicle acceleration in current time k, and obtain Subsequent time speed：
Similarly, predict that the speed at each moment in time domain can be calculated by last moment speed：
In formula：N≤P is each object time in prediction time domain.
Following speed can effectively be predicted under steady operating mode based on markovian Forecasting Methodology, but become soon Under operating mode can not effectively learner driver's behavior, cause its precision of prediction poor.So for fast variable working condition, present invention fortune It is theoretical with radial base neural net, the prediction of following speed is carried out by the study online to driver's driving behavior.
Radial base neural net is a kind of partial approximation network, compared to the neutral net of other forms, its convergence rate Fast and amount of calculation is small, is best suitable for being applied to the online speed prediction of motor vehicle driven by mixed power[20].Here, neural network model is defined Input NinFor driver pedal information and the speed of the past period：
In formula：HhFor past speed vector length.The output N of modeloutFor the prediction speed of following a period of time：
Nout=Vk+1,Vk+2,...,Vk+P(25)
Neuron uses Gaussian function as RBF in hidden layer：
In formula：yjExported for neutral net, ωijTo export weights, bfFor the default neuron threshold value of developer, x is god Through network inputs, ciFor neuron node center, σ is neuron RBF diffusion breadth, and h is hidden layer nodes.Such as This, you can obtain the nonlinear neural network model of speed prediction：
In formula：fnMapped for radial base neural net, its structure is as shown in Figure 5.
Set HhFor 9, i.e., history speed is 10 speed amounts in the past, then the radial base neural net input quantity is 11, Speed is predicted as following 5 seconds speed, then neutral net output quantity is 5, and neuron number is equal with input quantity number, i.e. h =10.In vehicle travel process, radial basis neural network foundation history speed and current driver's pedal information, in advance Measure following speed, meanwhile, it is current caused by information of vehicles be used as new historical information, by self-organizing center choose with Pseudoinverse technique determines that the method for weights realizes the adaptive on-line study of neutral net[21]。
In order to verify the validity of energy control method proposed by the present invention, emulation examination has been carried out under Matlab environment Test.Energy control method sampling time interval is set in emulation as 1 second, this can both ensure the stable control of system dynamic course System, can allow larger control amount of calculation again.It is P=5 seconds, battery SOC initial value and reference value to concurrently set prediction time domain It is all 0.65, simulation result is as shown in Figure 6.
Fig. 6 is shown during emulation online, for the comprehensive state of cyclic operation of a typical case, the operating mode category result judged. As can be seen from Figure, when car speed drastically changes, as 370s to 440s, 980s to 1030s and 1160s to 1220s it Between, operating mode is judged as fast variable working condition；And in the fluctuation of speed small range or the slow acceleration and deceleration of vehicle, such as 600s to 700s and For 1220s between 1900s, operating mode is judged as steady operating mode, it is possible thereby to illustrate the validity of operating mode classification determination methods.
Fig. 7 show the displaying directly perceived of speed prediction result, it can be seen that the car proposed by the present invention within the most of the time Fast Forecasting Methodology can be accurately predicted.
In order to further reasonably by data comparison evaluation and foreca result, introduce root-mean-square error (Root Mean here Square Error, RMSE) it is used as evaluation index.RMSE is by calculating the standard deviation table of difference between sample value and actual value Sample precision is levied, suitable for for contrasting predicted value and actual value, its calculation formula is as follows：
In formula：RMSE (k) is that root-mean-square error value of k-th of sampled point in prediction time domain, RMSE are in state of cyclic operation The root-mean-square error value of whole state of cyclic operation, NcFor whole state of cyclic operation sampled point number, Vc(k+i) it is kth in state of cyclic operation The true speed of ith sample point after individual sampled point.
Under same state of cyclic operation, the Forecasting Methodology for participating in contrast has：Prediction is kept, that is, predicts that speed keeps constant；Horse Er Kefu chains are predicted, i.e., whole to carry out speed prediction based on Markov Chain；Neural network prediction, i.e., it is whole to be based on radial direction base Neutral net carries out speed prediction；Integrated forecasting, i.e., overall speed Forecasting Methodology proposed by the present invention.The contrast of simulation result As shown in table 3.
The different Forecasting Methodology results contrasts of table 3
Holding prediction RMSE it can be seen from table as benchmark is higher, and precision of prediction is poor, and integrated forecasting combines The advantages of Markov Chain prediction and neural network prediction, its RMSE is minimum, and precision of prediction is optimal, and 31% is improved compared with benchmark Left and right.
Fig. 8 is shown in the state of cyclic operation of part, and integrated forecasting is predicted with Markov Chain and neural network prediction is each Following 5th second prediction error contrasts in sampling instant prediction time domain, and left side first is classified as state of cyclic operation 980s extremely in figure Prediction between 1030s, it can be seen that now operating mode is judged as fast variable working condition, and integrated forecasting uses neural network prediction, its As a result it is substantially better than Markov Chain prediction.Second is classified as state of cyclic operation 600s to the prediction between 700s in figure, now speed Fluctuation is smaller, and operating mode is judged as steady operating mode, and integrated forecasting predicts that its result is also significantly better than nerve using Markov Chain Neural network forecast.
Fig. 9 show speed, EVT Mode, the simulation result of battery SOC, as can be seen from Figure, actual vehicle speed substantially with Target vehicle speed is consistent, and battery SOC is able to maintain that near 0.65 and fluctuated, and illustrates that energy control method can be full first In the case of sufficient operator demand, battery SOC is maintained well, and hair is supplemented by the consumption of electric energy to a certain extent The energy that motivation provides less, preferably to adjust engine working point.
Figure 10 show engine under the operating mode, motor A and motor B rotating speed and torque, as can be seen from Figure, starts The machine fluctuation of speed is smaller, and this is due to that road surface decouples with engine, and the speed adjusting performance of motor is much better than engine again so that car The fluctuation of speed is mainly made up by the rotation speed change of two motors.And the torque of motor is due to relatively small, so being become by road surface The motor torque fluctuation changed and triggered can only be made up by motor to a certain extent.
Figure 11 show engine working point distribution map under the operating mode, it can be seen that energy control method can be preferable Adjust engine working point so that engine can be operated near optimal fuel-economy curve in most cases, hair Motivation operating efficiency is higher, and it is more excellent also to allow for vehicle economy.
It is proposed by the present invention based on improvement of the energy control method of PREDICTIVE CONTROL to vehicle performance in order to further verify, Simulation results of the multiple kinds of energy control method under different operating modes will be contrasted herein, wherein the strategy for participating in contrast has：In advance Observing and controlling system, i.e., energy control method proposed by the present invention；Control is kept, that is, assumes that speed keeps constant and uses PREDICTIVE CONTROL Energy control method；Dynamic Programming, that is, assume operating mode, it is known that globally optimal solution can be obtained；Rule, i.e., it is rule-based Energy control method, for as benchmark to be hoisted.Comparing result is as follows：
Figure 12 show the simulation result contrast of engine working point distribution, and as can be seen from Figure, PREDICTIVE CONTROL is with protecting It is similar to Dynamic Programming to hold the resulting result of control, and the distribution of the engine working point of PREDICTIVE CONTROL controls compared to holding High efficient district is more concentrated on, and the result obtained by rule is then poor, engine is then often operated in poorly efficient area.
Figure 13 show the simulation result contrast of battery SOC, and as can be seen from Figure, PREDICTIVE CONTROL is one with keeping control Kind real-time optimization, what is obtained is locally optimal solution, and its result is similar with the globally optimal solution obtained by dynamic rules, and regular Result SOC it is excessively stable, it is very big with the result difference of Dynamic Programming, effectively pass through the fluctuation regulation hair of battery SOC Motivation, control effect are poor.
Table 4 show simulation result contrast of the different-energy control method under different state of cyclic operations.Due to based on prediction The energy control method of control is a kind of real-time optimization, can not ensure the battery SOC and initial time of state of cyclic operation end time Identical, so for fair comparison, it is necessary to consider the consumption of electric energy simultaneously, foundation energy prices are electric by end time here Pond SOC is converted and is obtained equivalent fuel consumption,
In formula, Ec,s, Fc,sWith Δ SOCc,sRespectively correspond to the equivalent fuel consumption of state of cyclic operation and optimization method, fuel oil Consumption and battery SOC changing value,To convert electrical energy into the transforming factor of fuel oil, the valency of current diesel oil by energy prices Lattice are 5.54 yuan/liter, and the price of electric energy is 0.9 yuan/degree.For the improvement of vehicle economy, prediction it can be seen from table Control is relatively regular to have larger lifting close to Dynamic Programming, and better than control is kept, has compared to benchmark maximum reachable 18% lifting.
The different-energy control method simulation result of table 4 contrasts
Contrasted by the analysis to simulation result, demonstrate the validity of the speed prediction method of the present invention, accuracy carries 31% is risen, while also demonstrates the validity based on PREDICTIVE CONTROL energy control method, the relatively regular strategy of fuel economy has 18% lifting.
Certainly, described above is only one embodiment of the present invention, it should be pointed out that the common skill of the art For art personnel, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these are improved and profit Decorations are belonged within the protection domain of the claims in the present invention.

Claims (7)

1. a kind of producing condition classification Forecasting Methodology of vehicle driving-cycle, this method include：
Vehicle driving-cycle is categorized as steady operating mode and the fast class of variable working condition two under off-line state using K mean cluster algorithm, And it is presently in operating mode classification in on-line stage real-time judge vehicle；
For steady operating mode, using based on markovian speed prediction method, and fast variable working condition is directed to, using based on radially The speed prediction method of base neural net, the advantages of two methods are comprehensively utilized with this prediction effect to be optimal.
2. producing condition classification Forecasting Methodology according to claim 1, wherein according to following characteristics parameter：Peak acceleration (m/ S2), maximum deceleration (m/s2), average acceleration (m/s2), vehicle speed standard variance (km/h), max. speed and minimum speed it Poor (km/h), acceleration standard variance (m/s2) classify vehicle driving-cycle.
3. producing condition classification Forecasting Methodology according to claim 1, wherein using K mean cluster algorithm, by calculating sample Between close and distant degree carry out data classification, finally realize that the data in same class have larger characteristic similarity, inhomogeneity Between then differ greatly, specific operating mode judgment step is as follows：
Off-line phase：(1) multiple standard cycle operating modes are combined and forms sample；
(2) each sampling instant calculates the operating mode feature parameter of 10 seconds in the past in state of cyclic operation, obtains characteristic parameter sample number According to [x11,x12,...,x1m], [x21,x22,...,x2m] ... ..., [xn1,xn2,...,xnm], wherein m is characterized parameter ordinal number, n For state of cyclic operation length；
(3) K mean cluster algorithm is applied, randomly selects cluster centre c1=[c11,c12,...,c1m], c2=[c21,c22,..., c2m], the distance of all samples and cluster centre is calculated, and sample is grouped according to Nearest Neighbor Method, belong to different θm(k) cluster Domain, wherein k are iterations, then adjust cluster centre as the following formula：
<mrow> <msub> <mi>c</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <msub> <mi>n</mi> <mi>m</mi> </msub> </mfrac> <munder> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>&amp;Element;</mo> <msub> <mi>&amp;theta;</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </mrow> </munder> <mi>x</mi> <mo>,</mo> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mi>m</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>16</mn> <mo>)</mo> </mrow> </mrow>
If cm(k+1)≠cm(k), then continuing to adjust cluster centre, until cluster centre no longer changes, then it is assumed that classification is stable, Obtain the cluster centre c1 of steady operating mode and the cluster centre c2 of fast variable working condition；
On-line stage：(1) during vehicle actual travel, the operating mode feature parameter of 10 seconds is calculated in current sample time It is worth [x1,x1,...,xm]；
(2) characteristic ginseng value [x is calculated according to following formula1,x1,...,xm] to two cluster centres c1 and c2 distance d：
<mrow> <msub> <mi>d</mi> <mi>j</mi> </msub> <mo>=</mo> <msqrt> <mrow> <msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mn>1</mn> </msub> <mo>-</mo> <msub> <mi>c</mi> <mrow> <mi>j</mi> <mn>1</mn> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mn>2</mn> </msub> <mo>-</mo> <msub> <mi>c</mi> <mrow> <mi>j</mi> <mn>2</mn> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <mn>...</mn> <mo>+</mo> <msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>m</mi> </msub> <mo>-</mo> <msub> <mi>c</mi> <mrow> <mi>j</mi> <mi>m</mi> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>17</mn> <mo>)</mo> </mrow> </mrow>
In formula：J=1,2 correspond to two class operating modes；
(3) if d1≤d2, then current time is judged for steady operating mode, if d1＞ d2, then judge current time for fast variable working condition.
4. producing condition classification Forecasting Methodology according to claim 1, wherein assuming vehicle at each moment under steady operating mode Acceleration is unrelated with historical information, is only determined by current information, it is thus regarded that the acceleration change of vehicle is a kind of Markov Process, the changing rule of speed and acceleration is simulated using Markov chain model, and to following speed under steady operating mode It is predicted.
5. producing condition classification Forecasting Methodology according to claim 1, wherein for fast variable working condition, with radial base neural net Theory, the prediction of following speed is carried out by the study online to driver's driving behavior.
6. producing condition classification Forecasting Methodology according to claim 5, wherein radial basis neural network are according to history speed With current driver's pedal information, following speed is predicted, meanwhile, current caused information of vehicles is to believe as new history Breath, chosen by self-organizing center and pseudoinverse technique determines that the method for weights realizes the adaptive on-line study of neutral net.
7. a kind of producing condition classification Forecasting Methodology of hybrid car travel operating mode, this method uses any institutes of claim 1-6 The producing condition classification Forecasting Methodology of restriction.
CN201710645911.3A 2017-08-01 2017-08-01 A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power Pending CN107516107A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710645911.3A CN107516107A (en) 2017-08-01 2017-08-01 A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710645911.3A CN107516107A (en) 2017-08-01 2017-08-01 A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power

Publications (1)

Publication Number Publication Date
CN107516107A true CN107516107A (en) 2017-12-26

Family

ID=60722933

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710645911.3A Pending CN107516107A (en) 2017-08-01 2017-08-01 A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power

Country Status (1)

Country Link
CN (1) CN107516107A (en)

Cited By (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109159785A (en) * 2018-07-19 2019-01-08 重庆科技学院 A kind of automobile running working condition prediction technique based on Markov chain and neural network
CN109760523A (en) * 2019-01-21 2019-05-17 浙江工业大学 Composite power source energy management method based on BP neural network speed prediction
CN110503743A (en) * 2019-08-12 2019-11-26 浙江吉利汽车研究院有限公司 A kind of method for detecting abnormality of automatic driving vehicle, device and equipment
CN110533262A (en) * 2019-09-09 2019-12-03 合肥工业大学 A kind of operating condition rolling recognition methods of various dimensions
CN110562239A (en) * 2019-08-28 2019-12-13 武汉理工大学 Variable-domain optimal energy management control method and device based on demand power prediction
CN110598747A (en) * 2019-08-13 2019-12-20 广东工业大学 Road classification method based on self-adaptive K-means clustering algorithm
CN110682905A (en) * 2019-10-12 2020-01-14 重庆大学 Method for acquiring battery charge state reference variable quantity in time domain based on driving mileage
CN110827446A (en) * 2019-11-13 2020-02-21 北京理工大学 Method for predicting running state of electric automobile
CN111091249A (en) * 2019-12-30 2020-05-01 吉林大学 Method for realizing optimal distribution of vehicle global energy based on global domain-searching algorithm
CN111216736A (en) * 2020-01-09 2020-06-02 吉林大学 Driving condition-based self-adaptive adjustment method and system for auxiliary driving system
CN111267827A (en) * 2020-02-13 2020-06-12 山东中科先进技术研究院有限公司 Energy management method and system for hybrid electric vehicle
WO2020124779A1 (en) * 2018-12-17 2020-06-25 中国科学院沈阳自动化研究所 Working condition state modeling and model correction method
CN111597640A (en) * 2020-05-22 2020-08-28 上海海事大学 Method for predicting demand load of hybrid power ship under condition classification
CN112182962A (en) * 2020-09-24 2021-01-05 青岛大学 Hybrid electric vehicle running speed prediction method
CN112298155A (en) * 2020-10-30 2021-02-02 江苏紫琅汽车集团股份有限公司 Method for predicting energy consumption of hybrid power truck based on variable time domain model
CN112785845A (en) * 2020-12-30 2021-05-11 桂林电子科技大学 Vehicle speed prediction method based on K-means clustering and RBF neural network
CN113135113A (en) * 2020-01-19 2021-07-20 郑州宇通客车股份有限公司 Global SOC (System on chip) planning method and device
CN113479187A (en) * 2021-07-30 2021-10-08 南京航空航天大学 Layered different-step-length energy management method for plug-in hybrid electric vehicle
CN113552803A (en) * 2021-07-26 2021-10-26 桂林电子科技大学 Energy management method based on working condition identification
CN113859219A (en) * 2021-11-15 2021-12-31 重庆大学 Hybrid electric vehicle self-adaptive energy management method based on driving condition recognition
CN115675177A (en) * 2021-07-26 2023-02-03 比亚迪股份有限公司 Battery electric quantity management system, method and device of hybrid electric vehicle
US11794757B2 (en) 2018-06-11 2023-10-24 Colorado State University Research Foundation Systems and methods for prediction windows for optimal powertrain control
CN118135686A (en) * 2024-05-08 2024-06-04 成都赛力斯科技有限公司 Equivalent factor determination method and device

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103914985A (en) * 2014-04-25 2014-07-09 大连理工大学 Method for predicting future speed trajectory of hybrid power bus
CN106740829A (en) * 2017-03-23 2017-05-31 吉林大学 Based on the double semi-dragging truck riding stability automatic identifications of cluster analysis and early warning system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103914985A (en) * 2014-04-25 2014-07-09 大连理工大学 Method for predicting future speed trajectory of hybrid power bus
CN106740829A (en) * 2017-03-23 2017-05-31 吉林大学 Based on the double semi-dragging truck riding stability automatic identifications of cluster analysis and early warning system

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
CHANGLE XIANG等: ""Energy management of a dual-mode power-split hybrid electric vehicle based on velocity prediction and nonlinear model predictive control"", 《APPLIED ENERGY》 *
CHAO SUN等: ""Velocity predictors for predictive energy management in hybrid electric vehicles"", 《IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY》 *
杨盼盼: ""汽车未来行驶车速预测"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
王静: ""基于工况辨识的增程式电动城市客车最优能量分配策略"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
詹森: ""基于工况与驾驶风格识别的混合动力汽车能量管理策略研究"", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 *
邓丽君: ""插电式混合动力汽车未来运行工况预测"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Cited By (29)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11794757B2 (en) 2018-06-11 2023-10-24 Colorado State University Research Foundation Systems and methods for prediction windows for optimal powertrain control
CN109159785A (en) * 2018-07-19 2019-01-08 重庆科技学院 A kind of automobile running working condition prediction technique based on Markov chain and neural network
CN109159785B (en) * 2018-07-19 2020-05-01 重庆科技学院 Automobile driving condition prediction method based on Markov chain and neural network
WO2020124779A1 (en) * 2018-12-17 2020-06-25 中国科学院沈阳自动化研究所 Working condition state modeling and model correction method
CN109760523A (en) * 2019-01-21 2019-05-17 浙江工业大学 Composite power source energy management method based on BP neural network speed prediction
CN110503743A (en) * 2019-08-12 2019-11-26 浙江吉利汽车研究院有限公司 A kind of method for detecting abnormality of automatic driving vehicle, device and equipment
CN110598747A (en) * 2019-08-13 2019-12-20 广东工业大学 Road classification method based on self-adaptive K-means clustering algorithm
CN110562239A (en) * 2019-08-28 2019-12-13 武汉理工大学 Variable-domain optimal energy management control method and device based on demand power prediction
CN110562239B (en) * 2019-08-28 2020-10-30 武汉理工大学 Variable-domain optimal energy management control method and device based on demand power prediction
CN110533262A (en) * 2019-09-09 2019-12-03 合肥工业大学 A kind of operating condition rolling recognition methods of various dimensions
CN110682905A (en) * 2019-10-12 2020-01-14 重庆大学 Method for acquiring battery charge state reference variable quantity in time domain based on driving mileage
CN110827446A (en) * 2019-11-13 2020-02-21 北京理工大学 Method for predicting running state of electric automobile
CN111091249A (en) * 2019-12-30 2020-05-01 吉林大学 Method for realizing optimal distribution of vehicle global energy based on global domain-searching algorithm
CN111216736A (en) * 2020-01-09 2020-06-02 吉林大学 Driving condition-based self-adaptive adjustment method and system for auxiliary driving system
CN113135113A (en) * 2020-01-19 2021-07-20 郑州宇通客车股份有限公司 Global SOC (System on chip) planning method and device
CN113135113B (en) * 2020-01-19 2022-12-13 宇通客车股份有限公司 Global SOC (System on chip) planning method and device
US20220242390A1 (en) * 2020-02-13 2022-08-04 Shandong Institute Of Advanced Technology, Chinese Academy Of Sciences Co., Ltd Energy management method and system for hybrid electric vehicle
CN111267827A (en) * 2020-02-13 2020-06-12 山东中科先进技术研究院有限公司 Energy management method and system for hybrid electric vehicle
CN111267827B (en) * 2020-02-13 2021-07-16 山东中科先进技术研究院有限公司 Energy management method and system for hybrid electric vehicle
CN111597640A (en) * 2020-05-22 2020-08-28 上海海事大学 Method for predicting demand load of hybrid power ship under condition classification
CN112182962A (en) * 2020-09-24 2021-01-05 青岛大学 Hybrid electric vehicle running speed prediction method
CN112298155A (en) * 2020-10-30 2021-02-02 江苏紫琅汽车集团股份有限公司 Method for predicting energy consumption of hybrid power truck based on variable time domain model
CN112298155B (en) * 2020-10-30 2022-04-08 江苏紫琅汽车集团股份有限公司 Method for predicting energy consumption of hybrid power truck based on variable time domain model
CN112785845A (en) * 2020-12-30 2021-05-11 桂林电子科技大学 Vehicle speed prediction method based on K-means clustering and RBF neural network
CN113552803A (en) * 2021-07-26 2021-10-26 桂林电子科技大学 Energy management method based on working condition identification
CN115675177A (en) * 2021-07-26 2023-02-03 比亚迪股份有限公司 Battery electric quantity management system, method and device of hybrid electric vehicle
CN113479187A (en) * 2021-07-30 2021-10-08 南京航空航天大学 Layered different-step-length energy management method for plug-in hybrid electric vehicle
CN113859219A (en) * 2021-11-15 2021-12-31 重庆大学 Hybrid electric vehicle self-adaptive energy management method based on driving condition recognition
CN118135686A (en) * 2024-05-08 2024-06-04 成都赛力斯科技有限公司 Equivalent factor determination method and device

Similar Documents

Publication Publication Date Title
CN107516107A (en) A kind of driving cycle classification Forecasting Methodology of motor vehicle driven by mixed power
CN107688343A (en) A kind of energy control method of motor vehicle driven by mixed power
Hu et al. Reinforcement learning for hybrid and plug-in hybrid electric vehicle energy management: Recent advances and prospects
Lian et al. Rule-interposing deep reinforcement learning based energy management strategy for power-split hybrid electric vehicle
Lian et al. Cross-type transfer for deep reinforcement learning based hybrid electric vehicle energy management
Liu et al. Parallel reinforcement learning-based energy efficiency improvement for a cyber-physical system
Chen et al. Optimal strategies of energy management integrated with transmission control for a hybrid electric vehicle using dynamic particle swarm optimization
US7954579B2 (en) Adaptive control strategy and method for optimizing hybrid electric vehicles
Singh et al. Fuzzy logic and Elman neural network tuned energy management strategies for a power-split HEVs
Zhang et al. Route planning and power management for PHEVs with reinforcement learning
CN113554337B (en) Plug-in hybrid electric vehicle energy management strategy construction method integrating traffic information
Mozaffari et al. Multi-objective component sizing of a power-split plug-in hybrid electric vehicle powertrain using Pareto-based natural optimization machines
Yang et al. Energy management of the power-split hybrid electric city bus based on the stochastic model predictive control
CN109910863A (en) A kind of control method based on plug-in hybrid car dedicated transmission
Zhang et al. Integrated velocity optimization and energy management strategy for hybrid electric vehicle platoon: A multi-agent reinforcement learning approach
Peng et al. Ecological Driving Framework of Hybrid Electric Vehicle Based on Heterogeneous Multi-Agent Deep Reinforcement Learning
Yu et al. Real-time and hierarchical energy management-control framework for electric vehicles with dual-motor powertrain system
Liu et al. Parallel distance: A new paradigm of measurement for parallel driving
CN117184095B (en) Hybrid electric vehicle system control method based on deep reinforcement learning
Wang et al. Ecological cruising control of connected electric vehicle: a deep reinforcement learning approach
Urooj et al. Review of intelligent energy management techniques for hybrid electric vehicles
Gao et al. Multi-physical cooperative control of plug-in hybrid electric vehicles via cyber hierarchy and interactional network
CN106647279A (en) Locomotive intelligent control optimization calculating method based on fuzzy rules
Yuankai et al. Benchmarking deep reinforcement learning based energy management systems for hybrid electric vehicles
Mozaffari et al. Real-time immune-inspired optimum state-of-charge trajectory estimation using upcoming route information preview and neural networks for plug-in hybrid electric vehicles fuel economy

Legal Events

Date Code Title Description
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20171226

WD01 Invention patent application deemed withdrawn after publication