A deep learning-based adaptive hierarchical control method for direct current microgrid

Abstract

The application discloses a kind of based on deep learning's direct current microgrid self-adapting hierarchical control method, comprising: based on deep learning's direct current microgrid stable boundary online identification model, for real-time capture system dynamic characteristics;And the control strategy of distributed real-time in stable boundary and the adaptive stabilizing control strategy of stable boundary outside based on neural network.It is realized that the complete distributed real-time power coordination control based on bus voltage signal sharing in direct current microgrid stable boundary is realized by constructing multi-time scale hierarchical control framework, in combination with system dynamic characteristic analysis, while using deep learning method to actively stabilize management to abnormal operating state outside stable boundary.The application can improve the robustness and adaptability of system under variable operating conditions, ensure the safe and reliable operation of direct current microgrid, improve its flexibility and expansibility.

Description

Technical Field

[0001] This invention relates to the field of DC grid technology, and in particular to an adaptive hierarchical control method for DC microgrids based on deep learning, which combines system stabilization and power coordination optimization. Background Technology

[0002] With the development of renewable energy and energy storage technologies, DC microgrids, as a type of small-scale power system, are playing an increasingly important role in distributed energy integration and flexible power supply. Compared to traditional AC microgrids, DC microgrids have significant advantages in improving energy efficiency, simplifying structure, and reducing conversion losses. However, because DC microgrids primarily rely on power electronic devices for energy conversion, the system exhibits relatively weak inertia. Furthermore, the system is affected by various dynamic factors such as load fluctuations, the plug-and-play nature of distributed power sources, and the fast response characteristics of power electronic devices, further exacerbating the challenges in system stability and control strategies.

[0003] Currently, control strategies for DC microgrids can be categorized into centralized, distributed, and decentralized control modes based on different communication methods. In centralized control, the central controller collects and processes information from the power supply and load, and transmits power commands to local controllers via a communication network. While centralized control is effective, it has inherent limitations, including susceptibility to single-point failures, slow dynamic response due to communication delays, and poor system flexibility and scalability. Distributed control, although enhancing system fault tolerance, struggles to ensure globally optimal control performance. In contrast, decentralized control, which does not rely on communication networks or common signals, is becoming a more suitable solution for DC microgrids due to its plug-and-play nature, high flexibility, and good scalability. However, current research often neglects online identification of system stability boundaries, focusing excessively on real-time control within these boundaries while lacking effective management of operating states outside them. This limitation makes existing control strategies ill-suited to handle dynamic changes across multiple time scales, resulting in insufficient robustness under different operating conditions and increasing the risk of system instability.

[0004] With the rapid development of artificial intelligence technology, deep learning, as a powerful data-driven modeling tool, has demonstrated unique advantages in handling complex nonlinear systems. Applying deep learning to the control strategy design of DC microgrids enables in-depth mining and accurate modeling of the system's dynamic characteristics, providing new ideas for real-time identification of stability boundaries and stabilization control. However, there is currently a lack of a deep learning-based global adaptive control framework capable of identifying stability boundaries in complex dynamic environments and adjusting control strategies in real time to cope with internal and external disturbances.

[0005] Therefore, there is an urgent need to develop an adaptive hierarchical control strategy for DC microgrids based on deep learning, which combines deep learning to model system stability and identify boundaries, so as to realize decentralized real-time control within the stability boundary and adaptive stabilization control outside the stability boundary. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by proposing an adaptive hierarchical control method for DC microgrids based on deep learning. This method uses an artificial neural network to fit the implicit stability boundary of the system and provides stabilization control actions, thereby achieving adaptive distributed control of the system both inside and outside the stability boundary.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] An adaptive hierarchical control method for DC microgrids based on deep learning, characterized by including:

[0009] Step 1: Construct a fully decentralized control system based on shared bus voltage. Divide and control the operating modes of each unit within the system according to the fluctuation range of the DC microgrid bus voltage. The units within the system include photovoltaic units, which generate electricity under sunlight conditions; energy storage units, which store and release electricity to balance the power demand in the grid; and load units, which consume electricity from the grid.

[0010] Step 2: Construct a small-signal system model A with implicit stability boundaries:

[0011]

[0012] Among them, B1, B2, ... B n Let C be the small-signal model of n bus control units, and C be the equivalent small-signal model of the constant power load unit.

[0013] Step 3. Collect the droop coefficient and bus voltage value of the DC bus voltage control unit in the bottom-level fully distributed control model, calculate the system's static operating point, and input it into the small-signal model A to solve for the system's minimum damping ratio.

[0014]

[0015] According to the system instability criteria, the stability label is marked. When the minimum damping ratio corresponding to the system characteristic value exceeds 3%, the system is considered to be in a stable state; conversely, when the minimum damping ratio is less than 3%, the system is considered to be in an unstable state.

[0016] Step 4. Using the droop coefficient of each unit under bus voltage control and the bus voltage value as inputs and the stability label as outputs, train the artificial neural network 1 to fit the implicit stability boundary of the system.

[0017] Step 5. Based on the existing system of shared bus voltage signals, build an over / under voltage control strategy for the expansion unit to stabilize the system when it exceeds the stability boundary;

[0018] Step 6. Construct a small-signal model A' based on the structure and control strategy of the expanded system:

[0019]

[0020] Among them, D1, D2, ... D k These are small-signal models for k expansion control units;

[0021] Step 7: For cases exceeding the stability boundary, search for an optimal value within the range of the droop coefficient of the expansion unit, so that the expanded system can return to the stability boundary and the power borne by the expansion unit is minimized.

[0022] Step 8: Using the droop coefficient of each bus voltage unit, the DC bus voltage value, and the minimum damping ratio of the system before and after capacity expansion as inputs when the system exceeds the stability boundary, and the optimal value of the droop coefficient of the expansion unit as output, train the artificial neural network 2 to fit the relationship between the capacity expansion control effect and the droop coefficient, so as to guide the system to achieve real-time adaptive stabilization control.

[0023] Step 9: In the real-time control of the DC microgrid, the trained neural network 1 is used to determine the stability of the system in real time (whether it exceeds the stability boundary). If it does not exceed the stability boundary, the system operates in the fully distributed real-time control mode based on shared bus voltage before the expansion. If it exceeds the stability boundary, the artificial neural network 2 will adaptively guide the expansion unit to participate in the system stabilization control and restore the system stability.

[0024] Furthermore, the working modes in step 1 include:

[0025] Based on the rated value of the bus voltage ±5%, set the maximum allowable voltage fluctuation U. H2 Minimum value U L2 ;

[0026] Based on the rated value of the bus voltage ±3.33%, the maximum voltage U of the energy storage regulation range is set. H1 and minimum value U L1 ;

[0027] In mode 1, the voltage fluctuation range is U. H1 dc ≤U H2 The photovoltaic unit operates in droop mode to maintain bus voltage stability. Its output voltage and droop coefficient are expressed as follows:

[0028] U dc ​=U H2 -k pv P pv

[0029]

[0030] Where, k pv P is the droop factor of the photovoltaic unit. pv_max This represents the maximum output power of the photovoltaic system.

[0031] In mode 2, the voltage fluctuation range is U. N dc ≤U H1 The energy storage unit operates in droop charging mode to maintain bus voltage stability. Its output voltage and droop coefficient are expressed as follows:

[0032] U dc =U H1 +k b P bat

[0033]

[0034] Where, k b P is the droop factor of the energy storage unit. cha_lim This represents the maximum charging power of the energy storage unit.

[0035] In mode 3, the voltage fluctuation range is U. L1 dc ≤U N The energy storage unit operates in droop discharge mode to maintain bus voltage stability. Its output voltage and droop coefficient are expressed as follows:

[0036] U dc =U N +k b P bat

[0037]

[0038] In Mode 4, the bus voltage fluctuation range is U. L2 dc ≤U L1 Both the photovoltaic unit and the energy storage unit are in power-limited control mode, and the voltage fluctuation is dominated by the load unit.

[0039] Furthermore, step 2, which constructs a small-signal system model A with an implicit stability boundary, specifically includes:

[0040] ​​​Step 2.1: Equivalently model each part of the DC microgrid, and obtain the relationship between voltage and current based on the bidirectional buck-boost circuit structure of the DC bus voltage control unit in droop mode;

[0041]

[0042] Among them, C i L i R i and d i These represent the output regulating capacitor, filter inductor, equivalent resistance, and duty cycle of the i-th buck-boost converter, respectively; u oi i Li i oi and u dc This represents the converter output voltage, input-side inductor current, output current, and bus voltage.

[0043] Step 2.2: Based on the voltage and current dual closed-loop control, the DC bus voltage control unit adopts droop control to achieve coordinated power distribution among multiple units. Its corresponding control model can be expressed as:

[0044]

[0045] In the formula u refi k i Here, k represents the reference value for the output voltage and the droop factor of the i-th buck-boost converter. pi_i k ii_i u ir_i G i_i (s)=k pi_i +k ii_i / s represents the proportional-integral parameter integral term output and transfer function of the inner loop PI controller; k pu_i k iu_i u ur_i G u_i (s)=k pu_i +k iu_i / s represents the proportional-integral parameters, integral term output, and transfer function of the outer loop PI controller;

[0046] Step 2.3: The model for the control unit operating under power limiting and the multiple constant power loads connected in parallel is simplified as follows:

[0047]

[0048] Step 2.4: Combine the above equations and add a small-signal perturbation to obtain the small-signal model A of the system:

[0049]

[0050] Among them, B1, B2, ... B n Let C be the small-signal model of n bus control units, and C be the equivalent small-signal model of a constant power load unit.

[0051] When U dc ≥U H2 Within this range, the expansion unit converter operates in droop charging mode, the overvoltage controller reaches its output limit, and the droop coefficient of the expansion unit must meet the following requirements:

[0052]

[0053] WhenU dc ≤U L2 Within this range, the converter operates in droop discharge mode, the undervoltage controller reaches its output limit, and the droop coefficient of the expansion unit must meet the following requirements:

[0054]

[0055] When U L2 ≤U dc ≤U H2 At this point, both the overvoltage controller and the undervoltage controller exceed their output limits. Within this range, the expansion unit can adaptively adjust the charging and discharging power based on the droop coefficient. The current reference value, after being organized, is simplified as follows:

[0056]

[0057] Where I' cha_lim , I' dis_lim These correspond to the maximum charging current and the maximum discharging current for energy storage expansion, respectively.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1) This invention constructs an artificial neural network model to fit the implicit stability boundary of the system. While ensuring computational accuracy, it can calculate the stability margin of the system in real time and judge the stability of the system, which can be used to guide the subsequent stabilization control of the system.

[0060] 2) This invention constructs a fully distributed underlying control model based on shared bus voltage signals. When the system is within the stability boundary, distributed real-time control is achieved within the stability boundary under the proposed bus voltage signal sharing control. When the system exceeds the stability boundary, the expansion unit is guided by artificial neural network 2 to participate in the distributed control at the underlying level, achieving adaptive stabilization control outside the stability boundary and helping the system to restore stability.

[0061] 3) When training the artificial neural network 2 to fit the droop coefficient and stabilization control effect of the expanded control unit, this invention not only ensures the effectiveness of its stabilization control, but also considers the economic efficiency of the expansion cost of the expanded unit. Attached Figure Description

[0062] Figure 1 This is a flowchart illustrating an adaptive hierarchical control method for DC microgrids based on deep learning, according to an embodiment of the present invention.

[0063] Figure 2 This is the structural framework of the 600V low-voltage DC microgrid system in this embodiment;

[0064] Figure 3 This is a diagram showing the division of the working modes of each unit under the shared control of the bus voltage signal before the expansion of the DC microgrid in this embodiment;

[0065] Figure 4 This is a schematic diagram of the implicit stability boundary fitted by neural network 1 in this embodiment.

[0066] Figure 5 This is a block diagram of the over / undervoltage control strategy for the expansion unit in this embodiment.

[0067] Figure 6 This diagram illustrates the operational modes of the expansion unit at different voltage levels in this embodiment.

[0068] Figure 7 This is a flowchart of the data acquisition process for Neural Network 2 in this embodiment.

[0069] Figure 8 This is a schematic diagram of the adaptive hierarchical real-time control of a DC microgrid based on deep learning in an embodiment of the present invention; Detailed Implementation

[0070] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this does not limit the scope of protection of the present invention.

[0071] Please see Figure 1 This invention discloses a deep learning-based adaptive hierarchical control method for DC microgrids, comprising the following components: a deep learning-based implicit stability boundary neural network fitter, a deep learning-based stabilization control instructor, a fully decentralized control strategy based on bus voltage sharing within the stability boundary, and a capacity expansion control strategy based on over / under voltage controllers outside the stability boundary. Specifically, this invention is applied to a 600V low-voltage DC microgrid, and its structure is as follows: Figure 2As shown, the system consists of three energy storage units, one photovoltaic power generation unit, one capacity expansion unit, and several DC loads. The photovoltaic unit is connected to the DC bus via a Boost converter, while the energy storage unit and capacity expansion unit are connected to the DC bus via bidirectional DC-DC converters. The specific control strategy setup steps are as follows:

[0072] Step 1: Construct a fully distributed control system based on shared bus voltage, with the operating modes of each unit divided as follows: Figure 3 As shown, the specific division is as follows: In mode one, the bus voltage fluctuation range is U. H1 dc ≤U H2 The photovoltaic unit operates in droop mode, while the other units are under power limiting control. Voltage stability is mainly achieved by the photovoltaic unit, whose output voltage and droop coefficient are set as follows:

[0073] U dc =U H2 -k pv P pv

[0074]

[0075] In Mode 2, the bus voltage fluctuation range is U. N dc ≤U H1 The energy storage unit operates in droop charging mode, while the photovoltaic unit operates in power limiting control mode. Voltage stability is primarily achieved by the energy storage unit, whose output voltage and droop coefficient are set as follows:

[0076] U dc =U H1 +k b P bat

[0077]

[0078] In Mode 3, the bus voltage fluctuation range is U. L1 dc ≤U N The energy storage unit operates in droop discharge mode, while the photovoltaic unit remains in power-limited control mode. Voltage stability is primarily achieved by the energy storage unit, whose output voltage and droop coefficient are set as follows:

[0079] U dc =U N +k b P bat

[0080]

[0081] ​​​In Mode 4, the bus voltage fluctuation range is U. L2 dc ≤U L1 Both the photovoltaic unit and the energy storage unit are in power-limited control mode, and the voltage fluctuation is dominated by the load unit.

[0082] Step 2: Establish small-signal models for each part based on the above system network structure and control strategy.

[0083] In this embodiment, the system control unit can be divided into drooping working units and power-limited working units. The loads are all constructed as constant power loads. Since the power-limited working units are approximately equal to constant power loads, the small-signal model construction is mainly divided into two parts:

[0084] First, the bus voltage control unit, operating under drooping conditions, approximates an ideal voltage source, connected to the DC bus via a DC-DC converter to maintain bus voltage stability. Assuming a small switching ripple and a high fundamental frequency ratio, its state-space equation is:

[0085]

[0086] Among them, C i L i R i and d i These represent the output regulating capacitor, filter inductor, equivalent resistance, and duty cycle of the i-th buck-boost converter, respectively; u oi i Li i oi and u dc This represents the converter output voltage, input-side inductor current, output current, and bus voltage.

[0087] Then, linearizing the above equation, we obtain the small-signal model of the bus voltage control unit as follows:

[0088]

[0089] Among them, I Li U oi D i These represent the steady-state values ​​of the inductor current, output voltage, and duty cycle, respectively; 'Δ' represents the small-signal variable.

[0090] The bus voltage control unit operating in droop mode, based on dual-loop voltage and current control, employs droop control to achieve coordinated power distribution. Its converter control equation can be expressed as:

[0091]

[0092] In the formula u refi k​i Here, k represents the reference value for the output voltage and the droop factor of the i-th buck-boost converter. pi_i k ii_i u ir_i G i_i (s)=k pi_i +k ii_i / s represents the proportional-integral parameter integral term output and transfer function of the inner loop PI controller; k pu_i k iu_i u ur_i G u_i (s)=k pu_i +k iu_i / s represents the proportional-integral parameters, integral term output, and transfer function of the outer loop PI controller;

[0093] Then, linearizing the above equation, we obtain the small-signal model of the control equation for the bus voltage control unit converter:

[0094]

[0095] The equivalent state-space model of the constant power load unit is:

[0096]

[0097] In the formula C eq L eq R eq These represent the equivalent load input side voltage regulator capacitor, and the resistance and inductance values ​​of the equivalent circuit, respectively.

[0098] Then, linearizing the above equation, we obtain the small-signal model for a constant power load as follows:

[0099]

[0100] Finally, by combining the small-signal models of each part and defining the following state variables, the characteristic matrix A of the system is obtained:

[0101]

[0102] Where B1, B2, and B3 are the state-space matrices of the three bus control units, and C is the equivalent state-space matrix model of the constant power load unit.

[0103] Step 3: Collect the droop coefficient and bus voltage value of the bus voltage control unit in the underlying DC microgrid to calculate the static operating point and substitute them into the characteristic matrix A to solve for the minimum damping ratio of the system.

[0104]

[0105] In the formula, σ and ω are the real and imaginary parts of the eigenvalues ​​of the characteristic matrix A, respectively.

[0106] According to the system instability criteria, the stability label is marked. When the minimum damping ratio corresponding to the system characteristic value exceeds 3%, the system is judged to be in a stable state. Conversely, when the minimum damping ratio corresponding to the system characteristic value is less than 3%, the system is judged to be in an unstable state.

[0107] Step Four: As Figure 4 As shown, the droop coefficient and bus voltage value of each bus voltage control unit are used as inputs, and the stability label is used as the output. An implicit stability boundary based on classification learning artificial neural network is trained to fit the system for subsequent real-time stability evaluation of the system and to guide the stability control.

[0108] Step 5: Construct an over / undervoltage control strategy for the expansion unit under a distributed control strategy based on shared bus voltage. The control structure block diagram is as follows: Figure 5 As shown.

[0109] By setting the droop factor within a certain range, the operation of the energy storage expansion unit is ensured to meet the requirements when U dc >U H2 When the charging power reaches its upper limit, when U dc L2 The discharge power reaches the lower limit value at that time.

[0110] Based on the difference between the bus voltage and the maximum overvoltage and minimum undervoltage values, the operating mode of the expansion unit is as follows: Figure 6 As shown, it is divided into the following three parts:

[0111] When U dc ≥U H2 Within this range, the expansion unit converter is in maximum charging mode, the overvoltage controller reaches its output limit, and the expansion unit droop coefficient must meet the following requirements:

[0112]

[0113] When U dc ≤U L2 Within this range, the converter operates in maximum discharge mode, the undervoltage controller reaches its output limit, and the droop factor of the expansion unit must meet the following requirements:

[0114]

[0115] WhenU L2 ≤U dc ≤U H2 At this time, both the overvoltage controller and the undervoltage controller exceed their output limits, and the converter is in a droop charging / discharging state. The current reference value, after being adjusted, is simplified as follows: ​

[0116]

[0117] Among them I' cha_lim , I' dis_lim These correspond to the maximum charging current and the maximum discharging current for energy storage expansion, respectively.

[0118] Step 6: Build the small-signal model A' of the expanded system.

[0119]

[0120] Where D1 represents the small-signal model of the expansion control unit.

[0121] Step 7: As Figure 7 As shown, the system information (droop coefficients of each bus voltage control unit and bus voltage values, etc.) under the unstable condition determined in step 3 is recorded. The over / undervoltage expansion unit controller initializes the input expansion unit droop coefficient according to the voltage level within the droop coefficient range. Then, the system voltage deviation and stable state are checked to see if the voltage deviation is less than the maximum allowable value and whether the system has returned to the stable boundary. If both are satisfied, the economic efficiency of the expansion unit operation under this droop coefficient input is evaluated, and the droop coefficient value and the system information after expansion control are added to the dataset; if not satisfied, a new droop coefficient is re-entered, and the above operation is repeated.

[0122] Step 8: Using the dataset from Step 7, take the droop coefficient of the bus voltage control unit, the bus voltage value, the minimum damping ratio of the system under unstable conditions before capacity expansion, and the minimum damping ratio of the system after capacity expansion as inputs, and the droop coefficient of the expansion unit as outputs. Use artificial neural network 2 to fit the mapping between the droop coefficient of the expansion unit and its stabilization control effect.

[0123] Step 9: As Figure 8 As shown, in the real-time control of the system, the stable state of the system is determined by the artificial neural network 1 obtained in step four. If the system is determined to be within the stability boundary, it operates in a fully distributed real-time control based on shared bus voltage. If the system is determined to be outside the stability boundary, the droop coefficient k of the adaptive output expansion unit of the artificial neural network 2 obtained in step eight is used. pi This guides the over / undervoltage control strategy of the expansion unit, that is, it guides the system to perform adaptive stabilization control, so that the system returns to the stability boundary.

[0124] The above description is merely an exemplary example of the present invention. It is understood that the content described in this specification is only an enumeration of implementation forms of the inventive concept and is not intended to limit the invention in any way. Therefore, the scope of protection of the present invention should not be limited to the specific forms described in the examples, but should also include the technical means employed by those skilled in the art based on the inventive concept.

[0125] S1: Build a fully decentralized coordinated control strategy for DC microgrids based on the sharing of DC bus voltage signals.

[0126] S2: Based on the microgrid structure and control strategy constructed in S1, a small-signal stability analysis model for DC microgrids is built.

[0127] S3: Obtain the droop coefficient and DC bus voltage value of each bus voltage control unit in the DC microgrid, calculate the static operating point of the DC microgrid system under the corresponding state, and substitute it into the small signal analysis model obtained in S2 to calculate the system characteristic value.

[0128] S4: Based on the system eigenvalues ​​calculated in step S3, use the stability criterion to label the system with stability margin and stability label (stable or unstable).

[0129] S5: Take the droop coefficient and bus voltage value of each bus voltage control unit in step S3 as input, and the stability label obtained in step S4 as output, and fit the implicit stability boundary of the system through artificial neural network 1.

[0130] S6: Build a fully distributed coordinated control strategy for the expanded system based on the sharing of bus voltage signals.

[0131] S7: Based on the expanded system structure and control strategy constructed in S6, build a small-signal stability analysis model for DC microgrids.

[0132] S8: Substitute the droop coefficients and DC bus voltage values ​​of each bus voltage control unit as determined in step S44 under unstable conditions into the small-signal model of the expanded system, and calculate the droop coefficient that makes the system return to a stable state and minimizes the power borne by the expanded unit, and the system stability margin under this condition.

[0133] S9: The droop coefficient of the bus voltage control unit, the bus voltage value, the stability margin of the system under unstable conditions before capacity expansion, and the stability margin of the system after capacity expansion are taken as inputs, and the droop coefficient of the expansion unit is taken as output. The mapping between the droop coefficient of the expansion unit and the stabilization control effect is fitted by artificial neural network 2.

[0134] S10: The artificial neural network 1 obtained in step S5 determines the stable state of the system in real time. If the system is within the stable boundary, it operates under fully distributed real-time control. If the system exceeds the stable boundary, the droop coefficient k4 of the expansion unit is output according to the artificial neural network 2 obtained in step S9 to guide the microgrid system to perform adaptive stabilization control, so that the system returns to the stable boundary.

Claims

1. A deep learning-based adaptive hierarchical control method for DC microgrids, characterized in that, include: Step 1: Construct a fully decentralized control system based on shared bus voltage. Divide and control the operating modes of each unit within the system according to the fluctuation range of the DC microgrid bus voltage. The units within the system include photovoltaic units, which generate electricity under sunlight conditions; energy storage units, which store and release electricity to balance the power demand in the grid; and load units, which consume electricity from the grid. Step 2: Construct a small-signal system model A with implicit stability boundaries: Among them, B1, B2, ... B n Let C be the small-signal model of n bus control units, and C be the equivalent small-signal model of the constant power load unit. Step 3. Collect the droop coefficient and bus voltage value of the DC bus voltage control unit in the bottom-level fully distributed control model, calculate the system's static operating point, and input it into the small-signal model A to solve for the system's minimum damping ratio. According to the system instability criteria, the stability label is marked. When the minimum damping ratio corresponding to the system characteristic value exceeds 3%, the system is judged to be in a stable state. Conversely, when the minimum damping ratio corresponding to the system characteristic value is less than 3%, the system is judged to be in an unstable state. Step 4. Take the droop coefficient and bus voltage value of each unit under bus voltage control as input and the stability label as output, and train the artificial neural network 1 to fit the implicit stability boundary of the system. Step 5. Based on the existing system of bus voltage signal sharing, build overvoltage control strategy and undervoltage control strategy for expansion unit to stabilize the system when it exceeds the stability boundary; Step 6. Construct a small-signal model A' based on the structure and control strategy of the expanded system: Among them, D1, D2, ... D k These are small-signal models for k expansion control units; Step 7: For cases exceeding the stability boundary, search for an optimal value within the range of the droop coefficient of the expansion unit, so that the expanded system can return to the stability boundary and the power borne by the expansion unit is minimized. Step 8: Take the droop coefficient of each bus voltage unit, the DC bus voltage value, and the minimum damping ratio of the system before and after capacity expansion as inputs when the system exceeds the stability boundary, and the optimal value of the droop coefficient of the expansion unit as output, and train the artificial neural network 2 to fit the relationship between the capacity expansion control effect and the droop coefficient to facilitate subsequent guidance of the system to achieve real-time adaptive stabilization control. Step 9: In the real-time control of the DC microgrid, the trained neural network 1 is used to determine the stability of the system in real time, that is, to determine whether it exceeds the stability boundary. If it does not exceed the stability boundary, the system operates in the fully distributed real-time control mode based on shared bus voltage before the expansion. If it exceeds the stability boundary, the artificial neural network 2 will adaptively guide the expansion unit to participate in the system stabilization control and restore the system stability.

2. The deep learning-based adaptive hierarchical control method for DC microgrids according to claim 1, characterized in that: The working modes in step 1 include: Based on the rated value of the bus voltage ±5%, set the maximum allowable voltage fluctuation U. H2 Minimum value U L2 ; Based on the rated value of the bus voltage ±3.33%, the maximum voltage U of the energy storage regulation range is set. H1 and minimum value U L1 ; In mode 1, the voltage fluctuation range is U. H1 dc ≤U H2 The photovoltaic unit operates in droop mode to maintain bus voltage stability. Its output voltage and droop coefficient are expressed as follows:​ IN dc =U H2 -k pv P pv Where, k pv P is the droop factor of the photovoltaic unit. pv_max This represents the maximum output power of the photovoltaic system. In mode 2, the voltage fluctuation range is U. N dc ≤U H1 The energy storage unit operates in droop charging mode to maintain bus voltage stability. Its output voltage and droop coefficient are expressed as follows:​ IN dc =U H1 +k b P bat Where, k b P is the droop factor of the energy storage unit. cha_lim This represents the maximum charging power of the energy storage unit. In mode 3, the voltage fluctuation range is U. L1 dc ≤U N The energy storage unit operates in droop discharge mode to maintain bus voltage stability. Its output voltage and droop coefficient are expressed as follows:​ IN dc =U N +k b P bat In Mode 4, the bus voltage fluctuation range is U. L2 dc ≤U L1 Both the photovoltaic unit and the energy storage unit are in power-limited control mode, and the voltage fluctuation is dominated by the load unit.​ 3. The deep learning-based adaptive hierarchical control method for DC microgrids according to claim 1, characterized in that: Step 2, which involves constructing a small-signal system model A with an implicit stability boundary, specifically includes: Step 2.1: Equivalently model each part of the DC microgrid, and obtain the relationship between voltage and current based on the bidirectional buck-boost circuit structure of the DC bus voltage control unit in droop mode; Among them, C i L i R i and d i These represent the output regulating capacitor, filter inductor, equivalent resistance, and duty cycle of the i-th buck-boost converter, respectively; u oi i Li i oi and u dc This represents the converter output voltage, input-side inductor current, output current, and bus voltage. Step 2.2: Based on the voltage and current dual closed-loop control, the DC bus voltage control unit adopts droop control to achieve coordinated power distribution among multiple units. Its corresponding control model can be expressed as: In the formula k i Here are the reference values ​​for the output voltage and the droop factor of the i-th buck-boost converter; These are the proportional parameters, integral parameters, integral term output, and transfer function of the inner-loop PI controller. These are the proportional parameters, integral parameters, integral term output, and transfer function of the outer loop PI controller. Step 2.3: Simplify the model of the control unit operating under power limitation and the multiple constant power loads connected in parallel; Step 2.4: Combine the simplified equations of models 2.1, 2.2, and 2.3 and add small-signal perturbations to obtain the small-signal model A of the system.

4. The deep learning-based adaptive hierarchical control method for DC microgrids according to claim 2, characterized in that: WhenU dc ≥U H2 Within this range, the expansion unit converter operates in droop charging mode, the overvoltage controller reaches its output limit, and the droop coefficient of the expansion unit must meet the following requirements: When U dc ≤U L2 Within this range, the converter operates in droop discharge mode, the undervoltage controller reaches its output limit, and the droop coefficient of the expansion unit must meet the following requirements: WhenU L2 ≤U dc ≤U H2 At this point, both the overvoltage controller and the undervoltage controller exceed their output limits. Within this range, the expansion unit can adaptively adjust the charging and discharging power based on the droop coefficient. The current reference value, after being organized, is simplified as follows: Where I' cha_lim , I' dis_lim These correspond to the maximum charging current and the maximum discharging current for energy storage expansion, respectively.

Images