DAB Optimization Modulation Method Based on Artificial Intelligence and Circuit Transient Behavior

Abstract

The present invention provides a DAB optimization modulation method based on artificial intelligence and circuit transient behavior. The method of the present invention obtains an analytical expression suitable for the DAB data-driven model under full-mode UTPS control through numerical modeling of circuit physical laws and physical prototype sampling, and uses advanced artificial intelligence fitting technology, and obtains the reference value of the optimal static modulation angle at the next moment based on the trained first artificial neural network and obtains the reference value of the transient phase shift angle by solving the transient phase shift angle related expression, and then controls the dual active bridge converter based on the obtained reference value of the static modulation angle and the reference value of the transient phase shift angle. The method of the present invention accurately simulates the dynamic behavior of each state variable of the circuit, significantly enhances the dynamic performance and robustness of the system, and achieves a synchronous improvement in steady-state efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of power electronic converter control, and more specifically, to an adaptive optimization modulation method of a dual active bridge (DAB) triple phase shift angle based on a neural network sliding mode to improve the transmission efficiency and transient performance of the converter. Background Art

[0002] Under the "dual carbon" technology framework, new energy systems such as photovoltaic energy storage have become an important front for energy conservation and carbon reduction in the whole society. With the application of large-scale renewable energy and DC equipment, DC port power routers with high energy bidirectional conversion efficiency, new energy consumption rate, and flexible access capabilities for various loads have become a hot research topic today. It can realize the flexible connection of intermediate DC links such as photovoltaics and energy storage, reduce the power conversion loss after the introduction of the inverter, and improve energy utilization efficiency. Therefore, a multi-port power router with dual DC bus is an ideal solution.

[0003] As a power conversion unit and energy collection and coupling center with a three-level structure, the dual active bridge DC / DC converter carries the functions of current isolation, voltage conversion and bidirectional power transmission. Due to its advantages such as flexible bidirectional power transmission, high-frequency current isolation, wide voltage range and easy soft switching control, DAB has been widely used in energy storage systems, DC power grids and renewable energy generation.

[0004] Phase shift modulation is a classic control method for DAB converters. The easy-to-implement single phase shift (SPS) control can achieve zero voltage switching at unity gain, but under wide voltage conditions, it will have problems such as difficulty in soft switching, high circulating current, and low efficiency. Therefore, the most complex triple phase shift (TPS) control came into being, which can achieve minimum current stress, maximum transmission efficiency and soft switching range, but this modulation method relies on more complex analytical control theory.

[0005] The current mainstream analysis methods of DAB mainly include piecewise linear and harmonic analysis theories. Among them, piecewise linear theory relies on complex theoretical derivation and cannot obtain a unified expression applicable to all modes of DAB under TPS modulation; harmonic analysis is widely used in the field of steady-state performance optimization of artificial intelligence due to its global versatility, but its expression has high-order nonlinearity, which takes a long time to calculate the performance of different modulation strategies and is difficult to embed in the control program for terminal strategy deployment. In addition, since only a few low-order components can be selected in actual calculations, the model accuracy is not high. Therefore, a universal DAB analysis model that is applicable to all modes, simple in form, and does not require theoretical derivation is urgently needed.

[0006] At present, the research on DAB mainly focuses on two aspects, namely transient problems and steady-state problems. Compared with steady-state performance, there are relatively few studies on improving DAB transient performance. The main purpose of transient optimization is to eliminate the influence of non-ideal factors and improve the dynamic performance of the system. For example, natural switch surface control, feedforward control, and virtual direct power control are introduced to enhance the dynamic behavior of load disturbances and current or voltage reference changes, thereby improving the response speed and reducing overshoot when the working condition changes suddenly.

[0007] In order to ensure the dynamic performance in the actual control system, when the triple phase shift modulation scheme of DAB is formulated, two inner phase shift angles are usually generated by the steady-state optimization method, and the outer phase shift angle is generated by the PI control loop. This type of method is difficult to find the optimal PI control parameters, the outer phase shift angle has a weak ability to adjust the transient performance, and has the characteristics of slow output voltage tracking, large overshoot, and low system stability. Therefore, improved control strategies such as model prediction, dynamic matrix and sliding mode are proposed, which improve the dynamic performance to a certain extent, but all require complex theoretical derivation, lack of unified transient optimization control strategy for DAB full-mode triple phase shift modulation, and the method has low generalization. Due to the high nonlinearity of the DAB model under triple phase shift modulation, most of the above advanced control strategies are still limited to the single phase shift SPS control of DAB. Some literatures apply model predictive control to TPS control of DAB based on sub-modal derivation and convex optimization theory, but the model accuracy significantly affects the control performance, there is a certain system deviation, and it is difficult to obtain the optimal analytical solution of the control.

[0008] Although artificial intelligence has received extensive attention in power electronic equipment control in recent years due to its super nonlinear fitting and adaptive learning capabilities, the current application of artificial intelligence in DAB control is mostly concentrated in steady-state optimization, and transient control methods are almost all aimed at simple single-phase shift control. Since the mathematical model of DAB under three-phase shift modulation is highly nonlinear, complex artificial intelligence strategies are difficult to achieve real-time and rapid control in physical prototype systems. Therefore, there are few transient optimization and robust control methods for DAB under TPS control, and the application of AI technology is even more scarce.

[0009] From the above content, it can be seen that in practical applications, the transient control strategies of existing DAB three-phase shifted TPS all rely on complex multi-modal derivation and convex optimization theory solutions. A new robust control strategy that is universal to all modes, does not require theoretical derivation, and can achieve both transient performance and steady-state efficiency improvements is urgently needed. Summary of the invention

[0010] In view of the defects of the existing DAB analysis theory and transient control methods, the present invention proposes a DAB optimization modulation method based on artificial intelligence and circuit transient behavior. The method is based on a dual active bridge DAB unified triple phase shift (UTPS) hybrid optimization modulation method of neural network and circuit transient behavior, which accurately simulates the dynamic behavior of each state variable of the circuit, significantly enhances the dynamic performance and robustness of the system, and achieves a synchronous improvement in steady-state efficiency.

[0011] The technical solution adopted by the present invention is specifically as follows:

[0012] A DAB optimization modulation method based on artificial intelligence and circuit transient behavior, comprising:

[0013] Obtaining system parameters and operating parameters of the dual active bridge DC / DC converter, the system parameters include switching cycle, switching frequency, circuit output capacitance and transformation ratio; the operating parameters include input voltage, output voltage, load current at the current moment, reference value of output power at the next moment, and reference value of output voltage;

[0014] The input voltage at the current moment and the reference value of the output power and the reference value of the output voltage at the next moment are input into the trained first artificial neural network to predict the reference value of the optimal static modulation angle at the next moment. ;

[0015] Based on the obtained system parameters and operating parameters, the transient phase shift angle related expressions are solved to obtain the reference value of the transient phase shift angle. ;

[0016] Controlling the dual active bridge converter based on the obtained reference value of the static modulation angle and the reference value of the transient phase shift angle;

[0017] Wherein, the first artificial neural network is obtained by training based on the first training data set, with the input voltage, the reference value of the output power, the reference value of the output voltage under the specified working condition as input, and the reference value of the static modulation angle as output;

[0018] The transient phase-shift angle related expression is an expression of the secondary side H-bridge equivalent transmission current reference value at the next moment obtained by controlling the tracking error of the output voltage, and is obtained by inverse transformation in combination with the dual active bridge converter circuit characteristic formula and the analytical expression between the equivalent transmission current and the transient phase-shift angle per-unit value; the analytical expression between the equivalent output current and the transient phase-shift angle per-unit value is obtained by constructing a second artificial neural network based on a trained second artificial neural network; the second artificial neural network is obtained by training based on a second training data set, with the characteristics of each phase-shift angle per-unit value as input and the equivalent transmission current per-unit value as output.

[0019] Furthermore, the structure of the first artificial neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer is provided with at least two neurons.

[0020] Furthermore, the structure of the second artificial neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer is provided with only one neuron.

[0021] Furthermore, the input of the second artificial neural network contains at most The first term of .

[0022] Furthermore, the characteristic of the phase shift angle under the specified working condition is .

[0023] Furthermore, the first training data set is constructed by the following method:

[0024] A two-level optimization algorithm is used to minimize the inductor current peak and full soft switching operation as the optimization target to perform global optimization and obtain the given Optimal static phase shift angle modulation strategy under As each sample of the first training data set.

[0025] Furthermore, the method of controlling the tracking error of the output voltage includes adaptive sliding mode control and model predictive control.

[0026] Furthermore, the adaptive sliding mode control also includes introducing a nonlinear adaptive error compensation term, which is used to take effect after the operating condition suddenly changes for several cycles to eliminate potential output deviations.

[0027] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the DAB optimization modulation method based on artificial intelligence and circuit transient behavior is implemented.

[0028] A storage medium containing computer executable instructions, which, when executed by a computer processor, can implement the DAB optimization modulation method based on artificial intelligence and circuit transient behavior.

[0029] Beneficial effects of the present invention:

[0030] (1) The method of the present invention obtains an analytical expression for the DAB data-driven model under full-modal UTPS control through numerical modeling of circuit physics laws and sampling of physical prototypes, and uses advanced artificial intelligence fitting technology. It does not require complex sub-modal theoretical derivation, and the model is accurate and concise. Compared with the existing theoretical analysis model, it is easier to implement in actual control strategy deployment;

[0031] (2) The most advanced three-phase shift control strategy is adopted, which can achieve minimum current stress, maximum transmission efficiency and soft switching range, and has excellent steady-state performance;

[0032] (3) There is no need for complex derivation or application of optimization theory to calculate the optimal analytical solution to problems involving nonlinear objective functions and multiple constraints. The amount of calculation is small, and the control strategy has good generalization capabilities for different operating conditions and prototype parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0034] Figure 1 is a topological structure diagram of a dual active full bridge DAB in an embodiment of the present invention;

[0035] Figure 2 is a schematic diagram of TPS modulation of DAB in an embodiment of the present invention;

[0036] Figure 3 is a schematic diagram of a DAB equivalent circuit proposed in an embodiment of the present invention;

[0037] Figure 4 is a schematic diagram of the second training data set construction process in an embodiment of the present invention;

[0038] Figure 5 is a schematic diagram of the first artificial neural network construction process in an embodiment of the present invention;

[0039] Figure 6 It is a schematic diagram of a process flow of formulating a transient phase shift angle modulation strategy based on a neural network adaptive sliding mode control in an embodiment of the present invention;

[0040] Figure 7 is the efficiency curve of the DAB prototype connected to the resistance box load in the embodiment of the present invention ( );

[0041] Figure 8 is the efficiency curve of the DAB prototype connected to a constant power load in the embodiment of the present invention ( );

[0042] Fig. 9 It is a comparison diagram of the dynamic process of load switching of the DAB physical prototype resistor box under different control modes in the embodiment of the present invention; Fig. 9 (a) is a dynamic process diagram of the DAB prototype resistor box load switching under the control mode of Example 1. Fig. 9 (b) is a dynamic process diagram of the load switching of the DAB prototype resistor box under the control mode of Example 2. Fig. 9(c) is the dynamic process diagram of the load switching of the resistance box of the DAB prototype under PI control;

[0043] Fig.10 A comparison diagram of the dynamic process of reference output voltage switching when connected to a constant power load in different control modes in an embodiment of the present invention; Fig.10 (a) is a diagram showing the dynamic process of reference output voltage switching when a constant power load is connected under the control method of Example 1. Fig.10 (b) is a diagram showing the dynamic process of switching the reference output voltage when a constant power load is connected under the control method of Example 2. Fig.10 (c) in the figure is a dynamic process diagram of reference output voltage switching when connected to a constant power load under PI control. DETAILED DESCRIPTION

[0044] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited to these embodiments.

[0045] The present invention provides a DAB optimization modulation method based on artificial intelligence and circuit transient behavior, comprising the following steps:

[0046] Step 1: Obtain system parameters and operating parameters of the dual active bridge DC / DC converter. The system parameters include switching cycle, switching frequency, circuit output capacitance and transformation ratio; the operating parameters include input voltage, output voltage, load current at the current moment, reference value of output power at the next moment, and reference value of output voltage;

[0047] Step 2: Input the current input voltage, the reference value of the expected output power at the next moment, and the reference value of the output voltage into the trained first artificial neural network to predict the reference value of the optimal static modulation angle at the next moment. ;

[0048] Step 3: Based on the obtained system parameters and operating parameters, the transient phase shift angle related expression is used to solve the reference value of the transient phase shift angle. ;

[0049] Step 4: Control the dual active bridge converter based on the obtained reference value of the static modulation angle and the reference value of the transient phase shift angle.

[0050] In the present invention, the transient phase-shift angle related expression is obtained by inverse transformation of the expression of the output current reference value at the next moment obtained by controlling the tracking error of the output voltage, combined with the circuit characteristic formula of the dual active bridge converter and the analytical expression between the equivalent transmission current of the secondary side H bridge and the transient phase-shift angle per-unit value; the analytical expression between the equivalent transmission current and the transient phase-shift angle per-unit value is obtained based on the inverse transformation of the trained second artificial neural network spatial structure; the second artificial neural network is obtained by training based on the second training data set, with the features of the phase-shift angle per-unit value under the specified working conditions as input and the output current per-unit value as output.

[0051] The structure of the second artificial neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer has only one neuron and a tanh activation function is selected so as to obtain an analytical expression between the equivalent transmission current of the secondary H bridge and the per-unit value of the transient phase shift angle through the spatial structure and inverse transformation of the second artificial neural network. In a specific embodiment, the input of the second artificial neural network contains at most The first-order term is convenient for the optimal dynamic phase shift angle in subsequent transient control In another specific embodiment, in order to improve the fitting accuracy, the feature containing the phase shift angle under the specified working condition is greater than three dimensions, in addition to the original three input features In addition, it also contains other dimensional features, such as .

[0052] The second training data set is the equivalent output current of the secondary side of the circuit under different operating conditions and modulations The original data set, each sample of which contains the transient phase shift angle under the corresponding operating condition , static phase shift angle And the equivalent transmission current of the secondary side H bridge The second training data set can be obtained by performing a large number of numerical simulations based on the physical laws of the dual active bridge converter circuit and combining sampling of physical prototypes under multi-mode and multi-working conditions.

[0053] The first artificial neural network is obtained by training based on the first training data set, with the reference value of the output power and the reference value of the output voltage under the specified working condition as input and the reference value of the static modulation angle as output;

[0054] Each sample of the first training data set includes a given Optimal static phase shift angle modulation strategy under ,like Figure 5 As shown, it is constructed by the following method:

[0055] A two-level optimization algorithm is used to minimize the inductor current peak and full soft switching operation as the optimization target to perform global optimization and obtain the given Optimal static phase shift angle modulation strategy under As each sample of the first training data set.

[0056] The present invention uses the minimum output voltage Tracking deviation is the control target, that is, ,in The reference value of the DAB output voltage is used to obtain the expression of the reference value of the output current at the next moment, and the transient phase shift angle related expression is obtained by further combining the circuit characteristic formula of the dual active bridge converter and the analytical expression between the output current and the per-unit value of the transient phase shift angle. Among them, the control methods include adaptive sliding mode control, model predictive control (MPC), etc.

[0057] The present invention will be further described below in conjunction with specific embodiments and accompanying drawings:

[0058] Embodiment 1:

[0059] First, a DAB optimization modulation strategy based on artificial intelligence and circuit transient behavior is formulated, including:

[0060] Based on the TPS modulation principle and the physical laws of DAB circuits, through prototype sampling and numerical simulation, the second training data set of the changes in circuit state variables under different phase shift angle modulation under specific working conditions is obtained and passed to the second artificial neural network (Artificial Neural Network) architecture. The general data-driven model of DAB under full-modal modulation is obtained by network space deconstruction and inverse transformation, which provides a theoretical analysis basis for the subsequent modulation strategy formulation and takes into account both computational lightweight and model refinement.

[0061] In order to ensure the transient control margin, the present invention shifts the phase angle of the secondary side H bridge under the same working state. and outward phase angle As a static quantity, it is regarded as a constant, and only the phase shift angle in the primary H bridge is As a dynamic adjustment quantity in real-time control, the invention uses a double-layer nested heuristic algorithm, takes the minimum current stress and full soft switching (ZeroVoltage Switching, ZVS) operation as the optimization goal, and obtains the static phase shift angle reference value that ensures the optimal steady-state efficiency of the system. , obtain a first training data set, and combine artificial intelligence fitting technology, based on the first training data set, with the reference value of the output power and the reference value of the output voltage under the specified working condition as input, and the reference value of the static modulation angle as output to train and obtain a trained first artificial neural network. The trained first artificial neural network is a universal expression of the optimal static phase shift angle under any working condition, which is convenient for the deployment of the underlying control strategy.

[0062] Finally, with the minimum output voltage Tracking deviation is the control target, that is, ,in For the reference value of DAB output voltage, a nonlinear sliding mode surface and anti-jitter reaching law are preliminarily designed, and the real-time operating status of the equipment, adaptive compensation terms and optimal static phase shift angle are input into the DAB general data-driven model under full-mode modulation. Through neural network deconstruction and inverse transformation, a neural network-based adaptive sliding mode transient control strategy is formulated to achieve a significant improvement in dynamic performance and system robustness.

[0063] Specifically, the following steps are included:

[0064] Step 1: Construct a general data-driven model for DAB under full-modal modulation, including the following sub-steps:

[0065] Step 1.1: Construct the second training dataset:

[0066] The circuit topology of DAB is as follows Figure 1 As shown, the primary side H-bridge inverter and secondary side H-bridge rectifier Controllable switches and The switching cycle of the controllable switch is , the switching frequency is ; Circuit output capacitance Used to reduce output voltage ripple, The input (primary H-bridge) and output (secondary H-bridge) voltages are connected by inductors. Connected to an isolation transformer T with a transformation ratio of n:1, the primary voltage of the transformer is Size , v cd is the transformer secondary H-bridge input voltage; The initial opening time is defined as the starting point 0, and the TPS modulation waveform is as follows Figure 2 Phase shift control is the mainstream control method for DAB steady-state operation. for Hysteresis The duty cycle is used to control the direction and size of power flow, and the internal phase shift angle Used to reduce cycle power and expand soft switching ZVS range, external phase shift is the primary H-bridge output voltage and The phase shift between the primitive components, where , is half a switching cycle.

[0067] Based on this, we get Figure 3 The dual active full-bridge DAB equivalent circuit diagram shown in FIG. Figure 3 The DAB equivalent circuit and Kirchhoff's law show that the inductor current , the average value of DAB transmission power in a single cycle , output voltage , secondary output current The following dynamic equations are satisfied:

[0068] (1)

[0069] In the formula, is the equivalent transmission current of the secondary H bridge, that is The average value in one switching cycle, is the load current of DAB. For the interval time.

[0070] Furthermore, if Figure 4 As shown, the actual prototype is sampled under multi-mode and multi-condition Each set of data is then subjected to a large number of numerical simulations based on the above circuit physics laws to obtain the equivalent transmission current of the secondary side H bridge of the circuit under different operating conditions and triple phase shift modulation strategies. ; Then get the second training data set As a preferred solution, the equivalent transmission current of the H bridge on the secondary side of the circuit can also be estimated. Rated value ; then by dividing Will Normalized to variables that are independent of operating conditions , and use the function Further and Standardize and get the corresponding per-unit value as each sample, that is, get the second training data set .in:

[0071] (2)

[0072] Step 1.2: Construct the second artificial neural network to obtain the DAB general data-driven model under full-modal modulation:

[0073] Build the neural network architecture ANN2 of the DAB data-driven model to be used for the optimal transient phase shift angle reference value in the subsequent minimum output voltage tracking deviation problem When using a feedforward neural network to fit a nonlinear mapping relationship, only one neuron is set in the hidden layer, and the tanh activation function is selected. In order to improve the fitting accuracy, the original three input features, namely the phase shift angle of the secondary H bridge, , outward shift phase angle And the phase shift angle of the primary H bridge Expanded to multi-dimensional space, in this embodiment, the corresponding per-unit value is shown as an example, using , whose input elements contain at most the first-order term of the phase shift angle in the primary H-bridge, which is convenient for the optimal transient phase shift angle reference value in subsequent transient control Combined with the three-layer network structure of the feedforward neural network and the trained neuron connection weights and bias Parameters, obtained by neural network space deconstruction and inverse transformation, Phase shift angle per unit value The analytical expression between and further obtains Phase shift angle per unit value The analytical expression between is the general data-driven model of DAB under full-modal modulation, as follows:

[0074] (3)

[0075] In the formula, are coefficients of the second artificial neural network model, determined based on the second training data set. represents the connection weight between the neuron in the i+1th layer of the network and the jth neuron in the ith layer, for example represents the connection weight between the neuron in the hidden layer (the second layer) and the second neuron in the input layer (the first layer), is the bias of the i+1th layer of neurons. In this embodiment, the hidden layer and the output layer contain only one neuron, then , They represent the bias of neurons in the hidden layer (the second layer) and the output layer (the third layer), respectively.

[0076] Expand the previous continuous formula through Euler difference, then in discrete time, Output voltage at all times The nonlinear relationship with the three phase shift angles is as follows, where To control the cycle.

[0077] (4)

[0078] The DAB general data-driven model under full-mode modulation obtained above provides a complete theoretical analysis basis for the formulation of the subsequent optimal static phase shift angle and transient adaptive control strategy.

[0079] Step 2: Construct the first trained artificial neural network, including the following sub-steps:

[0080] Step 2.1: Build the first training dataset:

[0081] Determine the optimal static phase shift angle for minimum current stress and full soft switching operation.

[0082] In DAB three-phase shift modulation, current stress and soft switching are often used as important performance indicators. The phase shift angle in the primary H bridge is selected according to the adjustment margin of each phase shift angle on system performance. is the transient adaptive control quantity, Set to a steady-state constant static phase shift angle. When using the modulation strategy, the inductor current peak value is usually minimized. And full soft switching operation is the optimization goal to achieve the best steady-state operating efficiency, where the requirements for achieving full soft switching ZVS operation are as follows.

[0083] (5)

[0084] In order to reduce the burden of the algorithm in dealing with nonlinear constraints, complex constraints such as full soft switching operation and standard transmission power are imposed on the objective function by introducing penalty functions, and the original problem is transformed into an unconstrained single-objective optimization problem.

[0085] (6)

[0086] in, is the reference value of output power, is the active power penalty factor. After repeated experiments, Set to 200; when the modulation strategy meets the soft switching ZVS constraint, Set it to 1, otherwise to 10, to ensure the effective implementation of ZVS.

[0087] Then, a two-level optimization algorithm with strong global search capability and fast convergence speed is designed to pass the suboptimal solution obtained by the rapid convergence of the particle swarm to the lower-level genetic algorithm to ensure efficient global optimization, thereby obtaining the given Optimal static phase shift angle modulation strategy under .in, Indicates the output voltage reference value, Respectively Reference value; its specific implementation process is as follows Figure 5As shown, first give the required operating conditions , where the operating conditions can be set according to requirements, for example, The range is (0-1kW), The range is (150-300V), The range is (50-150V); the previous particle swarm algorithm PSO generates a preliminary modulation strategy And the speed corresponding to the three phase shift angle updates , the subscript i represents the number of particles in the particle swarm algorithm, the initial modulation strategy Substitute into the equivalent circuit obtained in step 1, combined with Figure 2 The modulation process and formula (1) are used to calculate the peak value of the inductor current under this strategy. , the average value of DAB transmission power in a single cycle and soft switching conditions, combined with formula (6) to achieve strategy performance Rapid evaluation; then update the current particle individual optimal strategy , particle swarm global optimal strategy , and adjust the function completion rate according to and strategies The update is repeated until the convergence condition is reached; then, the best strategy obtained by the previous particle swarm algorithm PSO is set as the initial population of the subsequent genetic algorithm, and the operation of the previous algorithm is repeated, and selection, crossover, mutation, and continuous update and iteration are performed to obtain the final global optimal static modulation angle as the reference value of the static modulation angle under the corresponding working condition. .

[0088] Subsequently, different operating conditions are traversed, and a large number of optimal modulation strategy data sets obtained by offline optimization of the above two-layer nested algorithm are used to obtain the first training data set. .

[0089] Step 2.2: Build the first artificial neural network:

[0090] Input the first training data set into the constructed first artificial neural network ANN1 for training to obtain the specified working condition and the optimal static modulation angle reference value The nonlinear mapping relationship between them is directly deployed into the underlying controller to realize the real-time optimal steady-state modulation of the underlying static modulator. In this embodiment, the structure of the first artificial neural network includes an input layer, a hidden layer and an output layer, wherein the number of input layer features is 3, which are input voltage , output voltage reference value and transmission power reference value , the hidden layer is set with 5 neurons, and the output result is two optimal static phase shift angle reference values .

[0091] Step 3: Formulate a transient phase shift angle modulation strategy based on neural network adaptive sliding mode control, such as Figure 6 shown.

[0092] In DAB transient control, the output voltage is usually Tracking is the goal, then the output voltage tracking error of the modulation strategy is rr for:

[0093] (7)

[0094] Therefore, the sliding surface s is designed as follows: is the proportional coefficient of each state variable of the sliding surface.

[0095] (8)

[0096] Then the time derivative of the sliding surface is As shown below:

[0097] (9)

[0098] From the circuit equivalent model, it can be found that DAB under UTPS modulation is a highly nonlinear system. In order to prevent the chattering phenomenon caused by the sign function, the sliding mode control selects the exponential reaching law and replaces the original sign function with the tanh activation function. Used to modulate the approach rate, It is used to adjust the switching rate of the sliding mode output control strategy, and the sliding surface reaching law The design is as follows:

[0099] (10)

[0100] Combining equations (9) and (10) and circuit characteristic equation (1) gives the first The secondary H-bridge output current reference value required at the moment for:

[0101] (11)

[0102] In the formula, Indicates the rate of change of the output voltage reference value at the kth moment.

[0103] Furthermore, the improvement of the output voltage jitter phenomenon may cause a small output voltage error for a period of time after the dynamic change, and the problem of slow bias decay. Therefore, in a preferred solution, a nonlinear adaptive error compensation term e is introduced. sIn this embodiment, it will take effect after the operating condition changes suddenly for ten cycles, so as to achieve rapid elimination of potential output deviation without introducing overshoot. is the adaptive compensation coefficient. The nonlinear adaptive error compensation term at the k+1th moment is expressed as follows:

[0104] (12)

[0105] In the formula, Indicates the output voltage reference value The moment when the mutation begins.

[0106] The equivalent output current of the secondary H-bridge obtained in step 1 Phase shift angle per unit value The analytical expression of the neural network data model between (3) shows that the dynamic phase shift angle per unit value is and and two static phase shift angles per unit value The following nonlinear relationship is satisfied.

[0107] (13)

[0108] Then the secondary side H bridge output current reference value in equation (11) is and the nonlinear adaptive error compensation term e in equation (12) s Substitute it into equation (13), and then get the optimal transient phase shift angle of the controller at the k+1th moment: .

[0109] (14)

[0110] Therefore, while accurately tracking the DAB output active power and voltage, the static and transient phase shift angles can be reasonably adjusted to achieve a dual improvement in the DAB transient performance and steady-state efficiency of the converter.

[0111] After formulating the DAB optimization modulation strategy based on artificial intelligence and circuit transient behavior of the present invention, the DAB optimization modulation method based on artificial intelligence and circuit transient behavior of the present invention can be carried out, which is specifically as follows:

[0112] Acquire the operating parameters of the dual active bridge converter in real time, including the input voltage, output voltage, load current at the current moment, the reference value of the output power at the next moment, and the reference value of the output voltage;

[0113] Input the current input voltage, the reference value of the expected output power at the next moment, and the reference value of the output voltage to the trained first artificial neural network deployed in the controller in step 2 to predict the reference value of the static modulation angle at the next moment. ;

[0114] Based on the obtained operating parameters and system parameters, the transient phase shift angle related expressions (including formula 14, formula 12 and formula 11) deployed in the controller are used to solve the reference value of the transient phase shift angle. ; Among them, the system parameters include the switching period, switching frequency, transformation ratio, circuit output capacitance, etc. of the dual active bridge converter.

[0115] Adaptive sliding mode control is performed on the dual active bridge converter based on the obtained reference value of the static modulation angle and the reference value of the transient phase shift angle.

[0116] Embodiment 2:

[0117] This embodiment is the same as the embodiment 1, except that in step 3 of the process of formulating a DAB optimization modulation strategy based on artificial intelligence and circuit transient behavior of the present invention, the adaptive sliding mode control is changed to model predictive control, and the remaining steps are the same as the embodiment 1.

[0118] When the model predictive control method is used to modulate the transient phase shift angle, the minimum output voltage error is also taken as the goal. To facilitate the calculation of the optimal solution, the control target is designed as Then the secondary side H bridge equivalent output current Phase shift angle per unit value The analytical expression (3) of the neural network data model ANN2 between and the circuit differential equation (4) of the output port of the DAB circuit are used as the prediction model and substituted into the control target. Then, the original output voltage control problem is transformed into the following convex optimization problem, where is the intermediate coefficient.

[0119] (15)

[0120] Then, for the above nonlinear convex optimization problem, the Karush-Kuhn-Tucker (KKT) condition is used to obtain The best transient phase shift angle reference value The following equation must be satisfied:

[0121] (16)

[0122] in, is the unknown coefficient that changes with the actual working conditions, and its specific calculation formula is:

[0123] (17)

[0124] in, is the connection weight between the hidden layer neurons and the output neurons of the second artificial neural network, is the bias of the output neuron of the second artificial neural network. Similarly, we can find The optimal solution when is , as shown in formula (18). Therefore, according to the above model predictive control method, formula (16) or formula (18) can be selected according to the actual value of k1 to realize the formulation of the transient phase shift angle reference value.

[0125] (18)

[0126] Further, after formulating the DAB optimization modulation strategy based on artificial intelligence and circuit transient behavior of the present invention, the DAB optimization modulation method based on artificial intelligence and circuit transient behavior of the present invention can be carried out, which is specifically as follows:

[0127] Acquire the operating parameters of the dual active bridge converter in real time, including the input voltage, load current, output voltage at the current moment, the reference value of the output power at the next moment, and the reference value of the output voltage;

[0128] Input the current input voltage, the reference value of the expected output power at the next moment, and the reference value of the output voltage to the trained first artificial neural network deployed in the controller in step 2 to predict the reference value of the static modulation angle at the next moment. ;

[0129] Based on the obtained operating parameters and system parameters, the transient phase shift angle related expressions deployed in the controller (including formula (16) and formula (17) or formula (18)) are used to solve the reference value of the transient phase shift angle. ; Among them, the system parameters include the switching period, switching frequency, transformation ratio, circuit output capacitance, etc. of the dual active bridge converter.

[0130] Adaptive sliding mode control is performed on the dual active bridge converter based on the obtained reference value of the static modulation angle and the reference value of the transient phase shift angle.

[0131] In order to further verify the effectiveness of the control strategy proposed in this invention, an inductor The output capacitance is 54mH The DAB prototype with a transformer ratio of 5:3 has a switching frequency of The control cycle is 100kHz. The control time of the control circuit of the present invention is 0.1 ms, the underlying microcontroller is DSP-TMS320F28335, and two load types, a resistor box and a constant power load with nonlinear impedance, are selected. The modulation method of the present invention (Example 1 and Example 2) is compared with a single-phase-shifted PI control (SPS-PI) and a unified three-phase-shifted PI control (UTPS-PI), wherein the control strategy of the present invention uses the same static phase shift angle as TPS-PI in steady-state control.

[0132] Figure 7 Input voltage The reference output voltage is 255V and When the voltage is 150V, the efficiency curve of DAB with a resistor box as the load at different powers is shown. It can be seen that the effect of the present invention is better than the triple phase-shift modulation efficiency under PI control, especially when the load power is 450W, the efficiency of the neural network adaptive sliding mode control strategy of Example 1 of the present invention can reach 98.03%, and under the condition that the output voltage gain is close to 1, the efficiency is still one percentage point higher than that of the traditional single phase-shift SPS control. When the load power is less than 300W, the efficiency of the strategy of the present invention is also significantly higher than the triple phase-shift modulation efficiency under PI control.

[0133] Figure 8 When connected to a constant power load, the input voltage The reference output voltage is 255V and It is the efficiency comparison result of four control strategies under 100V working condition. Among them, when the load power is greater than 180W, the modulation method of the embodiment of the present invention has excellent modulation efficiency, and the neural network sliding mode control strategy of Example 1 is basically at the highest efficiency point, much higher than SPS-PI control, and almost the same as the UTPS-PI method. After the load power is less than 180W, the efficiency advantage of the strategy of the present invention compared to SPS-PI gradually increases. It can be seen that the modulation method of the present invention has a higher steady-state efficiency advantage compared to the current mainstream advanced control strategy.

[0134] Fig. 9 Input voltage The reference output voltage is 255V and The dynamic experimental waveform of DAB load power switching when the voltage is 150V and the resistance box is connected. To switch the load power from 400W to 220W, By comparison, we can find that Fig. 9 (a) Fig. 9 (b) and Fig. 9 As shown in (c), the modulation method of Example 1 of the present invention has almost no steady-state voltage deviation and a small dynamic overshoot of the output voltage under this sudden change of operating conditions. In addition, it should be pointed out that compared with the MPC control of Example 2 with obvious steady-state voltage deviation and the PI control with large dynamic overshoot, the neural network adaptive sliding mode control in Example 1 has better transient performance.

[0135] Fig.10 Input voltage The dynamic process comparison diagram of DAB reference output voltage switching under 210V and 100W constant power load, where: The moment when the soft start ends and the 1.8A constant current mode is entered. The time limit current ends and the reference output voltage is Switch from 100V to 60V at any time. On the contrary, it switches from 60V to 100V. The input voltage switches from 210V to 270V. Fig.10 (a) Fig.10 (b) and Fig.10 From the dynamic results of (c) in the figure, it can be found that the method of the present invention can maintain good transient performance under complex mutation conditions when connected to a constant power load with negative impedance characteristics, with almost no overshoot and deviation, while the output voltage of PI control shows a larger dynamic overshoot and steady-state recovery time when connected to a constant power load. Therefore, the method of the present invention still has excellent control accuracy when there is a deviation in the model, and has good dynamic performance and robustness. In particular, Example 1 adopts a neural network-based adaptive sliding mode control method, which has better transient performance and system robustness than the traditional proportional-integral control and the MPC control of Example 2.

[0136] Corresponding to the aforementioned embodiment of a DAB optimization modulation method based on artificial intelligence and circuit transient behavior, the present invention also provides an electronic device, comprising one or more processors, for implementing the DAB optimization modulation method based on artificial intelligence and circuit transient behavior in the above embodiment.

[0137] The device embodiment can be implemented by software, hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor of any device with data processing capability in which it is located to read the corresponding computer program instructions in the non-volatile memory into the memory for execution. From the hardware level, it mainly includes a processor, memory, network interface, and non-volatile memory. In addition, any device with data processing capability in which the device in the embodiment is located can also include other hardware according to the actual function of the device with data processing capability, which will not be described in detail.

[0138] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0139] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Ordinary technicians in this field can understand and implement it without paying creative work.

[0140] An embodiment of the present invention further provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, a DAB optimization modulation method based on artificial intelligence and circuit transient behavior in the above embodiment is implemented.

[0141] The computer-readable storage medium may be an internal storage unit of any device with data processing capability described in any of the aforementioned embodiments, such as a hard disk or a memory. The computer-readable storage medium may also be any device with data processing capability, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), an SD card, a flash card (Flash Card), etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capability and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capability, and may also be used to temporarily store data that has been output or is to be output.

[0142] Those skilled in the art to which the present invention relates may make various modifications or additions to the specific embodiments described or replace them in a similar manner, but they will not deviate from the spirit of the present invention or exceed the scope defined by the attached claims.

Claims

1. A DAB optimization modulation method based on artificial intelligence and circuit transient behavior, characterized in that: include: Obtaining system parameters and operating parameters of the dual active bridge DC / DC converter, the system parameters include switching cycle, switching frequency, circuit output capacitance and transformation ratio; the operating parameters include input voltage, output voltage, load current at the current moment, reference value P of output power at the next moment, and reference value of output voltage; The internal phase shift angle D2 and the external phase shift angle D f As the static modulation amount, the phase shift angle D1 in the primary H bridge is used as the dynamic modulation amount; and the input voltage at the current moment and the reference value of the output power and the reference value of the output voltage at the next moment are input into the trained first artificial neural network to predict the internal phase shift angle D2 and the external phase shift angle D2 of the secondary H bridge at the next moment. f The reference value of the optimal static modulation angle is The superscript asterisk indicates the reference value of the corresponding symbol parameter; Based on the obtained system parameters and operating parameters, the transient phase shift angle related expression is used to solve the primary H-bridge internal phase shift angle D1, i.e. the reference value of the transient phase shift angle. Controlling the dual active bridge converter based on the obtained reference value of the static modulation angle and the reference value of the transient phase shift angle; The first artificial neural network is based on a first training data set, with an input voltage V1 and a reference value P of the output power under specified working conditions. * , output voltage reference value V2 * The reference value of the input and static modulation angle is obtained by training the output; The transient phase-shift angle related expression is an expression of the secondary side H-bridge equivalent transmission current reference value at the next moment obtained by controlling the tracking error of the output voltage, and is obtained by inverse transformation in combination with the dual active bridge converter circuit characteristic formula and the analytical expression between the equivalent transmission current and the transient phase-shift angle per-unit value; the analytical expression between the equivalent transmission current and the transient phase-shift angle per-unit value is obtained by constructing a second artificial neural network based on a trained second artificial neural network; the second artificial neural network is obtained by training based on a second training data set, with the characteristics of each phase-shift angle per-unit value as input and the equivalent transmission current per-unit value as output.

2. The method according to claim 1, characterized in that The structure of the first artificial neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer is provided with at least two neurons; the structure of the second artificial neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer is provided with only one neuron.

3. The method according to claim 1, characterized in that The input of the second artificial neural network contains at most a first-order term of the transient phase shift angle D1.

4. The method according to claim 1, characterized in that: The characteristic of each phase shift angle per unit value is 5. The method according to claim 1, characterized in that: The first training data set is constructed by the following method: using a double-layer optimization algorithm to minimize the inductor current peak value i Lmax and full soft switching operation as the optimization target to perform global optimization and obtain the given Optimal static phase shift angle modulation strategy under As each sample of the first training data set.

6. The method according to claim 1, characterized in that The methods of controlling the tracking error of the output voltage include adaptive sliding mode control and model predictive control.

7. The method according to claim 6, characterized in that The adaptive sliding mode control also includes introducing a nonlinear adaptive error compensation term, which is used to take effect after the operating condition suddenly changes for several cycles to eliminate potential output deviations.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the DAB optimization modulation method based on artificial intelligence and circuit transient behavior as described in any one of claims 1 to 7 is implemented.

9. A storage medium comprising computer executable instructions, wherein the computer executable instructions, when executed by a computer processor, implement the DAB optimization modulation method based on artificial intelligence and circuit transient behavior as claimed in any one of claims 1 to 7.

Images