Energy Management Method for Electric Vehicle Hybrid Energy Storage System Integrating Hippo Particle Swarm Optimization Algorithm (LSTM)

Abstract

This invention discloses an energy management method for electric vehicle hybrid energy storage systems that integrates the Hippodrome Particle Swarm Optimization (LSTM) algorithm. A multi-layer LSTM neural network controller is constructed to learn driving condition characteristics and minimize battery stress. A fractional-order integral derivative (FOID) controller is built to output a power correction component, which is then decomposed into a battery power reference component and a supercapacitor power reference component. Using an improved Hippodrome Particle Swarm Optimization (LSTM) algorithm, the LSTM weights and FOID controller parameters are jointly optimized through a multi-objective optimization function (including battery root mean square current and voltage error). This optimizes the weights and biases of the neural network controller, enabling real-time control of power distribution between the battery and supercapacitor in the electric vehicle hybrid energy storage system. A dynamic weighting mechanism adaptively corrects the LSTM power distribution results, controlling bus voltage fluctuations within ±3%. Finally, the power reference value is dynamically weighted and synthesized, significantly reducing the battery peak current.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle energy management technology, specifically to an energy management method for electric vehicle hybrid energy storage systems that integrates the Hippodrome Particle Swarm Optimization (LSTM) algorithm. It is particularly suitable for battery-supercapacitor hybrid energy storage systems, and achieves multi-scale power allocation through the synergistic optimization of fractional-order control and LSTM. Background Technology

[0002] The development of electric vehicles (EVs) is crucial for reducing carbon emissions in the transportation sector. However, improving their economic efficiency and driving range requires addressing challenges such as high battery costs and limited battery life. Batteries withstand high currents and rapid charge-discharge cycles during high-power transients caused by vehicle acceleration and braking. Combining batteries with supercapacitors to form a hybrid energy storage system, utilizing supercapacitors to handle power peaks while the battery provides continuous energy, can effectively reduce battery stress and extend its lifespan, representing a significant strategy for improving EV performance.

[0003] The energy management system is responsible for intelligently distributing power between batteries and supercapacitors. Offline optimization strategies can provide a design basis for online strategies, such as rule-based or learning-based strategies. Long Short-Term Memory (LSTM) networks have shown potential in energy management, being used to predict energy consumption or learn optimal control strategies. Existing PID controllers suffer from rigid frequency domain response and insufficient disturbance rejection capabilities when suppressing voltage fluctuations.

[0004] This invention introduces a fractional-order integral derivative (FOID) controller, which uses parameters... To achieve continuous switching of PI / PD characteristics and improve transient response and robustness, there is an urgent need to develop an energy management method and system for hybrid energy storage systems that can effectively reduce battery stress, extend battery life, ensure DC bus voltage stability, and is easy to optimize and deploy. Summary of the Invention

[0005] Purpose of the Invention: This invention aims to overcome the shortcomings of existing technologies and provide an energy management method for electric vehicle hybrid energy storage systems that integrates the Hippodrome Particle Swarm Optimization (LSTM) algorithm. It combines LSTM with a Fractional Integral Derivative (FOID) controller and utilizes a dynamic weighting mechanism. The system adaptively balances battery stress minimization and voltage stability objectives. By training a neural network with an improved fusion particle swarm optimization algorithm and optimizing FOID controller parameters, it achieves optimal (or suboptimal) power distribution between the battery and supercapacitor, significantly reducing the battery's root mean square current and peak current, alleviating battery stress, and extending its service life.

[0006] Technical Solution: This invention discloses an energy management method for electric vehicle hybrid energy storage systems that integrates the Hippodrome Particle Swarm Optimization (LSTM) algorithm, comprising the following steps:

[0007] Step 1: Obtain electric vehicle driving cycle data, including vehicle speed and load power ;

[0008] Step 2: Construct a multi-layer LSTM neural network controller to process the state data from historical moments at high speed. Load power Supercapacitor voltage DC bus voltage error As input, the time-dependent features of driving conditions are learned, and the multi-layer LSTM neural network controller learns the mapping from the input state to power allocation, outputting the battery power reference component. and supercapacitor power reference component ;

[0009] Step 3: Construct a fractional-order integral derivative FOID controller. The FOID controller uses the DC bus reference voltage. Compared with the measured voltage error As input, output power correction component ,Will Decomposed into battery power reference components and supercapacitor power reference component It is used to correct the power distribution results of the multilayer LSTM neural network controller and control the bus voltage fluctuation within ±3% of the reference value.

[0010] Step 4: Based on the optimization algorithm, a multi-objective joint optimization framework is adopted. The primary objective is to minimize the battery root mean square current, and the weights and biases of the multilayer LSTM neural network controller in Step 2 are optimized and trained. The secondary objective is to minimize the voltage error, and the FOID controller parameters are optimized accordingly. Optimization; the objective function is:

[0011] (1)

[0012] in, , This is a battery current time series, where T is the simulation time. The weighting coefficients for the primary and secondary objectives. , Let λ be the Laplace transform of the voltage error, where λ ∈ [0.1, 0.9].

[0013] Step 5: The power reference component obtained in Step 2 ( , ) and the power reference component of step 3 ( , Dynamic weighted summation yields the final battery power reference value. and supercapacitor power reference value The DC / DC converter inputs to batteries and supercapacitors.

[0014] Furthermore, the load power in step 1 The calculation method is as follows:

[0015] Multiple acceleration and constant speed deceleration intervals were set up, with acceleration intervals of 1~15s and 48~55s, constant speed intervals of 15~20s and 55~60s, and deceleration intervals of 20~30s and 60~65s, to simulate car driving and calculate the electrical load required for driving at each time point. The details are as follows:

[0016] (2)

[0017] in, As the driving force, This represents the tangent component of gravity (assuming the initial slope is 0). For air resistance, For rolling resistance, The mechanical power at the wheel, air density, This is the drag coefficient. The vehicle's frontal area. For electromechanical conversion power, For driving speed, This is the tire deformation dissipation coefficient.

[0018] Furthermore, step 4 utilizes an improved fusion particle swarm optimization algorithm, as detailed below:

[0019] Step 4.1: Initialize PSO algorithm parameters, including the number of particles / individuals. Maximum number of iterations Inertial weight u, initial inertia Final inertia value Set the catastrophic perturbation probability and the proportion of elite particles retained in the HO algorithm, with the upper and lower bounds of the particle position in each dimension as follows: , ;

[0020] Step 4.2: The position of each particle / individual represents a set of weights and biases of the multilayer LSTM neural network controller relative to FOID. Hippo particle swarm fusion formula initialization of particle swarm:

[0021] (3)

[0022] Where r is a random number between 0 and 1, and rand is a random number;

[0023] The nonlinear decaying inertia weight is:

[0024] (4)

[0025] The formula for updating the position of each particle / individual is:

[0026] (5)

[0027] in, For the optimal value of the particle, The global optimal value. As an acceleration factor, Social recognition coefficient;

[0028] Step 4.3: Compare the fitness obtained from the electrical simulation with the fitness corresponding to the best position in the particle's history, and set the calculation benchmark. If the particle gets stuck, use the HO algorithm to perform a local search and perturb the non-elite particles. If the particle gets stuck multiple times, keep the elite particles and reset the rest.

[0029] In the HO algorithm, elite particles represent male hippos, and the growth mechanism for non-elite particles is improved by introducing a new convergence factor T.

[0030] (6)

[0031] (7)

[0032] in, This indicates the position after the HO algorithm performs a local search, t is the current iteration number, iter is the maximum iteration number, A is a random number between 0 and 1, and B is a random number between 1 and 2. When T > 0.9, the naive hippo tends to perform a fine search (local development) near the optimal solution, thus improving the algorithm's accuracy.

[0033] Step 4.4: For each particle / individual position, decode into a set of weights and biases, FOID parameters for a multilayer LSTM neural network controller. ;

[0034] Step 4.5: Perform electrical simulation using a set of weights and biases of a multilayer LSTM neural network controller under either an acceleration / deceleration cycle or a target driving cycle, and use the parameters from the decoded FOID. Obtain the Laplace transform of the voltage error in the electrical simulation;

[0035] Step 4.6: Calculate the objective function value as the fitness;

[0036] Step 4.7: Update the optimal position and global optimal position of the particle / individual.

[0037] Step 4.8: Repeat steps 4.2-4.7 until the maximum number of iterations is reached or convergence is achieved. Output a set of weights and biases of the multilayer LSTM neural network controller corresponding to the global optimal position, as well as the parameters of FOID. .

[0038] Furthermore, the multi-layer LSTM neural network controller adopts a multi-layer structure, including an input layer. , , and A 50-cell LSTM layer, a 40-cell fully connected layer, and an output layer, outputting the battery power reference component. and supercapacitor power reference component .

[0039] Furthermore, the multi-layer LSTM neural network controller learns the time-dependent features of driving conditions, specifically as follows:

[0040] LSTM solves the dependency problem through a gating mechanism. The core formula is as follows:

[0041] Input Gate:

[0042] Forgotten Gate:

[0043] Candidate memory: Ĉt = tanh(Wc · ( )

[0044] Memory update: = ⊙ + ⊙ Ĉt

[0045] Output gate:

[0046] Hidden layer output:

[0047] For the current input, Let be the hidden state from the previous time step, ⊙ represent element-wise multiplication, and σ and tanh are the sigmoid activation functions. These are the input gate, forget gate, candidate memory, and output gate weights, respectively. These are the input gate, forget gate, candidate memory, and output gate bias, respectively. To hide the number of units and the dimension of the input vector, a state persistence mechanism is used to retain historical state information to solve the temporal dependency problem. The state update equation is as follows:

[0048] = ⊙ + ⊙ Ĉ t , (8)

[0049] in: Output for the forget gate. For input gate output, Ĉ t As candidate cell states, the system is trained using LSTM layers and fully connected layers, and finally outputs the battery power reference component. and supercapacitor power reference component .

[0050] Furthermore, the FOID controller in step 3 is specifically as follows:

[0051] Fractional operators of FOID controller Discretization was achieved using a 5th-order Oustaloup filter, and integral operators were constructed accordingly. ,correspond Differential operators ,correspond :

[0052]

[0053]

[0054]

[0055]

[0056] in, The sampling period is For the recursive calculation of poles / zeros, K is the amplitude compensation gain, optimized through pole-zero matching. This is the low-frequency cutoff frequency. This is the high-frequency cutoff frequency;

[0057] Input DC bus voltage Compared with the measured voltage error ;

[0058] Output power correction component:

[0059] ;

[0060] Decomposed into battery power reference component and supercapacitor power reference component: ;

[0061] in, and Used to adjust the dominance of the PID controller in power distribution between the battery and the supercapacitor. For the Laplace transform of voltage error, This is the integral gain, used for the integral term of the error, and is used to eliminate steady-state error. The differential gain is used for the differential term of the error, and is used for instantaneous fluctuations. For the inverse Laplace transform operator, It is an intermediate order fractional calculus, with a value range of [0.1, 0.9].

[0062] Furthermore, the output component of the FOID controller ( and ) respectively through weight constant ( After scaling, it participates in the dynamic weighted summation:

[0063] (12)

[0064] Among them, dynamic weights The sensitivity coefficient k=5.0, when When the voltage is >2V, η(t) > 0.88, and FOID dominates voltage regulation.

[0065] Beneficial effects:

[0066] This invention uses a multi-layer LSTM neural network controller to learn and train the time-dependent features of driving conditions, learning the mapping from input states to power allocation, and determining the battery power reference component and supercapacitor power reference component under the LSTM neural network. A fractional-order integral derivative FOID controller is used to determine the output power correction component, correcting the power allocation result of the multi-layer LSTM neural network controller, and determining the battery power reference component and supercapacitor power reference component under the FOID controller. Finally, a dynamic weighting mechanism is used... The adaptive balance between minimizing battery stress and achieving voltage stability enables optimal (or suboptimal) power distribution between the battery and the supercapacitor, significantly reducing the battery's root mean square current and peak current, alleviating battery stress, and extending its service life. Attached Figure Description

[0067] Figure 1 For the two acceleration events of the electric vehicle;

[0068] Figure 2 Composed of DC bus current;

[0069] Figure 3 This is the DC bus power configuration;

[0070] Figure 4 Generate a power reference calculation flow for LSTM;

[0071] Figure 5 This is the overall flowchart of the present invention. Detailed Implementation

[0072] In order to achieve the purpose of this application and make the technical solution and advantages clearer, the following describes this application in further detail with reference to the accompanying drawings and embodiments.

[0073] This invention discloses an energy management method for electric vehicle hybrid energy storage systems incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm, comprising the following steps:

[0074] Step 1: Obtain electric vehicle driving cycle data, including vehicle speed and load power .

[0075] Multiple acceleration, constant speed, and deceleration intervals were set up, with acceleration intervals of 1~15s and 48~55s, constant speed intervals of 15~20s and 55~60s, and deceleration intervals of 20~30s and 60~65s, to simulate car driving and calculate the electrical load required for driving at each time point. The details are as follows:

[0076] (2)

[0077] in, As the driving force, This represents the tangent component of gravity (assuming a slope of 0). For air resistance, For rolling resistance, The mechanical power at the wheel, air density, This is the drag coefficient. The vehicle's frontal area. For electromechanical conversion power, For driving speed, This is the tire deformation dissipation coefficient.

[0078] Step 2: Construct a multi-layer LSTM neural network controller to process the state data from historical moments at high speed. Load power Supercapacitor voltage DC bus voltage error As input, the time-dependent features of driving conditions are learned, and the multi-layer LSTM neural network controller learns the mapping from the input state to power allocation, outputting the battery power reference component. and supercapacitor power reference component .

[0079] The multi-layer LSTM neural network controller employs a multi-layer structure, including an input layer. , , and A 50-cell LSTM layer, a 40-cell fully connected layer, and an output layer, outputting the battery power reference component. and supercapacitor power reference component .

[0080] The multi-layer LSTM neural network controller learns the time-dependent features of driving conditions, as detailed below:

[0081] LSTM solves the dependency problem through a gating mechanism. The core formula is as follows:

[0082] Input Gate:

[0083] Forgotten Gate:

[0084] Candidate memory: Ĉt = tanh(Wc · ( )

[0085] Memory update: = ⊙ + ⊙ Ĉt

[0086] Output gate:

[0087] Hidden layer output:

[0088] For the current input, Let represent the hidden state from the previous time step, ⊙ denotes element-wise multiplication, and σ and tanh are the sigmoid activation functions. These are the input gate, forget gate, candidate memory, and output gate weights, respectively. These are the input gate, forget gate, candidate memory, and output gate bias, respectively. To hide the number of units and the dimension of the input vector, a state persistence mechanism is used to retain historical state information to solve the temporal dependency problem. The state update equation is as follows:

[0089] = ⊙ + ⊙ Ĉ t , (8);

[0090] in: Output for the forget gate. For input gate output, Ĉ t As candidate cell states, the system is trained using LSTM layers and fully connected layers, and finally outputs the battery power reference component. and supercapacitor power reference component .

[0091] Step 3: Construct a fractional-order integral derivative FOID controller. The FOID controller uses the DC bus reference voltage. Compared with the measured voltage error As input, output power correction component ,Will Decomposed into battery power reference components and supercapacitor power reference component It is used to correct the power distribution results of the multilayer LSTM neural network controller and control the bus voltage fluctuation within ±3% of the reference value.

[0092] The FOID controller is as follows:

[0093] Fractional operators of FOID controller Discretization was achieved using a 5th-order Oustaloup filter, and integral operators were constructed accordingly. ,correspond Differential operators ,correspond :

[0094]

[0095]

[0096]

[0097]

[0098] in, The sampling period is For the recursive calculation of poles / zeros, K is the amplitude compensation gain, optimized through pole-zero matching. This is the low-frequency cutoff frequency. This is the high-frequency cutoff frequency;

[0099] Input DC bus voltage Compared with the measured voltage error ;

[0100] Output power correction component:

[0101] ;

[0102] Decomposed into battery power reference component and supercapacitor power reference component: ;

[0103] in, and Used to adjust the dominance of the PID controller in power distribution between the battery and the supercapacitor. For the Laplace transform of voltage error, This is the integral gain, used for the integral term of the error, and is used to eliminate steady-state error. The differential gain is used for the differential term of the error, and is used for instantaneous fluctuations. For the inverse Laplace transform operator, It is an intermediate order fractional calculus, with a value range of [0.1, 0.9].

[0104] Step 4: Based on the optimization algorithm, a multi-objective joint optimization framework is adopted. The primary objective is to minimize the battery root mean square current, and the weights and biases of the multilayer LSTM neural network controller in Step 2 are optimized and trained. The secondary objective is to minimize the voltage error, and the FOID controller parameters are optimized accordingly. Optimization; the objective function is:

[0105] (1)

[0106] in, , This is a battery current time series, where T is the simulation time. The weighting coefficients for the primary and secondary objectives. , Let λ be the Laplace transform of the voltage error, where λ ∈ [0.1, 0.9].

[0107] The improved fusion particle swarm optimization algorithm is used as follows:

[0108] Step 4.1: Initialize PSO algorithm parameters, including the number of particles / individuals. Maximum number of iterations Inertial weight u, initial inertia Final inertia value Set the catastrophic perturbation probability and the proportion of elite particles retained in the HO algorithm, with the upper and lower bounds of the particle position in each dimension as follows: , ;

[0109] Step 4.2: The position of each particle / individual represents a set of weights and biases of the multilayer LSTM neural network controller relative to FOID. Hippo particle swarm fusion formula initialization of particle swarm:

[0110] (3)

[0111] Where r is a random number between 0 and 1, and rand is a random number;

[0112] The nonlinear decaying inertia weight is:

[0113] (4)

[0114] The formula for updating the position of each particle / individual is:

[0115] (5)

[0116] in, For the optimal value of the particle, The global optimal value. As an acceleration factor, Social recognition coefficient;

[0117] Step 4.3: Compare the fitness obtained from the electrical simulation with the fitness corresponding to the best position in the particle's history, and set the calculation benchmark. If the particle gets stuck, use the HO algorithm to perform a local search and perturb the non-elite particles. If the particle gets stuck multiple times, keep the elite particles and reset the rest.

[0118] In the HO algorithm, elite particles represent male hippos, and the growth mechanism for non-elite particles is improved by introducing a new convergence factor T.

[0119] (6)

[0120] (7)

[0121] in, This indicates the position after the HO algorithm performs a local search, t is the current iteration number, iter is the maximum iteration number, A is a random number between 0 and 1, and B is a random number between 1 and 2. When T > 0.9, the naive hippo tends to perform a fine search (local development) near the optimal solution, thus improving the algorithm's accuracy.

[0122] Step 4.4: For each particle / individual position, decode into a set of weights and biases, FOID parameters for a multilayer LSTM neural network controller. ;

[0123] Step 4.5: Perform electrical simulation using a set of weights and biases of a multilayer LSTM neural network controller under either an acceleration / deceleration cycle or a target driving cycle, and use the parameters from the decoded FOID. Obtain the Laplace transform of the voltage error in the electrical simulation;

[0124] Step 4.6: Calculate the objective function value as the fitness;

[0125] Step 4.7: Update the optimal position and global optimal position of the particle / individual.

[0126] Step 4.8: Repeat steps 4.2-4.7 until the maximum number of iterations is reached or convergence is achieved. Output a set of weights and biases of the multilayer LSTM neural network controller corresponding to the global optimal position, as well as the parameters of FOID. .

[0127] Step 5: The power reference component obtained in Step 2 ( , ) and the power reference component of step 3 ( , Dynamic weighted summation yields the final battery power reference value. and supercapacitor power reference value The DC / DC converter inputs to batteries and supercapacitors.

[0128] The output components of the FOID controller ( and ) respectively through weight constant ( After scaling, it participates in the dynamic weighted summation:

[0129] (12)

[0130] Among them, dynamic weights The sensitivity coefficient k=5.0, when When the voltage is >2V, η(t) > 0.88, and FOID dominates voltage regulation.

[0131] The above method and system will be experimentally verified below:

[0132] Figure 1 To simulate the two acceleration and deceleration events of an electric vehicle, multiple acceleration and constant speed deceleration intervals were set: acceleration intervals of 1-15s and 48-55s, constant speed intervals of 15-20s and 55-60s, and deceleration intervals of 20-30s and 60-65s, mimicking the operation of an electric vehicle. The required electrical load was calculated using the gradient integration method in MATLAB. .

[0133] The required electrical load As a result, the Hippo Particle Swarm Optimization method uses the battery root mean square current as the primary objective and the voltage error as the secondary objective, as the input layer of the multi-layer LSTM for training. It optimizes the LSTM weights, bias, and FIOD controller parameters, and then uses the LSTM controller and FIOD controller weighted sum to generate the final power reference for the battery and supercapacitor.

[0134] Depend on Figure 3 It is evident that during rapid acceleration and deceleration, the FOID-dominated mode is activated through a dynamic weighting mechanism, enabling the supercapacitor to instantaneously bear 82% of the peak power while limiting the battery power to 9.5kW, effectively reducing the battery peak current to 105A. Combined with the timing prediction capability of LSTM, voltage fluctuations are stabilized within ±2.3%, and power switching overshoot is reduced to 4.3%, verifying the significant advantages of the fusion algorithm in the synergistic optimization of battery stress minimization and voltage stability.

[0135] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An energy management method for electric vehicle hybrid energy storage systems incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm, characterized in that, Includes the following steps: Step 1: Obtain electric vehicle driving cycle data, including vehicle speed and load power ; Step 2: Construct a multi-layer LSTM neural network controller to process the state data from historical moments at high speed. Load power Supercapacitor voltage DC bus voltage error As input, the time-dependent features of driving conditions are learned, and the multi-layer LSTM neural network controller learns the mapping from the input state to power allocation, outputting the battery power reference component. and supercapacitor power reference component ; Step 3: Construct a fractional-order integral derivative FOID controller. The FOID controller uses the DC bus reference voltage. Compared with the measured voltage error As input, the output power correction component ,Will Decomposed into battery power reference components and supercapacitor power reference component It is used to correct the power distribution results of the multilayer LSTM neural network controller and control the bus voltage fluctuation within ±3% of the reference value. Step 4: Based on the optimization algorithm, a multi-objective joint optimization framework is adopted. The primary objective is to minimize the battery root mean square current, and the weights and biases of the multilayer LSTM neural network controller in Step 2 are optimized and trained. The secondary objective is to minimize the voltage error, and the FOID controller parameters are optimized accordingly. Optimization; the multi-objective optimization function is: (1)； in, , This is a battery current time series, where T is the simulation time. The weighting coefficients for the primary and secondary objectives. , Let λ be the Laplace transform of the voltage error, where λ ∈ [0.1, 0.9]. Step 5: The power reference component obtained in Step 2 ( , ) and the power reference component of step 3 ( , Dynamic weighted summation yields the final battery power reference value. and supercapacitor power reference value DC / DC converters that input to batteries and supercapacitors; The output component of the FOID controller ( and ) respectively through weight constant ( After scaling, it participates in the dynamic weighted summation: (12)； Among them, dynamic weights The sensitivity coefficient k=5.0, when When the voltage is >2V, η(t) > 0.88, and FOID dominates voltage regulation.

2. The energy management method for a hybrid energy storage system for electric vehicles incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm as described in claim 1, characterized in that, Load power in step 1 The calculation method is as follows: Multiple acceleration, constant speed, and deceleration intervals were set up, with acceleration intervals of 1~15s and 48~55s, constant speed intervals of 15~20s and 55~60s, and deceleration intervals of 20~30s and 60~65s, to simulate car driving and calculate the electrical load required for driving at each time point. The details are as follows: (2)； in, As the driving force, Let be the tangent component of gravity, and assume the initial slope is 0. For air resistance, For rolling resistance, The mechanical power at the wheel, air density, This is the drag coefficient. The vehicle's frontal area. For electromechanical conversion power, For driving speed, This is the tire deformation dissipation coefficient.

3. The energy management method for a hybrid energy storage system for electric vehicles incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm as described in claim 1, characterized in that, Step 4 utilizes an improved fusion particle swarm optimization algorithm, as detailed below: Step 4.1: Initialize PSO algorithm parameters, including the number of particles / individuals. Maximum number of iterations Inertial weight u, initial inertia Final inertia value Set the catastrophic perturbation probability and the proportion of elite particles retained in the HO algorithm, with the upper and lower bounds of the particle position in each dimension as follows: , ; Step 4.2: The position of each particle / individual represents a set of weights and biases of the multilayer LSTM neural network controller relative to FOID. Hippo particle swarm fusion formula initialization of particle swarm: (3) ； Where r is a random number between 0 and 1, and rand is a random number; The nonlinear decaying inertia weight is: (4)； The formula for updating the position of each particle / individual is: (5) ； in, For the optimal value of the particle, This is the globally optimal value. As an acceleration factor, Social recognition coefficient; Step 4.3: Compare the fitness obtained from the electrical simulation with the fitness corresponding to the best position in the particle's history, and set the calculation benchmark. If the particle gets stuck, use the HO algorithm to perform a local search and perturb the non-elite particles. If the particle gets stuck multiple times, keep the elite particles and reset the rest. In the HO algorithm, elite particles represent male hippos, and the growth mechanism for non-elite particles is improved by introducing a new convergence factor T. （6）； （7）； in, This indicates the position after the HO algorithm performs a local search, t is the current iteration number, iter is the maximum iteration number, A is a random number between 0 and 1, and B is a random number between 1 and 2. When T > 0.9, the naive hippo tends to perform a fine search near the optimal solution, thus improving the algorithm's accuracy. Step 4.4: For each particle / individual position, decode into a set of weights and biases, FOID parameters for a multilayer LSTM neural network controller. ; Step 4.5: Perform electrical simulation using a set of weights and biases of a multilayer LSTM neural network controller under either an acceleration / deceleration cycle or a target driving cycle, and use the parameters from the decoded FOID. Obtain the Laplace transform of the voltage error in the electrical simulation; Step 4.6: Calculate the objective function value as the fitness; Step 4.7: Update the optimal position and global optimal position of the particle / individual. Step 4.8: Repeat steps 4.2-4.7 until the maximum number of iterations is reached or convergence is achieved. Output a set of weights and biases of the multilayer LSTM neural network controller corresponding to the global optimal position, as well as the parameters of FOID. .

4. The energy management method for a hybrid energy storage system for electric vehicles incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm as described in claim 1, characterized in that, The multi-layer LSTM neural network controller adopts a multi-layer structure, including an input layer. , , and A 50-cell LSTM layer, a 40-cell fully connected layer, and an output layer, outputting the battery power reference component. and supercapacitor power reference component .

5. The energy management method for a hybrid energy storage system for electric vehicles incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm as described in claim 4, characterized in that, The multi-layer LSTM neural network controller learns the time-dependent features of driving conditions, as detailed below: LSTM solves the dependency problem through a gating mechanism. The core formula is as follows: Input Gate: ; Forgotten Gate: ; Candidate memory: Ĉt = tanh(Wc · ( ); Memory update: = ⊙ + ⊙ Ĉt; Output gate: ; Hidden layer output: ; For the current input, Let be the hidden state from the previous time step, ⊙ represent element-wise multiplication, and σ and tanh are the sigmoid activation functions. These are the input gate, forget gate, candidate memory, and output gate weights, respectively. These are the input gate, forget gate, candidate memory, and output gate bias, respectively. To hide the number of units and the dimension of the input vector, a state persistence mechanism is used to retain historical state information to solve the temporal dependency problem. The state update equation is as follows: = ⊙ + ⊙ C t , (8); in: Output for the forget gate. For input gate output, Ĉ t As candidate cell states, the system is trained using LSTM layers and fully connected layers, and finally outputs the battery power reference component. and supercapacitor power reference component .

6. The energy management method for a hybrid energy storage system for electric vehicles incorporating the Hippodrome Particle Swarm Optimization (LSTM) algorithm as described in claim 1, characterized in that, The FOID controller in step 3 is specifically as follows: Fractional operators of FOID controller Discretization was achieved using a 5th-order Oustaloup filter, and integral operators were constructed accordingly. ,correspond Differential operators ,correspond : ； ； ； ； in, The sampling period is For the recursive calculation of poles / zeros, K is the amplitude compensation gain, optimized through pole-zero matching. This is the low-frequency cutoff frequency. This is the high-frequency cutoff frequency; Input DC bus voltage Compared with the measured voltage error ; Output power correction component: ； Decomposed into battery power reference component and supercapacitor power reference component: ; in, and Used to adjust the dominance of the PID controller in power distribution between the battery and the supercapacitor. For the Laplace transform of voltage error, This is the integral gain, used for the integral term of the error, and is used to eliminate steady-state error. The differential gain is used for the differential term of the error and is used to suppress instantaneous fluctuations. For the inverse Laplace transform operator, It is an intermediate order fractional calculus, with a value range of [0.1, 0.9].

Images