A plug-in hybrid electric vehicle energy management system and method based on road segment information electric power distribution

Abstract

The application discloses a kind of based on road section information electric power distribution plug-in hybrid electric vehicle energy management system and method, including offline part: the acquisition of driving data, data are handled classification and synthesis new working condition.Data are trained support vector machine to identify working condition, combined new working condition is operated by DP algorithm, the result of DP algorithm is handled, and each working condition electric power distribution relationship is obtained by training BP network.Another part is the offline optimization of ECMS algorithm, and the offline optimization of equivalent factor is carried out for each working condition.Online part: each road section information and identification information obtained by traffic information acquisition module are obtained by BP neural network to obtain electric power distribution relationship, then plan SOC, and then select equivalent factor and operate according to working condition and SOC planning trajectory.The electric power distribution and the selection of equivalent factor are more reasonable in the application, and the fuel economy of plug-in hybrid electric vehicle can be improved.

Description

Technical Field

[0001] This invention belongs to the field of automotive energy management methods, and specifically relates to an energy management system and method for plug-in hybrid electric vehicles based on power allocation according to operating conditions. Background Technology

[0002] After years of development, the drawbacks of traditional gasoline-powered vehicles have become increasingly apparent: low energy efficiency and severe environmental pollution. Given the non-renewable nature of petroleum resources and the backdrop of resource scarcity, hybrid and pure electric vehicles are developing rapidly. The limitations of pure electric vehicle batteries prevent their widespread adoption, while hybrid vehicles are currently demonstrating their advantages. The contradiction between economic development and environmental pollution and energy shortages is becoming increasingly acute. Plug-in hybrid electric vehicles (PHEVs), as a transitional model between traditional gasoline-powered and pure electric vehicles, combine the advantages of both. Because PHEVs have both an engine and an electric motor as their power source, and use both fuel and electricity as their energy source, they address concerns about driving range associated with pure electric vehicles while reducing fuel consumption, making their advantages quite significant.

[0003] Plug-in hybrid electric vehicles (PHEVs) are a type of hybrid electric vehicle. They have the advantages of reducing fuel consumption and emissions. The extent to which fuel consumption is reduced depends mainly on the energy management strategy. Different energy management strategies have a significant impact on fuel economy, so research on the energy management of PHEVs is of great significance.

[0004] Equivalent fuel consumption strategy, as an instantaneous optimal algorithm, can be applied to plug-in hybrid electric vehicles. Its working principle is to equate electrical energy consumption to fuel consumption. The core of this strategy is the selection of the equivalent factor, which directly affects fuel economy. Factors influencing the equivalent factor mainly include operating conditions, driving distance, and available State of Charge (SOC). Reasonable selection of the equivalent factor can significantly improve fuel economy. With the development of connected information, real-time road segment information can be obtained from maps. This information can be used as an influence factor in the selection of the equivalent factor. By integrating road segment information into the selection of the equivalent factor, a reference SOC trajectory can be planned based on road segment information, indirectly affecting the selection of Energy Efficiency Factor (EF). Each road segment also includes the influence of operating condition category. Establishing an energy management strategy based on operating condition category-based energy allocation is of great significance for improving fuel economy. Summary of the Invention

[0005] The purpose of this invention is to plan the State of Charge (SOC) segment for different road segments under a total journey, so as to allocate the power more rationally and improve fuel economy. To this end, an energy management method for plug-in hybrid electric vehicles based on power allocation according to road segment information is provided.

[0006] The present invention achieves the above objectives by adopting the following technical solutions.

[0007] A plug-in hybrid electric vehicle energy management system based on road segment information-based power allocation, the adaptive energy management system comprising an offline part and an online part, the offline part comprising a DP module, a data analysis module, and a neural network module connected in sequence; the online part comprising a traffic information acquisition module, a working condition recognition module, a vehicle module, an ECMS algorithm module, and a SOC planning module, wherein the traffic information acquisition module and the working condition recognition module are connected, and the vehicle module, the ECMS algorithm module, and the SOC planning module are connected in sequence;

[0008] The system includes online and offline components for the neural network module, driving condition recognition module, and ECMS algorithm module. The offline component involves driver-collected and processed data. The neural network module uses this processed data to train the driving condition recognition module to identify road conditions. The processed data is then input into the DP module for calculation, and the results are fed into the data analysis module for further training of the neural network module, yielding the ratio of energy consumption per unit mileage under various driving conditions. Based on the driving conditions and energy consumption per unit mileage, the ECMS equivalence factor is optimized offline. The online component involves the traffic information acquisition module acquiring road segment mileage and average vehicle speed per segment. The system uses speed and average number of stops per road segment to identify different road segment conditions. The operating condition identification module employs a trained support vector machine to record the operating condition category for each segment. The system inputs the vehicle's current available State of Charge (SOC), the identified road segment operating condition category, and the corresponding road segment mileage information into the neural network module to obtain the ratio of power consumption per unit mileage for each operating condition. This ratio is then input into the SOC planning module to generate a reference trajectory. Based on the road segment operating condition category and the SOC reference trajectory, the ECMS algorithm module selects the equivalent factor EF. Finally, based on information such as the vehicle's required torque, the ECMS algorithm calculates the control parameters to perform vehicle control.

[0009] The present invention discloses an energy management method for plug-in hybrid electric vehicles based on road segment information-based power allocation, the method comprising the following steps:

[0010] (1) First, driver data is collected and the data is divided into four categories: congested, relatively smooth, smooth and high-speed conditions.

[0011] (2) Randomly combine each working condition of the classified segments into working conditions of different distances, calculate the average vehicle speed and average number of stops for each combined working condition, and use this data to train the support vector machine, which is used to identify the driving conditions of the road segment.

[0012] (3) DP programming was used to study the mutual influence of four operating conditions. A fixed distance of 60 km was used, with different proportions of each operating condition's mileage to the total mileage. Multiple calculations were performed, and the results of the DP dynamic programming were processed to obtain the ratio of SOC / km for each operating condition, denoted as SK1:SK2:SK3:SK4. The ratio of these values ​​for each operating condition was then used to determine the mutual influence between them. The proportion of electricity consumption for each operating condition was obtained. The calculation formula is as follows:

[0013] X1:X2:X3:X4=SK1:SK2:SK3:SK4

[0014] Where X1:X2:X3:X4 represents the ratio of power consumption per unit distance under four operating conditions; SK1 is the power consumption per kilometer (SOC / km) for operating condition category 1, SK2 is the power consumption per kilometer (SOC / km) for operating condition category 2, SK3 is the power consumption per kilometer (SOC / km) for operating condition category 3, and SK4 is the power consumption per kilometer (SOC / km) for operating condition category 4. The distances S1, S2, S3, and S4 for each operating condition are compared with the available SOC: S K As input, SK1, SK2, SK3, and SK4 are used as outputs to train a BP neural network. The neural network consists of three layers: an input layer, a hidden layer, and an output layer, which are used to allocate power according to road conditions based on information from the subsequent map.

[0015] S1, S2, S3, and S4 represent the road segment mileages for work condition categories 1, 2, 3, and 4, respectively. K This represents the current available SOC value for the vehicle.

[0016] (4) Road segment information obtained by the traffic information acquisition module: average vehicle speed, average number of stops, and road segment mileage S1, S2, S3, and S4 after support vector machine recognition for each road segment; vehicle available battery information: S K ; Transfer the data S1, S2, S3, S4, S K The input is fed into the BP neural network to obtain the output: the ratio of power consumption per unit distance under each working condition SK1:SK2:SK3:SK4;

[0017] (5) The BP neural network can obtain the operating condition information of each segment of the trip through the traffic information acquisition module to predict the power allocation of a trip. The output of the BP neural network is the ratio of power consumption per unit distance under each operating condition, SK1:SK2:SK3:SK4, which is used to calculate the SOC (State of Charge). The calculation formulas involved are as follows:

[0018] (SK1*S1+SK2*S2+SK3*S3+SK4*S4)x=S K

[0019] SOC ref =S K -(SK1*x*S 1cover )-(SK2*x*S 2cover )-(SK3*x*S 3cover )-(SK4*x*S 4cover )

[0020] S K Let x be the vehicle's current available battery power, and SOC be the set variable (used as an intermediate variable in the planning). ref SOC reference trajectory; S 1cover S represents the distance traveled in condition category 1. 2cover S represents the distance traveled in condition category 2. 3cover S represents the distance traveled in condition category 3. 4cover This represents the distance traveled so far in condition category 4.

[0021] (6) To reduce the impact of errors in condition identification on the reference trajectory, the average vehicle speed and average number of stops for different road segments obtained by the traffic information acquisition module are used to determine the condition of each road segment. To enhance the accuracy of identification, an online condition identification module for the current road segment is added. The identification time is the most recent driving data within the road segment in 300 seconds. If the identification time is inconsistent for more than 60 seconds, the SOC is re-planned according to the condition category identified online. At the same time, since the road segment information is constantly updated, it is updated every 60 seconds. The updated content includes road segment information, identification information and new SOC reference trajectory.

[0022] (7) The classified working conditions are combined and spliced ​​into four major categories: congested, relatively smooth, smooth and high-speed. For each category, the equivalent factor is optimized. Here, only the SOC change is optimized. Distance is not used as a variable. The only variable is SOC. The equivalent fuel consumption model is optimized. The minimum SOC is set to 0.3. The optimization is performed by GA genetic algorithm. The EF equivalent factor is the variable and the fuel consumption is the objective function. The distance for each working condition is set to 60km. The SOC gradually decreases from 1 to 0.3 at intervals of 0.05. The equivalent factor is optimized. The optimized EF is recorded and the corresponding SOC / Km value is also recorded. The recorded values ​​are plotted as an SOC / Km-EF map.

[0023] (8) The planned SOC corresponds to the working condition, calculate the unit mileage power consumption under each road segment, find the EF corresponding to the SOC / Km under the corresponding working condition from the Map, and make an equivalent factor; obtain the corresponding equivalent factor, and perform energy management of ECMS at the lower level. The input is the vehicle demand torque, and the output is the optimal engine motor torque to control the vehicle.

[0024]

[0025]

[0026]

[0027]

[0028] J min To minimize fuel consumption, t is the driving time under operating conditions, N is the final value of the driving time under operating conditions, Δt is the time interval set to 1 second, and m fc For engine fuel consumption, T fc Engine torque ω fc Engine speed, m fmc_eq P represents the equivalent fuel consumption of the electric motor, EF is the equivalence factor, and P is the equivalent fuel consumption of the electric motor. mc η is the motor power, η is the battery efficiency, Q is the fuel calorific value, and T is the fuel temperature. mc ω is the motor torque. mc The motor speed is m. total For optimal fuel consumption, This refers to engine torque. T represents the motor torque. req This represents the total required torque.

[0029] (9) During operation, fluctuations between the planned SOC curve and the actual SOC curve are inevitable. When a significant deviation from the planned SOC occurs, EF correction is performed to bring it closer to the planned curve. The method used is piecewise PI, which makes minor adjustments within a reasonable range of deviation from the planned curve and makes further adjustments when the deviation exceeds a threshold. The piecewise PI function is as follows:

[0030]

[0031] SOC ref The planned SOC trajectory is defined by SOC, where SOC is the vehicle's SOC at the current moment, and EF is the selected equivalent factor ΔSOC = |SOC|. ref -SOC|,K P K I is a coefficient.

[0032] Furthermore, the offline optimization DP dynamic programming data uses different proportions under different operating conditions as optimization data, and the proportions of a certain operating condition are gradually increased. The final optimization result of dynamic programming is the ratio of unit distance power consumption under different operating condition proportions under different available power.

[0033] Furthermore, the training data of the BP neural network is the data obtained by DP dynamic programming. The input is the vehicle's available power, the mileage of four different operating conditions, and the mileage of the four different operating conditions arranged in order. The output is the ratio of the unit power consumption of the four different operating conditions.

[0034] Furthermore, the SOC trajectory planning is based on the ratio of unit power consumption under four operating conditions output by the BP neural network and the available power of the vehicle to plan the SOC reference trajectory. Specifically, a functional relationship is established to allocate power and obtain the SOC planned trajectory.

[0035] Furthermore, the data for operating condition identification comes from the connected map, and the identification parameters are average vehicle speed and average number of stops. The result is the overall operating condition category of a certain road segment.

[0036] Furthermore, after identifying the road segment's operating condition category, an online identification of the current operating condition has been added. The identification time is the most recent driving data within the road segment within 300 seconds. If the identification is inconsistent, the SOC will be re-planned according to the online identified operating condition category. The condition for the operating condition category replacement is that the identification inconsistency time exceeds 60 seconds.

[0037] The outstanding benefits of this invention are as follows:

[0038] 1. Based on the road segment planning, SOC reference trajectories for different road segments are planned to make SOC allocation more reasonable;

[0039] 2. Select an equivalent factor that is more suitable for the current road segment based on the reference SOC of the road segment;

[0040] The method proposed in this invention allocates power and plans the SOC trajectory based on road segment information, making the power allocation and selection of equivalent factors more reasonable, and can better improve the fuel economy of plug-in hybrid electric vehicles. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the energy management method for allocating power according to road segment information of the present invention;

[0042] Figure 2 This is a schematic diagram illustrating the classification of operating conditions in this invention;

[0043] Figure 3 This is a schematic diagram of the working condition recognition based on support vector machine according to the present invention;

[0044] Figure 4 This is a schematic diagram of the power allocation based on the BP neural network of the present invention;

[0045] Figure 5 This is a schematic diagram of the SOC reference trajectory generation of the present invention;

[0046] Figure 6 This is a schematic diagram illustrating the selection of equivalence factors in this invention; Detailed Implementation

[0047] The following is a detailed description with reference to the accompanying drawings.

[0048] Figure 1 This is a schematic diagram of the energy management system structure to be adopted in this embodiment of the invention. It mainly includes a DP module, a data analysis module, a neural network module, a traffic information acquisition module, a working condition identification module, a vehicle module, an ECMS algorithm module, and a SOC planning module. The traffic information acquisition module acquires information such as the mileage of each road segment, the average speed of each segment, and the number of stops. The average speed and number of stops are transmitted to the working condition identification module to identify the working condition type of each road segment. The identification information from the working condition identification module, along with the mileage of the road segment and the SOC information from the vehicle module, are transmitted to the neural network module. The neural network module obtains the energy allocation ratio for each road segment's working condition based on the information and then transmits the information to the SOC planning module for planning. Subsequently, the ECMS algorithm module selects the equivalent factor based on the SOC planning information and performs ECMS algorithm calculations based on torque demand to allocate torque and control the vehicle.

[0049] Furthermore, in this embodiment, driver data is collected, and the data processing procedure is as follows: Figure 2 The eigenvalues ​​include: average vehicle speed, maximum vehicle speed, minimum vehicle speed, maximum acceleration, average acceleration, maximum deceleration, average deceleration, idling time, constant speed time ratio, acceleration time ratio, and deceleration time ratio. The dimensionality reduction method is PCA principal component analysis, and the classification method uses k-means clustering. The system is divided into four categories: congested, relatively smooth, smooth, and high-speed.

[0050] Furthermore, in this implementation, such as Figure 3 As shown, the clustered segments are combined, with each driving condition randomly combined into 300 sets of data at different distances, ranging from 1 km to 300 km with 1 km intervals. The average vehicle speed and average number of stops for each segment of these 300 combined data sets are calculated. This data is used to train a support vector machine, which is used to identify driving condition categories.

[0051] Furthermore, in this embodiment, the key to road segment condition identification lies in classifying the road segment information obtained from the map into specific types of conditions. This is specifically implemented by: obtaining road segment condition information from online maps for different road segments to identify the conditions, including average vehicle speed and average stopping time. The support vector machine then identifies the road segment conditions based on this information.

[0052] Furthermore, in this embodiment, the calculation results of DP programming are used to study the interaction relationship of four operating conditions. The DP programming algorithm uses a fixed distance of 60km for each operating condition. The available battery power of the hybrid vehicle and the proportion of each operating condition are variables. Specifically, the proportion is changed starting from the mileage of each condition accounting for one-quarter of the total driving distance, and then gradually changing only the proportion of one operating condition. The combined operating condition data with different proportions are used as the operating condition data for dynamic programming. After the dynamic programming calculation is completed, the result data is processed to obtain the SOC / km ratio of the four operating conditions. The ratios of the four operating conditions under a certain available battery power and a certain operating condition proportion are denoted as SK1:SK2:SK3:SK4. The ratio of the values ​​of each operating condition represents the interaction relationship of battery consumption among the operating conditions under a certain proportion of each operating condition with a certain amount of available battery power. The proportion law of each operating condition is obtained, and the battery consumption value of each road segment is calculated using a formula. The calculation formula is as follows:

[0053] X1:X2:X3:X4=SK1:SK2:SK3:SK4

[0054] The ratio X1:X2:X3:X4 represents the ratio of energy consumption per unit distance under the four operating conditions. SK1 is the energy consumption per kilometer (SOC / km) for operating condition category 1, SK2 is the energy consumption per kilometer (SOC / km) for operating condition category 2, SK3 is the energy consumption per kilometer (SOC / km) for operating condition category 3, and SK4 is the energy consumption per kilometer (SOC / km) for operating condition category 4.

[0055] Furthermore, in this implementation, for example, Figure 4 As shown, in this embodiment, the plug-in hybrid electric vehicle can use SOC: S K The BP neural network is trained using four types of working conditions mileage S1, S2, S3, and S4 as inputs and SK1, SK2, SK3, and SK4 as outputs. The neural network consists of three layers: an input layer, a hidden layer, and an output layer, which are used to allocate power according to the working conditions of the road segment based on the information from the subsequent map.

[0056] S1, S2, S3, and S4 represent the road segment mileages for work condition categories 1, 2, 3, and 4, respectively. K This represents the current available State of Charge (SOC) value for the vehicle.

[0057] Furthermore, in this embodiment, the traffic information acquisition module obtains the following road segment information: average vehicle speed, average number of stops, and mileage S1, S2, S3, S4 for each road segment; vehicle available battery power information: S K Data S1, S2, S3, S4, S KThe BP is input into the neural network to obtain the output: the ratio of power consumption per unit distance under each operating condition, given the current available power and the percentage of operating conditions.

[0058] S1, S2, S3, and S4 represent the road segment mileages for work condition categories 1, 2, 3, and 4, respectively. K This represents the current available State of Charge (SOC) value for the vehicle.

[0059] Furthermore, in this implementation, such as Figure 5 As shown, in this embodiment, the output of the BP neural network is to predict the power allocation for each segment of a trip, based on the operating conditions obtained through the traffic information acquisition module. The output of the BP neural network is the ratio of power consumption per unit distance under each operating condition (SK1:SK2:SK3:SK4), which is used to calculate the State of Charge (SOC). The calculation formulas involved are as follows:

[0060] (SK1*S1+SK2*S2+SK3*S3+SK4*S4)x=S K

[0061] SOC ref =S K -(SK1*x*S 1cover )-(SK2*x*S 2cover )-(SK3*x*S 3cover )-(SK4*x*S 4cover )

[0062] S K Let x be the vehicle's current available battery power, and SOC be the set variable (used as an intermediate variable in the planning). ref This is the SOC reference trajectory. 1cover S represents the distance traveled in condition category 1. 2cover S represents the distance traveled in condition category 2. 3cover S represents the distance traveled in condition category 3. 4cover This represents the distance traveled in condition category 4.

[0063] Furthermore, in this embodiment, to reduce the impact of errors in work condition identification on the reference trajectory, while using different road segment information obtained from the traffic information acquisition module to determine the work condition of a road segment, an online work condition identification module for the current road segment is added to enhance the accuracy of identification. The identification time is the most recent driving data within the road segment in the last 300 seconds. If the identification inconsistency time exceeds 60 seconds, the SOC is re-planned according to the work condition category identified online. At the same time, since the road segment information is constantly updated, it is updated every 60 seconds. The updated content includes road segment information, identification information, and the new SOC reference trajectory.

[0064] Furthermore, in this implementation, such as Figure 6 As shown, in this embodiment, the four clustered operating conditions are combined and spliced ​​into four major categories: congestion, slight congestion, relatively smooth traffic, smooth traffic, and high-speed traffic. For each category, equivalent factor optimization is performed. During this optimization, only the change in available State of Charge (SOC) is considered; the operating distance is not a variable. The only variable is available SOC, and an equivalent fuel consumption model is optimized. The minimum SOC is set to 0.3, and optimization is performed using a Genetic Algorithm (GA). The EF equivalent factor is the initial population size for the GA algorithm, and fuel consumption is the objective function. The offline optimization distance for each operating condition is set to 60km. The SOC gradually decreases from 1 to 0.3 at intervals of 0.05, continuously optimizing the equivalent factor. The optimized EF equivalent factor is recorded, along with the corresponding unit distance energy consumption (SOC / km). The recorded values ​​are plotted as an operating condition category-SOC / km-EF map.

[0065] Furthermore, in this embodiment, the planned reference SOC and its corresponding operating condition category are obtained, the unit mileage power consumption under each road segment is calculated, and the equivalent factor corresponding to the SOC / Km under the corresponding operating condition is found from the offline optimized equivalent factors according to the operating condition category and the unit mileage power consumption under that road segment.

[0066] Furthermore, in this embodiment, the corresponding equivalent factor, real-time vehicle torque demand, and speed are obtained. The lower layer performs energy management using ECMS, taking the vehicle torque demand as input and outputting the real-time optimal engine / motor torque to control the vehicle. The specific calculation formula is as follows:

[0067]

[0068]

[0069]

[0070]

[0071] J min To minimize fuel consumption, t is the driving time under operating conditions, N is the final value of the driving time under operating conditions, Δt is the time interval set to 1 second, and m fc For engine fuel consumption, T fc Engine torque ω fc Engine speed, m fmc_eq P represents the equivalent fuel consumption of the electric motor, EF is the equivalence factor, and P is the equivalent fuel consumption of the electric motor. mc η is the motor power, η is the battery efficiency, Q is the fuel calorific value, and T is the fuel temperature. mc ω is the motor torque. mc The motor speed is m. totalFor optimal fuel consumption, This refers to engine torque. T represents the motor torque. req This represents the total required torque.

[0072] Furthermore, in this embodiment, during operation, there will inevitably be a difference between the planned SOC reference trajectory and the actual SOC curve after operation, resulting in a deviation from the planned SOC. In this case, EF correction is performed to bring it closer to the planned curve. The method used is piecewise PID control, which makes small adjustments within a reasonable range of deviation from the planned curve and makes further adjustments when the deviation exceeds a threshold. The piecewise PID function is as follows:

[0073]

[0074] SOC ref The planned SOC trajectory is defined by SOC, where SOC is the vehicle's SOC at the current moment, and EF is the selected equivalent factor ΔSOC = |SOC|. ref -SOC|,K P K I is a coefficient.

Claims

1. A plug-in hybrid electric vehicle energy management system based on road segment information for power allocation, characterized in that, The energy management system includes an offline part and an online part. The offline part includes a DP module, a data analysis module, and a neural network module connected in sequence. The online component includes: a traffic information acquisition module, a work condition recognition module, a vehicle module, an ECMS algorithm module, and a SOC planning module. The traffic information acquisition module and the work condition recognition module are connected, while the vehicle module, the ECMS algorithm module, and the SOC planning module are connected in sequence. The system includes online and offline components for the neural network module, driving condition recognition module, and ECMS algorithm module. The offline component involves driver-collected and processed data. The neural network module uses this processed data to train the driving condition recognition module to identify road conditions. The processed data is then input into the DP module for calculation, and the results are fed into the data analysis module for further training of the neural network module, yielding the ratio of energy consumption per unit mileage under various driving conditions. Based on the driving conditions and energy consumption per unit mileage, the ECMS equivalence factor is optimized offline. The online component involves the traffic information acquisition module acquiring road segment mileage and average vehicle speed per segment. The system uses speed and average number of stops per road segment to identify different road segment conditions. The operating condition identification module employs a trained support vector machine to record the operating condition category for each segment. The system inputs the vehicle's current available State of Charge (SOC), the identified road segment operating condition category, and the corresponding road segment mileage information into the neural network module to obtain the ratio of power consumption per unit mileage for each operating condition. This ratio is then input into the SOC planning module to generate a reference trajectory. Based on the road segment operating condition category and the SOC reference trajectory, the ECMS algorithm module selects the equivalent factor EF. Finally, based on information such as the vehicle's required torque, the ECMS algorithm calculates the control parameters to perform vehicle control.

2. A method for energy management of plug-in hybrid electric vehicles based on road segment information-based power allocation, characterized in that, The method includes the following steps: (1) First, driver data is collected and the data is divided into four categories: congested, relatively smooth, smooth and high-speed conditions. (2) Randomly combine each working condition of the classified segments into working conditions of different distances, calculate the average vehicle speed and average number of stops for each combined working condition, and use this data to train the support vector machine, which is used to identify the driving conditions of the road segment. (3) Dynamic programming (DP) was used to study the interrelationships of four operating conditions. A fixed distance of 60 km was used, with different percentages of each operating condition relative to the total operating mileage. Multiple calculations were performed, and the DP dynamic programming results were processed to obtain the SOC / km ratio for each operating condition. : : : The values ​​for each operating condition are then compared to determine the interrelationships between them, thus revealing the proportion of electricity consumption for each condition. The calculation formula is as follows: ； in : : : The ratio represents the ratio of electricity consumption per unit distance under four different operating conditions; The value of SOC / km for power consumption under operating condition category 1. This represents the SOC / km energy consumption value for operating condition category 2. This represents the State of Charge (SOC) per kilometer (Km) for operating condition category 3. The SOC / km value is the energy consumption per kilometer for operating condition category 4. The mileage for each operating condition is... With available SOC: As input, , , , The output is used to train a BP neural network, which consists of three layers: an input layer, a hidden layer, and an output layer, and is used to allocate power based on the information from the map to the road section conditions. in The mileage of road sections in work condition categories 1, 2, 3, and 4. This represents the current available SOC value for the vehicle. (4) Road segment information obtained by the traffic information acquisition module: average vehicle speed, average number of stops, and road segment mileage after support vector machine recognition for each road segment. Vehicle available battery information: ; data , The input is fed into a BP neural network to obtain the output: the ratio of power consumption per unit distance under various operating conditions. : : : ; (5) The BP neural network can obtain the operating condition information of each segment of the trip through the traffic information acquisition module to predict the power allocation of a trip. The output obtained by the BP neural network is: the ratio of power consumption per unit distance under each operating condition. : : : The formulas involved in calculating the State of Computation (SOC) to obtain the relationships are as follows: ； ； Let x be the vehicle's current available battery power, and let x be a set variable (used as an intermediate variable in the planning). For SOC reference trajectory; This represents the distance traveled in condition category 1. This represents the distance traveled in condition category 2. This represents the distance traveled in working condition category 3. This represents the distance traveled so far in condition category 4. (6) To reduce the impact of errors in working condition identification on the reference trajectory, the average vehicle speed and average number of stops of different road segments obtained by the traffic information acquisition module are used to determine the working condition of each road segment. To enhance the accuracy of identification, an online identification module for the current road segment working condition is added. The identification time is the most recent driving data within the road segment in 300 seconds. If the identification time is inconsistent for more than 60 seconds, the SOC is re-planned according to the working condition category identified online. At the same time, since the road segment information is constantly updated, it is updated every 60 seconds. The updated content includes road segment information, identification information and new SOC reference trajectory. (7) The classified working conditions are combined and spliced ​​into four major categories: congested, relatively smooth, smooth and high-speed. For each category, the equivalent factor is optimized. Here, only the SOC change is optimized. Distance is not used as a variable. The only variable is SOC. The equivalent fuel consumption model is optimized. The minimum SOC is set to 0.

3. The optimization is performed by GA genetic algorithm. The EF equivalent factor is the variable and the fuel consumption is the objective function. The distance for each working condition is set to 60km. The SOC gradually decreases from 1 to 0.3 at intervals of 0.

05. The equivalent factor is optimized. The optimized EF is recorded and the corresponding SOC / Km value is also recorded. The recorded values ​​are plotted as an SOC / Km-EF map. (8) The planned SOC corresponds to the working condition, calculate the unit mileage power consumption under each road segment, find the EF corresponding to the SOC / Km under the corresponding working condition from the Map, and make an equivalent factor; obtain the corresponding equivalent factor, and perform energy management of ECMS at the lower level. The input is the vehicle demand torque, and the output is the optimal engine motor torque to control the vehicle. ； ； ； ； in To minimize fuel consumption, t represents the driving time under operating conditions, and N represents the final value of the driving time under operating conditions. The time interval is set to 1 second. This refers to engine fuel consumption. Engine torque For engine speed, EF represents the equivalent fuel consumption of the electric motor, and EF is the equivalence factor. Let η be the motor power, η be the battery efficiency, and Q be the fuel calorific value. For motor torque, This refers to the motor speed. For optimal fuel consumption, This refers to engine torque. This refers to the motor torque. This represents the total required torque. (9) During operation, fluctuations between the planned SOC curve and the actual SOC curve are inevitable. When a significant deviation from the planned SOC occurs, EF correction is performed to bring it closer to the planned curve. The method used is piecewise PI, which makes minor adjustments within a reasonable range of deviation from the planned curve and makes further adjustments when the deviation exceeds a threshold. The piecewise PI function is as follows: ； in The planned SOC trajectory is defined by SOC, where SOC is the vehicle's SOC at the current moment, and EF is the selected equivalent factor. = , , is a coefficient.

3. The method according to claim 2, characterized in that, The offline optimization DP dynamic programming data uses different proportions under different operating conditions as optimization data, and the proportions of a certain operating condition are gradually increased. The final optimization result of dynamic programming is the ratio of unit distance power consumption under different operating condition proportions under different available power.

4. The method according to claim 2, characterized in that, The training data of the BP neural network is the data obtained by DP dynamic programming. The input is the vehicle's available power, the mileage of four different operating conditions, and the mileage of the four different operating conditions arranged in order. The output is the ratio of the unit power consumption of the four different operating conditions.

5. The method according to claim 2, characterized in that, The SOC trajectory planning is based on the ratio of unit power consumption under four operating conditions output by the BP neural network and the available power of the vehicle to plan the SOC reference trajectory. Specifically, it involves establishing a functional relationship to allocate power and obtain the SOC planned trajectory.

6. The method according to claim 2, characterized in that, The data for condition identification comes from the connected map. The identification parameters are average vehicle speed and average number of stops. The result is the overall condition category of a certain road segment.

7. The method according to claim 2, characterized in that, After identifying the road condition category, an online identification of the current working condition was added. The identification time is the most recent driving data within the road segment in the last 300 seconds. If the identification is inconsistent, the SOC will be re-planned according to the online identified working condition category. The condition for replacing the operating condition category is that the time to identify inconsistencies exceeds 60 seconds.

Images