Hybrid vehicle energy management system and method based on traffic state, storage medium and computer program product

Abstract

The invention provides a hybrid vehicle energy management system and method based on a traffic state, a storage medium and a computer program product, and the system comprises the steps: constructing a congestion prediction model based on a deep learning model, the congestion prediction is used for predicting the traffic flow and the driving speed of a future time period according to the historical traffic flow data and the historical driving speed data; calculating a first congestion index according to the predicted traffic flow in the future period; calculating a second congestion index according to the predicted driving vehicle speed sequence of the future time period; performing weighted calculation on the first congestion index and the second congestion index to obtain a comprehensive traffic congestion index in a future time period; traffic jam types are divided according to the interval where the comprehensive traffic jam index is located; and determining an energy management mode and / or a target SOC of the hybrid vehicle based on the traffic jam type. According to the invention, multi-source data are collected and analyzed in real time, congestion is accurately quantified by using a deep learning algorithm, an energy distribution strategy of vehicles is dynamically adjusted, and the energy utilization efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of hybrid vehicle energy management technology, specifically relating to a hybrid vehicle energy management system, method, storage medium, and computer program product based on traffic conditions. Background Technology

[0002] Driving scenario prediction can be used for intelligent energy management, including but not limited to predictions of congestion and smooth traffic. Existing technologies mostly focus on identifying current traffic conditions, with limited research on predicting driving scenarios over multiple future time periods and using the prediction results for predictive energy management. This inability to adapt to the energy demands of different traffic scenarios in advance limits the energy utilization efficiency of hybrid vehicles. Summary of the Invention

[0003] To improve the energy management capabilities of hybrid vehicles, this invention proposes a hybrid vehicle energy management system, method, storage medium, and computer program product based on traffic conditions.

[0004] A traffic-state-based hybrid vehicle energy management system, which achieves one of the objectives of this invention, includes: Congestion prediction module: This module is used to construct a congestion prediction model based on a deep learning model. The congestion prediction is used to predict traffic flow and driving speed for future periods based on historical traffic flow data and historical driving speed data. The historical traffic flow data comes from a cloud-based traffic network system; the historical driving speed data comes from the vehicle's CAN bus, and the acquisition frequency is consistent with that of the historical traffic flow data. Both the historical traffic flow data and the historical driving speed data have undergone outlier removal processing, and missing data is filled in using linear interpolation. Congestion Index Calculation Module: Used to calculate a first congestion index based on the predicted traffic flow for the future period; calculate a second congestion index based on the predicted driving speed sequence for the future period; and perform a weighted calculation on the first and second congestion indices to obtain a comprehensive traffic congestion index for the future period. Energy management strategy adjustment module: used to determine the energy management mode and / or target SOC of hybrid vehicles based on the comprehensive traffic congestion index of the future time period; the energy management mode includes CS (Charge Sustaining Mode) and CD (Charge Depleting Mode).

[0005] Furthermore, the working process of the congestion prediction model module includes: normalizing historical traffic flow data and historical driving speed data and dividing them into training sets and test sets according to a certain ratio. The input samples of the training set and the test set are historical data for m consecutive time periods, and the output samples are the prediction data for the next n time periods. An LSTM network containing one input layer, two hidden layers and one output layer is constructed. The normalized training set is input into the congestion prediction based on the LSTM network for training. The prediction error is calculated by mean square error. The Adam optimizer is used to backpropagate and update the network weight matrix and offset until the maximum number of iterations is reached or the prediction error is lower than a preset threshold. The normalized test set is then input into the trained LSTM network for verification. After successful verification, the congestion prediction model is obtained.

[0006] The technical benefits include: data normalization eliminates the difference in dimensions between traffic flow and driving speed, preventing the data volume from affecting the LSTM network's ability to capture temporal features; proportionally dividing the training / test sets and independently validating them ensures that the model has good generalization ability, avoiding overfitting or underfitting; and the final output congestion prediction model can stably provide traffic flow / driving speed data for future periods, providing reliable input for subsequent congestion index calculation and energy strategy adjustment.

[0007] Furthermore, it also includes a data acquisition module, used to extract traffic flow data of road nodes from the cloud-based traffic network system to form historical traffic flow sequences. q 1. q 2、...、 q t (t represents the historical data time point), the acquisition frequency is set to 5 seconds / time; it is also used to acquire vehicle driving speed in real time from the vehicle CAN bus to form a historical driving speed sequence. v 1. v 2、...、 v t The data collection frequency is consistent with that of traffic flow data. The traffic flow data and driving speed data are cleaned, outliers are removed, and missing data are filled in using linear interpolation.

[0008] Furthermore, methods for calculating the first congestion index based on predicted traffic flow include: ; r1 represents the first congestion index; This indicates the predicted traffic flow. This indicates the actual capacity of the road.

[0009] The technical benefits include: combining future traffic flow predicted by LSTM to directly quantify the overall saturation of the road network from a macro-level perspective, avoiding the one-sidedness of relying solely on a single real-time traffic flow, and ensuring that subsequent congestion type classification does not deviate from the overall operating status of the road network.

[0010] Furthermore, methods for calculating the second congestion index based on predicted driving speeds include: ; r2 represents the first congestion index; Indicates the predicted driving speed. This indicates the maximum permissible driving speed when the road is clear.

[0011] The technical benefits include: directly reflecting the future driving status of a single vehicle by predicting driving speed, quantifying congestion from a micro-vehicle perspective, capturing individual driving experiences that cannot be covered by macro-road network indices, supplementing the shortcomings of macro-indices, allowing the comprehensive congestion index to simultaneously cover both the global road network and individual vehicle dimensions, making the congestion type classification more in line with actual driving scenarios, providing a basis for subsequent energy strategy adjustments that is accurate down to the individual vehicle level, and conforming to the core improvement of multi-source data collaborative prediction.

[0012] Furthermore, methods for determining the energy management mode and / or target SOC of hybrid vehicles include: When the overall traffic congestion index is in the first range, the energy management mode of the hybrid vehicle is the first mode that prioritizes the consumption of battery energy, and / or the target SOC maintains the vehicle's initial SOC value.

[0013] When the overall traffic congestion index is in the second range, the energy management mode of the hybrid vehicle is the first mode that prioritizes the consumption of battery energy. When the SOC drops to the lower limit of the first set range, the energy management mode switches from the first mode to the second mode that maintains the battery SOC in the first set range, so as to maintain the target SOC in the first set range.

[0014] When the overall traffic congestion index is in the third range, the energy management mode of the hybrid vehicle is the third mode that maintains the battery SOC in the second set range, where the minimum value of the second set range is greater than or equal to the maximum value of the first set range.

[0015] When the comprehensive traffic congestion index is in the fourth range, the energy management mode of the hybrid vehicle is the fourth mode, which maintains the battery SOC in the third set range. The minimum value of the third set range is greater than or equal to the maximum value of the second set range, in order to cope with future moderate traffic congestion.

[0016] When the comprehensive traffic congestion index is in the fifth interval, the energy management mode of the hybrid vehicle is the fifth mode, which maintains the battery SOC in the fourth set interval. The minimum value of the fourth set interval is greater than or equal to the maximum value of the third set interval, so that the vehicle mainly uses electricity when entering congested road conditions.

[0017] Furthermore, the preferred range for the first interval is [0, 0.2); the preferred range for the second interval is [0.2, 0.4); the preferred range for the third interval is [0.4, 0.6); the preferred range for the fourth interval is [0.6, 0.8); and the preferred range for the fifth interval is [0.8, 1].

[0018] The technical effects include: For the first section of the traffic flow scenario, the CD mode (power consumption mode) is used first. This mode can make full use of the battery's high-efficiency discharge range under smooth traffic conditions. For example, when driving at high speed, the battery discharge efficiency is higher than that of the engine, which reduces the energy loss caused by frequent engine intervention and improves energy utilization efficiency. The target SOC parameter can be maintained or reduced, which simplifies the control logic and avoids the battery redundant energy consumption caused by blindly increasing the target SOC. Since there is no frequent start-stop in the smooth traffic flow scenario, there is no need to store extra power, which is in line with the energy-saving concept of dynamic adaptation according to the congestion type in the briefing.

[0019] For the energy strategy in the second zone (basic smooth traffic scenario), the CD mode is still the main mode, continuing the energy-saving logic of the smooth traffic scenario and prioritizing the consumption of battery energy; when the battery level drops to the lower limit of the first set zone, the CS mode (battery maintenance mode) is switched to avoid the risk of insufficient battery power in the event of a sudden traffic jam, and reserve power for the subsequent traffic jam scenario. This not only makes good use of the efficient discharge opportunity during smooth traffic but also ensures the energy supply when switching operating conditions.

[0020] The energy strategy for the third zone (light traffic congestion scenario) can cope with the frequent acceleration and deceleration of the vehicle in light traffic congestion. If the battery is over-discharged, it will shorten its life. The CS mode maintains the SOC stability to avoid this problem and protect the battery. It stores more power to cope with the instantaneous energy demand of frequent start-stop, and reduces the energy consumption and noise of the engine caused by frequent start-stop to replenish power. This design not only meets the needs of light traffic congestion, but also continues the core of predictive energy management, avoiding the decline in driving experience caused by insufficient power, and balancing the goals of energy consumption, battery life and comfort.

[0021] The energy strategy for the fourth zone (moderate congestion scenario) can handle the more frequent start-stop conditions in moderate congestion. In moderate congestion, the vehicle accelerates and decelerates more frequently, and the instantaneous energy demand is greater. The SOC range is higher than that in light congestion, which can ensure sufficient battery power to meet more intensive power demand, such as frequent starts and low-speed driving. The engine is extremely inefficient in frequent start-stop situations. Maintaining a higher SOC can avoid the engine being forced to start due to insufficient battery power, reducing the number of times the engine intervenes, thereby further reducing energy consumption. At the same time, the control logic of this energy strategy to stabilize SOC can prevent battery damage during high-frequency charging and discharging, taking into account both energy consumption optimization and battery protection. It is fully consistent with the design concept of dynamically adjusting the energy strategy for different congestion levels in the disclosure document.

[0022] The energy strategy for the fifth zone (severe traffic congestion scenario) can handle situations where vehicles travel at low speeds for extended periods and frequently start and stop during heavy traffic. In this scenario, engine efficiency is at its lowest. The highest SOC range ensures that the vehicle prioritizes electric power, minimizing engine starts and significantly reducing energy consumption. High SOC reserves can also handle sudden short-term acceleration needs, such as brief escapes from traffic jams, avoiding power response delays due to insufficient battery power and improving driving comfort. Maintaining a stable high SOC also prevents deep battery discharge under the complex conditions of heavy traffic congestion, protecting battery cycle life. Ultimately, it achieves a balance between energy consumption optimization, comfort improvement, and battery protection, perfectly aligning with the invention's purpose of refined energy management based on prediction, as outlined in the disclosure document.

[0023] A second objective of this invention is a hybrid vehicle energy management method based on traffic conditions, comprising: A congestion prediction model based on a deep learning model is constructed. The congestion prediction is used to predict traffic flow and driving speed in future periods based on historical traffic flow data and historical driving speed data. The historical traffic flow data comes from a cloud-based traffic network system. The historical driving speed data comes from the vehicle CAN bus, and the acquisition frequency is consistent with that of the historical traffic flow data. Both the historical traffic flow data and the historical driving speed data have undergone outlier removal processing, and missing data are filled in using linear interpolation. A first congestion index is calculated based on the predicted traffic flow for the future period; a second congestion index is calculated based on the predicted driving speed sequence for the future period; and the first and second congestion indices are weighted to obtain a comprehensive traffic congestion index for the future period. Traffic congestion types are classified according to the range of the comprehensive traffic congestion index; the energy management mode and / or target SOC of hybrid vehicles are determined based on the traffic congestion type; the energy management mode includes CS (Charge Sustaining Mode) and CD (Charge Depleting Mode).

[0024] A non-transitory computer-readable storage medium for achieving the third objective of the present invention stores a computer program thereon, which, when executed by a processor, implements the steps of the traffic state-based hybrid vehicle energy management method.

[0025] A computer program product for achieving the fourth objective of the present invention includes a computer program / instructions that, when executed by a processor, implement the steps of the traffic state-based hybrid vehicle energy management method.

[0026] The beneficial effects of this invention include: 1. By collecting macro-level traffic flow data of the road network through the cloud platform and micro-level driving speed data through the vehicle-side CAN bus, a dual-source data collaborative input system is formed, taking into account both external road network and internal vehicle factors. At the same time, outlier removal and linear interpolation completion processing are performed to ensure data continuity and completeness, providing high-quality time-series data support for subsequent congestion prediction and avoiding the impact of poor-quality data on prediction accuracy.

[0027] 2. By employing an LSTM neural network to perform multi-step predictions on dual-source historical data, the limitations of existing technologies that only identify the current traffic state are overcome. Through processes such as data normalization, Adam optimizer error reduction, and training / test set validation, the model can stably output future traffic flow and vehicle speed, enabling early understanding of traffic state changes and providing core basis for predictive energy management.

[0028] 3. The comprehensive congestion index is calculated by combining macro traffic flow r1 and micro vehicle speed r2, which is more in line with actual driving scenarios than a single indicator; five types of congestion are divided according to the index, clearly defining the degree of congestion in different future time periods, avoiding energy strategy mismatch caused by vague congestion judgment.

[0029] 4. Dynamically adjust strategies based on different congestion types: When traffic is smooth, prioritize CD mode for efficient power consumption; when congested, use CS mode to stabilize SOC and prevent excessive discharge, with the target SOC gradually increasing as congestion worsens; at the same time, adapt to driving style (smooth traffic economy mode, congestion comfort mode), which not only improves the energy utilization efficiency of hybrid vehicles, but also reduces the energy consumption and impact of frequent engine start-stop, balancing energy saving, battery protection and driving comfort. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the vehicle congestion prediction model described in this invention; Figure 2 This is a schematic diagram of the system described in this invention. Detailed Implementation

[0031] The following detailed embodiments are provided to explain the technical solutions of the present invention, so that those skilled in the art can understand the present invention. The scope of protection of the present invention is not limited to the following specific embodiments. Any modifications or improvements made by those skilled in the art that incorporate the technical solutions of the present invention but differ from the following detailed embodiments are also within the scope of protection of the present invention.

[0032] This invention provides a method for hybrid vehicle energy management based on traffic conditions, such as... Figure 1 As shown, it specifically includes: 1. Data Extraction When making predictions about vehicle congestion, two aspects are considered: road network traffic flow data and vehicle speed.

[0033] 1.1 Extracting traffic network data Leveraging the transportation network, traffic data from different road nodes in the cloud is acquired, with a data collection frequency of 5 seconds per acquisition to ensure data timeliness. The data includes traffic flow (unit: vehicles / hour), traffic speed (unit: km / h), and vehicle density (unit: vehicles / km). Traffic flow is extracted as the core input parameter and denoted as the historical traffic flow sequence q1, q2, ..., q... t t represents the historical data time point, with each time point spaced 5 seconds apart. For example, q1 corresponds to the traffic flow in the first time period, q2 corresponds to the traffic flow in the second time period, and so on.

[0034] 1.2 Extract vehicle-side data In addition to traffic flow data from the road network, on the vehicle side, the vehicle's own driving speed is collected in real time via the vehicle's CAN bus. The collection frequency is consistent with that of traffic flow data, and the data is recorded as historical driving speed sequences v1, v2, ..., v t v t With q t Each pair represents a one-to-one correspondence, signifying time-series data at time period t. Both serve as the data source for driving scenario prediction. Furthermore, data cleaning is required after collection. The criteria remove outliers from traffic flow and driving speed data, such as zero values ​​caused by equipment failure or extreme values ​​far exceeding the normal range. At the same time, missing data is filled in using linear interpolation to confirm the integrity of the input data and provide reliable input for multi-step prediction.

[0035] 2. Construct and train a congestion prediction model Long Short-Term Memory Neural Network (LSTM) was used to process historical traffic flow data sequences. q 1. q 2、...、 q tand historical driving speed sequence v 1. v 2、...、 v t Multi-step prediction is performed, with the goal of obtaining the future. n Traffic flow values ​​for 5-second time intervals (consistent with the data collection frequency). , ... and driving speed , ... , and These correspond to the traffic flow and driving speed in the 5th second of the future, respectively. n can be set according to actual needs, such as 10 or 20. In this embodiment of the invention, n=12 is preferred, that is, to predict the traffic flow value and driving speed in the next 1 minute (e.g., 12 5-second intervals), covering the changes in traffic status in the short future period.

[0036] As mentioned above, traffic flow data comes from traffic network data in the cloud platform, and driving speed comes from vehicle-side driving data; both take into account external and internal factors, and both have continuous time series attributes of 5 seconds / time, which can better capture the time series change patterns through LSTM, thereby accurately predicting driving conditions in multiple time periods in the future.

[0037] The specific process of using LSTM networks to predict future traffic flow and vehicle speed includes: 2.1 Data Preprocessing: The acquired traffic flow dataset and driving speed dataset are divided into training and test sets according to a set ratio. The input samples of the training set are historical sequential data for m consecutive time periods, where m is the LSTM input time step. This means predicting data for the next n time periods based on data from the previous m collection cycles or time periods. Preferably, m=10, corresponding to the continuous historical data from the previous 10 collection cycles. The output samples of the training set are the predicted data for the next n time periods corresponding to the input samples. For example, if the input sample is... q t-9 ~ q t , v t-9 ~ v t Output predicted value ~ and ~ The input and output sample partitioning rules for the test set are consistent with those for the training set. Meanwhile, to avoid the influence of unit of measurement, the training and test set data are normalized separately, as shown in equations (1) and (2): in, and These represent the data before and after normalization in the training and test sets, respectively. and These represent the maximum and minimum values ​​of the data items, respectively.

[0038] 2.2 Building and Training an LSTM Network In one embodiment, an LSTM network is constructed, comprising one input layer, two hidden layers, and one output layer. The number of neurons in the input layer is consistent with the input time step m (i.e., 10 neurons). The first hidden layer contains 64 LSTM neurons, the second hidden layer contains 32 LSTM neurons, and the output layer contains n neurons, corresponding to the predicted values ​​at n future time steps. The network activation function is the Tanh function, and the forget gate, output gate, and output gate activation functions are sigmoid functions.

[0039] The training sets of normalized traffic flow data and driving speed data are respectively input into the LSTM network for training. The network training is shown in Equations (3) to (9). The forget gate is calculated as shown in Equation (3), the input gate is calculated as shown in Equation (4), the candidate memory cell is calculated as shown in Equation (5), the output gate is calculated as shown in Equation (6), the memory cell is updated as shown in Equation (7), the hidden state is updated as shown in Equation (8), and the network output is calculated as shown in Equation (9). in, This indicates the input of this layer. The data at any given moment, specifically the normalized traffic flow and vehicle speed at the current moment. This indicates the input of this layer. Hidden information in time , and These represent the forget gate, input gate, and output gate respectively. The state at any given moment, For memory cells, , For the corresponding weights, This is the offset. This represents the final output, which is the traffic flow and driving speed over multiple future time periods. To output the corresponding weights, and These represent the sigmoid and Tanh activation functions, respectively. This indicates that matrices are multiplied element by element; These represent the weight matrices corresponding to the forget gate, input gate, candidate memory cells, output gate, and network output, respectively. These are the corresponding offsets.

[0040] 2.3 Multi-step prediction iteration After training, predictions for the next n time periods are achieved through multi-step prediction: Input the data from the last m moments of the historical sequence. For example, if m=10, input the normalized AC flow data. q t-9 ~ q t and driving speed data v t-9 ~ v t Assuming n=5, the model outputs predicted values ​​for 5 future times. ~ , ~ .

[0041] 2.4. Inverse Normalization The normalized traffic flow and driving speed predictions for the next time period output by the network are converted into actual physical predictions, i.e., traffic flow and driving speed at time t+1 in the future, through the inverse normalization operation. 2.5 Error Calculation In one embodiment, mean squared error (MSE) is used as the model error evaluation index. The errors of the traffic flow predicted by LSTM compared with the actual traffic flow in the training set samples, and the errors of the predicted driving speed compared with the actual values ​​in the training set samples are calculated. Based on the error results, the weight matrix of the LSTM network is updated through backpropagation using the Adam optimizer. and offset The training process is repeated until the maximum number of iterations for LSTM network training is reached. The present invention preferably sets the maximum number of iterations to 500, or the error value is lower than a preset threshold, such as MSE < 0.01.

[0042] 2.6 Network Testing The normalized test set data is input into the trained LSTM neural network. The predicted values ​​of the test set are calculated by equations (3) to (9). After inverse normalization, the predicted values ​​are compared with the true values ​​of the test set. If the MSE error of the test set is lower than the preset threshold, the network model is verified to be feasible and the congestion prediction model is obtained. If the threshold is not met, the process returns to step (2) to adjust the network structure, such as increasing the number of hidden layer neurons or expanding the training set size to retrain the LSTM network.

[0043] 3. Intelligent Energy Management System 3.1 Calculate the traffic congestion index Traffic flow sequences for the next n time periods predicted by LSTM ~ and driving speed sequence ~ The traffic congestion index for each future time period is calculated separately to ensure that energy strategy adjustments cover multiple future time periods. On the one hand, this is based on the predicted traffic flow for each future time period. (k=1,2,...,5, corresponding to the 1st to 5th data collection periods in the future), calculate the first traffic congestion index r1 for this time period according to the following formula: in This indicates the predicted traffic flow. This indicates the actual capacity of the road (the maximum number of vehicles it can handle). The closer the value is to 1, the more congested the road is; the closer it is to 0, the smoother the traffic flow.

[0044] Based on the predicted driving speed for each future time period ( k =1,2,...,5, corresponding to the 1st to 5th data collection periods in the future), calculate the second traffic congestion index r2 for this time period: in Indicates the predicted driving speed. v This indicates the maximum driving speed allowed when the road is clear. The closer the value is to 1, the more congested the traffic; the closer it is to 0, the clearer the traffic.

[0045] Then, the traffic congestion index prediction results for each future time period, taking into account internal and external factors, are weighted to obtain the traffic congestion index used to classify the type for that time period, where k1 and k2 are each set to 0.5. 3.2 Traffic Congestion Type Classification: Based on the calculated traffic congestion index for each future time period, the traffic status is determined according to Table 1 below as one of the following: smooth, basically smooth, lightly congested, moderately congested, or heavily congested. This classification clarifies the specific congestion type for each of the next n time periods. The time periods are the same as the data collection period. For example, if the data collection period is 5 seconds, the specific congestion type within the next 5n seconds is predicted. Assuming n=5, if the next 5 seconds are basically smooth, the next 15-25 seconds will be lightly congested, providing a basis for adjusting energy strategies by time period.

[0046] Table 1 3.3. Based on varying traffic congestion indices for different time periods in the future, predict and identify driving scenarios, and adjust energy management strategies accordingly. This includes selecting CS (Charge Sustaining Mode), CD (Charge Depleting Mode), and adjusting the target SOC. CS mode maintains the battery SOC (State of Charge) within a stable range during vehicle operation, without intentionally consuming or replenishing the battery. CD mode prioritizes battery power until the SOC drops to a preset lower limit, at which point it switches to CS mode. Specific strategies can be tailored to different road conditions. When the traffic congestion index is in the range [0, 0.2), meaning that the road conditions will be smooth in a certain future time period: the CD mode will be the primary mode, prioritizing the consumption of battery energy; maintain or reduce the current target SOC; such as target SOC = 15%~20%; When the traffic congestion index is in the range [0.2, 0.4), meaning that the road conditions will be basically smooth in a certain period of time in the future: use CD mode as the main mode, and switch to CS mode when the SOC drops to the lower limit of the target SOC (e.g., 20%) to maintain the target SOC at 20%; When the traffic congestion index is in the range [0.4, 0.6), that is, when the road conditions are mildly congested in a certain period of time in the future: the CS mode is used as the main mode to avoid the battery from being over-discharged in congested road conditions with frequent start and stop. The target SOC is appropriately increased compared with the basic smooth working condition, and is set to 20%~25%. When the traffic congestion index is in the range [0.6, 0.8), that is, when the road conditions are moderately congested in a certain period of time in the future: the CS mode is used as the main mode, and the target SOC is appropriately increased compared with the light congestion mode to cope with the future moderate congestion road conditions, such as setting it to 25%~30%; When the traffic congestion index is in the range [0.8, 1.0], that is, when the road conditions are severely congested in a certain period of time in the future: the CS mode is the main mode, and the target SOC is appropriately increased compared with the moderate congestion mode to cope with the future severe congestion conditions, so that the main mode of use is electricity when entering congested road conditions. At this time, the target SOC is set to 30%~35%.

[0047] In addition, driving style selection can be adjusted based on the predicted congestion type to improve adaptability to diverse traffic scenarios. For example, the economy mode can be adapted to smooth traffic conditions, while the comfort mode can be adapted to congested traffic conditions to reduce acceleration and deceleration impact.

[0048] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0049] This invention also provides a hybrid vehicle energy management system based on traffic conditions, characterized in that, as Figure 2 As shown, it includes: Congestion prediction module: This module is used to construct a congestion prediction model based on a deep learning model. The congestion prediction is used to predict traffic flow and driving speed for future periods based on historical traffic flow data and historical driving speed data. The historical traffic flow data comes from a cloud-based traffic network system; the historical driving speed data comes from the vehicle's CAN bus, and the acquisition frequency is consistent with that of the historical traffic flow data. Both the historical traffic flow data and the historical driving speed data have undergone outlier removal processing, and missing data is filled in using linear interpolation. Congestion Index Calculation Module: Used to calculate a first congestion index based on the predicted traffic flow for the future period; calculate a second congestion index based on the predicted driving speed sequence for the future period; and perform a weighted calculation on the first and second congestion indices to obtain a comprehensive traffic congestion index for the future period. Energy management strategy adjustment module: used to determine the energy management mode and / or target SOC of hybrid vehicles based on the comprehensive traffic congestion index of the future time period; the energy management mode includes CS (Charge Sustaining Mode) and CD (Charge Depleting Mode).

[0050] In one embodiment, the working process of the congestion prediction model module includes: normalizing historical traffic flow data and historical driving speed data and dividing them into training sets and test sets according to a certain ratio. The input samples of the training set and the test set are historical data for m consecutive time periods, and the output samples are the prediction data for the next n time periods. An LSTM network containing one input layer, two hidden layers and one output layer is constructed. The normalized training set is input into the congestion prediction based on the LSTM network for training. The prediction error is calculated by mean square error. The Adam optimizer is used to backpropagate and update the network weight matrix and offset until the maximum number of iterations is reached or the prediction error is lower than a preset threshold. The normalized test set is then input into the trained LSTM network for verification. After successful verification, the congestion prediction model is obtained.

[0051] In one embodiment, the system further includes a data acquisition module for extracting traffic flow data of road nodes from a cloud-based traffic network system to form a historical traffic flow sequence. q 1. q 2、...、 q t (t represents the historical data time point), the acquisition frequency is set to 5 seconds / time; it is also used to acquire vehicle driving speed in real time from the vehicle CAN bus to form a historical driving speed sequence. v1. v 2、...、 v t The data collection frequency is consistent with that of traffic flow data. The traffic flow data and driving speed data are cleaned, outliers are removed, and missing data are filled in using linear interpolation.

[0052] In one embodiment, the method for calculating a first congestion index based on predicted traffic flow includes: ; r1 represents the first congestion index; This indicates the predicted traffic flow. q This indicates the actual capacity of the road.

[0053] In one embodiment, the method for calculating a second congestion index based on predicted driving speed includes: ; r2 represents the first congestion index; Indicates the predicted driving speed. v This indicates the maximum permissible driving speed when the road is clear.

[0054] In one embodiment, the method for determining the energy management mode and / or target SOC of a hybrid vehicle includes: When the comprehensive traffic congestion index is in the first interval [0, 0.2), the energy management mode of the hybrid vehicle is the first mode that prioritizes the consumption of battery energy, namely CD mode, to maintain or reduce the target SOC value, such as 15%~20%.

[0055] In one embodiment, when the comprehensive traffic congestion index is in the second interval [0.2, 0.4), the energy management mode of the hybrid vehicle is the first mode that prioritizes the consumption of battery energy, namely the CD mode. When the SOC drops to the lower limit of 20% of the first set interval, the energy management mode switches from the first mode to the second mode that maintains the battery SOC in the first set interval and does not deliberately consume or replenish the power, so as to maintain the target SOC in the first set interval.

[0056] In one embodiment, when the comprehensive traffic congestion index is in the third interval [0.4, 0.6), the energy management mode of the hybrid vehicle is to maintain the battery SOC in the second set interval of 20%~25%, without deliberately consuming or replenishing the power, i.e., the third mode, CD mode, in order to avoid the battery being over-discharged under congested conditions with frequent start-stop.

[0057] In one embodiment, when the comprehensive traffic congestion index is in the fourth interval [0.6, 0.8), the energy management mode of the hybrid vehicle is to maintain the battery SOC in the third set interval of 25%~30%, without deliberately consuming or replenishing the power, in order to cope with future moderate traffic congestion.

[0058] In one embodiment, when the comprehensive traffic congestion index is in the fifth interval [0.8,1], the energy management mode of the hybrid vehicle is to maintain the battery SOC in the fourth set interval of 30%~35%, and not to deliberately consume or replenish the power, so that the vehicle mainly uses electricity when entering congested road conditions.

[0059] This invention also provides a method for hybrid vehicle energy management based on traffic conditions, including: A congestion prediction model based on a deep learning model is constructed. The congestion prediction is used to predict traffic flow and driving speed in future periods based on historical traffic flow data and historical driving speed data. The historical traffic flow data comes from a cloud-based traffic network system. The historical driving speed data comes from the vehicle CAN bus, and the acquisition frequency is consistent with that of the historical traffic flow data. Both the historical traffic flow data and the historical driving speed data have undergone outlier removal processing, and missing data are filled in using linear interpolation. A first congestion index is calculated based on the predicted traffic flow for the future period; a second congestion index is calculated based on the predicted driving speed sequence for the future period; and the first and second congestion indices are weighted to obtain a comprehensive traffic congestion index for the future period. Traffic congestion types are classified according to the range of the comprehensive traffic congestion index; the energy management mode and / or target SOC of hybrid vehicles are determined based on the traffic congestion type; the energy management mode includes CS (Charge Sustaining Mode) and CD (Charge Depleting Mode).

[0060] This invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the various steps of the method described in this invention.

[0061] This invention also provides a non-transitory computer-readable storage medium storing a computer program. The computer program includes program instructions that, when executed by a processor, implement the various steps of the method described in this invention, which will not be elaborated further here.

[0062] The computer-readable storage medium can be the data transmission apparatus or the internal storage unit of a computer device provided in any of the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium can also be the external storage device of the computer device, such as the plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device.

[0063] Furthermore, the computer-readable storage medium may include both internal storage units and external storage devices of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium may also be used to temporarily store data that is to be output or has already been output.

[0064] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0066] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0067] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0068] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A hybrid vehicle energy management system based on traffic conditions, characterized in that, include: Congestion prediction module: used to build a congestion prediction model based on a deep learning model. The congestion prediction is used to predict traffic flow and driving speed in future periods based on historical traffic flow data and historical driving speed data. Congestion Index Calculation Module: Used to calculate a first congestion index based on the predicted traffic flow for the future period; calculate a second congestion index based on the predicted driving speed sequence for the future period; and perform a weighted calculation on the first and second congestion indices to obtain a comprehensive traffic congestion index for the future period. Energy Management Strategy Adjustment Module: Used to determine the energy management mode and / or target SOC of hybrid vehicles based on the comprehensive traffic congestion index for the future time period.

2. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, The working process of the congestion prediction model module includes: normalizing historical traffic flow data and historical driving speed data and dividing them into training set and test set according to the ratio. The input samples of the training set and the test set are historical data of m consecutive time moments, and the output samples are the prediction data corresponding to n future time moments; building an LSTM network containing 1 input layer, 2 hidden layers and 1 output layer; inputting the normalized training set into the congestion prediction based on the LSTM network for training until the number of training iterations reaches the maximum number of iterations or the prediction error is lower than the preset threshold; then inputting the normalized test set into the trained LSTM network for verification; after successful verification, the congestion prediction model is obtained.

3. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for calculating the first congestion index based on predicted traffic flow include: ； r1 represents the first congestion index; This indicates the predicted traffic flow. This indicates the actual capacity of the road.

4. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for calculating the second congestion index based on predicted driving speeds include: ； r2 represents the first congestion index; Indicates the predicted driving speed. This indicates the maximum permissible driving speed when the road is clear.

5. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for determining the energy management mode and / or target SOC of hybrid vehicles include: When the overall traffic congestion index is in the first range, the energy management mode of the hybrid vehicle is the first mode that prioritizes the consumption of battery energy, and / or the target SOC maintains the vehicle's initial SOC value.

6. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for determining the energy management mode and / or target SOC of hybrid vehicles include: When the overall traffic congestion index is in the second range, the energy management mode of the hybrid vehicle is the first mode that prioritizes the consumption of battery energy. When the SOC drops to the lower limit of the first set range, the energy management mode switches from the first mode to the second mode that maintains the battery SOC in the first set range, so as to maintain the target SOC in the first set range.

7. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for determining the energy management mode and / or target SOC of hybrid vehicles include: When the overall traffic congestion index is in the third range, the energy management mode of the hybrid vehicle is the third mode that maintains the battery SOC in the second set range, where the minimum value of the second set range is greater than or equal to the maximum value of the first set range.

8. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for determining the energy management mode and / or target SOC of hybrid vehicles include: When the overall traffic congestion index is in the fourth range, the energy management mode of the hybrid vehicle is the fourth mode, which maintains the battery SOC in the third set range.

9. The hybrid vehicle energy management system based on traffic conditions as described in claim 1, characterized in that, Methods for determining the energy management mode and / or target SOC of hybrid vehicles include: When the overall traffic congestion index is in the fifth range, the energy management mode of the hybrid vehicle is the fifth mode, which maintains the battery SOC in the fourth set range.

10. A traffic-state-based hybrid vehicle energy management method for the system as described in claim 1, characterized in that, include: A congestion prediction model based on a deep learning model is constructed, wherein the congestion prediction is used to predict traffic flow and driving speed in future periods based on historical traffic flow data and historical driving speed data. A first congestion index is calculated based on the predicted traffic flow for the future period; a second congestion index is calculated based on the predicted driving speed sequence for the future period; and the first and second congestion indices are weighted to obtain a comprehensive traffic congestion index for the future period. Traffic congestion types are classified according to the range of the comprehensive traffic congestion index; the energy management mode and / or target SOC of hybrid vehicles are determined based on the traffic congestion type.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the hybrid vehicle energy management method based on traffic conditions as described in claim 10.

12. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the traffic-state-based hybrid vehicle energy management method of claim 10.

Images