A data-driven hydrogen-electric hybrid energy management method and device

Abstract

This application discloses a data-driven hydrogen-electric hybrid energy management method and device, relating to the field of engineering control technology. The method includes: calculating the predicted hydrogen consumption of the fuel cell at time k+1 using the TDL method based on the actual hydrogen consumption of the fuel cell at time k and the historical actual SOC value of the lithium battery; calculating the reference SOC value of the lithium battery at time k+1; constructing an SOC prediction model in the prediction time domain at time k using the TDL method; and solving the power control increment sequence of the lithium battery in the control time domain at time k. Based on the first term in the power control increment sequence and the reference power of the lithium battery at time k-1, the reference power of the lithium battery at time k is calculated, and then the reference power of the fuel cell at time k is calculated, thereby achieving control of the actual output power of the lithium battery and the fuel cell at time k. This application can achieve optimal power allocation between the hydrogen fuel cell and the lithium battery, extending the driving range.

Description

Technical Field

[0001] This application relates to the field of engineering control technology, and in particular to a data-driven method and apparatus for hydrogen-electric hybrid energy management. Background Technology

[0002] With the development of the low-altitude economy, the application scenarios of drones are becoming increasingly widespread. Most existing drones use lithium batteries as their sole energy source, which has limited energy density, resulting in short flight times – a major challenge restricting drone development. Fuel cells (FCs), represented by hydrogen energy, have high energy density and can significantly improve flight endurance, alleviating the problem of insufficient drone battery life to some extent. However, hydrogen fuel cells are limited by their own performance; their response speed is slower than lithium batteries, and they cannot adapt to rapid load changes. Therefore, combining hydrogen fuel cells with lithium batteries to form a hybrid power system has become an effective solution. Since there are multiple power sources in the system, how to rationally allocate power demand becomes a key issue. Therefore, an energy management strategy needs to be designed to optimize power distribution between hydrogen fuel cells and lithium batteries, ensuring efficient system operation and stability.

[0003] Energy management strategies are divided into rule-based and optimization-based strategies. Rule-based strategies are simple in principle, easy to implement, and have low computational complexity. However, the thresholds of these control strategies are preset, largely dependent on experience, and cannot be changed during operation, resulting in poor adaptability. Optimization-based strategies, on the other hand, allocate power output according to optimal control theory under given operating conditions to achieve optimal system performance. Although computational complexity is higher, they offer flexible control and strong adaptability to different operating conditions.

[0004] Model predictive control (MPC) has been widely used in optimization-based energy management strategies. MPC obtains the current control action by solving a finite-time open-loop optimal control problem at each sampling time, exhibiting robustness and effectively handling multivariable and multi-constraint problems. However, in practice, it is often difficult to establish accurate models for systems with nonlinear, uncertain, or time-varying characteristics. To address this, data-driven predictive control methods have been proposed, but most existing data models lack clearly defined physical meanings for their model parameters. Furthermore, a large number of parameters in these data models require online identification. In practical applications, these two issues can lead to biases in parameter estimation during online identification and also impose additional computational burdens.

[0005] Therefore, there is an urgent need for an energy management strategy to overcome the difficulty in establishing predictive models for existing energy management strategies that employ MPC, while simultaneously optimizing power allocation between hydrogen fuel cells and lithium batteries. Summary of the Invention

[0006] The purpose of this application is to provide a data-driven hydrogen-electric hybrid energy management method and device, which solves the problem that it is difficult to establish a predictive model for existing energy management strategies that use MPC, and can accurately optimize the power distribution between hydrogen fuel cells and lithium batteries.

[0007] To achieve the above objectives, this application provides the following solution:

[0008] Firstly, this application provides a data-driven method for hydrogen-electric hybrid energy management, comprising:

[0009] Based on the actual hydrogen consumption of the fuel cell at time k and the historical actual SOC of the lithium battery, the TDL method is used to calculate the predicted hydrogen consumption of the fuel cell at time k+1.

[0010] The reference value of lithium battery SOC at time k+1 is calculated based on the predicted hydrogen consumption of fuel cell at time k+1, the reference value of lithium battery SOC at time k-1, and the historical actual value of lithium battery SOC.

[0011] At time k, the TDL method is applied to construct the SOC prediction model in the prediction time domain at time k. Combined with the lithium battery SOC reference value at time k+1 and the historical actual SOC value of the lithium battery, the lithium battery power increment sequence in the control time domain at time k is solved. The SOC prediction model is used to establish the relationship between the real-time measured SOC of the lithium battery and the output power of the lithium battery to predict the SOC value of the lithium battery at the next time.

[0012] The reference power of the lithium battery at time k is calculated based on the first term in the lithium battery power increment sequence and the reference power of the lithium battery at time k-1.

[0013] The reference power of the fuel cell at time k is calculated based on the reference power of the lithium battery at time k and the load power at time k; the reference power of the lithium battery and the fuel cell at time k is used to control the actual output power of the lithium battery and the fuel cell at time k.

[0014] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described data-driven hydrogen-electric hybrid energy management method.

[0015] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described data-driven hydrogen-electric hybrid energy management method.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0017] This application provides a data-driven energy management method and device for hydrogen-electric hybrid vehicles, and a data-driven energy management strategy. It applies the TDL method to construct a SOC prediction model in the time domain at time k. Based on the measured real-time SOC and lithium battery output power, the established SOC prediction model establishes the correlation between SOC and lithium battery output power by identifying time-varying parameters of the model. Based on the established SOC prediction model, it solves the power control increment sequence in the control time domain, rather than the traditional method of establishing a SOC prediction model based on physical mechanisms. This solves the problem of difficulty in establishing SOC prediction models in existing energy management strategies using MPC, and simultaneously achieves optimal power allocation between the hydrogen fuel cell and the lithium battery, extending the driving range. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an application environment diagram of a data-driven hydrogen-electric hybrid energy management method according to an embodiment of this application;

[0020] Figure 2 A flowchart illustrating a data-driven hydrogen-electric hybrid energy management method provided in an embodiment of this application;

[0021] Figure 3 A schematic diagram illustrating the technical concept of a data-driven hydrogen-electric hybrid energy management method provided in an embodiment of this application;

[0022] Figure 4 This is a schematic diagram illustrating the PI control effect according to an embodiment of this application;

[0023] Figure 5 This is a schematic diagram of the CFDL-APC control effect provided in an embodiment of this application;

[0024] Figure 6 This is a schematic diagram of the TDL-APC control effect provided in an embodiment of this application;

[0025] Figure 7 This is a schematic diagram showing the comparison between actual hydrogen consumption and predicted hydrogen consumption provided in an embodiment of this application;

[0026] Figure 8 This is a schematic diagram comparing the SOC reference value and the actual value provided in an embodiment of this application;

[0027] Figure 9 This is a schematic diagram of a battery power reference value curve provided in an embodiment of this application;

[0028] Figure 10 A schematic diagram of the actual power curve of an energy management strategy based on TDL-APC control provided in an embodiment of this application;

[0029] Figure 11 A schematic diagram of the actual power curve of a state machine-based energy management strategy provided in an embodiment of this application;

[0030] Figure 12 Hydrogen consumption and power fluctuation diagram of an energy management strategy based on the TDL-APC method provided in an embodiment of this application;

[0031] Figure 13 Hydrogen consumption and power fluctuation diagram of a state machine-based energy management strategy provided in an embodiment of this application;

[0032] Figure 14 This is a schematic diagram comparing hydrogen consumption values ​​under PI control, CFDL-APC control, and TDL-APC control, as provided in an embodiment of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] The data-driven hydrogen-electric hybrid energy management method provided in this application embodiment can be applied to, for example... Figure 1The application environment shown depicts a scenario where the terminal communicates with the server via a network. The data storage system stores the data the server needs to process. This data storage system can be configured independently, integrated into the server, or located in the cloud or on another server. The terminal can send the actual hydrogen consumption of the fuel cell at time k, the historical actual SOC of the lithium battery, and the reference SOC of the lithium battery at time k-1 to the server. After receiving the data, the server uses the TDL method to calculate the predicted hydrogen consumption of the fuel cell at time k+1 based on the actual hydrogen consumption of the fuel cell at time k and the historical actual SOC of the lithium battery. It also calculates the reference SOC of the lithium battery at time k+1 based on the predicted hydrogen consumption of the fuel cell at time k+1, the reference SOC of the lithium battery at time k-1, and the historical actual SOC of the lithium battery. At time k, the server uses the TDL method to construct a SOC prediction model in the prediction time domain at time k, and combines the reference SOC of the lithium battery at time k+1 and the historical actual SOC of the lithium battery to solve the lithium battery power increment sequence in the control time domain at time k. The server calculates the reference power of the lithium battery at time k based on the first term in the lithium battery power increment sequence and the reference power of the lithium battery at time k-1. Finally, the server calculates the reference power of the fuel cell at time k based on the reference power of the lithium battery at time k and the load power at time k. The reference power of the lithium battery and the fuel cell at time k is used to control the actual output power of the lithium battery and the fuel cell at time k. The server can feed back the reference power of the lithium battery and fuel cell at time k to the terminal. Furthermore, in some embodiments, the data-driven hydrogen-electric hybrid energy management method can also be implemented independently by the server or the terminal. For example, the terminal can directly perform data-driven hydrogen-electric hybrid energy management based on the actual hydrogen consumption of the fuel cell at time k-1, the historical actual SOC of the lithium battery, and the reference SOC of the lithium battery at time k-1. Alternatively, the server can obtain the actual hydrogen consumption of the fuel cell at time k-1, the historical actual SOC of the lithium battery, and the reference SOC of the lithium battery at time k-1 from the data storage system, and then perform data-driven hydrogen-electric hybrid energy management.

[0036] The terminal can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. The server can be a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0037] In one exemplary embodiment, such as Figure 2 and Figure 3 As shown, a data-driven energy management method for hydrogen-electric hybrid vehicles is provided. This method is executed by a computer device, specifically a terminal or server, or both. In this embodiment, the method is applied to... Figure 1The following steps, 101 to 105, are used as an example to illustrate the process of using a server in the example.

[0038] Step 101: Based on the actual hydrogen consumption of the fuel cell at time k and the historical actual SOC of the lithium battery, the TDL method is applied to calculate the predicted hydrogen consumption of the fuel cell at time k+1.

[0039] Step 102: Calculate the reference value of lithium battery SOC at time k+1 based on the predicted hydrogen consumption of fuel cell at time k+1, the reference value of lithium battery SOC at time k-1, and the historical actual value of lithium battery SOC.

[0040] Step 103: At time k, the TDL method is applied to construct the SOC prediction model within the prediction time domain at time k (Np steps from time k+1 to time k+Np). Combining the lithium battery SOC reference value at time k+1 and the historical actual SOC value of the lithium battery, the lithium battery power increment sequence within the control time domain at time k (Nv steps from k+1 to k+Nv-1) is solved. The SOC prediction model is used to establish the relationship between the real-time measured SOC of the lithium battery and the output power of the lithium battery to predict the SOC value of the lithium battery at the next time.

[0041] Step 104: Calculate the reference power of the lithium battery at time k based on the first term in the lithium battery power increment sequence and the reference power of the lithium battery at time k-1.

[0042] Step 105: Calculate the reference power of the fuel cell at time k based on the reference power of the lithium battery at time k and the load power at time k; the reference power of the lithium battery and the fuel cell at time k is used to control the actual output power of the lithium battery and the fuel cell at time k.

[0043] In another exemplary embodiment of this application, in step 101, hydrogen consumption prediction is based on TDL (Triangular Dynamic Linearization): using the TDL method, the hydrogen consumption change of the fuel cell and the SOC change of the lithium battery are correlated, and the hydrogen consumption value at the current time (time k) is predicted. The specific process is as follows:

[0044]

[0045] in, The hydrogen consumption value for the next time step is predicted from the current time step (time k), in liters per minute (lpm). This represents the hydrogen consumption value at the current moment. The time-varying TDL predictor parameters (referred to as triangular data model parameters, and the TDL data model is simply called the triangular data model) that reflect the dynamic characteristics of the system and are identified online are updated by the following formula:

[0046] (2)

[0047] in, For constant weight parameters, The change in hydrogen consumption is defined as:

[0048]

[0049] in, It consists of two parts. The first part is the most recent one at the current time. The second part involves summing the SOC changes with linearly increasing weights. arrive The earlier historical data (SOC change) are summed with linearly decreasing weights, as shown in the following expression:

[0050] (4)

[0051] in, The change in SOC is defined as follows:

[0052]

[0053] It is In the formula k Replace with k -1 is calculated; It is The formula is obtained by replacing k with k-1.

[0054] Using predicted hydrogen consumption values SOC reference value optimization is performed to obtain the SOC reference trajectory. generate.

[0055] In another exemplary embodiment of this application, in step 102, for the optimization of the SOC reference value, a reasonably planned SOC trajectory is equivalent to determining the trend of each power allocation. A quadratic programming (QP) problem is constructed, with the goal of obtaining the SOC reference trajectory in real time by solving the objective function.

[0056] Note that in hydrogen consumption prediction, the parameters , hour, It is only equal to the SOC change at the most recent moment, that is , and Let k and k-1 represent the actual SOC values ​​of the lithium battery at time k and k-1, respectively. Therefore, the hydrogen consumption prediction model at this time also establishes a prediction of hydrogen consumption. And the SOC model, i.e. Let the parameters for the above hydrogen consumption prediction be... , Construct the objective function as follows:

[0057] (6)

[0058] objective function It consists of 3 items, at each time step Perform a single-step optimization to find the SOC reference value. The first term is the SOC tracking term, which ensures the continuity between the SOC reference value and the current SOC state. The first term represents the reference SOC value of the lithium battery at time k+1, which is to be solved; the second term is the hydrogen consumption optimization term, which is minimized by the hydrogen consumption model. Obtain the SOC reference value. The third term is a smoothing term to avoid drastic changes in the reference value. Let k be the reference value of the lithium battery's SOC at time k. , , These are the weighting coefficients for each item; and These represent the maximum and minimum SOC values ​​of the lithium battery, respectively. The optimization problem is transformed into a constrained standard quadratic programming form for solving in solvers such as quadprog:

[0059] (7)

[0060] in,

[0061] (8)

[0062] As an optional implementation, an additional overload correction term is added to the optimization to correct for potential deviations in the SOC reference value of the lithium battery when dealing with high-power discharge. The approach is as follows:

[0063]

[0064] Minimize objective function A preliminary solution was obtained. Then, if at this time (time k) the actual power of the lithium battery Greater than a certain constant setting value Then, a soft constraint is used to correct the SOC reference value at time k+1, linearly adjusting it downwards, and finally mapping it to... Interval. For coefficients, For load power, This represents the actual power of the fuel cell. The solution yields... or This was then used as a SOC reference value. SOC tracking continued afterwards.

[0065] Based on the above, in step 102, the reference value of the lithium battery SOC at time k+1 is calculated according to the predicted hydrogen consumption of the fuel cell at time k+1, the reference value of the lithium battery SOC at time k-1, and the historical actual value of the lithium battery SOC. Specifically, this includes:

[0066] (2-1) Based on the predicted hydrogen consumption of the fuel cell at time k+1, the reference SOC of the lithium battery at time k-1, and the historical actual SOC of the lithium battery, construct the objective function for optimizing the SOC reference value.

[0067] (2-2) The objective function for optimizing the SOC reference value is transformed into a quadratic programming form.

[0068] (2-3) Solve the objective function for optimizing the converted SOC reference value to obtain the lithium battery SOC reference value at time k+1, specifically including:

[0069] (1) Solve the objective function of the converted SOC reference value to obtain the preliminary solution of the SOC reference value.

[0070] (2) Determine whether the actual output power of the lithium battery at time k is greater than the preset power value.

[0071] If not, the initial solution for the SOC reference value is the lithium battery SOC reference value at time k+1.

[0072] If so, the initial solution of the SOC reference value is corrected by using the actual output power of the lithium battery at time k, the load power at time k, the maximum SOC value of the lithium battery, and the minimum SOC value of the lithium battery. The corrected SOC reference value is then obtained, which is the lithium battery SOC reference value at time k+1.

[0073] In another exemplary embodiment of this application, in step 103, adaptive SOC tracking based on TDL is performed: an adaptive predictive controller (TDL-APC controller for short) based on a triangular data model is designed, and by... The tracking mechanism enables power distribution in a fuel cell hybrid system. The controller outputs a power reference value to the lithium battery to discharge or charge it in order to track the power distribution. Value. The controller employs a four-step recursive algorithm structure.

[0074] (1) Similar to hydrogen consumption prediction, based on the triangular data model, the predicted SOC and output power of the lithium battery at the next moment can be expressed as:

[0075]

[0076] Formula (10) is the SOC prediction model constructed using the TDL method. Wherein, the time-varying parameters... This reflects the relationship between changes in the state of charge (SOC) of a lithium battery and changes in its output power.

[0077] (11)

[0078] in, Estimate the weight parameters for constant parameters. For the extended form of control increment input, the first part is the most recent value at the current time. Changes in control inputs Perform linearly increasing weighted summation, the second part... arrive Sum the earlier historical data using linearly decreasing weights:

[0079] (12)

[0080] Defined as power control increment:

[0081]

[0082] in, This is the current reference power of the lithium battery. This is the reference power of the lithium battery at the previous moment.

[0083] (2) Construct a triangular data model predictor, with the prediction time domain being... The future control input sequence is defined as follows, containing the sequence from the current time step... To the future Control input at time (the power control increment sequence to be solved):

[0084]

[0085] in, express Time and Power control increment between moments; express Time and The power control increment between moments.

[0086] Redefine the history control input sequence, including from the past Time's up Actual control input at any given time (historical power control increment sequence):

[0087]

[0088] express The actual value of the power control increment at any given time; express The actual value of the power control increment at any given time.

[0089] Based on historical and future control input sequences, the future SOC of the lithium battery system is predicted using a triangular data model.

[0090]

[0091] Formula (16) represents the current time step (i). Predicted output of step It is formed by superimposing the weighted sum of the previous predicted value, the future control input, and the historical control input. Among them, ; The piecewise expression is as follows, and its value depends on Scope ( , or The coefficients within each segment are determined by... It consists of the product of a specific weight vector (including fractional terms and zero padding).

[0092] (17)

[0093] coefficient The piecewise expression is as follows, and its structure is similar to... Similar, but the weight vector and zero-padding method are different. When When the coefficients are zero, the coefficients are directly zero vectors.

[0094] (18)

[0095] Formula (16) can be written in the following compact form:

[0096]

[0097]

[0098] The SOC vector in the prediction time domain of the prediction output at time K. It consists of three superimposed parts: the current state item , A column vector consisting entirely of 1s , the current state Extending to the prediction time domain dimension Future control input weighting , The cumulative coefficient matrix for future control inputs; the weighted sum of historical control inputs. , This is the cumulative coefficient matrix for historical control inputs.

[0099]

[0100] In particular, if future control inputs are outside the control time domain If the value after the step is zero, then only truncation is needed. The former Column (denoted as) ) and future control inputs item ;

[0101] therefore,

[0102] (3) Construct a feedback corrector and define the feedback adjustment term. as follows:

[0103]

[0104] in, Indicates the SOC correction amount; For length is The weighted coefficient vector is used to adjust the compensation strength for prediction errors; For the current moment The actual SOC value; For the previous moment to the current moment The output single-step prediction value; the corrected prediction output vector is shown in equation (24), which represents the feedback adjustment term. Superimposed on the original predicted output vector The corrected prediction output is obtained above. This is used for the design of the subsequent optimal controller.

[0105]

[0106] (4) Construct the optimal controller to solve the power control increment sequence. The design objective of the controller is to make the SOC trajectory as close as possible to the reference trajectory, while maintaining the smoothness and reasonable amplitude of the control input. Define the objective function for solving the optimal controller problem. as follows:

[0107] (25)

[0108] in, The rate of change of SOC is defined as follows: , Sampling time, The rate of change weighting coefficient. To control the weighting coefficients, in equation (25)... Defined as the comprehensive error vector .

[0109] Will Substituting the expression (24) into :

[0110] (26)

[0111] Will Substituting the expression (22) into :

[0112] (27)

[0113] Let all and The irrelevant terms (i.e., the parts that can be considered constants) form a vector. Then the objective function becomes:

[0114] (28)

[0115] (29)

[0116] Merge control penalty items :

[0117] (30)

[0118] Let I be a unit vector. This can be rearranged into the standard form of a QP problem:

[0119]

[0120] Among them, the Hessian matrix for:

[0121] (32)

[0122] The coefficient vector of the first-order term is:

[0123] (33)

[0124] The power control increment sequence is obtained by minimizing the objective function Jsoc. Considering actual power constraints, the controller adopts... The first power control increment in Set constraints: ; / This represents the minimum / maximum output power of the lithium battery.

[0125] The power control increment is solved That is, the power control increment at the current moment, as expressed in equation (13). That is, the current reference power of the lithium battery is:

[0126]

[0127] Based on the above, in step 103, at time k, the TDL method is applied to construct the SOC prediction model in the prediction time domain at time k. Combining the lithium battery SOC reference value at time k+1 and the historical actual SOC value of the lithium battery, the lithium battery power increment sequence in the control time domain at time k is solved, specifically including:

[0128] (3-1) At time k, the TDL method is used to construct the SOC prediction model in the prediction time domain at time k.

[0129] (3-2) Define the power control increment sequence to be solved in the prediction time domain corresponding to time k.

[0130] (3-3) Obtain the historical power control increment sequence within the preset time period before time k.

[0131] (3-4) Based on the constructed SOC prediction model, the predicted value of lithium battery SOC in the prediction time domain at time k is calculated according to the power control increment sequence to be solved and the historical power control increment sequence in the prediction time domain at time k.

[0132] (3-5) Correct the predicted SOC value of the lithium battery in the prediction time domain at time k, and obtain the corrected predicted SOC value of the lithium battery in the prediction time domain at time k.

[0133] (3-6) Construct the power control incremental solution objective function based on the lithium battery SOC reference value at time k+1, the lithium battery SOC prediction correction value in the predicted time domain at time k, the actual SOC value of the lithium battery at time k, and the actual SOC value of the lithium battery at time k-1.

[0134] (3-7) The objective function for solving the power control increment is transformed into a standard quadratic programming form and solved to obtain the lithium battery power control increment sequence in the control time domain at time k.

[0135] This application uses a triangular dynamic linearization (TDL) data model, which contains only one time-varying parameter with a definite physical meaning. Based on online estimation of a single time-varying parameter, a novel data-driven adaptive control scheme is designed. This scheme addresses the problem that most existing data models lack clear physical meaning for their model parameters and that the large number of parameters in these models require online identification, leading to parameter estimation biases during online identification in practical applications.

[0136] The effectiveness of the method in this application is illustrated below based on simulation results:

[0137] (1) Verification of SOC tracking effect: settings With a fixed value of 67 and an initial SOC of 70, the control performance of the proposed TDL-based adaptive SOC tracker (the TDL-APC controller of this application) is compared with that of a PI (proportional-integral control) SOC tracker, a CFDL-APC (Compact Form Dynamic Linearization) SOC tracker, and a state machine-based SOC tracker under the same power load. Figures 4 to 6 As shown.

[0138] Table 1 PI Controller Parameter Table

[0139]

[0140] Table 2 CFDL-APC Controller Parameters

[0141]

[0142] Table 3. Eight states of a state machine

[0143]

[0144] Table 4 TDL-APC Controller Parameter Table

[0145]

[0146] (2) Verification of a complete energy management strategy: A complete energy management strategy under the combined action of a hydrogen consumption predictor, a SOC reference value optimizer, and a SOC tracker (same power load as in (1) verification of SOC tracking effect). The hydrogen consumption predictor effect diagram is shown below. Figure 7 As shown in Table 5, which is the parameter table for the hydrogen consumption predictor.

[0147] Table 5 Hydrogen Consumption Predictor Parameter Table

[0148]

[0149] The SOC reference value optimizer provides And the control effect of SOC tracker Figure 8 As shown, the SOC tracker calculates... like Figure 9 As shown in Table 6, which is the SOC reference value optimizer parameter table, and Table 7, which is the TDL-APC controller parameter table.

[0150] Table 6 SOC Reference Value Optimizer Parameter Table

[0151]

[0152] Table 7 TDL-APC Controller Parameter Table

[0153]

[0154] like Figure 10 The figure shows the actual power curves of the energy management strategy based on TDL-APC control, namely the total load power of the hybrid power system and the actual power load of each component. The red line represents the total load power. Blue represents the load power of the hydrogen fuel cell. Yellow indicates the actual power of the lithium battery. Green represents the power of the supercapacitor. Since the DC bus voltage is controlled by the battery converter, the power of the supercapacitor is not considered in the optimization problem. That is, once the supercapacitor discharges, it draws the same amount of energy from the battery system to recharge it. Therefore, within a given load cycle, the total load energy is shared solely by the fuel cell and the battery. Figure 11 The image shows the actual power curve of the state machine-based energy management strategy. (Comparison) Figure 10 and Figure 11 As can be seen, when faced with changes in load power, the TDL-APC strategy optimizes power allocation and dynamically adjusts, taking into account both hydrogen consumption and power output stability, and enabling load power to be allocated more flexibly among different power sources.

[0155] Figure 12 This is a graph showing hydrogen consumption and power fluctuations for an energy management strategy based on the TDL-APC method. It includes three sub-graphs: from top to bottom, the power curves for the fuel cell and lithium battery, the hydrogen consumption and equivalent hydrogen consumption curves, and the fuel cell power change rate curve. Equivalent hydrogen consumption is calculated proportionally by transferring the energy consumed by the battery to the fuel cell at a certain ratio. This means how much extra hydrogen would be consumed if the fuel cell provided all of this energy. The measured hydrogen consumption, equivalent hydrogen consumption, and other data under the complete energy management strategy are as follows: Hydrogen consumption: 33.08 g; Equivalent hydrogen consumption: 38.3052 g; Average change rate: 5.4478 W / s; Standard deviation: 164.4946 W / s; Maximum change rate: 487.4705 W / s; Minimum change rate: -530.8238 W / s. Figure 13The figure shows the hydrogen consumption and power fluctuation diagrams for the state machine-based energy management strategy. The measured final hydrogen consumption, equivalent hydrogen consumption, and other data are as follows: hydrogen consumption: 36.98 g; equivalent hydrogen consumption: 40.2629 g; average rate of change: 5.5812 W / s; standard deviation: 152.1116 W / s; maximum rate of change: 504.4154 W / s; minimum rate of change: -527.4108 W / s. The energy management strategy based on the TDL-APC method outperforms the state machine strategy in terms of hydrogen consumption: hydrogen consumption is reduced by 3.9 g, saving 10.54% of hydrogen. Equivalent hydrogen consumption is reduced by 1.96 g, saving 4.86% of hydrogen. The reduction in hydrogen consumption and equivalent hydrogen consumption indicates that the TDL-APC strategy has higher energy utilization efficiency in fuel cells, can better optimize hydrogen fuel use, and extend the driving range. The average power change rate of the TDL-APC strategy is 5.4478 W / s, slightly lower than the 5.5812 W / s of the state machine strategy, indicating that the TDL-APC strategy is generally more stable in terms of power change. The maximum power change rate of the TDL-APC strategy is 487.4705 W / s, lower than the 504.4154 W / s of the state machine strategy, indicating that the TDL-APC strategy has less fluctuation during power abrupt changes. The standard deviation of power fluctuation of the state machine strategy is slightly lower (152.1116 W / s vs. 164.4946 W / s), indicating that the state machine strategy has a slight advantage in the dispersion of power fluctuations, but the difference is small. Therefore, the energy management strategy based on the TDL-APC method can not only extend the driving range but also improve the operational stability of the fuel cell system.

[0156] like Figure 14 As shown, the hydrogen consumption values ​​were verified when the SOC reference value was set to a fixed value (i.e., 67). Specifically, the hydrogen consumption was 38.75g when using the PI method to track the SOC setpoint, 38.01g when using the CFDL-APC method, and 36.53g when using the TDL-APC method proposed in this application. Tracking the SOC setpoint was achieved by using three controllers (PI, CFDL-APC, and TDL-APC) to control the discharge of the lithium battery. Figure 14 It can be concluded that the proposed TDL-APC controller can stabilize the lithium battery SOC at 67% under power load, has the best tracking effect, and the lowest hydrogen consumption value. The other two controllers have different degrees of error.

[0157] In this application, a SOC prediction model is established and controlled based on input and output data. The established SOC prediction model is based on the measurement of real-time SOC and lithium battery power, and the parameters are identified. Instead of the traditional approach of establishing a SOC prediction model based on physical mechanisms, this application establishes a correlation between the two and constructs an objective function to obtain the power control increment. It addresses the difficulty in establishing a SOC prediction model for hybrid power systems by employing a data-driven modeling approach and utilizes a TDL data model to address the shortcomings of existing data models.

[0158] This application also provides an application scenario in which the above-described data-driven hydrogen-electric hybrid energy management method is applied. Specifically, the data-driven hydrogen-electric hybrid energy management method provided in this embodiment can be applied to the hydrogen-electric hybrid energy management scenario of unmanned aerial vehicles (UAVs). This scenario includes a data acquisition stage, a power allocation calculation stage, and a power output control stage. The data acquisition stage is used to acquire the actual hydrogen consumption value of the fuel cell at time k, the historical actual SOC value of the lithium battery, and the reference SOC value of the lithium battery at time k-1. The power allocation calculation stage is used to calculate the reference power of the lithium battery and fuel cell at time k based on the acquired data. The power output control stage is used to control the actual output power of the lithium battery and fuel cell at time k based on the reference power of the lithium battery and fuel cell at time k. The data-driven hydrogen-electric hybrid energy management method provided in this embodiment belongs to the power allocation calculation stage. In addition, the above-described data-driven hydrogen-electric hybrid energy management method can be applied to any device involving hydrogen-electric hybrid energy management.

[0159] In one exemplary embodiment, this application provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the data-driven hydrogen-electric hybrid energy management method in the above-described method embodiments.

[0160] In one exemplary embodiment, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the data-driven hydrogen-electric hybrid energy management method described in the above method embodiments.

[0161] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0162] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0163] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0164] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0165] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A data-driven hydrogen electric hybrid energy management method, characterized in that, include: Based on the actual hydrogen consumption of the fuel cell at time k and the historical actual SOC of the lithium battery, the TDL method is used to calculate the predicted hydrogen consumption of the fuel cell at time k+1. The reference value of lithium battery SOC at time k+1 is calculated based on the predicted hydrogen consumption of fuel cell at time k+1, the reference value of lithium battery SOC at time k-1, and the historical actual value of lithium battery SOC. At time k, the TDL method is applied to construct the SOC prediction model in the prediction time domain at time k. Combined with the lithium battery SOC reference value at time k+1 and the historical actual SOC value of the lithium battery, the lithium battery power increment sequence in the control time domain at time k is solved. The SOC prediction model is used to establish the relationship between the real-time measured SOC of the lithium battery and the output power of the lithium battery to predict the SOC value of the lithium battery at the next time. The reference power of the lithium battery at time k is calculated based on the first term in the lithium battery power increment sequence and the reference power of the lithium battery at time k-1. The reference power of the fuel cell at time k is calculated based on the reference power of the lithium battery and the load power at time k; the reference power of the lithium battery and the fuel cell at time k is used to control the actual output power of the lithium battery and the fuel cell at time k. The formula for calculating the predicted hydrogen consumption of the fuel cell at time k+1 is as follows: ； wherein ; ； In the formula, The hydrogen consumption value of the fuel cell at time k+1 is predicted at time k. The actual hydrogen consumption of the fuel cell at time k; These are weight parameters; ; ; Let be the change in hydrogen consumption at time k and time k-1; for Time and The change in SOC at time t; It is In the formula k Replace with k -1 is calculated; For online identification of time-varying TDL predictor parameters that reflect the dynamic characteristics of the system; It is The formula is obtained by replacing k with k-1.

2. The data-driven hydrogen-electric hybrid energy management method according to claim 1, characterized in that, The reference value of lithium battery SOC at time k+1 is calculated based on the predicted hydrogen consumption of the fuel cell at time k+1, the reference value of lithium battery SOC at time k-1, and the historical actual value of lithium battery SOC. Specifically, it includes: Based on the predicted hydrogen consumption of the fuel cell at time k+1, the reference SOC of the lithium battery at time k-1, and the historical actual SOC of the lithium battery, an objective function for optimizing the SOC reference value is constructed. The objective function for optimizing the SOC reference value is transformed into a quadratic programming form; The objective function for optimizing the converted SOC reference value is solved to obtain the SOC reference value of the lithium battery at time k+1.

3. The data-driven hydrogen-electric hybrid energy management method according to claim 2, characterized in that, The expression for the objective function for optimizing the SOC reference value is: ； in, ; ； In the formula, Let SOC be the reference value of the lithium battery at time k+1, which is to be solved. and These represent the actual SOC values ​​of the lithium battery at time k and time k-1, respectively. This represents the reference SOC value of the lithium battery at time k; and These represent the maximum and minimum SOC values ​​of the lithium battery, respectively. , , These are the weighting coefficients for each item.

4. The data-driven hydrogen-electric hybrid energy management method according to claim 2 or 3, characterized in that, The objective function for optimizing the converted SOC reference value is solved to obtain the lithium battery SOC reference value at time k+1, specifically including: The objective function for optimizing the transformed SOC reference value is solved to obtain a preliminary solution for the SOC reference value; Determine whether the actual output power of the lithium battery at time k is greater than the preset power value; If not, the initial solution for the SOC reference value is the SOC reference value of the lithium battery at time k+1; If so, the initial solution of the SOC reference value is corrected by using the actual output power of the lithium battery at time k, the load power at time k, the maximum SOC value of the lithium battery, and the minimum SOC value of the lithium battery. The corrected SOC reference value is then obtained, which is the lithium battery SOC reference value at time k+1.

5. The data-driven hydrogen-electric hybrid energy management method according to claim 4, characterized in that, The revised formula for calculating the SOC reference value is as follows: ； In the formula, This indicates the corrected SOC reference value; This indicates the SOC reference value before correction; and These represent the maximum and minimum SOC values ​​of the lithium battery, respectively. This represents the actual output power of the lithium battery at time k; This represents the load power at time k; Represents the coefficient.

6. The data-driven hydrogen-electric hybrid energy management method according to claim 1, characterized in that, At time k, the TDL method is applied to construct a SOC prediction model in the prediction time domain at time k. Combining the lithium battery SOC reference value at time k+1 and the historical actual SOC value of the lithium battery, the power increment sequence of the lithium battery in the control time domain at time k is solved, specifically including: At time k, the TDL method is applied to construct the SOC prediction model in the prediction time domain at time k; Define the power control increment sequence to be solved in the prediction time domain corresponding to time k; Obtain the historical power control increment sequence within a preset time period prior to time k; The predicted SOC value of the lithium battery in the prediction time domain at time k is calculated based on the constructed SOC prediction model, the power control increment sequence to be solved in the prediction time domain at time k, and the historical power control increment sequence. The predicted SOC value of the lithium battery in the prediction time domain at time k is corrected to obtain the corrected predicted SOC value of the lithium battery in the prediction time domain at time k. The objective function for incremental power control is constructed based on the reference SOC value of the lithium battery at time k+1, the predicted correction value of the lithium battery SOC in the predicted time domain at time k, the actual SOC value of the lithium battery at time k, and the actual SOC value of the lithium battery at time k-1. The objective function for solving the power control increment is transformed into a standard quadratic programming form and solved to obtain the lithium battery power control increment sequence in the control time domain at time k.

7. The data-driven hydrogen-electric hybrid energy management method according to claim 6, characterized in that, The expression for the objective function of power control increment is: ； in, ; ； ； In the formula, This represents the reference SOC value of the lithium battery at time k+1; This represents the predicted SOC correction value for the lithium battery at time k+1; Represents a column vector consisting entirely of 1s; This represents the actual SOC value of the lithium battery at time k; This represents the actual SOC value of the lithium battery at time k-1; Sampling time; The rate of change weighting coefficient; To control the weighting coefficients; Indicates the SOC correction amount; The plan to be solved at time k is to take the first term of the sequence at a future time point and apply a length of k+. -1 power control increment sequence; To predict values ​​in the time domain; This represents the rate of change of SOC.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the data-driven hydrogen-electric hybrid energy management method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the data-driven hydrogen-electric hybrid energy management method according to any one of claims 1-7.

Images