Energy management method and device for dual-source powered vehicle, computer device and medium

Abstract

The application relates to the technical field of dual-source power supply, in particular to an energy management method and device of a dual-source power supply vehicle, computer equipment and a medium. The method comprises the following steps: S1, acquiring multiple types of running data of the dual-source power supply vehicle in a sliding time window; S2, based on the multiple types of running data, using a trained energy consumption prediction model to predict the predicted energy consumption of the dual-source power supply vehicle in a prediction time domain; S3, determining the availability information of the catenary for supplying power to the dual-source power supply vehicle in the prediction time domain, the battery constraint condition of the dual-source power supply vehicle and the pantograph constraint condition; S4, taking the predicted energy consumption, the availability information, the battery constraint condition and the pantograph constraint condition as constraint conditions of an objective function, and generating an energy configuration strategy by minimizing the objective function; the energy configuration strategy comprises an energy distribution strategy of the on-board battery and the pantograph, a pantograph operation strategy and a charging and discharging strategy of the on-board battery. The method can timely switch the energy.

Description

Technical Field

[0001] This application relates to the field of dual-source power supply technology, and in particular to an energy management method, device, computer equipment, and medium for a dual-source power supply vehicle. Background Technology

[0002] As policies are further implemented, vehicles with dual-source power supply are becoming an important direction for replacing traditional fuel vehicles. Among them, "dual-source power supply technology," which combines the advantages of flexible range and efficient power replenishment with the dual-source power supply of battery and pantograph, is widely used in high-intensity working conditions such as ports, mining areas, and intercity short-haul transportation.

[0003] Currently, dual-powered vehicles, as described in patent CN120663811A, often determine whether to use a battery or a pantograph to power the vehicle's motor based on whether there is an overhead contact line above the vehicle. This requires starting to use the power grid only after "entering" the contact line area and switching to the battery only after "leaving" the contact line, making it impossible to prepare in advance based on road conditions ahead. This results in delayed energy switching and high energy consumption. Summary of the Invention

[0004] Therefore, it is necessary to provide an energy management method, device, computer equipment, and medium for dual-source powered vehicles to address the aforementioned technical problems.

[0005] An energy management method for a dual-source powered vehicle, the method comprising:

[0006] S1. Acquire various types of operational data for dual-powered vehicles within a sliding time window;

[0007] Preferably, the dual-source power supply vehicle is a vehicle powered by both the overhead contact line and an on-board battery;

[0008] Preferably, the length of the sliding time window is fixed, and the end time of the sliding time window is the current time;

[0009] Preferably, the various types of operational data include the operational status data, environmental data, and driver operation data of the dual-powered vehicle;

[0010] S2. Based on the aforementioned multi-type operational data, use the trained energy consumption prediction model to predict the predicted energy consumption of the dual-source power supply vehicle in the prediction time domain.

[0011] S3. Determine the availability information of the contact network supplying power to the dual-source power supply vehicle in the predicted time domain, the battery constraints of the dual-source power supply vehicle, and the pantograph constraints.

[0012] Preferably, the availability information of the overhead contact system in the prediction time domain refers to whether the overhead contact system is in normal working condition in the prediction time domain and whether it can provide stable and reliable power to vehicles powered by dual sources.

[0013] Preferably, the battery constraints include, but are not limited to, state of charge constraints, charge / discharge power constraints, charge / discharge rate constraints, and battery discharge cycles constraints.

[0014] Preferably, the pantograph constraint conditions include, but are not limited to, raising and lowering constraints, pantograph dwell constraints, pantograph hysteresis constraints, and limits on the number of raising and lowering times and time of the pantograph.

[0015] S4. Using the predicted energy consumption, the availability information, the battery constraints, and the pantograph constraints as the objective function, an energy allocation strategy is generated by minimizing the objective function; the energy allocation strategy includes an energy distribution strategy for the vehicle battery and the pantograph, a pantograph operation strategy, and a vehicle battery charging and discharging strategy.

[0016] In one embodiment, step S4 includes:

[0017] Establish an objective function; the constraints of the objective function include the predicted energy consumption, the availability information, the battery constraints, and the pantograph constraints.

[0018] An energy allocation strategy is generated by minimizing the objective function;

[0019] The expression for the objective function Z is:

[0020]

[0021] Among them, c grid [k] represents the overhead contact line electricity price at time step k; P grid [k] represents the pantograph power at time step k; P batt [k] represents the battery power at time step k; SOC[k] represents the battery state of charge at time step k; SOC ref c is the desired state of charge; deg,1 and c deg,1 As a weighting of battery aging costs; P batt,max The maximum allowable charge and discharge power of the battery; λ SOC λ is the weight of the state-of-charge tracking error; sm λ is the weight of the rate of change of power; ρ ρ[k] represents the soft constraint penalty weight; ρ[k] represents the soft constraint relaxation at time step k; Δt represents the time step size; and N represents the number of time steps.

[0022] In one embodiment, step S4 is followed by:

[0023] When the current moment is in the predicted time domain, the motor of the dual-source power supply vehicle is powered according to the energy configuration strategy, and the energy utilization rate, response time and charge state stability of the dual-source power supply vehicle are obtained.

[0024] Based on the energy utilization rate, response time, and state of charge stability of the dual-source power supply vehicle, the constraints and weights of the objective function are optimized to generate a new energy configuration strategy based on the optimized objective function.

[0025] In one embodiment, the multiple types of operational data include the speed and load of the dual-powered vehicle, the gradient of the route traveled by the dual-powered vehicle, operational information for the dual-powered vehicle, and the ambient temperature along the route.

[0026] In one embodiment, the dual-powered vehicle includes a display and interaction unit for displaying the energy configuration strategy and receiving user operation commands.

[0027] In one embodiment, step S1 further includes:

[0028] When the sampling times of each type of operational data are inconsistent among the multiple types of operational data, any one type of operational data shall be used as the reference data.

[0029] Determine the reference sampling time of the reference data, and determine the data collected at a time adjacent to the reference sampling time from various types of non-reference data;

[0030] Based on the parameters collected at the nearest time, the data of each type of non-reference data at the reference sampling time are calculated;

[0031] Denoising and missing data repair processes are performed on the non-reference data and the reference data at the reference sampling time to obtain multiple types of preprocessed data;

[0032] Based on multiple types of preprocessed data, the time-series characteristics, frequency domain characteristics, and vehicle-derived characteristics of the dual-source power supply vehicle are statistically analyzed and normalized to predict energy consumption.

[0033] In one embodiment, the process of determining the prediction time domain in step S2 includes:

[0034] Multiple candidate prediction time domains and filtering conditions can be preset;

[0035] Based on the historical predicted energy consumption and historical actual energy consumption of each candidate prediction time domain, the deviation, interval coverage and interval width of each candidate prediction time domain are calculated.

[0036] The candidate prediction time domains that meet the screening conditions are determined as the initial prediction time domains. Based on the deviation, the interval coverage and the interval width of the initial prediction time domains, the reliability curves of each initial prediction time domain are obtained.

[0037] Based on the reliability curves of each initial prediction time domain, determine the target indicators for reliability scoring among the deviation, the interval coverage, and the interval width;

[0038] Based on the target index of the initial prediction time domain, a reliability score is obtained for each of the initial prediction time domains;

[0039] The initial prediction time domain corresponding to the maximum value in the reliability score is determined as the prediction time domain.

[0040] An energy management device for a dual-source powered vehicle, the device comprising:

[0041] The data acquisition module is used to acquire various types of operating data of dual-source powered vehicles within a sliding time window;

[0042] The energy consumption prediction module is used to predict the predicted energy consumption of the dual-source power supply vehicle in the prediction time domain based on the multi-type operating data and the trained energy consumption prediction model.

[0043] The condition determination module is used to determine the availability information of the overhead contact line supplying power to the dual-source power supply vehicle in the prediction time domain, the battery constraints of the dual-source power supply vehicle, and the pantograph constraints.

[0044] The strategy acquisition module is used to generate an energy configuration strategy by minimizing the objective function, which is defined by the predicted energy consumption, the availability information, the battery constraints, and the pantograph constraints. The energy configuration strategy includes an energy allocation strategy for the vehicle battery and the pantograph, a pantograph operation strategy, and a vehicle battery charging and discharging strategy.

[0045] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0046] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0047] The aforementioned energy management method, device, computer equipment, and medium for dual-powered vehicles acquire various types of operational data from the dual-powered vehicle within a sliding time window. Based on this data, a trained energy consumption prediction model predicts the vehicle's energy consumption in the prediction time domain. It also determines the availability information of the overhead contact line supplying the vehicle, the battery constraints, and the pantograph constraints. Using predicted energy consumption, availability information, battery constraints, and pantograph constraints as the objective function, an energy configuration strategy is generated by minimizing this objective function. This strategy includes energy allocation strategies for the onboard battery and pantograph, pantograph operation strategies, and onboard battery charging and discharging strategies. This allows for timely switching of the vehicle's power supply path based on the generated strategy, avoiding passive switching only when the onboard battery is nearly depleted or when the vehicle load changes abruptly, thus saving energy. Furthermore, the energy configuration strategy can coordinate the dynamic allocation between the onboard battery and pantograph, preventing redundant power supply or energy waste and improving energy utilization. Attached Figure Description

[0048] Figure 1 This is an application environment diagram of an energy management method for a dual-source powered vehicle in one embodiment;

[0049] Figure 2 This is a flowchart illustrating an energy management method for a dual-source powered vehicle in one embodiment.

[0050] Figure 3 This is a schematic diagram of the data collection and energy consumption prediction process in one embodiment;

[0051] Figure 4 This is a schematic diagram of the control response flow in one embodiment;

[0052] Figure 5 This is a schematic diagram of the overall process of an energy management method for a dual-source powered vehicle in one embodiment. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] The energy management method for dual-source powered vehicles provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, the dual-powered vehicle includes a control module. This module acquires various types of operational data from the dual-powered vehicle within a sliding time window. Based on this data, it uses a trained energy consumption prediction model to predict the vehicle's energy consumption in the prediction time domain. It also determines the availability information of the overhead contact line supplying the vehicle, the battery constraints, and the pantograph constraints. Using the predicted energy consumption, availability information, battery constraints, and pantograph constraints as the objective function, an energy allocation strategy is generated by minimizing this objective function. This strategy includes an energy distribution strategy for the onboard battery and pantograph, a pantograph operation strategy, and an onboard battery charging / discharging strategy.

[0055] In some embodiments, the control module is also used to perform fault detection, thereby enabling fault-tolerant switching of energy management under abnormal conditions.

[0056] In some embodiments, the control module may be a VCU (Vehicle Control Unit) or an integrated VMS (Vehicle Management System).

[0057] In one embodiment, such as Figure 2 As shown, an energy management method for a dual-source powered vehicle is provided, which can be applied to... Figure 1 Taking the control module in the example, the explanation includes the following steps:

[0058] S1. Acquire various types of operational data for dual-powered vehicles within a sliding time window;

[0059] Dual-source power supply vehicles combine power from both overhead contact lines and onboard energy storage. Examples include dual-source electrified heavy-duty trucks that combine overhead contact line power and onboard battery power, and hybrid electric multiple units that combine overhead contact line power and onboard battery power.

[0060] In some embodiments, the length of the sliding time window is fixed, and the end time of the sliding time window is the current time. For example, if the sliding time window is 30 minutes and the current time is 13:00, then the sliding time window is from 12:31 to 13:00.

[0061] In some embodiments, the length of the sliding time window is fixed, and the start time of the sliding time window is updated once at a preset interval. For example, if the preset interval is 2 minutes, the current time is 13:00, and the sliding time window is from 12:31 to 13:00, then the start time of the sliding time window will be updated to 12:33 to 13:02 at 13:02.

[0062] The various operational data include operating status data, environmental data, and driver operation data for dual-powered vehicles. These data can be collected by sensors and then directly acquired by the control module. Furthermore, different types of operational data are collected using different sensors. For example, temperature data in the environmental data is collected using a temperature sensor, and vehicle speed data in the operating status data is collected using a vehicle speed sensor.

[0063] In some embodiments, various types of operational data can also be obtained from OBD (On-Board Diagnostics) via the CAN (Controller Area Network Bus) bus.

[0064] The acquired operational data includes data from at least one time point. For example, with a sliding time window from 12:31 to 13:00, the acquired temperature data includes temperatures at 12:31, 12:35, 12:41, 12:50, and 12:55. Furthermore, the time points for the acquired operational data can be determined based on the sensor's sampling time point.

[0065] S2. Based on multiple types of operational data, use the trained energy consumption prediction model to predict the predicted energy consumption of dual-source power supply vehicles in the prediction time domain.

[0066] The trained energy consumption prediction model is a model capable of predicting the energy consumption required by a dual-powered vehicle over a future period (prediction time domain). Energy consumption prediction models include, but are not limited to, LSTM (Long Short-Term Memory), Transformer models, GRU (Gated Recurrent Unit), 1D-CNN (One-Dimensional Convolutional Neural Network), and XGBoost (eXtreme Gradient Boosting).

[0067] The forecast time domain refers to a future time range, such as the next hour or the next 20 minutes. The forecast time domain includes multiple time periods. For example, if the total length of the forecast time domain is one hour, it can be divided into 60 time periods with a step size of one minute, each with a corresponding forecast energy consumption. The forecast energy consumption can be an interval value or a specific value.

[0068] Each time the start time of the sliding time window is updated, the various types of operational data will also be updated synchronously. At the same time, the updated operational data will be used to predict the predicted energy consumption in the prediction time domain.

[0069] Predicted energy consumption refers to the equivalent electrical power or energy required by a dual-source powered vehicle under specific transportation tasks and driving conditions to overcome various resistances and maintain operation while simultaneously meeting the demands of vehicle accessory loads. It can also be the net energy extracted from the battery after the output energy is fed back to the battery, and the net energy extracted from the pantograph after the input energy is fed back to the grid. Vehicle accessory loads include air conditioning, air compressors, electric power steering, etc.

[0070] In some embodiments, after obtaining the predicted energy consumption, prediction error analysis and uncertainty assessment are performed based on the predicted and actual energy consumption. The prediction error analysis and uncertainty assessment employ a five-layer mechanism: online error monitoring, confidence interval estimation, uncertainty decomposition, online calibration, and policy style modulation.

[0071] 1. Error monitoring: Perform root mean square error, mean absolute error, mean absolute percentage error, negative log-likelihood, and calibration error statistics on the prediction residuals within the sliding time window, and perform drift detection, outputting the drift detection results;

[0072] 2. Uncertainty estimation: Combine model ensemble or use MCDropout (Monte Carlo Dropout) to estimate cognitive uncertainty, and use heteroscedastic regression, quantile regression, or conformal prediction to estimate noise uncertainty;

[0073] 3. Interval or quantile output: Provides the prediction interval for energy consumption, so that the coverage of the prediction interval can be effectively controlled at the set confidence level;

[0074] 4. Online calibration and rollback: Temperature scaling or order-preserving regression is used to calibrate the prediction interval of predicted energy consumption online. When the confidence level is low or the error increases sharply, conformal prediction is used to dynamically adjust the width of the prediction interval.

[0075] 5. Strategy Coupling: Variance, prediction interval width, and coverage are used as risk signals input to reinforcement learning and the controller to achieve opportunity constraint and risk avoidance.

[0076] Temperature scaling scales the variance of Gaussian predictions or calibrates the probability output with temperature to make the nominal coverage approximately equal to the actual coverage.

[0077] Ordinal-preserving regression performs monotonic mapping calibration on quantile or probability outputs to eliminate systematic bias.

[0078] Conformal prediction calculates the width of the non-parametric confidence interval online based on the prediction residual distribution of the sliding time window, ensuring the coverage of a limited number of samples at a set confidence level.

[0079] S3. Determine the availability information of the overhead contact line supplying power to the dual-source power supply vehicle in the prediction time domain, as well as the battery constraints and pantograph constraints of the dual-source power supply vehicle.

[0080] The overhead contact system is a facility that provides external power to vehicles powered by dual power sources. The availability information of the overhead contact system in the prediction time domain refers to whether the overhead contact system is in normal working order and whether it can provide stable and reliable power to vehicles powered by dual power sources.

[0081] Battery constraints include, but are not limited to, state of charge (SCC) constraints, charge / discharge power constraints, charge / discharge rate constraints, and battery discharge cycle constraints. SCC constraints ensure the battery's SCC remains within a safe range, protecting battery life and preventing damage. Charge / discharge power constraints, also known as power slope limits, prevent overheating or performance degradation due to excessive current. Charge / discharge rate constraints ensure the battery operates within a safe current range. Battery discharge cycle constraints, also known as battery aging constraints, limit the number of deep discharges, ensuring the battery remains in a healthy and safe state and preventing rapid aging.

[0082] Pantograph constraints include, but are not limited to, raising and lowering constraints, pantograph dwelling constraints, pantograph hysteresis constraints, and limits on the number and duration of pantograph raising and lowering. Raising and lowering constraints mean that the pantograph can only draw power from the overhead contact line when it is raised and the dual-source powered vehicle is located in a covered section. In areas without overhead contact lines, tunnels, switches, or fault sections, the pantograph must be lowered. Pantograph hysteresis constraints mean that when a dual-source powered vehicle enters a covered section, the pantograph will not immediately raise, but will only raise if hysteresis conditions are met. For example, the pantograph will only raise to begin drawing power when the vehicle's battery state of charge drops below a first threshold (e.g., 30%). Pantograph dwelling constraints mean that when a dual-source powered vehicle is about to leave a covered section, the pantograph will not immediately lower, but will continue to draw power until the battery state of charge reaches a second threshold (e.g., 80%) or it is confirmed that the vehicle has completely left the covered section before lowering the pantograph.

[0083] S4. Using predicted energy consumption, availability information, battery constraints, and pantograph constraints as the objective function, an energy allocation strategy is generated by minimizing the objective function. The energy allocation strategy includes the energy distribution strategy for the vehicle battery and pantograph, the pantograph operation strategy, and the vehicle battery charging and discharging strategy.

[0084] In generating energy allocation strategies, the energy management problem can be constructed as a constrained mathematical optimization problem. Within each sliding time window, the global optimal solution is solved using predicted energy consumption, thus generating energy allocation strategies in a rolling manner. Alternatively, the predicted curve can be used as a constraint, and real-time energy allocation strategies within the safety boundary can include energy allocation strategies for the vehicle battery and pantograph, pantograph operation strategies, and vehicle battery charging and discharging strategies, which are then optimized cumulatively through a reward function. In the default implementation, the predicted curve is the sequence of energy consumption demand predicted in the future prediction time domain at fixed step sizes.

[0085] In some embodiments, minimizing the objective function can be performed using PPO (Proximal Policy Optimization), DDPG (Deep Deterministic Policy Gradient) reinforcement learning model, DQN (Deep Q-Network), or dual-delay deep deterministic policy gradient algorithm, thereby generating the optimal energy allocation strategy under the current operating conditions.

[0086] The energy distribution strategy for the vehicle-mounted battery and pantograph includes how much power the vehicle-mounted battery should provide and how much power the pantograph should draw from the overhead contact line. The pantograph operation strategy includes when the pantograph can be raised and when it can be lowered. The vehicle-mounted battery charging and discharging strategy includes when the vehicle-mounted battery should be charged and when it should be discharged.

[0087] The aforementioned energy management method, device, computer equipment, and medium for dual-powered vehicles acquire various types of operational data from the dual-powered vehicle within a sliding time window. Based on this data, a trained energy consumption prediction model predicts the vehicle's energy consumption in the prediction time domain. It also determines the availability information of the overhead contact line supplying the vehicle, the battery constraints, and the pantograph constraints. Using predicted energy consumption, availability information, battery constraints, and pantograph constraints as the objective function, an energy configuration strategy is generated by minimizing this objective function. This strategy includes energy allocation strategies for the onboard battery and pantograph, pantograph operation strategies, and onboard battery charging and discharging strategies. This allows for timely switching of the vehicle's power supply path based on the generated strategy, avoiding passive switching only when the onboard battery is nearly depleted or when the vehicle load changes abruptly, thus saving energy. Furthermore, the energy configuration strategy can coordinate the dynamic allocation between the onboard battery and pantograph, preventing redundant power supply or energy waste and improving energy utilization.

[0088] In one embodiment, step S4 includes:

[0089] Establish the objective function; the constraints of the objective function include predicted energy consumption, availability information, battery constraints, and pantograph constraints.

[0090] Energy allocation strategies are generated by minimizing the objective function;

[0091] The expression for the objective function Z is:

[0092]

[0093] Among them, c grid [k] represents the overhead contact line electricity price at time step k; P grid [k] represents the pantograph power at time step k; P batt [k] represents the battery power at time step k; SOC[k] represents the battery state of charge at time step k; SOC ref c is the desired state of charge; deg,1 and c deg,1 As a weighting of battery aging costs; P batt,max The maximum allowable charge and discharge power of the battery; λ SOC λ is the weight of the state-of-charge tracking error; sm λ is the weight of the rate of change of power; ρ ρ[k] represents the soft constraint penalty weight; ρ[k] represents the soft constraint relaxation at time step k; Δt represents the time step size; and N represents the number of time steps.

[0094] A time step can be understood as a time period within the prediction time domain. For example, if the total length of the prediction time domain is 1 hour, and the prediction time domain is divided into 60 time periods with a step size of 1 minute, then the first time period is time step 1.

[0095] Pantograph power P at time step k grid [k], Battery power P at time step k batt [k] represents the variables that need to be solved. Among them, the pantograph power P grid If [k] is greater than 0, it indicates that the pantograph needs to be raised, and the pantograph power P grid If [k] equals 0, it means the pantograph needs to be lowered, and the battery power P... batt If [k] is greater than 0, it indicates that the vehicle battery needs to discharge, and the battery power P batt If [k] is less than 0, it means that the vehicle battery needs to be charged.

[0096] When minimizing the objective function, the generated energy allocation strategy must satisfy all the constraints of the objective function.

[0097] In this embodiment, an energy configuration strategy is generated by minimizing the objective function. This enables multi-objective collaborative optimization, breaks the limitation of "single-point optimality", reduces overall operating costs, and upgrades energy management from "passive response" to "proactive planning" and from "experience-driven" to "data and model-driven". Ultimately, it achieves safe, efficient, economical, and long-life operation, realizes sustainable optimization and autonomous learning, and has strong adaptability.

[0098] In one embodiment, the various types of operational data include the speed and load of the dual-powered vehicle, the gradient of the route traveled by the dual-powered vehicle, operational information for the dual-powered vehicle, and the ambient temperature along the route.

[0099] Among these, speed can be collected by a speed sensor, load by a load sensor, gradient by a gradient sensor, ambient temperature by a temperature sensor, and operational information by an operational information acquisition device. Operational information refers to the various operations performed by the operator in the dual-powered vehicle, such as pantograph raising and lowering, battery charging, and battery discharging. After the dual-powered vehicle starts, various sensors begin data acquisition. Furthermore, after acquiring multiple types of operational data, edge processors deployed at the network edge (close to the data source or terminal device) can aggregate and preprocess this data, and then send the processed data to the energy consumption prediction model deployed in the prediction module for prediction. A specific schematic diagram is shown below. Figure 3 As shown, the prediction module can be deployed in an in-vehicle edge computing unit or a cloud-based real-time scheduling system, supporting real-time modeling and inference.

[0100] In this embodiment, multiple types of operational data include the speed and load of the dual-powered vehicle, the gradient of the dual-powered vehicle's route, operational information for the dual-powered vehicle, and the ambient temperature along the route. This can significantly improve prediction accuracy and enhance the system's robustness and anti-interference capabilities.

[0101] In one embodiment, step S4 is followed by:

[0102] When the current moment is in the prediction time domain, the motor of the dual-source power supply vehicle is powered according to the energy configuration strategy, and the energy utilization rate, response time and charge state stability of the dual-source power supply vehicle are obtained.

[0103] Based on the energy utilization rate, response time, and state of charge stability of dual-source powered vehicles, the constraints and weights of the objective function are optimized to generate a new energy configuration strategy based on the optimized objective function.

[0104] Specifically, "powering the motors of a dual-source power supply vehicle according to the energy allocation strategy while the current time is within the prediction time domain" means that once the current time has rolled into the prediction time domain, the motors of the dual-source power supply vehicle are powered according to the energy allocation strategy at the current time. For example, if the energy allocation strategy from 12:01 to 12:30 is obtained at 12:00, then at 12:01, the motors of the dual-source power supply vehicle are powered according to the energy allocation strategy of 12:01 to drive the dual-source power supply vehicle.

[0105] When powering the motor of a dual-powered vehicle according to the energy configuration strategy, the BMS (Battery Management System), motor controller, and pantograph lifting controller work together to achieve specific energy allocation. Specifically, after receiving the energy configuration strategy, the control module sends control signals to the BMS controller or pantograph lifting controller. When the BMS controller receives the control signal, it controls the battery to provide power to the motor and sends control signals to the motor to drive the dual-powered vehicle. When the pantograph lifting controller receives the control signal, it provides power to the motor through the pantograph and sends control signals to the motor to drive the dual-powered vehicle. The specific control response flowchart is as follows: Figure 4 As shown. The control module is communicatively connected to the BMS, pantograph lifting controller, and motor controller. The control module supports bidirectional flow control, enabling seamless switching between the pantograph and battery.

[0106] Since the predicted energy consumption is updated as the sliding time window moves, the energy allocation strategy is also updated accordingly. This means that although an energy allocation strategy corresponding to the predicted time domain is generated, the strategy is not fully executed. For example, if the energy allocation strategy for 12:01 to 12:30 is obtained at 12:00, then at 12:01, the motor of the dual-powered vehicle is powered according to the 12:01 energy allocation strategy to drive the vehicle. If the energy allocation strategy for 12:02 to 12:31 is obtained at 12:01, then at 12:02, the motor of the dual-powered vehicle is powered according to the newly generated 12:02 energy allocation strategy.

[0107] Since the current moment is already within the prediction time domain, and the motors of the dual-source power supply vehicle have been powered according to the energy configuration strategy, it means that the energy utilization rate, response time, and charge state stability of the dual-source power supply vehicle can be obtained.

[0108] Energy efficiency, also known as system energy efficiency, measures the efficiency of converting "net input energy on the power supply side" into "effective traction output" or energy consumption per unit mileage. State of charge (SOC) stability measures the fluctuation and slope (charge / discharge rate) of SOC around a target band / reference value, reflecting the quality of control. Response time measures the delay and settling time from the issuance of an energy configuration strategy to the execution target and the achievement of the configuration planned by the energy configuration strategy.

[0109] The energy utilization rate, response time, and state of charge stability of dual-source power supply vehicles can be obtained through standard bus signals from the BMS, inverter, motor controller, pantograph controller, VCU, energy meter, and DC bus power meter.

[0110] When optimizing the constraints and weights of the objective function based on the energy utilization rate, response time, and state of charge stability of dual-powered vehicles, the following adjustments are made to the prediction time domain, power slope limit, pantograph dwell constraint, pantograph hysteresis constraint, and weights within the short cycle. Within the medium cycle, the desired state of charge is adjusted, the weights of the overhead contact line price and battery aging cost are adjusted, and the fluctuation range of parameters is updated. Within the long cycle, energy utilization rate, response time, and state of charge stability are used to trigger retraining, calibration, and threshold re-estimation. The duration of the short cycle is shorter than that of the medium cycle, and the duration of the medium cycle is shorter than that of the long cycle.

[0111] In this embodiment, while the current time is in the prediction time domain, the motors of the dual-source power supply vehicle are powered according to the energy configuration strategy. The energy utilization rate, response time, and state-of-charge stability of the dual-source power supply vehicle are obtained. Based on these parameters, the constraints and weights of the objective function are optimized. A new energy configuration strategy is then generated based on the optimized objective function. This approach considers the actual operating conditions of the dual-source power supply vehicle and the response of the executing object, making the optimized constraints and weights closer to real-world operating requirements. This avoids the problem of "theoretically optimal but poor in practice," enhancing the engineering feasibility of the strategy. Furthermore, this application can also reduce control response time.

[0112] In one embodiment, the dual-powered vehicle includes a display and interaction unit for displaying energy configuration strategies and receiving user operation commands.

[0113] In this embodiment, by displaying the generated energy configuration strategy in the display interaction unit, the operator can directly obtain the energy configuration strategy and operate the dual-source power supply vehicle according to the energy configuration strategy.

[0114] In one embodiment, step S1 further includes:

[0115] When the sampling time of each type of operational data is inconsistent among multiple types of operational data, any type of operational data shall be used as the reference data.

[0116] Determine the reference sampling time for the reference data, and identify data collected at times adjacent to the reference sampling time from various types of non-reference data;

[0117] Based on parameters collected at nearby times, calculate the data of various non-reference data at the reference sampling time;

[0118] Denoising and missing data processing are performed on various types of non-benchmark data at the benchmark sampling time and benchmark data to obtain multiple types of preprocessed data;

[0119] Based on multiple types of preprocessed data, the time-series characteristics, frequency domain characteristics, and vehicle-derived characteristics of dual-source power supply vehicles are statistically analyzed and normalized to predict energy consumption.

[0120] The baseline data can be any type of operational data. For example, if vehicle speed is the baseline data, then all other data besides vehicle speed are non-baseline data.

[0121] Data for various non-benchmark data at the benchmark sampling time can be calculated using interpolation.

[0122] Denoising methods include, but are not limited to, exponential smoothing, Kalman filtering, and wavelet denoising.

[0123] Missing data repair primarily addresses the repair of "new missing data" or "data corruption" introduced during the denoising process. Repair methods include, but are not limited to, interpolation repair and random forest repair.

[0124] Time-series characteristics and frequency-domain characteristics refer to the mean, variance, trend, and FFT (Fast Fourier Transform) spectrum of various types of operational data. Vehicle-derived characteristics refer to features obtained based on the physical laws of the vehicle. For example, instantaneous power is determined by multiplying the voltage and current of the onboard battery; the gradient of the dual-powered vehicle's route is determined by the difference between the actual acceleration and the theoretical acceleration on a flat road; and regenerative braking efficiency is determined based on the ratio of energy recovered to energy lost by the dual-powered vehicle.

[0125] Normalization methods include, but are not limited to, min-max normalization, standard deviation normalization, and log normalization.

[0126] After normalization, the normalized time-series features, frequency domain features, and vehicle-derived features are input into the energy consumption prediction model for energy consumption prediction.

[0127] In this embodiment, when the sampling times of each type of operational data are inconsistent across multiple types of operational data, a reference sampling time is determined by using any one type of operational data as the reference data. Data collected near the reference sampling time is then identified from various types of non-reference data. Based on the parameters collected near the reference sampling time, the data for each type of non-reference data at the reference sampling time is calculated. This solves the problem of asynchronous multi-source data and achieves spatiotemporal alignment. Denoising and missing data repair are performed on the data at the reference sampling time for each type of non-reference data and the reference data, improving data quality and enhancing model robustness. Based on multiple types of preprocessed data, the temporal, frequency, and vehicle-derived features of dual-source powered vehicles are statistically analyzed and normalized. Energy consumption prediction is then performed using these normalized features, transforming the original data into higher-level features with greater physical meaning and predictive value. This constructs a multi-dimensional, information-rich feature set, enabling the energy consumption prediction model to more comprehensively understand energy consumption influencing factors and significantly improve prediction accuracy.

[0128] In one embodiment, the process of determining the prediction time domain in step S2 includes:

[0129] Multiple candidate prediction time domains and filtering conditions can be preset;

[0130] Based on the historical predicted energy consumption and historical actual energy consumption of each candidate prediction time domain, the deviation, interval coverage and interval width of each candidate prediction time domain are calculated.

[0131] Candidate prediction time domains that meet the screening criteria are determined as initial prediction time domains. Based on the deviation, interval coverage, and interval width of the initial prediction time domains, the reliability curves of each initial prediction time domain are obtained.

[0132] Based on the reliability curves of each initial prediction time domain, determine the target indicators for reliability scoring among deviation, interval coverage, and interval width;

[0133] Based on the target indicators in the initial prediction time domain, the reliability score for each initial prediction time domain is obtained.

[0134] The initial prediction time domain corresponding to the maximum value in the reliability score is determined as the prediction time domain.

[0135] Each candidate prediction time domain corresponds to a screening condition, which includes a deviation threshold. Interval coverage threshold and interval width threshold These thresholds are derived from quantile thresholds obtained from offline backtesting statistics and can be calibrated online or adaptively adjusted. Since both historical predicted energy consumption and historical actual energy consumption are interval values, the deviation is the root mean square error between the historical predicted energy consumption and the historical actual energy consumption. The selection criteria for each candidate prediction time domain can be consistent or inconsistent.

[0136] Interval coverage refers to the overlap ratio between historical predicted energy consumption and historical actual energy consumption. For example, if the historical actual energy consumption is (A1, A2) and the historical predicted energy consumption is (B1, B2), then the interval coverage is Overlap / (A2-A1), and the Overlap is min(A2,B2)−max(A1,B1). Interval width is the result of subtracting the minimum value from the maximum value of historical actual energy consumption. For example, continuing the above example, the interval width is the result of subtracting A1 from A2.

[0137] The initial prediction time domain that meets the screening criteria refers to the candidate prediction time domain that meets at least one of the following conditions: deviation is less than the deviation threshold, interval coverage is less than the interval coverage threshold, and interval width is less than the interval width threshold.

[0138] The reliability curve is a two-dimensional curve with the candidate prediction time domain as the horizontal axis and the deviation, interval coverage, and interval width as the vertical axis.

[0139] Furthermore, based on the thresholds in the screening criteria, the target indicators are normalized, and based on the normalized target indicators, the reliability scores for each initial prediction time domain are obtained.

[0140] The formula for reliability scoring is: After obtaining R h Afterwards, smoothing processing is required to obtain the final reliability score for each initial prediction time domain. Here, w1(h), w2(h), w3(h), and w4(h) are all weights of the initial prediction time domain h, and S... err (h) is the deviation score calculated based on the normalized deviation, S cov (h) is the interval coverage calculated based on the normalized interval coverage, S w (h) is the interval width calculated based on the normalized interval width. Reliability scoring formula. Characterization bias, interval width, and interval coverage are all target indicators. , , , RMSE h The root mean square error (RMSE) of the initial prediction time domain h is given by rmse. best (h) is the preset optimal reference value, ε rmse (h) is the initial prediction time domain deviation threshold h, CoverGaph The initial prediction interval coverage of time domain h, W is the initial interval coverage threshold for the prediction time domain h. h The initial prediction time domain interval width, As the initial prediction interval width threshold h, DriftScore h This is the drift index value.

[0141] In some embodiments, a comprehensive reliability score can be calculated based on the deviation score, interval width score, interval coverage score, and drift reliability score. The drift reliability score is the score obtained from drift detection.

[0142] In this embodiment, by presetting multiple candidate prediction time domains and screening conditions, and based on the historical predicted energy consumption and historical actual energy consumption of each candidate prediction time domain, the deviation, interval coverage, and interval width of each candidate prediction time domain are calculated. The candidate prediction time domains that meet the screening conditions are determined as the initial prediction time domains. Based on the deviation, interval coverage, and interval width of the initial prediction time domains, the reliability curves of each initial prediction time domain are obtained. According to the reliability curves of each initial prediction time domain, the target indicators for reliability scoring in deviation, interval coverage, and interval width are determined. Based on the target indicators of the initial prediction time domains, the reliability scores of each initial prediction time domain are obtained. The initial prediction time domain corresponding to the maximum value in the reliability scores is determined as the prediction time domain. This achieves adaptive optimization of the prediction time domain, and significantly improves the accuracy, robustness, and overall energy efficiency of the energy management strategy for dual-source powered vehicles while ensuring system safety.

[0143] This application also provides an application scenario in which the above-described energy management method for a dual-source powered vehicle is applied. Specifically, the application of the energy management method for a dual-source powered vehicle in this scenario is as follows:

[0144] Sensors are used to collect various types of operational data from the dual-powered vehicle within a sliding time window. This data is then cleaned and preprocessed, and features are extracted. Based on these features, an energy consumption prediction model is used to predict energy consumption in the prediction time domain. Error analysis and uncertainty assessment are performed on the predicted energy consumption, and the results are fed back to the energy consumption prediction model to optimize it. The availability information of the overhead contact line supplying the dual-powered vehicle, the battery constraints of the dual-powered vehicle, and the pantograph constraints are determined within the prediction time domain. Using the predicted energy consumption, availability information, battery constraints, and pantograph constraints as the objective function, an energy allocation strategy is generated by minimizing the objective function. The prediction and strategy output flowchart is shown below. Figure 5 As shown.

[0145] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0146] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0147] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0148] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0149] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0150] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0151] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An energy management method for a dual-source powered vehicle, characterized in that, The method includes: S1. Acquire various types of operational data for dual-powered vehicles within a sliding time window; S2. Based on the aforementioned multi-type operational data, use the trained energy consumption prediction model to predict the predicted energy consumption of the dual-source power supply vehicle in the prediction time domain. S3. Determine the availability information of the overhead contact line supplying power to the dual-source power supply vehicle in the predicted time domain, the battery constraints of the dual-source power supply vehicle, and the pantograph constraints; the battery constraints include state of charge constraints, charge and discharge power constraints, charge and discharge rate constraints, and battery discharge count constraints; the pantograph constraints include lifting constraints, pantograph dwell constraints, pantograph hysteresis constraints, and lifting count and time limits for the pantograph. S4. Using the predicted energy consumption, the availability information, the battery constraints, and the pantograph constraints as the objective function, an energy allocation strategy is generated by minimizing the objective function; the energy allocation strategy includes an energy distribution strategy for the vehicle battery and pantograph, a pantograph operation strategy, and a vehicle battery charging and discharging strategy; step S4 includes: Establish an objective function; the constraints of the objective function include the predicted energy consumption, the availability information, the battery constraints, and the pantograph constraints. An energy allocation strategy is generated by minimizing the objective function; The expression for the objective function Z is: ； The objective function is used to determine the pantograph power P at time step k. grid [k] and battery power P at time step k batt [k];c grid [k] represents the overhead contact line electricity price at time step k; SOC[k] represents the battery state of charge at time step k; SOC ref c is the desired state of charge; deg,1 and c deg,2 As a weighting of battery aging costs; P batt,max The maximum allowable charge and discharge power of the battery; λ SOC λ is the weight of the state-of-charge tracking error; sm λ is the weight of the rate of change of power; ρ ρ[k] represents the soft constraint penalty weight; ρ[k] represents the soft constraint relaxation at time step k; Δt represents the time step size; N represents the number of time steps. Step S4 is followed by: When the current moment is in the predicted time domain, the motor of the dual-source power supply vehicle is powered according to the energy configuration strategy, and the energy utilization rate, response time and charge state stability of the dual-source power supply vehicle are obtained. Based on the energy utilization rate, response time, and state of charge stability of the dual-source power supply vehicle, the constraints and weights of the objective function are optimized to generate a new energy configuration strategy based on the optimized objective function.

2. The method according to claim 1, characterized in that, The various types of operational data include the speed and load of the dual-powered vehicle, the gradient of the route traveled by the dual-powered vehicle, the operational information of the dual-powered vehicle, and the ambient temperature along the route.

3. The method according to claim 1, characterized in that, The dual-source power supply vehicle includes a display and interaction unit, which is used to display the energy configuration strategy and receive user operation commands.

4. The method according to claim 1, characterized in that, Step S1 also includes: When the sampling times of each type of operational data are inconsistent among the multiple types of operational data, any one type of operational data shall be used as the reference data. Determine the reference sampling time of the reference data, and determine the data collected at a time adjacent to the reference sampling time from various types of non-reference data; Based on the parameters collected at the nearest time, the data of each type of non-reference data at the reference sampling time are calculated; Denoising and missing data repair processes are performed on the non-reference data and the reference data at the reference sampling time to obtain multiple types of preprocessed data; Based on multiple types of preprocessed data, the time-series characteristics, frequency domain characteristics, and vehicle-derived characteristics of the dual-source power supply vehicle are statistically analyzed and normalized to predict energy consumption.

5. The method according to claim 1, characterized in that, The process of determining the prediction time domain in step S2 includes: Multiple candidate prediction time domains and filtering conditions can be preset; Based on the historical predicted energy consumption and historical actual energy consumption of each candidate prediction time domain, the deviation, interval coverage and interval width of each candidate prediction time domain are calculated. The candidate prediction time domains that meet the screening conditions are determined as the initial prediction time domains. Based on the deviation, the interval coverage and the interval width of the initial prediction time domains, the reliability curves of each initial prediction time domain are obtained. Based on the reliability curves of each initial prediction time domain, determine the target indicators for reliability scoring among the deviation, the interval coverage, and the interval width; Based on the target index of the initial prediction time domain, a reliability score is obtained for each of the initial prediction time domains; The initial prediction time domain corresponding to the maximum value in the reliability score is determined as the prediction time domain.

6. An energy management device for a dual-source powered vehicle, used to perform the method according to any one of claims 1-5, characterized in that, The device includes: The data acquisition module is used to acquire various types of operating data of dual-source powered vehicles within a sliding time window; The energy consumption prediction module is used to predict the predicted energy consumption of the dual-source power supply vehicle in the prediction time domain based on the multi-type operating data and the trained energy consumption prediction model. The condition determination module is used to determine the availability information of the overhead contact line supplying power to the dual-source power supply vehicle in the prediction time domain, the battery constraints of the dual-source power supply vehicle, and the pantograph constraints. The strategy acquisition module is used to generate an energy configuration strategy by minimizing the objective function, which is defined by the predicted energy consumption, the availability information, the battery constraints, and the pantograph constraints. The energy configuration strategy includes an energy allocation strategy for the vehicle battery and the pantograph, a pantograph operation strategy, and a vehicle battery charging and discharging strategy.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 5.

Images