Fuel cell hybrid electric vehicle energy management method considering energy recovery efficiency

Abstract

The application discloses a fuel cell hybrid electric vehicle energy management method considering energy recovery efficiency, and belongs to the field of fuel cell hybrid electric vehicle control. The application aims at improving the rule-based energy management strategy to solve the problem that the rule-based energy management strategy has energy saving space. The application needs to calculate the required power, draw the optimal output power of the power battery generated by the PMP algorithm into a graph, obtain the working mode data under different power and SOC, divide the power battery output power and power battery SOC relation graph into different modes through the clustering method, and distribute the output power of the fuel cell and the power battery according to the set distribution scheme according to the value of the required power, so as to obtain the output power of the fuel cell and the power battery. The application uses the Pontryagin maximum principle to guide the rule-based energy management strategy, so that the optimization effect of the improved energy management strategy is close to the optimal effect, and the braking energy recovery rate is expressed through the rational fraction fitting method, which is more in line with the actual energy recovery condition.

Description

Technical Field

[0001] This invention belongs to the field of fuel cell hybrid electric vehicle control. Background Technology

[0002] With the rapid development of the global economy, the number of cars on the road has increased rapidly, leading to a surge in our energy demand. Currently, energy consumption is primarily based on non-renewable energy sources such as oil, natural gas, and coal. The large-scale consumption of these energy sources leads to resource shortages, environmental pollution, and the greenhouse effect. To reduce the consumption of non-renewable resources and protect our environment, developing new renewable energy sources has become an urgent problem for all countries, and new energy vehicles have become a key research focus.

[0003] Fuel cell hybrid electric vehicles (FCEVs), with fuel cells as the primary power source and batteries as the auxiliary power source, represent a new direction that countries are actively exploring. Fuel cells convert the chemical energy from the reaction of hydrogen and oxygen into electrical energy, producing only water as a pollution-free byproduct, making them environmentally friendly. While fuel cells have high power density, their dynamic performance is poor and their response speed is slow. To compensate for these shortcomings, batteries are typically used as an auxiliary power source. Batteries have a faster response speed; when the load changes significantly, batteries can provide power that fuel cells cannot provide in a timely manner, thus meeting the vehicle's power demands. Furthermore, when the vehicle is braking, the electric motor can convert some of its kinetic energy into electrical energy and store it in the batteries, reducing energy waste and improving energy utilization.

[0004] Fuel cell hybrid electric vehicles have multiple power sources with different characteristics and complex energy flow, thus requiring the design of energy management strategies to improve the overall performance and fuel economy of the vehicle.

[0005] Patent CN113401009A discloses an energy management strategy that uses a fuel cell hybrid vehicle as the controlled object, allocating the output power of the fuel cell and the power battery according to the state of charge (SOC) of the power battery and the required power. The method used in this patent to allocate the output power of the fuel cell and the power battery is based on experience and cannot fully explore the potential for energy optimization, leaving room for further optimization. Furthermore, this patent does not involve research on energy recovery.

[0006] Patent CN113071372A discloses a semi-following power energy management strategy, which allocates the output power of the fuel cell and the power battery based on the state of charge (SOC) of the power battery and the motor power. This strategy divides the power into different ranges based on the power battery's charge compensation power, and different allocation schemes are applied when the motor power falls within different ranges. However, this strategy also fails to fully explore energy-saving potential, and the patent still does not address research related to energy recovery, leaving room for improvement.

[0007] In summary, the main problems with rule-based energy management strategies for fuel cells are as follows: 1. Rule-based energy management strategies are primarily designed based on the experience of the policymakers and are closely linked to personal habits and understanding of energy management. This type of strategy cannot fully tap into energy-saving potential and has significant room for improvement.

[0008] 2. Existing energy management strategies often assume that energy recovery is ideal, but the efficiency of energy recovery changes with the vehicle's operating speed. Existing data provides some information on energy recovery, but it is difficult to establish a mathematical formula to calculate the value of the recovery efficiency.

[0009] Rule-based energy management strategies rely heavily on the experience of the policymakers, depending solely on their understanding of energy allocation and the power limitations of fuel cells and batteries. This approach falls short of optimal energy management and has room for improvement. Furthermore, there is currently a lack of detailed research on energy recovery efficiency. Summary of the Invention

[0010] The purpose of this invention is to address the issue of energy-saving potential in rule-based energy management strategies by providing an improved energy management method for fuel cell hybrid electric vehicles that considers energy recovery efficiency.

[0011] The steps of this invention are: S1. By analyzing the forces, the magnitude of the traction force is determined, and thus the required power of the motor is calculated: (1) (2) (3) (4) (5) in It's about the quality of the car. It is gravitational acceleration. It's the slope of the road surface. It is the frontal area of ​​the car. It is the rolling friction coefficient of a car tire. It is a rotational mass parameter. It is the air drag coefficient. It is air density; S2. Based on the data, using the least squares method to fit the relationship between braking energy recovery efficiency and speed, fit a fourth-degree polynomial function: (6) in It refers to the braking energy recovery efficiency; Suppose the function to be solved is in the form shown below: (7) in There are six coefficients to be determined. It is a vector composed of parameters; there are 26 sets of sample values ​​respectively. and ;get (8) In the above formula It is a constant matrix. It is a constant vector. The coefficient vector is to be determined. We set it to 1, thus giving the fraction a unique coefficient vector, as shown below: (9) The coefficient vector to be determined is obtained using the least squares method: (10) In the formula of Yes, it is a generalized inverse matrix; Finally, the coefficients are obtained, and when substituted into equation (7), the final braking energy recovery efficiency is obtained. ; S3. When the power demand of the car is less than 0, the car is in a braking state, and the power demand is as follows: (13) in This is the required power after adding the recovery factor. It is the braking energy recovery rate; The formula for calculating the State of Charge (SOC) of a power battery is shown below: (14) S4. The specific steps of the workflow are as follows: (1) Calculate the required power of the car motor based on the current speed of the car and the road conditions, determine the sign of the required power, and decide whether to add braking energy recovery rate. (2) The optimal power distribution method of the power battery is obtained by using the Pontryagin maximum; S5. Constraints include constraints on the output power of the fuel cell and the power battery, equality constraints between the output power of the fuel cell and the output power of the power battery, and constraints on the SOC of the power battery. The constraint conditions are in the following form: (15) in That is the minimum output power of the power battery. That is the maximum output power of the power battery. That is the minimum output power of a fuel cell. That is the maximum output power of the fuel cell. It is the minimum SOC of the power battery. It is the maximum value of the SOC of the power battery. It is the output power of the fuel cell; S6. The objective function is shown below: (16) in It is a function of hydrogen consumption rate. It is the initial moment. It is the final value time. This is the total hydrogen consumption; S7. The Hamiltonian function is shown in the following formula: (17) in It is a costate variable; The output power of the fuel cell when the Hamiltonian function reaches its extreme value is the optimal output power. At this point, the output power of the corresponding power battery can be calculated, which is the optimal output power of the power battery. S8. Clustering using the K-means method: (1) The sample Divide into s clusters, and randomly select s samples as the centers of the clusters, denoted as s. ; (2) Calculate the Euclidean distance from each point in the sample to the center point. For a specific sample In other words, it is at the center point The Euclidean distance is as follows: (18) (3) Based on the obtained Euclidean distance The value determines which cluster the sample belongs to, and the sample is assigned to the cluster with the nearest centroid. (4) The new centers of each cluster after the transformation are obtained by finding the minimum value of the loss function, which is shown below: (19) in To represent different clusters, It is the number of different clusters. It is the total number of samples. It is the overall loss function. It is the number of samples in each cluster, satisfying Find the minimum value of the loss function and expand it as follows: (20) Simplifying the above equation, each summation can be expressed in turn using... express, If is the loss function for each cluster, then the above equation can be simplified to: (twenty one) by For example, For the center point of the cluster Find the partial derivative and set it to 0, as shown in the following equation: (twenty two) Therefore, we can find Value: (twenty three) Similarly, it can be done by... Find the partial derivatives to determine the center point of each cluster. The value; As can be seen from the above, when the samples contained in different clusters do not change, the center point of each cluster also does not change. When the center point of the calculated cluster changes, repeat steps (2) to (4). When the value of the center point of each cluster does not change, the resulting clusters are the final clustering results. By classifying the obtained power battery output power and SOC data through the above steps (1) to (4), the final pattern diagram is obtained; S9. In the above pattern diagram, find the output power of the power battery corresponding to the current SOC based on the required power, calculate the output power of the fuel cell using the required power, and calculate the SOC value of the power battery at the next moment using the above two power values.

[0012] This invention uses the Pontryagin maximum principle to guide a rule-based energy management strategy, making the optimization effect of the improved energy management strategy close to the optimal effect. Furthermore, it expresses the braking energy recovery rate through rational fraction fitting, which is more consistent with the actual energy recovery situation. Attached Figure Description

[0013] Figure 1 This is a schematic diagram illustrating the calculation of the power demand of a car. Figure 2 This is a diagram of braking energy recovery efficiency fitted using the least squares method; Figure 3 This is a diagram of braking energy recovery efficiency fitted by a rational fraction; Figure 4 This is the equivalent circuit diagram of the SOC of a power battery; Figure 5 This is a schematic diagram of the mode division based on the optimal output power of the power battery obtained from the Pontryagin maximum; Figure 6 This is a flowchart for optimizing energy management strategies; Figure 7 This is a comparison chart of hydrogen consumption between the rule-based energy management strategy and the improved energy management strategy of this invention. Detailed Implementation

[0014] Rule-based energy management strategies are formulated based on the experience of the strategist. While these strategies are simple in structure, they often deviate significantly from optimal allocation methods due to their reliance on personal experience. This invention uses the Pontryagin maximum principle to guide rule-based energy management strategies, bringing them closer to the optimal allocation method.

[0015] The technical solution of the present invention is as follows: 1. Power Requirement Calculation: When a fuel cell vehicle is in normal operation, the motor needs to output a certain amount of traction to ensure the vehicle can operate normally. Therefore, to ensure the vehicle can run normally, the output power of the fuel cell and power battery must meet the power requirement of the motor. Thus, the power requirement needs to be calculated.

[0016] Pontryagin's maximum principle (PMP) is a method in optimal control theory that finds the optimal control input to move the system from its initial state to its ideal terminal state, given constraints on input control or system state. This method can handle optimization problems. Our optimization objective is to minimize the hydrogen consumption of the vehicle while satisfying the power constraints of the fuel cell and the power battery, as well as the terminal constraints. The control variable is the output power of the power battery, and the state variable is the state of charge (SOC) of the power battery. This problem can be solved using the PMP algorithm to obtain the optimal output power of the power battery under different power demands and power battery SOCs. The optimal output power of the power battery generated by the PMP algorithm is plotted to obtain data on the operating modes under different power and SOCs. Clustering is then used to divide the relationship between the power battery output power and the power battery SOC into different patterns.

[0017] After allocating the output power of the fuel cell and the power battery according to the pre-set distribution scheme based on the required power value, the output power of the fuel cell and the power battery is obtained. The SOC value of the power battery at the next moment is then calculated using the allocated power battery power and the SOC value at that moment.

[0018] The following is a detailed description of the rule-based energy management strategy for incorporating the Pontryagin maximum in this invention, with reference to the accompanying drawings: This invention improves upon rule-based energy management strategies by incorporating Pontryagin maximum and regenerative braking efficiency. First, it is necessary to derive expressions for the motor's required power and the battery's state of charge (SOC). Figure 1 This is a schematic diagram for calculating the required power. The magnitude of the traction force is determined through force analysis, and then the required power of the motor is calculated. The formula is as follows: (1) (2) (3) (4) (5) in It's about the quality of the car. It is gravitational acceleration. It's the slope of the road surface. It is the frontal area of ​​the car. It is the rolling friction coefficient of a car tire. It is a rotational mass parameter. It is the air drag coefficient. It refers to air density.

[0019] Power batteries play a role in recovering braking energy, but in reality, not all braking energy can be recovered, leading to the issue of braking energy recovery rate. The braking energy recovery rate function in this invention is obtained through identification, with the least squares method being the most commonly used fitting method.

[0020] Figure 2 The curve representing the relationship between regenerative braking efficiency and speed is obtained by fitting data using the least squares method. In this identification method, the fitting function is a fourth-degree polynomial function, and its expression is as follows: (6) in It is the regenerative braking efficiency, from Figure 2 It can be seen that the curves under this identification method do not conform to the trend of the data, and the fitting accuracy cannot meet the needs of energy recovery calculation. Therefore, other fitting methods need to be used to fit the data.

[0021] The data trends indicate that rational fraction fitting can be used. While more complex than polynomial fitting, rational fraction fitting is more efficient in certain aspects. For singularities and peaks, rational fraction fitting yields results with lower order and is simpler. The principle of rational fraction fitting is explained below. Let the function to be solved be as follows: (7) in There are six coefficients to be determined. It is a vector composed of parameters. Currently, there are 26 sets of data, resulting in 26 sample values, which are... and .

[0022] Multiplying the denominator by the fraction on the left, we get the following equation: (8) In the formula is constant matrix, It is a constant vector. The coefficient vector to be determined is... Set it to 1, so that the fraction has a unique coefficient vector, as shown below: (9).

[0023] The coefficient vector to be determined is obtained using the least squares method, as shown in the following formula: (10) In the formula yes The generalized inverse matrix.

[0024] The final coefficients obtained through the above calculations are shown below. (11).

[0025] The results are as follows Figure 3 As shown, the expression is as follows: (12).

[0026] pass Figure 2 and Figure 3 The comparison shows that the fractional function has a better fitting effect; therefore, the fractional function is used as the fitting function. It can be seen that the braking energy recovery rate is related to the vehicle's speed; different speeds correspond to different recovery rates.

[0027] When the power demand of a car is less than 0, the car is in a braking state, and the power demand is shown in the following formula: (13) in This is the required power after adding the recovery factor. It is the braking energy recovery rate.

[0028] Figure 4 This is the equivalent circuit diagram of a power battery. The formula for calculating the SOC of a power battery is shown below: (14).

[0029] Figure 6 The workflow of this invention is as follows, and the specific implementation steps are as follows: 1. Calculate the required power of the car motor based on the car's current speed and road conditions, determine whether the required power is positive or negative, and thus decide whether to add regenerative braking.

[0030] 2. The optimal power distribution method for the power battery is obtained through Pontryagin maximum. When using Pontryagin maximum, it is necessary to determine the constraints, control variables, state variables, objective function, and Hamiltonian function. First, the state of charge (SOC) of the power battery is selected as the state variable, and the output power of the fuel cell is selected as the control variable. The constraints include constraints on the output power of the fuel cell and the power battery, equality constraints between the fuel cell output power and the power battery output power, and constraints on the SOC of the power battery.

[0031] The constraint conditions are in the following form: (15) in That is the minimum output power of the power battery. That is the maximum output power of the power battery. That is the minimum output power of a fuel cell. That is the maximum output power of the fuel cell. It is the minimum SOC of the power battery. It is the maximum value of the SOC of the power battery. It is the output power of the fuel cell.

[0032] The objective function is as follows: (16) in It is a function of hydrogen consumption rate. It is the initial moment. It is the final value time. This is the total hydrogen consumption.

[0033] The Hamiltonian function is shown in the following equation: (17) in These are costate variables. The fuel cell output power when the Hamiltonian function reaches its extreme value is the optimal output power. At this point, the corresponding output power of the power battery can be calculated, i.e., the optimal output power of the power battery. Then, clustering is used to divide the existing data between the optimal output power and SOC into different patterns, such as... Figure 5 As shown. This invention employs the K-means clustering algorithm. K-means is a partitioning clustering method that uses distance as the basis for similarity partitioning; samples that are close together belong to the same group, while samples that are far apart are divided into different groups.

[0034] The process is as follows: 1) Sample Divide into s clusters, and randomly select s samples as the centers of the clusters, denoted as s. .

[0035] 2) Calculate the Euclidean distance from each point in the sample to the center point. For a specific sample... In other words, it is at the center point The formula for calculating the Euclidean distance is as follows: (18).

[0036] 3) Based on the obtained Euclidean distance The value determines which cluster the sample belongs to, and the sample is assigned to the cluster with the nearest center point.

[0037] 4) The termination condition of the K-means clustering algorithm is whether the centers of each cluster have changed. When they have not changed, the resulting clusters are the ones we want. The new centers of each cluster after the change are obtained by finding the minimum value of the loss function. The loss function is shown below: (19) in To represent different clusters, It is the number of different clusters. It is the total number of samples. It is the overall loss function. It is the number of samples in each cluster, satisfying .

[0038] Find the minimum value of the loss function and expand it as follows: (20).

[0039] Simplifying the above equation, each summation can be expressed in turn using... express, If is the loss function for each cluster, then the above equation can be simplified to: (twenty one).

[0040] Maximum likelihood estimation is used to determine the value of the centroid of each cluster at which the loss function reaches its minimum, and this is taken as... example, For the center point of the cluster Find the partial derivative and set it to 0, as shown in the following equation: (twenty two).

[0041] Therefore, we can find The values ​​are shown below: (twenty three).

[0042] Similarly, it can be done by... Find the partial derivatives to determine the center point of each cluster. The value of . Through the above calculation process, it can be seen that when the samples contained in different clusters do not change, the center point of each cluster also does not change. When the calculated center point of a cluster changes, repeat steps 2) to 4). When the value of the center point of each cluster does not change, the resulting clusters are the final clustering results. By classifying the obtained power battery output power and SOC data through steps 1) to 4), the final pattern diagram is obtained, as shown below. Figure 5 As shown.

[0043] 3. Based on the power demand in the above diagram, find the output power of the power battery corresponding to the current SOC. Then, calculate the output power of the fuel cell using the power demand. Finally, calculate the SOC value of the power battery at the next moment using these two power values. Thus, the power that the fuel cell and power battery should output is obtained. The output power signal sequence of the fuel cell and power battery is sent to the actuator, causing the fuel cell and power battery to operate according to the algorithm results and output the corresponding power.

[0044] Figure 7 The graph shows a comparison of hydrogen consumption between the rule-based energy management strategy and the improved energy management strategy of this invention. As can be seen from the graph, compared with the rule-based energy management strategy, the improved energy management strategy of this invention consumes less hydrogen under three different operating conditions. Therefore, the improved energy management strategy has better performance.

[0045] This invention uses a combination of Pontryagin maximum and K-means clustering analysis to guide rule-based energy management strategies and improves these strategies. It also uses the fractional function method to fit the recycling efficiency and incorporates it into the battery model during vehicle deceleration or braking, enabling the battery model to more accurately reflect real-life situations and more precisely calculate the battery's SOC value during operation.

[0046] The design of an energy management strategy and design process for a fuel cell hybrid electric vehicle incorporating the Pontryagin maximum was presented; a battery SOC model considering energy recovery efficiency was constructed.

[0047] Simulation process This invention uses three different operating conditions—low-speed SC03, high-speed US06, and urban suburban EUDC—to collect data from various driving scenarios. These three conditions encompass low-speed driving, high-speed driving, and frequent start-stop cycles, meeting the simulation requirements and verifying the algorithm's reliability under different driving conditions. Under these three conditions, an improved energy management strategy and a rule-based energy management strategy are used to allocate the output power of the fuel cell and the power battery, respectively, to obtain simulation results. The comparison results show that under SC03 operating conditions, the improved energy management strategy consumes 50.75 grams of hydrogen, while the rule-based energy management strategy consumes 110 grams, resulting in a 53% fuel saving. Under US06 operating conditions, the improved energy management strategy consumes 133.2 grams of hydrogen, while the rule-based strategy consumes 161.2 grams, resulting in a 17% fuel saving. Under EUDC operating conditions, the improved energy management strategy consumes 59.05 grams of hydrogen, while the rule-based strategy consumes 116.7 grams, resulting in a 49% fuel saving. Figure 7 As shown, the improved energy management strategy of this invention consumes less hydrogen under all three operating conditions, thus exhibiting better performance.

[0048] Summary of symbols involved in this invention: SOC (State of Charge); Pontryagin's Maximum Principle (PMP); Power required; The speed of the car; Transmission efficiency; The drag generated by the car's own weight; The rolling friction of a car's wheels; Air resistance; The acceleration resistance of a car; The initial state of charge (SOC) value of the power battery; The open-circuit voltage of the equivalent circuit of the power battery; The output power of the power battery; Open-circuit resistance of the equivalent circuit of the power battery; The capacity of the power battery; The state of charge (SOC) value of the power battery at the next moment; The quality of the car; Gravitational acceleration; The slope of the road surface; The frontal area of ​​a car; The coefficient of rolling friction of car tires; Rotational mass parameters; Air drag coefficient; air density; Power requirement after adding the recovery factor; Braking energy recovery efficiency; The coefficients to be determined; Parameter vector; The specific values ​​of the 26 parameter vectors; The recovery efficiency values ​​corresponding to the 26 groups and parameter vectors; Constant matrix; Constant vector; The coefficient vector to be determined; The generalized inverse matrix; The transpose of the matrix; Minimum output power of the power battery; The maximum output power of the power battery; Minimum output power of a fuel cell; The maximum output power of a fuel cell; Minimum SOC of power battery; The maximum SOC of the power battery; Fuel cell output power; Hydrogen consumption rate function; Initial moment; Total hydrogen consumption; Costate variables; Total sample size; s: number of distinct center points; The selected center point; Specific samples; The specific center point; The Euclidean distance between a sample point and a center point; Indicates different clusters; The number of different clusters; Total number of samples; The overall loss function; The number of samples in each cluster; The loss function for each cluster.

Claims

1. An energy management method for fuel cell hybrid electric vehicles considering energy recovery efficiency, characterized in that: The steps are as follows: S1. By analyzing the forces, the magnitude of the traction force is determined, and thus the required power of the motor is calculated: F i =mg sinθ (2) F j =mgμcosθ (3) Where m is the mass of the car, g is the acceleration due to gravity, θ is the slope of the road surface, A is the frontal area of ​​the car, μ is the rolling friction coefficient of the car tires, δ is the rotational mass parameter, and C... D ρ is the air drag coefficient, and ρ is the air density. S2. Based on the data, using the least squares method to fit the relationship between braking energy recovery efficiency and speed, fit a fourth-degree polynomial function: eta=-3.273×10 -8 V 4 +9.147×10 -6 V 3 -0.0009121V 2 +0.0401V+0.09005 (6) Where eta is the braking energy recovery efficiency; Suppose the function to be solved is in the form shown below: Where a0, a1, a2, b0, b1, b2 are six coefficients to be determined, and N = [n, n 2 [] is a vector composed of parameters; there are 26 sets of sample values ​​N(1), N(2)...N(26) and eta(1), eta(2),...eta(26); This yields... Xa=y (8) In the above equation, X is a constant matrix, y is a constant vector, and a is the coefficient vector to be determined. Setting b² to 1 ensures that the fraction has a unique coefficient vector, as shown below: a=[a0, a1, a2, b0, b1] (9) The coefficient vector to be determined is obtained using the least squares method: a=X + y=(X T X) -1 X T and (10) X in the formula + It is the generalized inverse matrix of X; Finally, the coefficient is obtained, and the coefficient is substituted into equation (7) to obtain the final braking energy recovery efficiency eta. S3. When the power demand of the car is less than 0, the car is in a braking state, and the power demand is as follows: P req =etaP e (13) Where P req It is the required power after adding the recovery factor, and eta is the braking energy recovery rate; The formula for calculating the State of Charge (SOC) of a power battery is shown below: S4. The specific steps of the workflow are as follows: (1) Calculate the required power of the car motor based on the current speed of the car and the road conditions, determine the sign of the required power, and decide whether to add a braking energy recovery rate. (2) The optimal power distribution method of the power battery is obtained by using the Pontryagin maximum; S5. Constraints include constraints on the output power of the fuel cell and the power battery, equality constraints between the output power of the fuel cell and the output power of the power battery, and constraints on the SOC of the power battery. The constraint conditions are in the following form: Where P batmin It is the minimum output power of the power battery, P batmax It is the maximum output power of the power battery, P fcmin This is the minimum output power of a fuel cell, P. fcmax That is the maximum output power of the fuel cell, SOC min It is the minimum SOC of the power battery. max It is the maximum value of the SOC of the power battery, P fc It is the output power of the fuel cell; S6. The objective function is shown below: Where f is the hydrogen consumption rate function, t0 is the initial time, and t f At the final moment, J represents the total hydrogen consumption; S7. The Hamiltonian function is shown in the following formula: Where λ is a costate variable; The output power of the fuel cell when the Hamiltonian function reaches its extreme value is the optimal output power. At this point, the output power of the corresponding power battery can be calculated, which is the optimal output power of the power battery. S8. Clustering using the K-means method: (1) Select samples y1, y2, y3, ..., y r The samples are divided into s clusters, and s samples are randomly selected as the centers of the clusters, denoted as t1, t2, t3, ..., t4. s ; (2) Calculate the Euclidean distance from each point in the sample to the center point. For a specific sample d w In other words, its distance to the center point t z The Euclidean distance is as follows: L wz =||d w -t z || (18) (3) Based on the obtained Euclidean distance L wz The value determines which cluster the sample belongs to, and the sample is assigned to the cluster with the nearest centroid. (4) The new centers of each cluster after the transformation are obtained by finding the minimum value of the loss function, which is shown below: Where p g (g = 1, 2, ..., s) represents different clusters, S is the number of different clusters, r is the total number of samples, Q is the overall loss function, and h1, h2, ..., hs are the number of samples in each cluster, satisfying r = h1 + h2 + h3 + ... + hs. Find the minimum value of the loss function and expand it as follows: Simplifying the above equation, each summation can be represented by Q1, Q2, Q3, ..., Q... s Indicates Q1, Q2, Q3, ..., Q s If is the loss function for each cluster, then the above equation can be simplified to: min Q=min(Q1+Q2+Q3+...+Q s )=min Q1+min Q2+min Q3+...+min Q s (21) Taking Q1 as an example, the partial derivative of Q1 with respect to the center point t1 of the cluster is taken, and the result of the partial derivative is set to 0, as shown in the following equation: The value of t1 can then be calculated: Similarly, this can be achieved by considering Q1, Q2, Q3, ..., Q... s Find the partial derivatives and locate the center points t2, t3, ..., t of each cluster. s The value; As can be seen from the above, when the samples contained in different clusters do not change, the center point of each cluster also does not change; when the center point of the calculated cluster changes, repeat steps (2) to (4). When the value of the center point of each cluster does not change, the resulting clusters are the final clustering results; by classifying the obtained power battery output power and SOC data through the above steps (1) to (4), the final pattern diagram is obtained. S9. In the above pattern diagram, find the output power of the power battery corresponding to the current SOC based on the required power, calculate the output power of the fuel cell using the required power, and calculate the SOC value of the power battery at the next moment using the above two power values.

Images