Electric vehicle energy management method and system based on vehicle speed prediction

Abstract

The invention discloses an electric vehicle energy management method and system based on vehicle speed prediction, and relates to the technical field of electric vehicle energy management, and the method comprises the steps: collecting vehicle driving data to construct a state vector, constructing a quantile LSTM vehicle speed distribution predictor, and predicting the quantile vehicle speed track of each step in the future; mapping the quantile vehicle speed track into traction power distribution through longitudinal dynamics, and carrying out battery power, current and SOC quantile rolling prediction; predicting and defining predicted time domain energy consumption based on the battery power quantile so as to construct an MPC optimization objective function of the CVaR risk constraint; constructing a deterministic constraint set based on a quantile prediction trajectory, defining a terminal risk cost according to a predicted terminal state, and merging the terminal risk cost into an MPC optimization target to obtain a final target function; the robustness and adaptability of the electric vehicle energy management system are obviously improved, especially in complex and uncertain road and traffic environments.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle energy management technology, and in particular to an electric vehicle energy management method and system based on vehicle speed prediction. Background Technology

[0002] With the rapid development of the electric vehicle industry, energy management strategies, as core technologies for improving energy efficiency, ensuring battery safety, and extending driving range, have always been a research focus for both academia and industry. Traditional energy management methods typically rely on rules or optimization algorithms, such as dynamic programming, Pontryagin minimum principle, and model predictive control (MPC). In recent years, with the improvement of vehicle sensing and computing capabilities, data-driven predictive energy management has become the mainstream direction. The core of this type of method lies in using historical and real-time driving data (such as vehicle speed, acceleration, and gradient information) to predict future driving needs, thereby proactively optimizing power allocation.

[0003] However, existing technologies still have several key limitations. First, most predictive methods only provide a single-point estimate of future vehicle speed or other states, ignoring the inherent and irreducible uncertainties in real driving environments (such as sudden changes in traffic flow and randomness in driver behavior). This leads to subsequent energy optimization based on the assumption of a single vehicle speed prediction. When actual operating conditions deviate from the prediction, it may cause battery power, current, or state of charge (SOC) to exceed the safety boundary, thereby affecting the optimization effect of vehicle energy management strategies. Second, existing methods lack the ability to describe and control risks (especially extreme tail risks) in a refined and quantifiable way. They cannot describe the probability of battery overload current exceeding the safety threshold and the average severity of such safety issues once it occurs. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an electric vehicle energy management method and system based on vehicle speed prediction, which solves the limitations of existing electric vehicle energy management methods that are based on single vehicle speed prediction, cannot quantify and control extreme tail risks, and have insufficient energy management accuracy.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an electric vehicle energy management method based on vehicle speed prediction, comprising, Collect vehicle driving data to construct a state vector, build a quantile LSTM vehicle speed distribution predictor, and predict the quantile vehicle speed trajectory for each future step. The quantile vehicle speed trajectory is mapped to the traction power distribution through longitudinal dynamics, and the battery power, current and SOC quantile rolling prediction is performed. Based on the prediction of battery power quantiles, the predicted time-domain energy consumption is defined to construct the MPC optimization objective function with CVaR risk constraints; A set of deterministic constraints is constructed based on the quantile-predicted trajectory, and the terminal risk cost is defined according to the predicted terminal state. The terminal risk cost is then incorporated into the MPC optimization objective to obtain the final objective function. Based on the final objective function, the set of deterministic constraints, and the current state vector, optimization is performed to obtain the optimal control vector, control is executed, and data is recorded.

[0007] As a preferred embodiment of the electric vehicle energy management method based on vehicle speed prediction described in this invention, the step of collecting vehicle driving data to construct a state vector, building a quantile LSTM vehicle speed distribution predictor, and predicting the quantile vehicle speed trajectory for each future step includes: Collect vehicle driving data, including vehicle speed. acceleration Battery SOC Road slope Speed ​​of the vehicle in front acceleration and distance ; When there is no vehicle in front, and Let it be 0. It is positive infinity; The collected vehicle driving data is concatenated into a state vector. ; Output the current state vector and a sequence of historical states of length L. ; A quantile LSTM vehicle speed distribution predictor is constructed using quantile LSTM, with the prediction step size N and sampling period set. Define the set of quantiles ,in , and Representing the low quantile, middle quantile, and high quantile respectively, the historical state sequence is... As input, the quantile LSTM vehicle speed distribution predictor is used to predict the quantile vehicle speed trajectory for each future step; Construct the quantile loss function; The predictor is trained using historical data, and gradient descent optimization is performed using the quantile loss function and Adam optimizer to update the predictor parameters. If the loss of the predictor no longer decreases significantly during continuous iteration, the iteration is stopped and the trained quantile LSTM vehicle speed distribution predictor is output. The newly collected historical state sequence is input into the trained quantile LSTM vehicle speed distribution predictor, which outputs a quantile vehicle speed trajectory of length N.

[0008] As a preferred embodiment of the electric vehicle energy management method based on vehicle speed prediction described in this invention, the step of mapping the quantile vehicle speed trajectory to a traction power distribution through longitudinal dynamics and performing rolling predictions of battery power, current, and SOC quantiles includes: For each quantile The acceleration quantile sequence is calculated using the difference method to obtain the quantile acceleration trajectory; Traction demand is calculated using longitudinal dynamics based on quantile vehicle speed trajectory and quantile acceleration trajectory. Calculate the traction power distribution based on traction force requirements; Based on the vehicle's current SOC and battery model parameters, and considering drive and recovery efficiency, the traction power is converted into battery power. For each quantile trajectory of the battery power distribution, the battery current is derived using the zero-order equivalent circuit, and the vehicle SOC is recursively updated based on the battery current. The final output will show the future battery power, current, and SOC distribution trajectory.

[0009] As a preferred embodiment of the electric vehicle energy management method based on vehicle speed prediction described in this invention, the step of defining the predicted time-domain energy consumption based on battery power quantile prediction to construct an MPC optimization objective function with CVaR risk constraints includes, Calculate the predicted time-domain energy consumption for each quantile trajectory of the battery power distribution. ; Construct CVaR risk terms based on predicted time-domain energy consumption; Based on the predicted time-domain energy consumption and CVaR risk term, an MPC optimization objective function is constructed, which includes the expected term, risk term, and smoothness term.

[0010] As a preferred embodiment of the electric vehicle energy management method based on vehicle speed prediction described in this invention, wherein: the construction of a deterministic constraint set based on quantile-predicted trajectory includes, Based on the quantile distribution trajectories of battery power and current, power and current risk constraints are constructed. Construct SOC opportunity constraints based on SOC quantile distribution trajectories; Based on the acceleration quantile distribution trajectory, feasible and comfort constraints are constructed; By merging power and current risk constraints, SOC opportunity constraints, and realizability and comfort constraints, a set of deterministic constraints is ultimately output. .

[0011] As a preferred embodiment of the electric vehicle energy management method based on vehicle speed prediction described in this invention, the step of defining a terminal risk cost based on the predicted terminal state, incorporating the terminal risk cost into the MPC optimization objective, and obtaining the final objective function includes: Based on all predicted trajectories, the result of the Nth step of the predicted sequence is defined as the terminal distribution state. ; Based on historical terminal distribution and from state The actual tail energy consumption occurring within a fixed future time window Establish a historical experience database ; Terminal distribution status Calculate the Euclidean distance between the sample and the historical experience database, and select the K nearest samples as the neighbor set. ; Based on neighbor set The terminal risk cost is defined as the CVaR of the neighbor's tail energy consumption; By incorporating the terminal risk cost into the MPC optimization objective function, we obtain the final objective function. .

[0012] As a preferred embodiment of the electric vehicle energy management method based on vehicle speed prediction described in this invention, the step of optimizing the solution based on the final objective function, the set of deterministic constraints, and the current state vector to obtain the optimal control vector, executing control, and recording data includes: Taking the final objective function, the set of deterministic constraints, and the current state vector as input, MPC optimization is performed under the premise of satisfying the set of deterministic constraints to obtain the optimal control sequence. ; According to the rolling time domain principle, only the first control variable is executed. The process proceeds to the actuator and enters the next sampling cycle, resampling and rolling optimization, recording each step of the control operation already performed. And the corresponding actual response data, to update the historical experience database.

[0013] Secondly, the present invention provides an electric vehicle energy management system based on vehicle speed prediction, comprising, The vehicle speed prediction module collects real-time driving data of the vehicle, including vehicle speed, acceleration, battery SOC, road gradient, following distance and the status of the vehicle in front, to construct a state vector. Based on the historical state sequence, it uses an LSTM network to predict the vehicle speed distribution (quantile trajectory) for each future step. The quantile mapping module converts the predicted vehicle speed distribution into a traction power distribution through a longitudinal dynamics model, calculates the battery power distribution based on the traction power distribution, and predicts changes in battery current and SOC based on the battery power. An optimization objective building module is constructed based on battery power quantile prediction, defining energy consumption in the prediction time domain, and constructing an optimization objective function with CVaR risk constraints for risk control in energy management; The constraint set construction module constructs a set of deterministic constraints related to energy management based on quantile predicted trajectories, including constraints such as SOC, power, and current, to ensure that the control strategy meets the physical and safety requirements of the system. The terminal risk definition module defines the terminal risk cost based on the predicted terminal status and incorporates it into the MPC optimization objective to reduce tail risk. The optimization solution module performs MPC solution based on the final optimization objective function, the set of deterministic constraints, and the current state to obtain the optimal control vector, execute control commands, and record real-time response data.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the electric vehicle energy management method based on vehicle speed prediction as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the electric vehicle energy management method based on vehicle speed prediction as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: A quantile LSTM speed predictor provides accurate modeling of vehicle speed uncertainty, further predicting quantile changes in battery power, current, and SOC. This enables the system to assess energy consumption under different vehicle speed distributions in real time. Combined with the MPC optimization objective function based on CVaR risk constraints, comprehensive control over parameters such as battery power and SOC is achieved, ensuring efficient energy consumption and stable operation of the electric vehicle energy management system in complex environments. By constructing a deterministic constraint set, strict safety and physical limitations are set for the control strategy. By incorporating terminal risk costs, the system can proactively control potential future risks. By solving the optimization problem based on vehicle speed prediction and risk control, accurate energy management decisions are provided for electric vehicles. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the electric vehicle energy management method based on vehicle speed prediction in Example 1.

[0019] Figure 2 This is a structural diagram of the electric vehicle energy management system based on vehicle speed prediction in Example 1.

[0020] Figure 3 This is a flowchart of the distribution propagation and the construction of the deterministic constraint set in Example 1. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1, referring to Figure 1 , Figure 2 and Figure 3 This is the first embodiment of the present invention, which provides an electric vehicle energy management method based on vehicle speed prediction, including the following steps: S1. Collect vehicle driving data to construct a state vector, build a quantile LSTM vehicle speed distribution predictor, and predict the quantile vehicle speed trajectory for each future step. Specifically, vehicle driving data is collected, including the vehicle's speed. acceleration Battery SOC Road slope Speed ​​of the vehicle in front acceleration and distance ; When there is no vehicle in front, and Let it be 0. It is positive infinity; The collected vehicle driving data is concatenated into a state vector. , is represented as: ; Output the current state vector and a sequence of historical states of length L. ; A quantile LSTM vehicle speed distribution predictor is constructed using quantile LSTM, with the prediction step size N and sampling period set. Define the set of quantiles ,in , and Representing the low quantile, middle quantile, and high quantile respectively, the historical state sequence is... As input, the quantile LSTM vehicle speed distribution predictor is used to predict the quantile vehicle speed trajectory for each future step, represented as: ; in, This indicates that the q-quantile predicts the vehicle speed at the i-th step. This represents a quantile LSTM network; Construct the quantile loss function, expressed as: ; ; in, This represents the quantile loss function. This represents the quantile regression loss. Indicates the actual speed label. This indicates the predicted vehicle speed at the q-th quantile. Indicates error; The predictor is trained using historical data, and gradient descent optimization is performed using the quantile loss function and Adam optimizer to update the predictor parameters. If the loss of the predictor no longer decreases significantly during continuous iteration, the iteration is stopped and the trained quantile LSTM vehicle speed distribution predictor is output. The newly collected historical state sequence is input into the trained quantile LSTM vehicle speed distribution predictor, which outputs a quantile vehicle speed trajectory of length N.

[0025] By collecting vehicle driving data and constructing a state vector, the system can build a dynamic and complete state vector based on the real-time state of the vehicle, the driving conditions of the vehicle in front, and road gradient. This state vector provides comprehensive and accurate input for subsequent vehicle speed distribution prediction. A quantile LSTM vehicle speed distribution predictor is constructed. The LSTM model can not only predict future vehicle speed point values, but also provide low, medium, and high quantile predictions of vehicle speed distribution. Optimization through the quantile loss function ensures that the model can predict vehicle speed distribution as accurately as possible during training, thereby improving the accuracy of future vehicle speed prediction. By outputting vehicle speed trajectories of low, medium, and high quantiles, the system can provide the vehicle energy management system with full-range uncertainty information on vehicle speed. This allows the prediction results to reflect the possible range of future vehicle speeds, avoiding the uncertainty problems caused by predicting only point values ​​in traditional methods. This makes subsequent energy management more reliable and safer, especially in uncertain environments, effectively avoiding problems such as extreme energy consumption or over-braking, thereby improving energy utilization efficiency and enhancing the robustness of the system.

[0026] S2. Map the quantile vehicle speed trajectory to the traction power distribution through longitudinal dynamics, and perform rolling prediction of battery power, current and SOC quantiles; Specifically, for each quantile The acceleration quantile sequence is calculated using the difference method, and the quantile acceleration trajectory is obtained, which is represented as: ; in, This represents the q-quantile predicted acceleration for the i-th future step; The traction demand is calculated using longitudinal dynamics based on the quantile vehicle speed trajectory and the quantile acceleration trajectory, and is expressed as follows: ; in, This indicates that the q quantile for the i-th step in the future predicts traction demand. Indicates vehicle mass. This represents the aerodynamic drag coefficient. Represents gravitational acceleration. Indicates the rolling resistance coefficient. This represents the road gradient at the i-th future step, obtained by taking N steps forward from the current position on the map to obtain the gradient prediction sequence. The traction power distribution is calculated based on the traction demand and is expressed as follows: ; in, This represents the q-quantile prediction of traction power at the i-th step. Based on the vehicle's current SOC and battery model parameters, and considering drive and recovery efficiency, the traction power is converted into battery power, expressed as follows: ; in, This represents the predicted battery power based on the q-quantile at the i-th step. Indicates drive power. Indicates the recovery power; For each quantile trajectory of the battery power distribution, the battery current is derived using the zero-order equivalent circuit, and the vehicle's SOC is recursively updated based on the battery current, as shown below: ; ; ; in, This represents the q-quantile for predicting the battery current at the i-th step. This represents the open-circuit voltage corresponding to SOC. Indicates internal resistance. Indicates the future number The q-quantile of the step predicts the battery SOC. Indicates the nominal capacity; The final output will show the future battery power, current, and SOC distribution trajectory.

[0027] By mapping vehicle speed distribution to traction power distribution through longitudinal dynamics, and utilizing information such as vehicle acceleration, speed, and road gradient, the system can calculate the traction force demand at different quantile speeds, and further calculate the traction power from the traction force. The beneficial effect of this step is that by transforming the quantile prediction of vehicle speed into the traction power distribution, energy management not only considers the expected value of vehicle speed but also handles the uncertainty in power prediction. Furthermore, the quantile rolling prediction of battery power, current, and SOC ensures accurate prediction of battery energy consumption, enabling the system to evaluate energy consumption under different vehicle speed distributions in real time. It combines power demand, energy consumption, and battery state to comprehensively predict battery state and updates the battery state through dynamic recursion of the battery internal resistance model and SOC. Through quantile rolling prediction, the system can dynamically adjust the energy management strategy under conditions of high uncertainty, avoiding energy efficiency losses caused by ignoring uncertainty in traditional methods, and achieving more efficient energy utilization and longer driving range.

[0028] S3. Define the predicted time-domain energy consumption based on battery power quantile prediction to construct the MPC optimization objective function with CVaR risk constraints; Specifically, the predicted time-domain energy consumption is calculated for each quantile trajectory of the battery power distribution. , is represented as: ; in, quantiles Predicted time-domain energy consumption; Based on the predicted time-domain energy consumption, a CVaR risk term is constructed, expressed as: ; in, Indicates the confidence level. Indicates the CVaR risk term. This represents the VaR approximate threshold variable. Indicates the number of quantiles; Based on the predicted time-domain energy consumption and CVaR risk term, an MPC optimization objective function is constructed, which includes an expected term, a risk term, and a smoothness term, and is expressed as follows: ; in, This represents the objective function for MPC optimization. Indicates expected energy consumption. Indicates tail risk. Indicates the smoothness control term. and Indicates weight, This indicates the control increment.

[0029] This paper calculates the CVaR (Conditional Value at Risk) of the predicted time-domain energy consumption based on battery power quantile prediction and constructs an MPC (Multi-Purpose Controlled Calculation) objective function with CVaR risk constraints. This results in a risk-benefit balanced optimization objective. By analyzing energy consumption within the predicted time domain, this step proposes using CVaR for risk assessment and control. This method not only optimizes the expected energy consumption but also effectively controls tail risk. Specifically, when energy consumption exceeds a certain threshold in the worst-case scenario, the system can take corresponding measures to adjust. By combining risk, expected, and smoothness terms into the MPC objective function, the system achieves a comprehensive optimization objective that considers both maximizing energy utilization and ensuring system stability and safety in complex environments. This technology avoids the problem of excessive energy consumption caused by traditional mean-based optimization strategies in extreme cases, effectively ensuring that the battery SOC fluctuates within the predicted range, avoiding unnecessary overcharging and discharging and power spikes, improving battery lifespan, and reducing energy loss.

[0030] S4. Construct a set of deterministic constraints based on the quantile predicted trajectory, define the terminal risk cost according to the predicted terminal state, incorporate the terminal risk cost into the MPC optimization objective, and obtain the final objective function. Specifically, based on the quantile distribution trajectories of battery power and current, power and current risk constraints are constructed, expressed as follows: ; ; in, and These represent the high-quantile predicted battery power and the high-quantile predicted battery current, respectively. and These represent the upper limit of power and the upper limit of current, respectively. Based on the SOC quantile distribution trajectory, an SOC opportunity constraint is constructed, expressed as: ; ; in, and These represent the predicted battery SOC at the low quantile and high quantile, respectively. and These represent the lower and upper limits of battery SOC safety, respectively. Based on the acceleration quantile distribution trajectory, realizability and comfort constraints are constructed, as follows: ; ; in, and These represent the lower and upper limits of acceleration, respectively. This represents the median predicted acceleration. This indicates the upper limit of jerk, where jerk refers to the rate of change of acceleration; By merging power and current risk constraints, SOC opportunity constraints, and realizability and comfort constraints, a set of deterministic constraints is ultimately output. ; Furthermore, based on all predicted trajectories, the result of the Nth step of the predicted sequence is defined as the terminal distribution state. , is represented as: in, This indicates the terminal state of the predicted vehicle speed at the median quantile. Indicates the final state of the road slope. This indicates the median predicted battery SOC terminal state. Indicates the distance to the vehicle in front at the end of the journey; Based on historical terminal distribution and from state The actual tail energy consumption occurring within a fixed future time window Establish a historical experience database ; Terminal distribution status Calculate the Euclidean distance between the sample and the historical experience database, and select the K nearest samples as the neighbor set. ; Based on neighbor set The terminal risk cost is defined as the CVaR of the neighbor's tail energy consumption, expressed as: ; in, Indicates the cost of terminal risks. The VaR approximate threshold variable representing the terminal risk cost, Indicates the confidence level; By incorporating the terminal risk cost into the MPC optimization objective function, we obtain the final objective function. , is represented as: ; in, Describes the final objective function. The weights representing the terminal risk costs are determined through cross-validation.

[0031] By constructing a set of deterministic constraints, the system can set strict safety and physical limitations for the control strategy, such as power, current, SOC, and acceleration constraints. These constraints ensure the safety and stability of the vehicle in actual operation. By combining these constraints with the terminal risk cost, the system further enhances the prediction and control of future risks. By introducing the risk cost of the terminal state (tail energy consumption based on historical data and neighbor samples) into the optimization objective, the system increases the consideration of terminal risks, enabling it to dynamically adjust the control strategy based on the experience of historical tail energy consumption and control potential future risks in advance, rather than relying solely on the optimization of the current state. This design makes the optimization process more in line with reality, avoiding the neglect of long-term safety due to over-reliance on short-term predictions, thereby enhancing the long-term operational stability and safety of the system, especially in complex and uncertain road and traffic environments.

[0032] S5. Based on the final objective function, the set of deterministic constraints, and the current state vector, perform optimization to obtain the optimal control vector, execute control, and record the data; Specifically, the final objective function, the set of deterministic constraints, and the current state vector are taken as inputs, and MPC optimization is performed under the premise of satisfying the set of deterministic constraints to obtain the optimal control sequence. , is represented as: ; in, Represents the optimal control sequence; According to the rolling time domain principle, only the first control variable is executed. The process proceeds to the actuator and enters the next sampling cycle, resampling and rolling optimization, recording each step of the control operation already performed. And the corresponding actual response data, to update the historical experience database.

[0033] The optimal control vector is obtained by optimizing the final objective function and the set of deterministic constraints. By solving the optimization problem based on vehicle speed prediction and risk control, the optimal control strategy is obtained and applied to the vehicle control system. This process involves not only optimized energy scheduling but also real-time feedback and control strategy updates. By executing control and recording actual response data, the system can continuously adjust and optimize the control strategy, thereby improving control accuracy in each cycle. By dynamically updating the experience base, the system can continuously learn and adapt to new driving scenarios and traffic environments during execution, further improving prediction accuracy and control efficiency. In this process, the system achieves real-time adaptation to unknown future conditions and can continuously optimize energy management strategies during execution, improving the electric vehicle's range and operating efficiency, and continuously enhancing the accuracy and robustness of subsequent prediction and control decisions.

[0034] This embodiment also provides an electric vehicle energy management system based on vehicle speed prediction, including: The vehicle speed prediction module collects real-time driving data of the vehicle, including vehicle speed, acceleration, battery SOC, road gradient, following distance and the status of the vehicle in front, to construct a state vector. Based on the historical state sequence, it uses an LSTM network to predict the vehicle speed distribution (quantile trajectory) for each future step. The quantile mapping module converts the predicted vehicle speed distribution into a traction power distribution through a longitudinal dynamics model, calculates the battery power distribution based on the traction power distribution, and predicts changes in battery current and SOC based on the battery power. An optimization objective building module is constructed based on battery power quantile prediction, defining energy consumption in the prediction time domain, and constructing an optimization objective function with CVaR risk constraints for risk control in energy management; The constraint set construction module constructs a set of deterministic constraints related to energy management based on quantile predicted trajectories, including constraints such as SOC, power, and current, to ensure that the control strategy meets the physical and safety requirements of the system. The terminal risk definition module defines the terminal risk cost based on the predicted terminal status and incorporates it into the MPC optimization objective to reduce tail risk. The optimization solution module performs MPC solution based on the final optimization objective function, the set of deterministic constraints, and the current state to obtain the optimal control vector, execute control commands, and record real-time response data.

[0035] This embodiment also provides a computer device applicable to the electric vehicle energy management method based on vehicle speed prediction, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the electric vehicle energy management method based on vehicle speed prediction as proposed in the above embodiment. The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0036] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the electric vehicle energy management method based on vehicle speed prediction as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0037] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for electric vehicle energy management based on speed prediction, characterized in that: The method comprises the following steps: Collecting vehicle driving data to construct a state vector, constructing a quantile LSTM vehicle speed distribution predictor, and predicting a quantile vehicle speed trajectory for each future step; Mapping the quantile vehicle speed trajectory to a traction power distribution through longitudinal dynamics, and performing rolling prediction of battery power, current, and SOC quantiles; Defining a prediction horizon energy consumption based on the battery power quantile prediction to construct an MPC optimization objective function with CVaR risk constraints; Based on the quantile prediction trajectory, a set of deterministic constraints is constructed, and a terminal risk cost is defined according to the predicted terminal state, which is incorporated into the MPC optimization objective to obtain a final objective function; According to the final objective function, the set of deterministic constraints and the current state vector, optimization is performed to obtain the optimal control vector, and the control is executed and the data is recorded.

2. The electric vehicle energy management method based on vehicle speed prediction as claimed in claim 1, wherein: The method comprises the following steps: Collecting vehicle driving data, including host vehicle speed , acceleration , battery SOC , road slope , front vehicle speed , acceleration , and vehicle distance ; When there is no preceding vehicle, let and be 0, let be positive infinity; Concatenating collected vehicle driving data into a state vector ; Output the current state vector, as well as the history state sequence of length L ; A quantile LSTM is used to construct a quantile LSTM vehicle speed distribution predictor, set the prediction step N and the sampling period , set a quantile set , , and respectively represent the low quantile, the middle quantile and the high quantile, and use the quantile LSTM vehicle speed distribution predictor to predict the quantile speed trajectory of each step in the future with the historical state sequence as input; Constructing a quantile loss function; Using historical data to train the predictor, using the quantile loss function and the Adam optimizer to perform gradient descent optimization, updating the parameters of the predictor, and stopping iteration when the loss of the predictor no longer decreases significantly in the continuous iteration process to output the trained quantile LSTM vehicle speed distribution predictor; Input the newly collected historical state sequence into the trained quantile LSTM vehicle speed distribution predictor to output a quantile vehicle speed trajectory with a length of N.

3. The electric vehicle energy management method based on vehicle speed prediction as claimed in claim 2, wherein: The method comprises the following steps: Including, For each quantile The acceleration quantile sequence is calculated by difference calculation to obtain the quantile acceleration trajectory. Based on the quantile vehicle speed trajectory and the quantile acceleration trajectory, the longitudinal dynamics is used to calculate the traction force demand; According to the traction force demand, the traction power distribution is calculated; According to the current SOC of the vehicle and the battery model parameters, considering the driving and recovery efficiency, the traction power is converted into battery power; For each quantile trajectory of the battery power distribution, the battery current is derived using the zero-order equivalent circuit, and the vehicle SOC is recursively updated according to the battery current; Finally, the future battery power, current, and SOC distribution trajectories are output.

4. The electric vehicle energy management method based on vehicle speed prediction as claimed in claim 3, wherein: The method comprises the following steps: Calculating predicted time-energy expenditure for each quantile trajectory of battery power distribution ; According to the predicted horizon energy consumption, a CVaR risk term is constructed; According to the predicted horizon energy consumption and the CVaR risk term, an MPC optimization objective function is constructed, which includes an expectation term, a risk term, and a smoothness term.

5. The electric vehicle energy management method based on vehicle speed prediction as claimed in claim 4, wherein: The method comprises the following steps: Based on the battery power and current quantile distribution trajectories, power and current risk constraints are constructed; Based on the SOC quantile distribution trajectory, an SOC chance constraint is constructed; Based on the acceleration quantile distribution trajectory, realizability and comfort constraints are constructed; combining the power and current risk constraints, the SOC chance constraints, and the achievability and comfort constraints, to output a final set of deterministic constraints .

6. The electric vehicle energy management method based on vehicle speed prediction as claimed in claim 5, wherein: The method comprises the following steps: According to all predicted trajectories, define the result of the Nth step of the prediction sequence as the terminal distribution state ; According to historical terminal distribution state and actual tail energy consumption occurred in future fixed time window from state , establish historical experience library ;​ Terminal distribution state Calculate the Euclidean distance between the sample and the historical experience library, and select the K nearest samples as the neighbor set ; According to the neighbor set , define the terminal risk cost as the CVaR of the neighbor tail energy consumption; incorporating the terminal risk cost into the MPC optimization objective function to obtain a final objective function .

7. The electric vehicle energy management method based on vehicle speed prediction as claimed in claim 6, wherein: The method comprises the following steps: The final target function, the set of deterministic constraints and the current state vector are taken as inputs, and the optimal control sequence is obtained by performing MPC optimal solving under the premise of meeting the set of deterministic constraints ; According to the receding horizon principle, only the first control quantity is executed to the actuator, and the next sampling period is entered, the receding optimization is resampled, the executed control at each step and the corresponding actual response data are recorded to update the historical experience library. and the corresponding actual response data are recorded to update the historical experience library.

8. A speed prediction based electric vehicle energy management system based on the speed prediction based electric vehicle energy management method of any one of claims 1-7. Including, A vehicle speed prediction module collects real-time driving data of the vehicle, including vehicle speed, acceleration, battery SOC, road slope, following distance and front vehicle state, to construct a state vector, and predicts the vehicle speed distribution (quantile trajectory) of each step in the future based on the historical state sequence through an LSTM network; A quantile mapping module converts the predicted vehicle speed distribution into traction power distribution through a longitudinal dynamics model, calculates the battery power distribution according to the traction power distribution, and predicts the changes of battery current and SOC based on the battery power; An optimization objective construction module defines the energy consumption in the prediction time domain based on the battery power quantile prediction, and constructs an optimization objective function with CVaR risk constraints for risk control in energy management; A constraint set construction module constructs a set of deterministic constraints related to energy management based on the quantile prediction trajectory, including SOC, power, current and other constraints, to ensure that the control strategy meets the physical and safety requirements of the system; A terminal risk definition module defines the terminal risk cost according to the predicted terminal state, and integrates it into the MPC optimization objective to reduce the tail risk; An optimization solving module performs MPC solving according to the final optimization objective function, the set of deterministic constraints and the current state, obtains the optimal control vector, executes the control instruction, and records the real-time response data. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the electric vehicle energy management method based on vehicle speed prediction according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the electric vehicle energy management method based on vehicle speed prediction according to any one of claims 1-7.

Images