Micro-grid droop control voltage frequency adjusting method based on power compensation

Through the recursive Petri fuzzy neural network control method, the problem of inverter voltage frequency regulation in the microgrid is solved, and fast and accurate voltage and frequency compensation is achieved when load changes, improving the system's power quality.

CN118920511APending Publication Date: 2024-11-08HUAIYIN INSTITUTE OF TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411041914.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In microgrids, traditional inverter control strategies are difficult to effectively adjust the voltage frequency, especially when the load changes, which cannot quickly and accurately compensate for voltage and frequency offsets, resulting in a decrease in power quality.

Method used

The control method based on recursive Petri fuzzy neural network is adopted, and the output power is calculated through abc-dq coordinate transformation, and the recursive Petri fuzzy neural network compensation controller is introduced. The network weight is adjusted by using gradient descent method to realize the adaptive response to the three-phase inverter system. Control, quickly compensate for voltage and frequency changes.

Benefits of technology

This method does not require the establishment of an inverter accurate mathematical model, it has strong robustness and adaptability, and can quickly stabilize the voltage and frequency when the load changes suddenly, and improve the system's power quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118920511A_ABST
    Figure CN118920511A_ABST
Patent Text Reader

Abstract

The invention discloses a micro-grid droop control voltage frequency adjusting method based on power compensation, which comprises the following steps: firstly, calculating active power and reactive power output by an inverter, generating given values of voltage and frequency through droop control, designing frequency error and integral thereof, and voltage amplitude error and integral thereof as input; and a recursive Petri fuzzy neural network compensation controller is designed, the output of the controller is used as compensation of active power and reactive power, and finally, an obtained voltage reference value is sent to a voltage and current double closed-loop control system. The recursive Petri fuzzy neural network compensation controller designed by the invention not only comprises the capability of processing time-varying input or output by the recursive network, but also integrates the capability of processing a complex discrete system by the Petri network. Moreover, the controller does not depend on a detailed mathematical model of the system, and can effectively reduce the deviation between the output voltage and frequency and the rated value when the load suddenly changes and the system parameters change, thereby improving the quality of the output electric energy of the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of inverter control, and in particular to a method for regulating voltage and frequency by droop control of a microgrid based on power compensation. Technical Background

[0002] Renewable Energy Sources (RESs) have received a great deal of attention due to concerns about global warming, rising electricity consumption, and depletion of fossil fuels. Microgrids are a relatively new concept that provides a way to integrate RESs into existing distribution networks.

[0003] In order to ensure the power quality of microgrids, various advanced control strategies are constantly applied to inverter control. Control strategy is the core to ensure the excellent output performance of inverters. Therefore, designing an adaptive control strategy in microgrid inverters that can meet the changes in working conditions within a certain range can make full use of the flexibility of power electronic converters and effectively reduce the cost of microgrid system design, which has important theoretical research value and engineering practice significance.

[0004] When the load of the microgrid changes, the voltage frequency deviates from the reference value. Although traditional fuzzy neural networks (FNNs) have been widely used in industrial applications, fixed fuzzy operators cannot optimize fuzzy logic reasoning. Moreover, there are not enough free variables to modify the fuzzy operators; Compensatory fuzzy neural networks can solve the shortcomings of traditional neural networks; In addition, the cyclic connection of the recursive neural network gives the network more memory for system information, so the recursive neural network has the ability to process time-varying inputs or outputs and the ability to process complex spatiotemporal patterns; Petri network is a mathematical and graphical tool suitable for deadlock analysis, scheduling and control in automated manufacturing systems, with the ability to analyze, model and simulate, as well as the ability to control various discrete event systems. Summary of the invention

[0005] Purpose of the invention: In order to solve the problems existing in the prior art, this paper provides a method for regulating voltage and frequency by droop control of a microgrid based on power compensation, which uses neural networks that do not rely on system models and have superior dynamic response performance, and has strong robustness and adaptability to changes in system parameters. The recursive Petri fuzzy neural network algorithm designed in this paper can achieve real-time adjustment of output weights through adaptive rules when the load changes, to compensate for the interference of the three-phase inverter system, reduce changes in voltage and frequency, and enable the controller to achieve faster and more precise control.

[0006] Technical solution: The present invention discloses a method for regulating voltage and frequency by droop control of a microgrid based on power compensation, comprising the following steps:

[0007] Step 1: According to the load current and filter capacitor voltage of the three-phase full-bridge LC inverter, the output active power P and reactive power Q are calculated by abc-dq coordinate transformation;

[0008] Step 2: Based on the P and Q calculated in step 1, observe the changes in frequency and voltage through droop control;

[0009] Step 3: According to the droop control in step 2, a recursive Petri fuzzy neural network is introduced to construct a recursive Petri fuzzy neural network compensation controller, wherein the recursive Petri fuzzy neural network compensation controller restores voltage and frequency, and uses a gradient descent method to perform adaptive adjustment of network weights;

[0010] Step 4: According to step 3, recursively Petri fuzzy neural network compensation controller is used to obtain the output of the controller, compensate voltage and frequency, and finally send the obtained voltage reference value to the voltage and current dual closed-loop control system to control the full-bridge switch of the three-phase full-bridge LC inverter through PWM signal.

[0011] Furthermore, the abc-dq coordinate transformation in step 1 calculates the output active power P and reactive power Q as follows:

[0012]

[0013] P=1.5(V d I d +V q I q )

[0014] Q=1.5(V q I d -V d I q )

[0015] Among them, V d and V q are the voltage on the AC side capacitor on the d-axis and q-axis respectively, I d and I q are the AC side load current on the d-axis and q-axis respectively, v a ,v b ,v c are the voltages on the capacitors on the AC side of the abc axis, i a ,i b ,i c are the AC side load currents on the abc axes respectively, and P and Q are the active power and reactive power respectively.

[0016] Furthermore, in step 2, the changes in frequency and voltage are observed through droop control;

[0017] Droop Control:

[0018] ω=ω * -k p (PP * )

[0019] E=E * -k q (QQ * )

[0020] Among them, k p and k q are the droop coefficients of frequency and voltage, P * and Q * are active power reference value and reactive power reference value respectively, ω * and E * are the frequency reference value and the voltage reference value respectively, and ω and E represent the frequency and voltage respectively.

[0021] Furthermore, in step 3, a recursive Petri fuzzy neural network is introduced to restore the voltage and frequency, as follows:

[0022] Define the input of the recursive Petri fuzzy neural network:

[0023]

[0024] The output of the recursive Petri fuzzy neural network design is used as the compensation amount of active power and reactive power, which is designed as:

[0025]

[0026] The recursive compensatory Petri fuzzy neural network algorithm includes input layer, membership layer, rule layer, Petri layer, and recursive layer. N is the total number of rules, y1 6 and represents the output layer output, and represents the output layer network, represents the lth output of the recurrent layer, and shows the adjustable weights between the recurrent layer and the output layer;

[0027] Furthermore, the gradient descent method is used to adaptively adjust the network weights, and the adaptive rates of the output weights, recursive layer weights, and membership layer parameters are designed as follows:

[0028]

[0029] in: and are the center point and width of the i-th input and j-th Gaussian function to the membership layer node, respectively. is the amount of self-update of the adjustable weights between the recurrent layer and the output layer, is the recursive weight self-update amount, and They are respectively the self-updated amount of the center point and width of the i-th input and j-th Gaussian function to the membership layer node.

[0030] Beneficial effects:

[0031] 1. The present invention adopts a recursive Petri fuzzy neural network compensation control method, which does not require the establishment of an accurate mathematical model of the inverter and is particularly suitable for the control of high-frequency inverters with complex structures and time-varying parameters.

[0032] 2. When the load changes suddenly, the present invention can reduce the deviation of the output voltage and frequency from the rated value, thereby improving the power quality of the system.

[0033] 3. The present invention utilizes the complementary advantages of fuzzy control and neural network control, has superior dynamic response performance, and has strong robustness and adaptability to parameter changes and load disturbances in the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the main circuit topology and control circuit structure of a three-phase full-bridge LC inverter;

[0035] Figure 2 It is a schematic diagram of voltage frequency compensation based on recursive Petri fuzzy neural network;

[0036] Figure 3 The present invention simulates the change of frequency in the initial state;

[0037] Figure 4 The present invention simulates the change of voltage in the initial state;

[0038] Figure 5 is the change of frequency under the simulation algorithm of the present invention;

[0039] Figure 6 This is the voltage change under the simulation algorithm of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0041] The present invention discloses a method for regulating voltage and frequency by droop control of a microgrid based on power compensation, which mainly comprises the following steps:

[0042] Step 1. Establishment of mathematical model of three-phase LC inverter

[0043] Figure 1 FIG. 1 is a structural diagram of a three-phase LC inverter in an island mode used in the present invention. dc is the DC bus voltage; the inverter unit is a three-phase bridge structure; the switching device adopts an IGBT with a commutation freewheeling diode, the filtering unit adopts an LC filter, and the output load.

[0044] First, the voltage and current are sampled and the coordinate transformation is performed. The voltage and current abc-dq coordinate transformation is:

[0045]

[0046] Among them, V d and V q are the voltage on the AC side capacitor on the d-axis and q-axis respectively, I d and I q are the AC side load current on the d-axis and q-axis, I ld and I lq are the AC side output current on the d-axis and q-axis respectively; v a ,v b ,v c are the voltages on the capacitors on the AC side of the abc axis, i a ,i b ,i c are the AC side load currents on the abc axis, i la ,i lb ,i lc They are the AC side output current on the abc axis respectively;

[0047] Then calculate the output active power and reactive power:

[0048] P=1.5(V d I d +V q I q )

[0049] Q=1.5(V q I d -V d I q )

[0050] Among them, P and Q are active power and reactive power respectively;

[0051] According to droop control, when the load changes, the voltage and frequency of the system will change:

[0052] The droop control is:

[0053] ω=ω * -k p (PP *)

[0054] E=E * -k q (QQ * )

[0055] Among them, k p and k q are the droop coefficients of frequency and voltage, P * and Q * are active power reference value and reactive power reference value respectively, ω * and E * They are frequency reference value and voltage reference value respectively;

[0056] Step 2. Design of recursive Petri fuzzy neural network compensation controller based on gradient descent method

[0057] Six-layer recursive Petri fuzzy neural network with gradient descent method Figure 2 As shown, it includes input layer, membership layer, rule layer, Petri layer, recursive layer and output layer.

[0058] (1) The network and output of the neuron input layer are:

[0059]

[0060] in, and are the input and output of the ith neuron, N is the total number of rules,

[0061] (2) The membership layer is represented by the following Gaussian membership function. The network and output of this layer are:

[0062]

[0063] in, and are the center point and width of the i-th input and j-th Gaussian function to the membership layer node, respectively. Represents the output of the jth neuron in the membership layer.

[0064] (3) The rule layer implements the fuzzy reasoning mechanism. Each node in this layer multiplies the input signal and outputs the product. The network and output of this layer are:

[0065]

[0066] in, is the weight between the membership layer and the rule layer, assumed to be 1.

[0067] (4) Petri layer is based on PN theory, using competition to select appropriate excitation nodes to generate tokens in Petri layer. PN consists of two types of nodes: transition nodes and position nodes. When creating a token at the input position, the transition is enabled. Therefore, the conditions for triggering or not triggering the transition can be described as:

[0068]

[0069] in, is the transition amount, α and β are positive constants, d th is a dynamic threshold, which changes with the function H. The function H is:

[0070]

[0071] When a transition is triggered, the required tokens are removed from their input positions and new tokens are generated at each output position. On the other hand, when a transition is not triggered, the tokens remain at the original input positions. Therefore, the network and output of this layer are described as:

[0072]

[0073] in, is the output of the Petri layer.

[0074] (5) Recurrent layer, the output of each node in this layer, represented by addition, is obtained by multiplying the input signal and producing the multiplication result. In addition, according to the cyclic characteristics, the feedback loop is used to memorize past information. Therefore, due to the existence of these cyclic connections, the network is endowed with the ability to store information. The network and output of this layer are expressed as:

[0075]

[0076] in, is the recursive weight, is the output of the recurrent layer.

[0077] (6) Output layer, each node is the sum of the product of the rule layer output and the adjustable weight between the rule layer and the output layer. The network and output of this layer are:

[0078]

[0079] y 6 (N) = net 6 (N)

[0080] in, represents the connection weight between the output layer and the loop and rule layers, y 6 (N) is the output of the recursive Petri fuzzy neural network compensation controller.

[0081] In order to improve the online learning ability of the neural network, the gradient descent method is used to train the weights online. Taking the frequency recursive Petri fuzzy neural network compensation controller as an example (the voltage recursive Petri fuzzy neural network compensation controller is the same), the minimum error is constructed, and the error function E is:

[0082]

[0083] The adaptive rate of the recursive Petri fuzzy neural network parameters is as follows:

[0084]

[0085] is the amount of self-update of the adjustable weights between the recurrent layer and the output layer, η l is the learning rate.

[0086]

[0087] is the self-update amount of the adjustable weights of the recursive layer, η rl is the learning rate.

[0088]

[0089]

[0090] and are the self-updating amounts of the center point and width of the i-th input and j-th Gaussian function to the membership layer node respectively; η ij and η ij2 is the learning rate.

[0091] See also Figure 3-Figure 6 , Figure 3 and Figure 4 The present invention simulates the change of frequency and voltage in the initial state without compensation. It can be seen from the figure that when the load is added for 0.1 seconds, both the voltage and frequency change, and deviate from the rated value when stable; Figure 5 and Figure 6 The recursive Petri fuzzy neural network compensation control method is adopted. It can be seen that when the load changes, the voltage and frequency quickly stabilize at the rated value, indicating that the control method of the present invention has strong robustness and adaptability to load disturbances.

[0092] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode, but also include technical schemes composed of any combination of the above-mentioned technical features. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also regarded as the protection scope of the present invention.

Claims

1. A method for regulating voltage and frequency by droop control of a microgrid based on power compensation, characterized in that: The steps include: Step 1: According to the load current and filter capacitor voltage of the three-phase full-bridge LC inverter, the output active power P and reactive power Q are calculated by abc-dq coordinate transformation; Step 2: Based on the P and Q calculated in step 1, observe the changes in frequency and voltage through droop control; Step 3: According to the droop control in step 2, a recursive Petri fuzzy neural network is introduced to construct a recursive Petri fuzzy neural network compensation controller, wherein the recursive Petri fuzzy neural network compensation controller restores voltage and frequency, and uses a gradient descent method to perform adaptive adjustment of network weights; Step 4: According to step 3, recursively Petri fuzzy neural network compensation controller is used to obtain the output of the controller, compensate voltage and frequency, and finally send the obtained voltage reference value to the voltage and current dual closed-loop control system to control the full-bridge switch of the three-phase full-bridge LC inverter through PWM signal.

2. The method for regulating voltage and frequency by droop control of a microgrid based on power compensation according to claim 1 is characterized in that: In step 1, the abc-dq coordinate transformation is used to calculate the output active power P and reactive power Q, as follows: P=1.5(V d I d +V q I q ) Q=1.5(V q I d -V d I q ) Among them, V d and V q are the voltage on the AC side capacitor on the d-axis and q-axis respectively, I d and I q are the AC load current on the d-axis and q-axis, respectively, and v a ,v b ,v c are the voltages on the capacitors on the AC side of the abc axis, i a ,i b ,i c are the AC side load currents on the abc axes respectively, and P and Q are the active power and reactive power respectively.

3. The method for regulating voltage and frequency by droop control of a microgrid based on power compensation according to claim 1, characterized in that: In step 2, the frequency and voltage changes are observed through droop control; Droop Control: oh = oh * -k p (PP * ) E=E * -k q (Q-Q * ) Among them, k p and k q are the droop coefficients of frequency and voltage, P * and Q * are active power reference value and reactive power reference value respectively, ω * and E * are the frequency reference value and the voltage reference value respectively, and ω and E represent the frequency and voltage respectively.

4. The method for regulating voltage and frequency by droop control of a microgrid based on power compensation according to claim 1 is characterized in that: In step 3, a recursive Petri fuzzy neural network is introduced to restore voltage and frequency, as follows: Define the input of the recursive Petri fuzzy neural network: The output of the recursive Petri fuzzy neural network design is used as the compensation amount of active power and reactive power, which is designed as: The recursive compensatory Petri fuzzy neural network algorithm includes input layer, membership layer, rule layer, Petri layer, and recursive layer. N is the total number of rules. and represents the output layer output, and represents the output layer network, represents the lth output of the recurrent layer, and Shows the adjustable weights between the recurrent layer and the output layer.

5. The method for regulating voltage and frequency by droop control of a microgrid based on power compensation according to claim 4 is characterized in that: The gradient descent method is used to adaptively adjust the network weights. The adaptive rates of the output weights, recursive layer weights, and membership layer parameters are designed as follows: in: and are the center point and width of the i-th input and j-th Gaussian function to the membership layer node, respectively. is the amount of self-update of the adjustable weights between the recurrent layer and the output layer, is the recursive weight self-update amount, and They are respectively the self-updated amount of the center point and width of the i-th input and j-th Gaussian function to the membership layer node.