New energy vehicle super capacitor and power battery hybrid energy storage control method and system

Abstract

The application provides a new energy vehicle super capacitor and power battery hybrid energy storage control method and system, relates to the technical field of new energy vehicles, and comprises the following steps: acquiring current working condition parameters and energy storage state variables, and constructing a multi-objective optimization model; based on an energy distribution sequence, performing robustness constraint according to a sliding surface function, introducing an adaptive reaching law to adjust a switching gain; rolling out in a prediction time domain to estimate a future response trajectory; constructing a segmented continuous variable characteristic curve, and adjusting the switching characteristics of the sliding surface function according to different working conditions. The application realizes efficient collaborative control of the hybrid energy storage system, improves the energy utilization rate, and prolongs the service life of the battery.

Description

Technical Field

[0001] This invention relates to new energy vehicle technology, and more particularly to a method and system for controlling hybrid energy storage of supercapacitors and power batteries in new energy vehicles. Background Technology

[0002] With the increasing severity of the energy crisis and environmental problems, new energy vehicles, as a clean and efficient means of transportation, have become an important direction for the development of the automotive industry. Among the many key technologies of new energy vehicles, energy storage systems are one of the core components, directly affecting vehicle performance and range. Currently, power batteries are widely used in new energy vehicles due to their high energy density, but their relatively low power density makes it difficult to meet the performance requirements of vehicles under high-power demand conditions such as acceleration and hill climbing. Supercapacitors, with their advantages of high power density, high charge and discharge efficiency, and long cycle life, are gradually being introduced into the energy storage systems of new energy vehicles, forming a hybrid energy storage system of supercapacitors and power batteries. This system comprehensively utilizes the advantages of both to improve overall vehicle performance and energy efficiency.

[0003] Traditional energy management strategies are mostly based on simple rule-based or fuzzy control, which cannot effectively handle complex and ever-changing vehicle operating conditions and struggle to achieve optimal energy allocation between supercapacitors and power batteries, resulting in low overall system efficiency. Existing control methods lack robustness to changes in system parameters and external disturbances, easily leading to reduced control accuracy and system chattering in practical applications, affecting the stability and lifespan of energy storage systems. Most energy management strategies lack the ability to predict future operating conditions, making control decisions solely based on the current state, thus failing to achieve globally optimal energy allocation. This is particularly problematic in urban road conditions with frequently changing operating conditions, making it difficult to fully leverage the advantages of hybrid energy storage systems. Summary of the Invention

[0004] This invention provides a method and system for controlling the hybrid energy storage of supercapacitors and power batteries in new energy vehicles, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a method for controlling hybrid energy storage of supercapacitors and power batteries in new energy vehicles, comprising:

[0006] The system acquires the current operating parameters of the vehicle, the state variables of the supercapacitor and the power battery. Based on the vehicle's power demand and energy flow constraints, it constructs a multi-objective optimization model for the hybrid energy storage system. The system then uses state-space equations to collaboratively model the power distribution between the supercapacitor and the power battery, thereby obtaining the energy distribution sequence.

[0007] Based on the energy allocation sequence, robustness constraints are applied according to the sliding mode surface function, and an adaptive reaching law is introduced to dynamically adjust the switching gain, thereby generating a power allocation command to suppress chattering.

[0008] Based on the power allocation command, the power allocation is rolled out in the prediction time domain. The future response trajectories of the supercapacitor and the power battery are predicted by the energy allocation sequence in the prediction time domain, and a power allocation strategy is obtained.

[0009] The sliding surface function is constructed as a piecewise continuous variable characteristic curve, and the power allocation strategy is used as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding surface function are adjusted according to the state variable deviation in different operating conditions to form the final energy allocation optimization sequence.

[0010] Based on vehicle power demand and energy flow constraints, a multi-objective optimization model for a hybrid energy storage system is constructed. A state-space equation is used to collaboratively model the power distribution between the supercapacitor and the power battery, resulting in an energy distribution sequence including:

[0011] Acquire power demand data of the vehicle under real-time driving conditions, establish power demand characteristic curves including acceleration power, braking power and cruising power, and calculate the power distribution coefficient for each operating condition range based on the power demand characteristic curves.

[0012] Combining the power demand characteristic curve and the power allocation coefficient, the charging and discharging power limits, state of charge variation ranges, and temperature constraints of the supercapacitor and power battery in each operating condition range are extracted to construct an energy flow constraint set.

[0013] Based on the power allocation coefficient and the energy flow constraint set, a multi-objective optimization model is constructed with the goal of maximizing system energy utilization efficiency and minimizing power loss. The dynamic weighting coefficient of the optimization objective in each operating condition interval is calculated by an adaptive weight adjustment algorithm.

[0014] Substituting the dynamic weighting coefficients and the energy flow constraint set into the state-space equation of the hybrid energy storage system, the charging and discharging power of the supercapacitor and the power battery are modeled collaboratively using the rolling time-domain prediction method. By solving the state-space equation, the energy allocation sequence that satisfies the multi-objective optimization model is obtained.

[0015] Substituting the dynamic weighting coefficients and the energy flow constraint set into the state-space equation of the hybrid energy storage system, and using a rolling time-domain prediction method to collaboratively model the charging and discharging power of the supercapacitor and the power battery, the energy allocation sequence satisfying the multi-objective optimization model is obtained by solving the state-space equation, including:

[0016] Substituting the dynamic weighting coefficients and energy flow constraints into the state space equations of the hybrid energy storage system, a set of state equations is constructed with supercapacitor voltage, power battery current and state of charge as state variables and charging and discharging power as control variables.

[0017] A rolling time-domain prediction framework is constructed based on the state equation set. A power distribution prediction sequence for supercapacitors and power batteries is established in each prediction interval. According to the evolution law of the state equation set, the end state of each prediction interval is set as the initial state of the next prediction interval.

[0018] The state response trajectories of the supercapacitor and the power battery are calculated using the state equation set and the power allocation prediction sequence. The state response trajectories are combined with the dynamic weighting coefficients to correct the power allocation prediction sequence, thereby generating a corrected prediction sequence that satisfies the energy flow constraint set.

[0019] Substituting the corrected prediction sequence into the state equations and solving them yields an energy allocation sequence that satisfies the multi-objective optimization requirements.

[0020] Based on the energy allocation sequence, robust constraints are applied according to the sliding mode surface function, and an adaptive reaching law is introduced to dynamically adjust the switching gain, generating a power allocation command to suppress chattering, including:

[0021] Based on the energy allocation sequence, a sliding mode surface function is constructed, and the power allocation deviation and power change rate of the hybrid energy storage system are set as state parameters. The boundary conditions of the robustness constraint are determined according to the state parameters.

[0022] An adaptive reaching law is established using the boundary conditions, and the rate and magnitude of change of the power allocation deviation are used as adaptive parameters to dynamically calculate the adjustment amount of the switching gain. The adaptive reaching law and the adjustment amount of the switching gain are substituted into the sliding surface function, and a power allocation command to suppress chattering is generated based on the real-time changes of the state parameters.

[0023] Based on the power allocation command, a rolling expansion is performed within the prediction time domain. The future response trajectories of the supercapacitor and the power battery are predicted using the energy allocation sequence within the prediction time domain, resulting in a power allocation strategy including:

[0024] The power allocation command is rolled out in the prediction time domain to generate a command sequence containing multiple control time points. Based on the dynamic response characteristics of the hybrid energy storage system, the command sequence is divided into multiple sub-intervals according to the operating conditions, and a power allocation benchmark is established in each sub-interval.

[0025] Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. The piecewise linearization method is used to calculate the energy allocation state in each sub-interval. The state transition at the boundary of the sub-interval is used as the critical transition constraint. A state transition equation considering the switching of operating conditions is established.

[0026] The future response trajectories of the supercapacitor and the power battery are predicted by using the energy distribution state and the state transition equation. The future response trajectory is used to characterize the dynamic characteristics of the energy storage unit during the switching process under different operating conditions, and a power distribution strategy for the hybrid energy storage system is generated.

[0027] Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. A piecewise linearization method is used to calculate the energy allocation state within each sub-interval. State transitions at the boundaries of sub-intervals are used as critical transition constraints. State transition equations considering operating condition switching are established, including:

[0028] Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. The prediction time domain is divided into acceleration power range, braking power range and steady power range. According to the power change characteristics of each range, a corresponding energy allocation sub-sequence is established.

[0029] Substitute the energy allocation subsequence into the piecewise linearization calculation framework, and use the rate of change of the state of charge of the supercapacitor and the rate of change of the output power of the power battery as linearization coefficients to establish a piecewise state equation characterizing the dynamic characteristics of charging and discharging power.

[0030] The state transition characteristics at the boundaries of the sub-intervals are analyzed using the piecewise state equations. The energy transfer efficiency and power fluctuation amplitude of the supercapacitor and the power battery during the switching of operating conditions are calculated. The energy transfer efficiency and the power fluctuation amplitude are combined to construct critical transfer constraints. The state transition equations of the hybrid energy storage system are established based on the critical transfer constraints.

[0031] The sliding mode surface function is constructed as a piecewise continuous variable characteristic curve, and the power allocation strategy is used as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding mode surface function are adjusted according to the state variable deviation in different operating condition ranges to form the final energy allocation optimization sequence, including:

[0032] The sliding surface function is constructed as a piecewise continuous variable characteristic curve. The power allocation strategy is used as the control input variable. The piecewise interval of the sliding surface function is established by the power change rate and energy transfer efficiency of the power allocation strategy. The initial switching boundary of the sliding surface is set based on the piecewise interval.

[0033] Based on the segmented interval division results, state variables are collected in real time during the charging and discharging process of the supercapacitor and the power battery. The state-of-charge deviation of the supercapacitor, the output power deviation of the power battery, and the energy transfer efficiency of the energy storage system are used as evaluation indicators. The state variable deviation in each operating condition interval is calculated, and the state variable deviation is mapped to the segmented interval.

[0034] The switching characteristic parameters of the sliding surface function in each segmented interval are calculated using the state variable deviation. The switching characteristic parameters are combined with the initial switching boundary to establish an adaptive switching law. The dynamic characteristics of the sliding surface function in different segmented intervals are adjusted by the adaptive switching law.

[0035] The power allocation strategy is modified based on the dynamically adjusted sliding surface function, and an energy allocation optimization sequence considering the robustness constraints of the charging and discharging process is constructed based on the modification results.

[0036] A second aspect of the present invention provides a hybrid energy storage control system for supercapacitors and power batteries in new energy vehicles, comprising:

[0037] The first unit is used to obtain the current operating parameters of the vehicle, the state variables of the supercapacitor and the power battery. Based on the vehicle's power demand and energy flow constraints, a multi-objective optimization model of the hybrid energy storage system is constructed. The power distribution of the supercapacitor and the power battery is modeled collaboratively through state-space equations to obtain the energy distribution sequence.

[0038] The second unit is used to perform robust constraints based on the energy allocation sequence and the sliding surface function, and to introduce an adaptive reaching law to dynamically adjust the switching gain, thereby generating a power allocation command to suppress chattering.

[0039] The third unit is used to perform rolling expansion in the prediction time domain based on the power allocation command, and to predict the future response trajectories of the supercapacitor and the power battery through the energy allocation sequence in the prediction time domain, so as to obtain the power allocation strategy.

[0040] The fourth unit is used to construct the sliding surface function into a piecewise continuous variable characteristic curve, and take the power allocation strategy as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding surface function are adjusted according to the state variable deviation in different operating conditions to form the final energy allocation optimization sequence.

[0041] A third aspect of the present invention provides an electronic device, comprising:

[0042] processor;

[0043] Memory used to store processor-executable instructions;

[0044] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0045] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0046] The beneficial effects of this application are as follows:

[0047] By constructing a multi-objective optimization model for a hybrid energy storage system and using state-space equations for collaborative modeling, the power distribution between supercapacitors and power batteries becomes more scientific and reasonable, giving full play to their respective characteristics and improving overall energy utilization efficiency.

[0048] Robustness constraints are applied based on sliding mode surface functions, and an adaptive reaching law is introduced to dynamically adjust the switching gain. This effectively solves the "chattering" problem in traditional control methods and improves system stability and reliability.

[0049] By performing a rolling unfolding method within the prediction time domain, the future response trajectories of the supercapacitor and power battery are predicted, enabling forward-looking power allocation and enhancing the system's adaptability to complex operating conditions and control accuracy.

[0050] By constructing the sliding surface function as a piecewise continuous variable characteristic curve and dynamically adjusting the switching characteristics of the sliding surface function according to the state variable deviation in different operating conditions, optimized control of energy distribution under various operating conditions is achieved, further improving the flexibility and adaptability of energy management. Attached Figure Description

[0051] Figure 1 This is a schematic flowchart of the hybrid energy storage control method for supercapacitors and power batteries in new energy vehicles according to an embodiment of the present invention.

[0052] Figure 2 This is a flowchart illustrating the adaptive switching parameter optimization and adjustment process of the hybrid energy storage system according to an embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0055] Figure 1 This is a flowchart illustrating the hybrid energy storage control method for supercapacitors and power batteries in new energy vehicles according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0056] The system acquires the current operating parameters of the vehicle, the state variables of the supercapacitor and the power battery. Based on the vehicle's power demand and energy flow constraints, it constructs a multi-objective optimization model for the hybrid energy storage system. The system then uses state-space equations to collaboratively model the power distribution between the supercapacitor and the power battery, thereby obtaining the energy distribution sequence.

[0057] Based on the energy allocation sequence, robustness constraints are applied according to the sliding mode surface function, and an adaptive reaching law is introduced to dynamically adjust the switching gain, thereby generating a power allocation command to suppress chattering.

[0058] Based on the power allocation command, the power allocation is rolled out in the prediction time domain. The future response trajectories of the supercapacitor and the power battery are predicted by the energy allocation sequence in the prediction time domain, and a power allocation strategy is obtained.

[0059] The sliding surface function is constructed as a piecewise continuous variable characteristic curve, and the power allocation strategy is used as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding surface function are adjusted according to the state variable deviation in different operating conditions to form the final energy allocation optimization sequence.

[0060] In one optional implementation, a multi-objective optimization model of the hybrid energy storage system is constructed based on vehicle power demand and energy flow constraints. The power allocation between the supercapacitor and the power battery is then collaboratively modeled using state-space equations, resulting in an energy allocation sequence including:

[0061] Acquire power demand data of the vehicle under real-time driving conditions, establish power demand characteristic curves including acceleration power, braking power and cruising power, and calculate the power distribution coefficient for each operating condition range based on the power demand characteristic curves.

[0062] Combining the power demand characteristic curve and the power allocation coefficient, the charging and discharging power limits, state of charge variation ranges, and temperature constraints of the supercapacitor and power battery in each operating condition range are extracted to construct an energy flow constraint set.

[0063] Based on the power allocation coefficient and the energy flow constraint set, a multi-objective optimization model is constructed with the goal of maximizing system energy utilization efficiency and minimizing power loss. The dynamic weighting coefficient of the optimization objective in each operating condition interval is calculated by an adaptive weight adjustment algorithm.

[0064] Substituting the dynamic weighting coefficients and the energy flow constraint set into the state-space equation of the hybrid energy storage system, the charging and discharging power of the supercapacitor and the power battery are modeled collaboratively using the rolling time-domain prediction method. By solving the state-space equation, the energy allocation sequence that satisfies the multi-objective optimization model is obtained.

[0065] The energy distribution of the hybrid energy storage system is initiated by the onboard data acquisition unit, which continuously monitors vehicle operating parameters. This unit connects to the vehicle controller via a CAN bus interface, with a sampling frequency set to 10 Hz. It synchronously reads key parameters such as vehicle speed sensor signals, accelerator pedal position sensor signals, brake pedal pressure sensor signals, engine speed signals, and transmission gear signals. The raw data is converted from analog to digital and stored in a circular buffer with a capacity of 1000 data points. When the buffer is full, a first-in, first-out (FIFO) strategy is used to update the data. The signal preprocessing module performs digital filtering on the acquired data using a Butterworth low-pass filter with a cutoff frequency set to 2 Hz and a filter order of 4, effectively suppressing sensor noise and electromagnetic interference.

[0066] The power demand characteristic curve is constructed based on the instantaneous power demand value calculated from filtered operating condition data. Acceleration power demand is calculated using a vehicle longitudinal dynamics model, multiplying parameters such as vehicle mass, air resistance coefficient, rolling resistance coefficient, and gradient resistance by the measured acceleration to obtain the required traction force, which is then multiplied by the vehicle speed to obtain the acceleration power demand. Braking power demand is calculated by multiplying the brake pedal pressure sensor signal by the braking system transfer function, and a negative value indicates energy recovery. Cruise power demand equals the power required to overcome driving resistance, including the sum of air resistance power and rolling resistance power. The characteristic curves are stored using a three-dimensional lookup table structure. The vehicle speed axis is divided into a range of 0 to 150 km / h with a step size of 5 km / h, the power axis is divided into a range of -80 kW to +120 kW with a step size of 2 kW, and the time axis records the duration and frequency of each operating condition point.

[0067] The power allocation coefficient calculation module determines the operating condition type based on the power demand change rate. This change rate is obtained by dividing the difference between the current power demand and the power demand of the previous sampling period by the sampling time interval. The threshold for high-dynamic operating conditions is set at a power change rate greater than 15 kW per second. In this case, the supercapacitor allocation coefficient is set to 0.75, and the power battery allocation coefficient is set to 0.25, fully utilizing the fast response characteristics of the supercapacitor. For medium-dynamic operating conditions, the power change rate is between 3 and 15 kW per second. The allocation coefficient is calculated using a linear interpolation method, with the supercapacitor coefficient linearly decreasing from 0.75 to 0.35 and the power battery coefficient linearly increasing from 0.25 to 0.65. For low-dynamic operating conditions, the power change rate is less than 3 kW per second. The supercapacitor allocation coefficient is set to 0.3, and the power battery allocation coefficient is set to 0.7, prioritizing the use of power batteries with higher energy density. The allocation coefficient switching uses a first-order inertial element for smooth transition, with a time constant set to 0.5 seconds to avoid system oscillation.

[0068] The energy flow constraint set encompasses all operational limitations of supercapacitors and power batteries. The supercapacitor's charging and discharging power constraints exhibit a non-linear relationship with its state of charge (SOC). Within the SOC range of 30% to 70%, the maximum charging and discharging power remains at the rated value of 40 kW. When the SOC falls below 30%, the discharging power decreases according to a quadratic function with a decay coefficient of 0.8. Similarly, when the SOC exceeds 70%, the charging power decreases according to a quadratic function. The supercapacitor's temperature constraint requires the operating temperature to be maintained within the range of -30°C to 65°C. The temperature sensor uses a thermistor type with a measurement accuracy of ±1°C. When the temperature deviates from the optimal operating range by 10°C to 40°C, the power output capability decreases linearly, with a 20% reduction in power for every 10°C deviation.

[0069] The constraints on the power battery include a state of charge (SOC) constraint setting the operating range to 20% to 90%. Charge and discharge power limits are described using a piecewise function; power is unrestricted within the SOC range of 35% to 85%, maintaining a rated power of 30 kW; outside this range, power linearly decreases to 60% of the rated power. The power battery temperature management system requires operating temperatures between 5°C and 50°C, equipped with a liquid cooling system. The coolant flow rate is automatically adjusted based on temperature feedback, and power limitation protection is triggered when the temperature exceeds the safe range. The constraint set also includes boundary conditions such as single charge / discharge depth limits, continuous operating time limits, and power ramp-up rate limits. All constraints form an inequality constraint matrix for subsequent optimization calculations.

[0070] The multi-objective optimization model aims to maximize energy utilization efficiency and minimize power loss. Energy utilization efficiency is defined as the ratio of the effective output power of the energy storage system to the total input power, encompassing multiple aspects such as energy storage device efficiency, converter efficiency, and transmission efficiency. Power loss calculation covers all loss components, including supercapacitor internal resistance loss, power battery internal resistance loss, DC-DC converter switching loss, and wire resistance loss. The optimization variables are set as the power allocation ratio at each moment in the prediction time domain, with the variable dimension equal to the number of prediction steps multiplied by the number of energy storage devices. The constraint matrix includes linear and nonlinear constraints such as power balance constraints, energy storage device physical constraints, and system stability constraints.

[0071] The adaptive weight adjustment algorithm dynamically calculates the weight coefficients of the optimization target based on real-time operating conditions. The algorithm's input parameters include multi-dimensional feature vectors such as the rate of change in power demand, current state of charge, temperature status, and historical operating condition statistics. Weight calculation employs fuzzy logic reasoning, establishing a fuzzy rule base of input features and weight outputs, containing 81 fuzzy rules covering all operating condition combinations. The membership function is defined using a trigonometric function. The efficiency target weight output ranges from 0.2 to 0.8, and the loss target weight is its complement, ensuring a weight sum of 1. The weight update frequency is set to once per second, and an exponentially weighted moving average algorithm is used to smooth the weights. A smoothing factor of 0.15 is set to avoid drastic weight fluctuations affecting system stability.

[0072] State-space equation modeling describes the hybrid energy storage system as a multi-input multi-output discrete-time system. The state vector includes five state variables: supercapacitor state of charge (SOC), power battery SOC, supercapacitor temperature, power battery temperature, and output power of each energy storage device. The input vector includes two control inputs: power demand command and allocation coefficient adjustment command. The output vector includes four output variables: total system output power, SOC change, efficiency index, and loss index. The state transition matrix is ​​established based on the electrochemical and thermodynamic characteristics of the energy storage devices. The SOC transition coefficient for the supercapacitor is set to 0.98 to reflect its extremely low self-discharge rate, while the SOC transition coefficient for the power battery is set to 0.995 to reflect its superior charge retention capability. The input and output matrices are determined based on the characteristics of the power electronic converter and sensors. The converter efficiency curve is fitted using a cubic polynomial with a fitting accuracy exceeding 99.5%.

[0073] The rolling time-domain predictive control strategy sets the prediction time domain length to 25 seconds, the control time domain length to 8 seconds, the sampling period to 0.5 seconds, the prediction steps to 50 steps, and the control steps to 16 steps. The prediction algorithm is implemented based on a neural network model trained on historical operating condition data. The network structure uses a Long Short-Term Memory (LSTM) network with two hidden layers, each containing 128 neurons. The ReLU activation function is used. The training dataset contains 10,000 typical operating condition sequences, and the training error converges to a root mean square error of less than 5%. The prediction model takes historical power demand data from the 10 seconds prior to the current moment as input and outputs a predicted power demand sequence for the next 25 seconds, achieving a prediction accuracy of over 90% under standard operating conditions.

[0074] In the collaborative modeling process, the power allocation between the supercapacitor and the power battery is coordinated and controlled through a coupled optimization algorithm. The algorithm considers the dynamic response differences between the two energy storage devices. The supercapacitor's response time constant is set to 0.05 seconds to handle the high-frequency components of the power demand, while the power battery's response time constant is set to 1.5 seconds to handle the low-frequency components. Frequency domain decomposition is achieved using a Fast Fourier Transform (FFT) with 1024 transform points, a frequency resolution of 0.02 Hz, and a frequency division point of 0.8 Hz. Power components above the frequency division point are preferentially allocated to the supercapacitor, while those below are primarily handled by the power battery. The power allocation result is then used to recover the time-domain signal through an inverse Fourier transform, serving as the power command output for each energy storage device.

[0075] The state-space equations are solved using a predictive control algorithm framework, transforming the multi-objective optimization problem into a constrained quadratic programming problem. The objective function is designed as a weighted sum of efficiency and loss objectives, with weight coefficients updated in real-time by an adaptive adjustment algorithm. Constraints include equality constraints in the state-space equations and inequality constraints related to the physical constraints of the energy storage device. The quadratic programming solver employs the interior-point method, with a convergence accuracy set to 1e-6, a maximum iteration count limited to 100, and a solution time controlled within 50 milliseconds to meet real-time control requirements.

[0076] In one optional implementation, the dynamic weighting coefficients and the set of energy flow constraints are substituted into the state-space equation of the hybrid energy storage system. A rolling time-domain prediction method is used to collaboratively model the charging and discharging power of the supercapacitor and the power battery. By solving the state-space equation, the energy allocation sequence satisfying the multi-objective optimization model is obtained, including:

[0077] Substituting the dynamic weighting coefficients and energy flow constraints into the state space equations of the hybrid energy storage system, a set of state equations is constructed with supercapacitor voltage, power battery current and state of charge as state variables and charging and discharging power as control variables.

[0078] A rolling time-domain prediction framework is constructed based on the state equation set. A power distribution prediction sequence for supercapacitors and power batteries is established in each prediction interval. According to the evolution law of the state equation set, the end state of each prediction interval is set as the initial state of the next prediction interval.

[0079] The state response trajectories of the supercapacitor and the power battery are calculated using the state equation set and the power allocation prediction sequence. The state response trajectories are combined with the dynamic weighting coefficients to correct the power allocation prediction sequence, thereby generating a corrected prediction sequence that satisfies the energy flow constraint set.

[0080] Substituting the corrected prediction sequence into the state equations and solving them yields an energy allocation sequence that satisfies the multi-objective optimization requirements.

[0081] The dynamic weighting coefficients and energy flow constraints are substituted into the state-space equations of the hybrid energy storage system to construct a set of state equations. In this step, the supercapacitor voltage Vsc, the battery current Ibat, and the state of charge (SOC) are used as state variables, and the charging / discharging power P is used as the control variable to construct the set of state equations. Specifically, the voltage state equation of the supercapacitor can be expressed as Vsc at the next moment equals the current voltage minus an amount proportional to the discharge power, where the proportionality coefficient is related to the supercapacitor's capacitance. The battery current state equation is expressed as Ibat at the next moment equals the current plus an increment related to the battery power change, where the increment is determined by the battery's internal resistance and capacity. The battery's SOC is calculated based on the principle of current integration, relating it to the battery capacity and charging / discharging efficiency. Simultaneously, constraints such as voltage upper and lower limits, current upper and lower limits, SOC safety range, and power change rate limits, which are included in the energy flow constraints, are embedded into the set of state equations.

[0082] Based on the constructed state equations, a rolling time-domain prediction framework is established. Under this framework, the entire energy allocation process is divided into multiple consecutive prediction intervals, each containing a specific number of time steps. For example, the prediction time domain can be set to T=10, indicating that each prediction covers the energy allocation situation for the next 10 time steps. Within each prediction interval, power allocation prediction sequences P_sc(t) and P_bat(t) for supercapacitors and power batteries are established, where t represents the time index within the interval. According to the evolution law of the state equations, the state values ​​[Vsc(T), Ibat(T), SOC(T)] at the end of each prediction interval are set as the initial states [Vsc(0), Ibat(0), SOC(0)] of the next prediction interval, ensuring the continuity of state variables between adjacent prediction intervals.

[0083] Within the rolling prediction framework, the state response trajectories of the supercapacitor and the power battery are calculated using a set of state equations and a power allocation prediction sequence. Specifically, the initially predicted power allocation sequence is substituted into the set of state equations, and the evolution trajectory of each state variable within the next T time steps is iteratively calculated. The calculated supercapacitor voltage trajectory Vsc_traj, power battery current trajectory Ibat_traj, and SOC trajectory SOC_traj constitute a complete set of state response trajectories. These state response trajectories are combined with dynamic weighting coefficients to correct the power allocation prediction sequence. The dynamic weighting coefficients are dynamically adjusted according to the current system state; the battery protection weight is increased when the SOC approaches the lower limit, and the energy recovery efficiency weight is increased when the supercapacitor voltage approaches the upper limit. The comprehensive performance index of the current prediction sequence is calculated through a weighted evaluation function, and the power allocation sequence is corrected based on the evaluation results, generating corrected prediction sequences P_sc_revised(t) and P_bat_revised(t) that satisfy the energy flow constraint set.

[0084] The correction process is implemented through iterative optimization algorithms, such as using gradient descent to adjust the power allocation value in each iteration until the evaluation function converges to the optimal value or reaches the preset number of iterations. During the correction process, if any constraint violation is found, such as the supercapacitor voltage exceeding the safe range, a penalty factor is introduced to correct the power allocation value that violates the constraint, ensuring that the corrected power allocation sequence fully satisfies the system constraints.

[0085] The corrected prediction sequence is substituted into the state equations for solution, yielding an energy allocation sequence that meets the multi-objective optimization requirements. In practice, only the power allocation value corresponding to the first time step in the prediction sequence is used as the current control command. Then, the process is repeated to the next time step, updating the system state and recurring the prediction correction process. As time progresses, through continuous rolling prediction and control, complete energy allocation sequences P_sc_optimal and P_bat_optimal are eventually generated, achieving optimal energy management of the hybrid energy storage system.

[0086] In electric vehicle applications, this method demonstrates significant advantages. For example, during acceleration, the algorithm automatically allocates a larger power demand to the supercapacitor, utilizing its high power density to respond quickly while keeping the battery current within a small range, effectively reducing transient stress on the battery. During the regenerative braking phase, the algorithm intelligently allocates regenerative power based on the supercapacitor's current voltage state and the amount of regenerative energy, prioritizing charging the supercapacitor. This avoids battery losses at high charging rates, extends battery life, and improves energy recovery efficiency, reserving energy for the next acceleration.

[0087] In one optional implementation, based on the energy allocation sequence, robust constraints are applied according to the sliding mode surface function, and an adaptive reaching law is introduced to dynamically adjust the switching gain, generating a power allocation command to suppress chattering, including:

[0088] Based on the energy allocation sequence, a sliding mode surface function is constructed, and the power allocation deviation and power change rate of the hybrid energy storage system are set as state parameters. The boundary conditions of the robustness constraint are determined according to the state parameters.

[0089] An adaptive reaching law is established using the boundary conditions, and the rate and magnitude of change of the power allocation deviation are used as adaptive parameters to dynamically calculate the adjustment amount of the switching gain. The adaptive reaching law and the adjustment amount of the switching gain are substituted into the sliding surface function, and a power allocation command to suppress chattering is generated based on the real-time changes of the state parameters.

[0090] Based on the energy allocation sequence, a sliding mode surface function is constructed. In a hybrid energy storage system, power allocation deviation and power change rate are set as state parameters. The power allocation deviation represents the difference between the actual allocated power and the ideal allocated power, and the power change rate represents the change in power per unit time. The state parameter vector is defined as x = [x1, x2]. T Where x1 represents the power distribution deviation and x2 represents the power change rate. The sliding surface function can be expressed as:

[0091] s = cx1 + x2;

[0092] Here, 'c' is a positive constant used to adjust the slope of the sliding surface. By appropriately selecting the value of 'c', the motion characteristics of the system on the sliding surface can be made to meet the expected requirements.

[0093] Based on the state parameters, the boundary conditions for robustness constraints are determined. Hybrid energy storage systems are subject to uncertainties such as parameter disturbances and measurement errors during actual operation. To ensure the robustness of the control system, boundary conditions need to be established to constrain these uncertainties. Assuming the upper limit of system disturbance is D, the boundary conditions for robustness constraints can be expressed as |d(t)| ≤ D, where d(t) is the time-varying disturbance.

[0094] Based on the above boundary conditions, an adaptive reaching law is established. Traditional sliding mode control uses a fixed switching gain, which easily leads to system chattering. To solve this problem, an adaptive reaching law is introduced to dynamically adjust the switching gain. The design of the adaptive reaching law considers the rate and magnitude of change of the power distribution deviation, expressed as:

[0095] K(t) = K0 + ΔK(t);

[0096] Where K0 is the basic switching gain, and ΔK(t) is the dynamic adjustment amount of the switching gain. ΔK(t) is calculated as follows:

[0097] ΔK(t) = α|x1|β|x2|;

[0098] In the formula, α and β are positive constants used to adjust the adaptive strength. When the power distribution deviation or power change rate is large, the switching gain will automatically increase to accelerate the system toward the sliding surface; when the state is close to the sliding surface, the switching gain will automatically decrease, thereby reducing chattering.

[0099] Substituting the adaptive reaching law and the adjustment amount of the switching gain into the sliding mode surface function, we can obtain the power distribution control law for suppressing chattering:

[0100] u = -f(x) - K(t)·sgn(s) - ks;

[0101] Where f(x) is the deterministic part of the system, sgn(s) is the sign function, and k is the rate of attainment coefficient. To further suppress chattering, the sign function sgn(s) can be replaced with the continuous saturation function sat(s / φ), where φ is the boundary layer thickness. By adjusting the value of φ, the relationship between chattering suppression and control accuracy can be balanced.

[0102] In practical applications, the control output is dynamically calculated based on the real-time state parameters of the hybrid energy storage system. For example, when a sudden increase in power distribution deviation is detected, the adaptive mechanism will quickly increase the switching gain to accelerate the system back to equilibrium. When the system approaches equilibrium, the switching gain will automatically decrease to reduce the oscillation of the control signal, thereby achieving a smooth transition.

[0103] The power allocation command generation process is as follows: First, real-time parameters of the hybrid energy storage system are collected, including the state of charge and output power of each energy storage unit; then, the ideal power allocation value is determined according to the energy allocation sequence; next, the power allocation deviation and power change rate are calculated to form a state parameter vector; then, the sliding mode surface function value is calculated based on the state parameters; then, the current switching gain is calculated based on the adaptive reaching law; finally, the power allocation control command is generated and sent to the power controller of each energy storage unit.

[0104] Furthermore, the physical constraints of each energy storage unit must be considered during power allocation, such as the charge / discharge rate limits of batteries and the capacity limitations of supercapacitors. When generating power allocation commands, the calculation results need to be constrained to ensure that the control commands do not exceed the physical limitations of the energy storage units.

[0105] The above method enables precise control of power distribution in hybrid energy storage systems, effectively suppressing system chatter and improving system operational stability and energy utilization efficiency. This method boasts advantages such as strong adaptability, good anti-interference capability, and fast response speed, making it suitable for power management in various hybrid energy storage systems.

[0106] In one optional implementation, the power allocation command is rolled out over the prediction time domain, and the future response trajectories of the supercapacitor and the power battery are predicted using the energy allocation sequence within the prediction time domain, resulting in a power allocation strategy including:

[0107] The power allocation command is rolled out in the prediction time domain to generate a command sequence containing multiple control time points. Based on the dynamic response characteristics of the hybrid energy storage system, the command sequence is divided into multiple sub-intervals according to the operating conditions, and a power allocation benchmark is established in each sub-interval.

[0108] Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. The piecewise linearization method is used to calculate the energy allocation state in each sub-interval. The state transition at the boundary of the sub-interval is used as the critical transition constraint. A state transition equation considering the switching of operating conditions is established.

[0109] The future response trajectories of the supercapacitor and the power battery are predicted by using the energy distribution state and the state transition equation. The future response trajectory is used to characterize the dynamic characteristics of the energy storage unit during the switching process under different operating conditions, and a power distribution strategy for the hybrid energy storage system is generated.

[0110] The power allocation commands are rolled out over the prediction time domain to generate a command sequence containing multiple control time points. The prediction time domain can be set to T seconds, and within this time domain, it is divided into N sampling points according to a fixed sampling interval Δt, i.e., N = T / Δt. For each sampling point tk (k = 1, 2, ..., N), a corresponding power allocation command Pk is generated based on the current operating conditions and system state. These power allocation commands constitute the command sequence {P1, P2, ..., PN}, which is used to guide the power allocation between the supercapacitor and the power battery in the hybrid energy storage system.

[0111] Based on the dynamic response characteristics of the hybrid energy storage system, the command sequence is divided into multiple sub-intervals according to operating condition characteristics. Operating condition characteristics may include acceleration, constant speed, and braking conditions. By analyzing the rate of change, amplitude, and duration of power commands at adjacent sampling points, the switching points of operating conditions can be identified. For example, when the power command suddenly changes from a positive value to a negative value, it indicates a switch from acceleration to braking; when the power command fluctuates within a relatively stable range, it indicates a constant speed condition. Assume the entire prediction time domain is divided into M sub-intervals {[t0, t1], [t1, t2], ..., [tM-1, tM]}, where t0 is the initial time and tM is the end time of the prediction time domain.

[0112] A power allocation benchmark is established within each sub-interval. For the sub-interval [ti-1, ti], the average power demand Pavg,i and the power fluctuation range ΔPi within that interval are calculated. Based on the characteristics that supercapacitors are suitable for handling high-frequency power fluctuations and power batteries are suitable for providing continuous and stable power, the supercapacitors are assigned to bear the power fluctuation portion, while the power batteries bear the basic load. Specifically, within sub-interval i, the power allocation benchmark PB,i of the power battery is set to the average power Pavg,i of that interval, and the power allocation benchmark PSC,i of the supercapacitor is set to zero, meaning that the energy of the supercapacitor is required to remain relatively balanced within this interval.

[0113] Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed, and the energy allocation state in each sub-interval is calculated using a piecewise linearization method. For any time t in the sub-interval [ti-1, ti], the energy state ESC(t) of the supercapacitor and the energy state EB(t) of the power battery can be calculated as follows: ESC(t) = ESC(ti-1) + ∫[ti-1, t]PSC(τ)dτ, EB(t) = EB(ti-1) + ∫[ti-1, t]PB(τ)dτ, where PSC(τ) and PB(τ) are the power outputs of the supercapacitor and the power battery at time τ, respectively.

[0114] Considering various constraints in actual operation, such as the power limits, energy limits, and efficiency characteristics of supercapacitors and power batteries, the initial allocation needs to be adjusted. For example, when the state of charge of the supercapacitor is close to its upper limit, its charging power should be reduced; when the discharge power of the power battery exceeds the safe range, some power demand should be transferred to the supercapacitor.

[0115] State transitions at the boundaries of sub-intervals are used as critical transition constraints to establish state transition equations considering operating condition switching. At the operating condition switching point *ti*, the energy states of the supercapacitor and the power battery must satisfy the continuity condition: *ESC(ti+)* = *ESC(ti-)* and *EB(ti+)* = *EB(ti-)*, where *ti-* and *ti+* represent the extreme moments before and after the switching point, respectively. Simultaneously, considering the change in power allocation strategy brought about by the operating condition switching, additional transition constraints need to be applied to ensure the stability and safety of system operation.

[0116] The future response trajectories of supercapacitors and power batteries are predicted using energy allocation state and state transition equations. For each sub-interval, the state trajectory of the energy storage unit within that interval is calculated based on the initial state and power allocation strategy. By connecting the state trajectories within each sub-interval, the complete response trajectories of supercapacitors and power batteries over the entire prediction time domain are obtained.

[0117] By characterizing the dynamic characteristics of energy storage units during switching between different operating conditions using future response trajectories, a power allocation strategy for the hybrid energy storage system can be generated. Based on the predicted response trajectories, the performance of the current power allocation strategy can be evaluated, including energy utilization efficiency, energy storage unit lifetime impact, and system responsiveness. If the predicted performance does not meet the requirements, the power allocation benchmark can be adjusted or the state transition constraints can be modified to regenerate the power allocation strategy until the system performance requirements are met.

[0118] In practical applications, as the vehicle's operating status changes in real time, the power allocation command needs to be updated periodically and the above process needs to be re-executed. This rolling optimization approach ensures that the power allocation strategy can adapt to dynamically changing operating conditions, achieving efficient management of the hybrid energy storage system.

[0119] In one optional implementation, an energy allocation sequence in the prediction time domain is constructed based on the power allocation benchmark. A piecewise linearization method is used to calculate the energy allocation state within each sub-interval. State transitions at the boundaries of sub-intervals are used as critical transition constraints. A state transition equation considering operating condition switching is established, including:

[0120] Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. The prediction time domain is divided into acceleration power range, braking power range and steady power range. According to the power change characteristics of each range, a corresponding energy allocation sub-sequence is established.

[0121] Substitute the energy allocation subsequence into the piecewise linearization calculation framework, and use the rate of change of the state of charge of the supercapacitor and the rate of change of the output power of the power battery as linearization coefficients to establish a piecewise state equation characterizing the dynamic characteristics of charging and discharging power.

[0122] The state transition characteristics at the boundaries of the sub-intervals are analyzed using the piecewise state equations. The energy transfer efficiency and power fluctuation amplitude of the supercapacitor and the power battery during the switching of operating conditions are calculated. The energy transfer efficiency and the power fluctuation amplitude are combined to construct critical transfer constraints. The state transition equations of the hybrid energy storage system are established based on the critical transfer constraints.

[0123] The energy allocation sequence construction module generates a control sequence for energy storage devices within the prediction time domain based on a power allocation benchmark. The prediction time domain length is set to 10 seconds, the time step is set to 0.2 seconds, and it includes 50 discrete time nodes. The power allocation benchmark consists of two components: a supercapacitor power benchmark and a power battery power benchmark. The benchmark values ​​are obtained through statistical analysis of historical operating data. The supercapacitor power benchmark covers a range of -40 kW to +60 kW, and the power battery power benchmark covers a range of -20 kW to +40 kW. The sequence construction adopts a rolling window strategy. Each control cycle predicts the energy allocation demand 10 seconds in advance. After executing the allocation command at the current moment, the window slides 0.2 seconds to proceed to the next round of prediction. The prediction accuracy requirements are: power allocation error less than 5%, state of charge prediction error less than 3%, and prediction calculation time controlled within 50 milliseconds to meet real-time control requirements.

[0124] The predicted time domain is divided based on the changing characteristics of load power demand to determine the interval boundaries. The acceleration power interval corresponds to the stage of positive power demand growth; the power change rate is greater than 5 kWh / s when entering this interval, and the power exhibits a monotonically increasing trend within it. The braking power interval corresponds to the load braking and recovery stage; the power demand is negative and its absolute value increases when entering this interval; the power change rate is less than -3 kWh / s when entering this interval, and the recovered power gradually increases within it. The stable power interval corresponds to the stage of relatively stable power demand; the absolute value of the power change rate is less than 2 kWh / s when entering this interval, and the power fluctuation amplitude within this interval is controlled within 10% of the average value. The interval switching judgment uses hysteresis logic to avoid frequent switching; the hysteresis width is set to 20% of the boundary threshold, i.e., the acceleration interval entry threshold is 5 kWh / s, and the exit threshold is 4 kWh / s; the braking interval entry threshold is -3 kWh / s, and the exit threshold is -2.4 kWh / s.

[0125] The energy allocation subsequence establishes specialized allocation strategies for different power ranges. In the acceleration power range subsequence, the high power density of supercapacitors is prioritized, with supercapacitors handling 70% to 80% of the total power demand, and the power battery handling the remaining 20% ​​to 30%. The allocation ratio is dynamically adjusted based on acceleration intensity. In the braking power range subsequence, charging of supercapacitors is prioritized, utilizing their high charge acceptance capacity. Supercapacitors accept 80% to 90% of the recovered power, and the power battery accepts the remaining 10% to 20%. The charging allocation ratio is adjusted based on the state of charge (SBC) difference. The steady-state power range subsequence aims for optimal energy efficiency, determining the optimal allocation ratio based on the current SBC and load power. When the SBC difference is less than 5%, a uniform allocation strategy is adopted; when the difference is greater than 5%, allocation is tilted towards energy storage devices with lower SBCs.

[0126] The piecewise linearization calculation framework approximates the nonlinear characteristics of energy storage devices linearly within each sub-interval. The linearization accuracy requirement is an error of less than 3%. When the error exceeds this range, the interval division is refined to improve the approximation accuracy. The rate of change of the supercapacitor's state of charge (SOC) serves as a linearization coefficient, reflecting its charging and discharging dynamic characteristics. The rate of change is calculated using a forward difference method, i.e., the current SOC minus the previous SOC divided by the time step, with the result expressed as a percentage per second. The rate of change of the power battery's output power serves as another linearization coefficient, describing its dynamic power output response characteristics. The rate of change is also calculated using a forward difference method, i.e., the current power minus the previous power divided by the time step, with the result expressed as kilowatts per second. The linearization coefficient update frequency is consistent with the control cycle at 5 times per second. The coefficient calculation uses a moving average filter to suppress the influence of measurement noise, and the sliding window length is set to 5 sampling points, i.e., a 1-second time window.

[0127] The piecewise state equations are established based on linearization coefficients to describe the dynamic characteristics of the energy storage device's charging and discharging power. The state equations are expressed in discrete-time state-space form, and the state variables include four physical quantities: the supercapacitor's state of charge (SOC), the power battery's SOC, the supercapacitor's terminal voltage, and the power battery's terminal voltage. The elements of the state transition matrix are determined based on the linearization coefficients. The diagonal elements are 1 minus the ratio of the time step to the discharge time constant, and the off-diagonal elements are the product of the time step and the coupling coefficient, which reflects the degree of mutual influence between different energy storage devices. The input matrix describes the influence of control inputs on the state variables. The matrix elements are determined based on the power response characteristics of the energy storage device; the supercapacitor's corresponding element is 0.9, reflecting its fast response capability, while the power battery's corresponding element is 0.7, reflecting its relatively slow response characteristic. The state equation parameters are updated every 2 seconds to adapt to changes in operating conditions. Parameter updates are achieved online using a recursive least squares method, and a forgetting factor of 0.95 is set to balance tracking capability and noise resistance.

[0128] State jump analysis at sub-interval boundaries identifies abrupt changes in the energy storage device's state during operating condition switching. State jump detection employs a sliding variance method, calculating the variance of the state variable over five consecutive sampling points. A state jump is defined as a value exceeding three times the variance during normal operation. Supercapacitor state-of-charge jumps primarily occur during the transition between acceleration and braking zones, typically ranging from 2% to 5% of the state of charge, with a duration of approximately 0.5 to 1 second. Power battery output power jumps mainly occur during the transition between stable and acceleration zones, ranging from 10% to 25% of the rated power, with a recovery time of approximately 1 to 2 seconds depending on the battery's internal resistance. The state jump detection module is configured with adjustable threshold parameters to adapt to different application scenarios. The threshold range is 2 to 5 times the normal variance, with a default setting of 3 times the normal variance. The threshold is lowered when the detection sensitivity is too high and raised when the false alarm rate is too high.

[0129] Energy transfer efficiency calculations encompass all energy loss stages, including internal losses in the energy storage device, power converter losses, and control circuit losses. The efficiency calculation employs a real-time power measurement method, where output power is divided by input power to obtain instantaneous efficiency. The energy transfer efficiency of supercapacitors is primarily affected by the charge / discharge rate; efficiency can reach over 95% at low rates, but drops to around 85% at high rates, showing an inverse relationship with charge / discharge current. The energy transfer efficiency of power batteries is mainly affected by temperature and state of charge (SOC). The optimal efficiency operating point is an ambient temperature of 25 degrees Celsius and a SOC of 50%, at which point the efficiency reaches 92%. Efficiency gradually decreases as the operating point deviates from the optimal point. The efficiency measurement accuracy is required to reach 1%, with the sampling frequency set to 100 Hz. The sampled data undergoes low-pass filtering to eliminate high-frequency noise interference, and the filter cutoff frequency is set to 10 Hz.

[0130] Power fluctuation amplitude calculation describes the degree of fluctuation in the output power of an energy storage device. The calculation method is the standard deviation of power within a statistical time window, with the time window length set to 2 seconds and containing 10 power sampling points. The power fluctuation amplitude of supercapacitors is typically 5% to 15% of the average power, mainly caused by changes in load power demand, with fluctuation frequencies concentrated in the range of 0.5 Hz to 5 Hz. The power fluctuation amplitude of power batteries is typically 2% to 8% of the average power, with relatively smaller fluctuations due to their slower response speed, which plays a smoothing role; the fluctuation frequencies are mainly distributed in the range of 0.1 Hz to 2 Hz. Power fluctuation monitoring uses a real-time sliding calculation method, updating the fluctuation amplitude once per sampling period. The calculation results are used to evaluate the operational stability and control quality of the energy storage device.

[0131] The critical transfer constraint construction combines energy transfer efficiency and power fluctuation amplitude to form a constraint condition, which is the intersection of the lower efficiency limit constraint and the upper fluctuation limit constraint. The lower efficiency limit constraint requires the supercapacitor efficiency to be no less than 80% and the power battery efficiency to be no less than 85%. When the constraint is violated, a working condition switching delay mechanism is triggered, with a delay time of 1 to 3 seconds depending on the degree of violation. The upper fluctuation limit constraint requires the supercapacitor power fluctuation amplitude to not exceed 20% of the average power and the power battery power fluctuation amplitude to not exceed 12% of the average power. When the constraint is violated, a power smoothing control algorithm is activated to suppress power fluctuation by increasing the control gain. The constraint check frequency is set to 20 times per second, and the constraint status is recorded through Boolean variables. A true value indicates that the constraint is satisfied, and an false value indicates that the constraint is violated. When the constraint status changes, the corresponding control strategy adjustment is triggered.

[0132] The state transition equations are established based on critical transition constraints to describe the state evolution of the hybrid energy storage device under different operating conditions. A Markov chain model is used to describe the state transition probability characteristics. The state space is defined as containing nine discrete states, corresponding to combinations of three power ranges and three constraint states: acceleration satisfying constraints, acceleration violating constraints, braking satisfying constraints, braking violating constraints, stationary satisfying constraints, and stationary violating constraints. The state transition probability matrix is ​​obtained through statistical analysis of historical operating data. Matrix elements represent the probability of transitioning from the current state to the next state, with probability values ​​ranging from 0 to 1, and the sum of probabilities in each row equal to 1. The transition probability calculation uses the maximum likelihood estimation method, with a statistical sample size of no less than 1000 state transition events, achieving an estimation accuracy within 5% to meet control accuracy requirements.

[0133] In one optional implementation, the sliding mode surface function is constructed as a piecewise continuous variable characteristic curve, and the power allocation strategy is used as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding mode surface function are adjusted according to the state variable deviation in different operating condition ranges to form the final energy allocation optimization sequence, including:

[0134] The sliding surface function is constructed as a piecewise continuous variable characteristic curve. The power allocation strategy is used as the control input variable. The piecewise interval of the sliding surface function is established by the power change rate and energy transfer efficiency of the power allocation strategy. The initial switching boundary of the sliding surface is set based on the piecewise interval.

[0135] Based on the segmented interval division results, state variables are collected in real time during the charging and discharging process of the supercapacitor and the power battery. The state-of-charge deviation of the supercapacitor, the output power deviation of the power battery, and the energy transfer efficiency of the energy storage system are used as evaluation indicators. The state variable deviation in each operating condition interval is calculated, and the state variable deviation is mapped to the segmented interval.

[0136] The switching characteristic parameters of the sliding surface function in each segmented interval are calculated using the state variable deviation. The switching characteristic parameters are combined with the initial switching boundary to establish an adaptive switching law. The dynamic characteristics of the sliding surface function in different segmented intervals are adjusted by the adaptive switching law.

[0137] The power allocation strategy is modified based on the dynamically adjusted sliding surface function, and an energy allocation optimization sequence considering the robustness constraints of the charging and discharging process is constructed based on the modification results.

[0138] like Figure 2 As shown, the method includes:

[0139] The sliding surface function construction module converts the dynamic characteristics of the power allocation strategy into piecewise continuous variable characteristic curves. The power allocation strategy, as the control input variable, includes two components: the supercapacitor power allocation ratio and the power battery power allocation ratio, with values ​​ranging from 0 to 1, and their sum always equal to 1. The variable characteristic curves are defined using a three-segment structure: the low-power range corresponds to a power change rate of less than 5 kW / s; the medium-power range corresponds to a power change rate between 5 kW / s and 20 kW / s; and the high-power range corresponds to a power change rate greater than 20 kW / s. Within each segment, the sliding surface function is expressed as a quadratic polynomial. The polynomial coefficients are determined based on the energy transfer efficiency characteristics within that range. The coefficients for the low-power range are set to 0.8, 0.15, and 0.05, corresponding to the quadratic, linear, and constant terms, respectively. The coefficients for the medium-power range are adjusted to 1.2, 0.25, and 0.1, and the coefficients for the high-power range are further adjusted to 1.8, 0.4, and 0.2 to enhance control responsiveness.

[0140] The segmented interval division is based on a judgment criterion established by the coupling relationship between power change rate and energy transfer efficiency. The power change rate is calculated by dividing the difference in power demand between the current sampling period and the previous sampling period by the sampling time interval of 0.1 seconds. Energy transfer efficiency is defined as the ratio of the output power to the input power of the energy storage device, including all loss factors such as internal losses of the energy storage unit, power converter losses, and line transmission losses. The efficiency calculation module is configured with a lookup table structure to store efficiency data under different operating conditions. The row index of the table is the power change rate, the column index is the current power level, and the table value is the corresponding energy transfer efficiency. The lookup table precision is set to a grid resolution of 1 kW / s for the power change rate and 2 kW for the power level.

[0141] The initial switching boundaries are determined based on the boundary power change rate values ​​of each segment interval. The switching boundary for the low-power and medium-power intervals is set at 5 kW / s, and the switching boundary for the medium-power and high-power intervals is set at 20 kW / s. Buffer zones are set near the switching boundaries to avoid frequent switching. The buffer width is set to 10% of the boundary value, i.e., 0.5 kW / s and 2 kW / s. When the power change rate enters the buffer zone, a hysteresis judgment logic is used; the segmented switching operation is only performed if the rate of change completely crosses the buffer boundary. The sliding surface function maintains continuity at the boundaries of each segment interval. Boundary constraints ensure the continuity of function values ​​and derivative values. The Lagrange multiplier method is used to optimize parameters when solving for boundary constraints.

[0142] The state quantity acquisition module monitors the operating state parameters of the supercapacitor and the power battery in real time. The sampling frequency is set to 20 Hz, and the acquired parameters include eight basic physical quantities: supercapacitor voltage, current, temperature, and state of charge (SOC); and power battery voltage, current, temperature, and SOC. The SOC calculation employs a fusion estimation strategy combining the ampere-hour integral method and the open-circuit voltage method. The ampere-hour integral method has a weighting coefficient of 0.7, and the open-circuit voltage method has a weighting coefficient of 0.3. The fusion result is optimized for state estimation using a Kalman filter, with the filter process noise covariance set to 0.01 and the observation noise covariance set to 0.05. Temperature measurement uses a thermistor sensor with a measurement accuracy of ±0.5 degrees Celsius. The sensor is positioned at the center of the energy storage unit and on the surface of the heat sink. The internal temperature distribution of the energy storage unit is calculated using the temperature gradient.

[0143] The state quantity deviation calculation compares the actual measured value with the expected set value to obtain the deviation signal. The expected state of charge of the supercapacitor is set to 50%, with an allowable deviation range of ±10%. When the deviation exceeds this range, the deviation signal is amplified linearly, with an amplification factor set to 2.0. The expected output power of the power battery is dynamically determined according to the power allocation strategy. The power deviation is calculated as a percentage deviation, where the deviation value equals the difference between the actual power and the expected power divided by the expected power and then multiplied by 100%. The expected energy transfer efficiency of the energy storage device is set to 90%. The efficiency deviation is calculated as an absolute deviation, with a deviation threshold set to 5%. When the threshold is exceeded, an efficiency optimization adjustment mechanism is triggered.

[0144] The deviation mapping process allocates the calculated state variable deviations to corresponding segmented intervals. The mapping rule determines the interval based on the power change rate under the current operating condition, and the deviation value within the interval serves as the adjustment input for the segmented sliding surface function. The mapping function is implemented using a piecewise linear interpolation method, with interpolation nodes set at the midpoints of each segmented interval, and the interpolation accuracy controlled within 1%. The deviation signal undergoes digital filtering to eliminate high-frequency noise interference. The filter is a second-order Butterworth low-pass filter with a cutoff frequency set to 1 Hz and a phase delay compensation coefficient set to 0.05 seconds.

[0145] The switching characteristic parameter calculation dynamically adjusts the switching behavior of the sliding surface function based on the mapped state variable deviation. The parameter calculation employs fuzzy logic inference to achieve adaptive adjustment. The fuzzification process converts the three input quantities—charge state deviation, power deviation, and efficiency deviation—into fuzzy sets. The membership function is defined using a triangular function, and each input quantity is divided into five fuzzy subsets: negative large, negative small, zero, positive small, and positive large. The inference rule base contains 125 rules covering all input combinations. The consequent of each rule is the adjustment amount of the switching characteristic parameter, with the parameter adjustment range set to ±20%. The defuzzification process uses the centroid method to calculate the final parameter adjustment value, maintaining a calculation accuracy of 0.1%.

[0146] The adaptive handover law combines calculated handover characteristic parameters with the initial handover boundary to form a dynamically adjustable handover judgment criterion. The handover law adopts an exponential reaching law form, with the reaching speed controlled by the handover characteristic parameters. Larger parameter values ​​result in faster reaching speeds, and the parameter adjustment range is limited to 0.5 to 2.0 times the initial value to prevent over-adjustment. The handover law also includes a chatter suppression mechanism, reducing high-frequency chatter by replacing the sign function with a saturation function. The saturation function boundary layer thickness is set to 0.1, and a linear function transition is used within the boundary layer to ensure smooth handover. The handover characteristic parameter update frequency is set to 10 times per second, and the parameter update uses a first-order inertial filter for gradual adjustment, with a time constant set to 0.2 seconds.

[0147] The dynamic adjustment process of the sliding surface function corrects the function parameters within each segment interval based on the output signal of the adaptive switching law. The adjustment mechanism employs a proportional-integral (PI) control strategy to optimize parameters. The proportional control section directly adjusts the function parameters based on the current deviation signal, with the proportional coefficient set to 0.5. The integral control section performs long-term adjustment based on the cumulative effect of the deviation signal, with the integral coefficient set to 0.1 and the integral time window length set to 10 seconds. The parameter adjustment range is limited to ±30% of the initial value; amplitude limiting protection is used to prevent parameter divergence when the range is exceeded. After adjustment, the sliding surface function undergoes a continuity check to ensure the continuity of function values ​​and derivative values ​​at segment boundaries. If the check fails, parameter optimization is performed again.

[0148] The power allocation strategy correction module outputs a correction signal based on the dynamically adjusted sliding surface function. This correction signal is calculated using a sliding mode control law, which comprises two components: an equivalent control term and a switching control term. The equivalent control term maintains the sliding surface function value at zero; this is achieved by solving the partial derivative of the sliding surface function with respect to the power allocation strategy. The switching control term overcomes modeling errors and external interference; the control gain is dynamically adjusted based on the upper bound of the interference, with the gain adjustment range set to 1.0 to 5.0 times the base value. The corrected power allocation strategy must satisfy physical constraints: the supercapacitor power allocation ratio is limited to 0.1 to 0.9, and the power battery power allocation ratio is its complement, ensuring the ratio sum is always equal to 1.

[0149] The robustness constraints are designed to account for uncertainties during charging and discharging, and include multiple aspects such as parameter uncertainty constraints, external disturbance constraints, and measurement noise constraints. Parameter uncertainty constraints model the variation range of parameters such as the energy storage device's capacity, internal resistance, and efficiency, with the uncertainty range set at ±15% of the nominal value. External disturbance constraints consider the random fluctuations in load power demand, with the upper limit of the disturbance amplitude set at 10% of the rated power, and the frequency range limited to 0.1 Hz to 10 Hz. Measurement noise constraints are determined based on sensor accuracy specifications, with the standard deviation of voltage measurement noise set at 0.1% of full scale, and the standard deviation of current measurement noise set at 0.2% of full scale.

[0150] The energy allocation optimization sequence construction transforms the revised power allocation strategy into a specific control command sequence for energy storage devices. The sequence generation period is set to 0.1 seconds, and the prediction time domain length is set to 5 seconds, containing 50 control steps. Sequence optimization employs a rolling optimization strategy, recalculating the optimal control sequence for each control cycle. After executing the first control variable, the time domain window slides to perform the next round of optimization. The control command sequence includes control signals such as supercapacitor power commands, power battery power commands, and charge / discharge mode switching commands. The command numerical accuracy is maintained at the 1-watt level, and the response time is controlled within 10 milliseconds.

[0151] A second aspect of the present invention provides a hybrid energy storage control system for supercapacitors and power batteries in new energy vehicles, comprising:

[0152] The first unit is used to obtain the current operating parameters of the vehicle, the state variables of the supercapacitor and the power battery. Based on the vehicle's power demand and energy flow constraints, a multi-objective optimization model of the hybrid energy storage system is constructed. The power distribution of the supercapacitor and the power battery is modeled collaboratively through state-space equations to obtain the energy distribution sequence.

[0153] The second unit is used to perform robust constraints based on the energy allocation sequence and the sliding surface function, and to introduce an adaptive reaching law to dynamically adjust the switching gain, thereby generating a power allocation command to suppress chattering.

[0154] The third unit is used to perform rolling expansion in the prediction time domain based on the power allocation command, and to predict the future response trajectories of the supercapacitor and the power battery through the energy allocation sequence in the prediction time domain, so as to obtain the power allocation strategy.

[0155] The fourth unit is used to construct the sliding surface function into a piecewise continuous variable characteristic curve, and take the power allocation strategy as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding surface function are adjusted according to the state variable deviation in different operating conditions to form the final energy allocation optimization sequence.

[0156] A third aspect of the present invention provides an electronic device, comprising:

[0157] processor;

[0158] Memory used to store processor-executable instructions;

[0159] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0160] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0161] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0162] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling the hybrid energy storage of supercapacitors and power batteries in new energy vehicles, characterized in that, include: The system acquires the current operating parameters of the vehicle, the state variables of the supercapacitor and the power battery. Based on the vehicle's power demand and energy flow constraints, it constructs a multi-objective optimization model for the hybrid energy storage system. The system then uses state-space equations to collaboratively model the power distribution between the supercapacitor and the power battery, thereby obtaining the energy distribution sequence. Based on the energy allocation sequence, robustness constraints are applied according to the sliding mode surface function, and an adaptive reaching law is introduced to dynamically adjust the switching gain, thereby generating a power allocation command to suppress chattering. Based on the power allocation command, the power allocation is rolled out in the prediction time domain. The future response trajectories of the supercapacitor and the power battery are predicted by the energy allocation sequence in the prediction time domain, and a power allocation strategy is obtained. The sliding surface function is constructed as a piecewise continuous variable characteristic curve, and the power allocation strategy is used as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding surface function are adjusted according to the state variable deviation in different operating conditions to form the final energy allocation optimization sequence.

2. The method according to claim 1, characterized in that, Based on vehicle power demand and energy flow constraints, a multi-objective optimization model for a hybrid energy storage system is constructed. A state-space equation is used to collaboratively model the power distribution between the supercapacitor and the power battery, resulting in an energy distribution sequence including: Acquire power demand data of the vehicle under real-time driving conditions, establish power demand characteristic curves including acceleration power, braking power and cruising power, and calculate the power distribution coefficient for each operating condition range based on the power demand characteristic curves. Combining the power demand characteristic curve and the power allocation coefficient, the charging and discharging power limits, state of charge variation ranges, and temperature constraints of the supercapacitor and power battery in each operating condition range are extracted to construct an energy flow constraint set. Based on the power allocation coefficient and the energy flow constraint set, a multi-objective optimization model is constructed with the goal of maximizing system energy utilization efficiency and minimizing power loss. The dynamic weighting coefficient of the optimization objective in each operating condition interval is calculated by an adaptive weight adjustment algorithm. Substituting the dynamic weighting coefficients and the energy flow constraint set into the state-space equation of the hybrid energy storage system, the charging and discharging power of the supercapacitor and the power battery are modeled collaboratively using the rolling time-domain prediction method. By solving the state-space equation, the energy allocation sequence that satisfies the multi-objective optimization model is obtained.

3. The method according to claim 2, characterized in that, Substituting the dynamic weighting coefficients and the energy flow constraint set into the state-space equation of the hybrid energy storage system, and using a rolling time-domain prediction method to collaboratively model the charging and discharging power of the supercapacitor and the power battery, the energy allocation sequence satisfying the multi-objective optimization model is obtained by solving the state-space equation, including: Substituting the dynamic weighting coefficients and energy flow constraints into the state space equations of the hybrid energy storage system, a set of state equations is constructed with supercapacitor voltage, power battery current and state of charge as state variables and charging and discharging power as control variables. A rolling time-domain prediction framework is constructed based on the state equation set. A power distribution prediction sequence for supercapacitors and power batteries is established in each prediction interval. According to the evolution law of the state equation set, the end state of each prediction interval is set as the initial state of the next prediction interval. The state response trajectories of the supercapacitor and the power battery are calculated using the state equation set and the power allocation prediction sequence. The state response trajectories are combined with the dynamic weighting coefficients to correct the power allocation prediction sequence, thereby generating a corrected prediction sequence that satisfies the energy flow constraint set. Substituting the corrected prediction sequence into the state equations and solving them yields an energy allocation sequence that satisfies the multi-objective optimization requirements.

4. The method according to claim 1, characterized in that, Based on the energy allocation sequence, robust constraints are applied according to the sliding mode surface function, and an adaptive reaching law is introduced to dynamically adjust the switching gain, generating a power allocation command to suppress chattering, including: Based on the energy allocation sequence, a sliding mode surface function is constructed, and the power allocation deviation and power change rate of the hybrid energy storage system are set as state parameters. The boundary conditions of the robustness constraint are determined according to the state parameters. An adaptive reaching law is established using the boundary conditions, and the rate and magnitude of change of the power allocation deviation are used as adaptive parameters to dynamically calculate the adjustment amount of the switching gain. The adaptive reaching law and the adjustment amount of the switching gain are substituted into the sliding surface function, and a power allocation command to suppress chattering is generated based on the real-time changes of the state parameters.

5. The method according to claim 1, characterized in that, Based on the power allocation command, a rolling expansion is performed within the prediction time domain. The future response trajectories of the supercapacitor and the power battery are predicted using the energy allocation sequence within the prediction time domain, resulting in a power allocation strategy including: The power allocation command is rolled out in the prediction time domain to generate a command sequence containing multiple control time points. Based on the dynamic response characteristics of the hybrid energy storage system, the command sequence is divided into multiple sub-intervals according to the operating conditions, and a power allocation benchmark is established in each sub-interval. Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. The piecewise linearization method is used to calculate the energy allocation state in each sub-interval. The state transition at the boundary of the sub-interval is used as the critical transition constraint. A state transition equation considering the switching of operating conditions is established. The future response trajectories of the supercapacitor and the power battery are predicted by using the energy distribution state and the state transition equation. The future response trajectory is used to characterize the dynamic characteristics of the energy storage unit during the switching process under different operating conditions, and a power distribution strategy for the hybrid energy storage system is generated.

6. The method according to claim 5, characterized in that, Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. A piecewise linearization method is used to calculate the energy allocation state within each sub-interval. State transitions at the boundaries of sub-intervals are used as critical transition constraints. State transition equations considering operating condition switching are established, including: Based on the power allocation benchmark, an energy allocation sequence in the prediction time domain is constructed. The prediction time domain is divided into acceleration power range, braking power range and steady power range. According to the power change characteristics of each range, a corresponding energy allocation sub-sequence is established. Substitute the energy allocation subsequence into the piecewise linearization calculation framework, and use the rate of change of the state of charge of the supercapacitor and the rate of change of the output power of the power battery as linearization coefficients to establish a piecewise state equation characterizing the dynamic characteristics of charging and discharging power. The state transition characteristics at the boundaries of the sub-intervals are analyzed using the piecewise state equations. The energy transfer efficiency and power fluctuation amplitude of the supercapacitor and the power battery during the switching of operating conditions are calculated. The energy transfer efficiency and the power fluctuation amplitude are combined to construct critical transfer constraints. The state transition equations of the hybrid energy storage system are established based on the critical transfer constraints.

7. The method according to claim 1, characterized in that, The sliding mode surface function is constructed as a piecewise continuous variable characteristic curve, and the power allocation strategy is used as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding mode surface function are adjusted according to the state variable deviation in different operating condition ranges to form the final energy allocation optimization sequence, including: The sliding surface function is constructed as a piecewise continuous variable characteristic curve. The power allocation strategy is used as the control input variable. The piecewise interval of the sliding surface function is established by the power change rate and energy transfer efficiency of the power allocation strategy. The initial switching boundary of the sliding surface is set based on the piecewise interval. Based on the segmented interval division results, state variables are collected in real time during the charging and discharging process of the supercapacitor and the power battery. The state-of-charge deviation of the supercapacitor, the output power deviation of the power battery, and the energy transfer efficiency of the energy storage system are used as evaluation indicators. The state variable deviation in each operating condition interval is calculated, and the state variable deviation is mapped to the segmented interval. The switching characteristic parameters of the sliding surface function in each segmented interval are calculated using the state variable deviation. The switching characteristic parameters are combined with the initial switching boundary to establish an adaptive switching law. The dynamic characteristics of the sliding surface function in different segmented intervals are adjusted by the adaptive switching law. The power allocation strategy is modified based on the dynamically adjusted sliding surface function, and an energy allocation optimization sequence considering the robustness constraints of the charging and discharging process is constructed based on the modification results.

8. A hybrid energy storage control system for supercapacitors and power batteries in new energy vehicles, used to implement the method described in any one of claims 1-7, characterized in that, include: The first unit is used to obtain the current operating parameters of the vehicle, the state variables of the supercapacitor and the power battery. Based on the vehicle's power demand and energy flow constraints, a multi-objective optimization model of the hybrid energy storage system is constructed. The power distribution of the supercapacitor and the power battery is modeled collaboratively through state-space equations to obtain the energy distribution sequence. The second unit is used to perform robust constraints based on the energy allocation sequence and the sliding surface function, and to introduce an adaptive reaching law to dynamically adjust the switching gain, thereby generating a power allocation command to suppress chattering. The third unit is used to perform rolling expansion in the prediction time domain based on the power allocation command, and to predict the future response trajectories of the supercapacitor and the power battery through the energy allocation sequence in the prediction time domain, so as to obtain the power allocation strategy. The fourth unit is used to construct the sliding surface function into a piecewise continuous variable characteristic curve, and take the power allocation strategy as input. During the charging and discharging process of the supercapacitor and the power battery, the switching characteristics of the sliding surface function are adjusted according to the state variable deviation in different operating conditions to form the final energy allocation optimization sequence.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images