CN116581770A - Micro-grid system VSG double-droop control method based on self-adaptive neural network - Google Patents

Abstract

The invention discloses a micro-grid system VSG double droop control method based on an adaptive neural network. The micro-grid system architecture is mainly composed of a photovoltaic power generation unit, a storage battery energy storage unit, an inverter module, an LC filter module, and a VSG control module. , where the front stage of the photovoltaic power generation unit uses a DC‑DC boost converter to achieve maximum power point tracking, and the rear stage is a DC‑AC grid-connected inverter, which is connected in parallel with the energy storage unit on the DC side, and VSG is introduced to provide inertia and Damping, establish an adaptive neural network controller, obtain the angular frequency offset and the change slope of the angular frequency in real time through the neural network controller, combine the set virtual parameter selection rules to obtain the virtual parameters J and D, and control it through the adaptive neural network The virtual parameters output by the controller adjust the frequency and voltage of the microgrid system to complete the double droop control. Through the invention, the double droop control strategy can be realized, and the stability of the microgrid when connected to the grid is improved.

Description

VSG double droop control method for microgrid system based on adaptive neural network

technical field

The invention relates to the field of clean energy, in particular to a microgrid system VSG double droop control method based on an adaptive neural network.

Background technique

In recent years, new energy power generation such as photovoltaics and wind power has been widely used in large power grids due to its cleanliness, high penetration rate, and sustainable utilization. At the same time, it also brings some serious challenges. Due to the randomness and volatility of new energy power generation, it has a great impact on the stability of the system. The proposal of the microgrid aims to solve these drawbacks and has become a research hotspot today. However, due to the lack of damping and inertia of the system when the microgrid is connected to the grid, it is easy to cause frequency and voltage collapse when it is disturbed. Therefore, in order to make the frequency and voltage adapt to system changes and ensure the safe operation of the system, virtual synchronous power generation is introduced into the microgrid. Machine control technology (Virtual Synchronous Generator, VSG), which has the inertia and damping characteristics of synchronous generators, is applied to the microgrid to provide inertia and damping, thereby ensuring the stability of the system output frequency and voltage. The inertia and damping parameters of the VSG are flexible and variable, so in order to improve the stability control of the virtual synchronous machine, an adaptive method can be used to improve the VSG. The prior art proposes an adaptive parameter control strategy based on fuzzy control, which adapts the inertia and damping parameters according to the frequency deviation and rate of change, but does not consider the influence of the droop control characteristics. The existing technology fully considers the relationship between power angle and active power and the angular frequency oscillation process, and gives the selection principles of moment of inertia and damping coefficient, which effectively improves the stability of photovoltaic grid connection. However, the effect of voltage is not considered. In the existing technology, a multi-parameter adaptive control strategy such as frequency and virtual inertia is proposed to meet the requirements of power grids for frequency and moment of inertia under different working conditions, which realizes frequency modulation and on/off grid switching, but does not consider damping impact of changes on it. In the existing technology, in order to solve the voltage and current impact problem generated when connecting to the grid, a model predictive control method is adopted, and the weight coefficient adaptive rule of the model predictive control is formulated by using the change of frequency, which greatly improves the pre-synchronization and operation. The dynamic response when switching between modes, but the adjustment of the frequency is not mentioned in the text. The existing technology adopts the output speed feedback to adjust the damping, and uses the power angle characteristic to propose a parameter adaptive control strategy, which reduces the frequency deviation and suppresses the power overshoot, but does not consider the influence of virtual inertia on the system. The prior art [9] proposes a dual-machine parallel VSG control strategy based on adaptive feedforward control. Adaptively increase the power deficit of the feedforward compensation system through the change of angular frequency, improve the response speed and power distribution accuracy of the system, suppress the system frequency fluctuation and active power oscillation, and ensure the rapid stability and safety of the system, but Q-U droop Controls are not considered. The existing technology is based on the use of fuzzy algorithm to realize the adaptive adjustment of VSG virtual inertia and damping coefficient. This strategy can reasonably suppress the fluctuation of VSG frequency and power in the transient process and maintain the stable operation of the power grid. However, the fuzzy control method leads to parameter adjustment. Not exactly.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art and provide a micro-grid system VSG double droop control method based on an adaptive neural network. The micro-grid system architecture is mainly composed of a photovoltaic power generation unit, a battery energy storage unit, and an inverter module , LC filter module, and VSG control module. Among them, the front stage of the photovoltaic power generation unit adopts a DC-DC boost converter to achieve maximum power point tracking, and the rear stage is a DC-AC grid-connected inverter, which is connected in parallel with the energy storage unit in DC On the other hand, VSG is introduced to provide inertia and damping for the system, an adaptive neural network controller is established, and the angular frequency offset and angular frequency change slope are obtained in real time through the neural network controller, and the virtual parameter J is obtained by combining the set virtual parameter selection rules. and D, adjust the frequency and voltage of the microgrid system through the virtual parameters output by the adaptive neural network controller, and complete the double droop control.

Further, in the establishment of an adaptive neural network controller, the angular frequency offset and the change slope of the angular frequency are obtained in real time through the neural network controller, and the virtual parameters J and D are obtained in combination with the set virtual parameter selection rules, including:

Firstly, the angular frequency offset and the change slope of the angular frequency are obtained in real time through the neural network controller, namely:

Secondly, the neural network controller obtains the virtual parameters J and D according to the given virtual parameter selection rules, namely:

In the formula: J ₀ and D ₀ represent the inertia and damping parameters adjusted and output by the neural network; K _j and K _d are the adjustment coefficients of inertia and damping respectively; T _j and T _d are the upper and lower limits of parameter changes.

Further, said adjusting the voltage of the microgrid system through the virtual parameter output by the adaptive neural network controller includes:

The voltage is adjusted by the virtual parameters output by the neural network:

V＝(L _f C _f ) ^-1 [ω _i i _q L _f -r _f i _d -v _odi -i _odi L _f +v _id ]

In the formula: v _odi is the d-axis voltage component of the VSG output, L _f , C _f , r _f represent the inductance, capacitance and resistance of the LC filter respectively, i _odi is the d-axis current component of the VSG output, i _d , i _q d and q-axis current components input by the VSG, v _id is the d-axis current component input by the VSG, and ω _i is the angular frequency output by the VSG;

The tracking error is:

In the formula:

Another error variable e _2i is expressed as:

e _2i ＝d _i v _odi -α _ij

In the formula: α _ij is a virtual control quantity.

Further, said adjusting the angular frequency of the microgrid system through the virtual parameter output by the adaptive neural network controller includes:

From droop control:

ω _i =ω _ni -k _p p _i

In the formula: ω _i is the angular frequency output by the VSG, ω _ni is the per unit value of the angular frequency, and P _i is the output active power after filtering.

The adjusted error is:

The angular frequency output is:

The beneficial effects of the present invention are: the technical solution proposed by the present invention can control the frequency, voltage, active power and reactive power adaptively, realize the double droop control strategy, and improve the stability of the microgrid when connected to the grid.

Detailed ways

The technical solution of the present invention will be further described in detail below, but the protection scope of the present invention is not limited to the following description.

In order to make the object, technical solution and advantages of the present invention clearer, the present invention is further described in detail. It should be understood that the specific embodiments described here are only used to explain the present invention, and are not intended to limit the present invention, that is, the described embodiments are only some of the embodiments of the present invention, but not all of the embodiments.

Accordingly, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts belong to the protection scope of the present invention. It should be noted that relative terms such as the terms "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any relationship between these entities or operations. There is no such actual relationship or order between them.

Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed other elements of, or also include elements inherent in, such a process, method, article or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The characteristics and performance of the present invention will be described in further detail below in conjunction with the examples.

A microgrid system VSG double droop control method based on an adaptive neural network. The microgrid system architecture is mainly composed of a photovoltaic power generation unit, a battery energy storage unit, an inverter module, an LC filter module, and a VSG control module. The front stage of the power generation unit adopts a DC-DC boost converter to achieve maximum power point tracking, and the rear stage is a DC-AC grid-connected inverter, which is connected in parallel with the energy storage unit on the DC side. VSG is introduced to provide inertia and damping for the system, and an automatic Adapt to the neural network controller, obtain the angular frequency offset and the change slope of the angular frequency in real time through the neural network controller, combine the set virtual parameter selection rules to obtain the virtual parameters J and D, and output the virtual parameters through the adaptive neural network controller The parameters adjust the frequency and voltage of the microgrid system to complete the double droop control.

Further, in the establishment of an adaptive neural network controller, the angular frequency offset and the change slope of the angular frequency are obtained in real time through the neural network controller, and the virtual parameters J and D are obtained in combination with the set virtual parameter selection rules, including:

Firstly, the angular frequency offset and the change slope of the angular frequency are obtained in real time through the neural network controller, namely:

Secondly, the neural network controller obtains the virtual parameters J and D according to the given virtual parameter selection rules, namely:

In the formula: J ₀ and D ₀ represent the inertia and damping parameters adjusted and output by the neural network; K _j and K _d are the adjustment coefficients of inertia and damping respectively; T _j and T _d are the upper and lower limits of parameter changes.

Further, said adjusting the voltage of the microgrid system through the virtual parameter output by the adaptive neural network controller includes:

The voltage is adjusted by the virtual parameters output by the neural network:

V＝(L _f C _f ) ^-1 [ω _i i _q L _f -r _f i _d -v _odi -i _odi L _f +v _id ]

In the formula: v _odi is the d-axis voltage component of the VSG output, L _f , C _f , r _f represent the inductance, capacitance and resistance of the LC filter respectively, i _odi is the d-axis current component of the VSG output, i _d , i _q d and q-axis current components input by the VSG, v _id is the d-axis current component input by the VSG, and ω _i is the angular frequency output by the VSG;

The tracking error is:

In the formula:

Another error variable e _2i is expressed as:

e _2i ＝d _i v _odi -α _ij

In the formula: α _ij is a virtual control quantity.

Further, said adjusting the angular frequency of the microgrid system through the virtual parameter output by the adaptive neural network controller includes:

From droop control:

ω _i =ω _ni -k _p p _i

In the formula: ω _i is the angular frequency output by the VSG, ω _ni is the per unit value of the angular frequency, and P _i is the output active power after filtering.

The adjusted error is:

The angular frequency output is:

Specifically, the microgrid system architecture is mainly composed of photovoltaic power generation units, battery energy storage units, inverter modules, LC filter modules, and VSG control modules. Among them, the front stage of photovoltaic power generation units uses DC-DC boost converters to achieve maximum power. Point tracking, the rear stage is a DC-AC grid-connected inverter, which is connected in parallel with the energy storage unit on the DC side, and the VSG is introduced to provide inertia and damping for the system to achieve double droop stability control and improve system stability.

The outer loop control of the VSG is a power control loop, and the inner loop is a double droop control. By adjusting the virtual inertia and damping, the control of the frequency and voltage of the VSG is affected, and the deviation is reduced to the allowable range, thereby affecting the double droop control. net target.

VSG realizes frequency control through the rotor motion equation. In order to facilitate the control of the converter inverter, the model is processed by reducing the order, and the rotor motion equation can be obtained by modeling the virtual synchronous machine:

In the formula: J represents the moment of inertia of the virtual synchronous machine, ω is the instantaneous angular frequency of the rotor, ω ₀ is the rated angular frequency of the rotor, D is the damping parameter of the VSG, δ is the power angle of the VSG, and P _m represents the mechanical power. The virtual moment of inertia J can reduce the frequency deviation, and the damping coefficient D can suppress the voltage fluctuation and reduce its deviation. According to the QU droop control characteristics of VSG, the voltage control equation of VSG can be obtained as:

U _q =K _q (Q _ref -Q)+U _n (2)

In the formula: U _q represents the desired target voltage after system optimization, K _q represents the reactive power droop coefficient, Q _ref represents the system reactive power reference value, Q is the real-time reactive power of VSG, U _n is the rated voltage.

The frequency control equation can be obtained from the f-p droop control characteristic of VSG:

In the formula: f is the target frequency to be adjusted, f _n is the rated frequency, p and P _ref are the reference value and actual value of active power respectively, and K _p is the droop control coefficient of active power.

In order to ensure the stability of the output voltage, the excitation regulation is adopted, and the control equation of the excitation regulation is [17-18]:

E＝k _v ∫[K _q (Q _ref -Q)+(U ₀ -U _q )] (4)

Combining formulas (1), (2), (3), and (4), the double-sag closed-loop function can be obtained

The traditional virtual synchronous machine uses the rotor motion equation of formula (1) to adjust the output parameters, and adjusts the frequency and voltage through the virtual parameters selected in advance. This method has great limitations and cannot respond to system changes in real time. Making the best regulation will cause large errors in the frequency and voltage of the system and affect the stability of grid connection. Therefore, this paper studies a method of adaptive virtual parameters to solve this problem.

According to the classical control theory, the angular frequency and damping ratio of the VSG are:

In the formula:

Through equations (5) and (6), we can know the influence of virtual moment of inertia J and damping coefficient D on the system. When J takes a fixed value, the pole will change with the change of damping coefficient D. When D gradually increases, the pole will change from negative The imaginary axis is gradually approaching the negative real axis, and the stability of the system is gradually improving at this time, and when the damping coefficient D (assuming D=30) takes a fixed value, the change of the virtual inertia J also brings a great change to the pole Influence, if the J value is small, there will be an underdamped situation, which will reduce the stability of the system. However, if a larger J is used, although it will play a certain role in supporting the frequency, this choice will produce a huge frequency deviation, which will easily lead to a system crash.

Through the analysis, it can be concluded that both the virtual inertia J and the damping coefficient D will affect the stability of the system. Therefore, in order for the system to cope with frequency fluctuations caused by various emergencies, J and D should be adjusted at the same time to improve the stability of the system.

The active power presents a change law of oscillation attenuation, and its critical stable point is also constantly changing. The change of angular frequency can be divided into four stages. The first stage: the change slope of angular frequency at time [t ₁ -t ₂ ] changes from a positive value to gradually decreases to zero, resulting in a gradual increase in the angular frequency, so the error with the rated angular frequency is also increasing, and the corresponding power angle changes from point a to point b, and the active power is also increasing, which affects the system Grid-connected stability, so it is necessary to increase the values of inertia J and damping D to suppress the unfavorable factors brought by angular frequency deviation and rate of change, the second stage: the angular velocity change slope starts from zero at time [t ₂ -t ₃ ] Gradually decreases to a negative value, and the value of the angular velocity gradually approaches the rated value, but there is still an error. In this process, the power angle changes from point b to point c, the rate of change of active power decreases from positive to zero, and the power reaches the maximum value, so in order to quickly drop the angular frequency to the rated angular velocity, the value of the inertia should be reduced. If the angular frequency offset is too large, the damping D can be increased to further suppress the deviation. The changes in the remaining two stages are the same as the first two The changes in stages are the same. From this analysis, the selection rules of inertia J and damping D can be obtained, as shown in Table 1.

Table 1 Selection rules for inertia J and damping D

Artificial neural network can be understood as composed of a large number of neurons connected by weights, capable of information storage, parallel processing, spontaneous learning and other operations. The algorithm of the artificial neural network is mainly the BP (Back Propagation) algorithm, which is also called the back propagation algorithm. The BP algorithm can be regarded as a function, which is composed of non-linear and changeable units. It has some nonlinear capabilities, such as mapping capabilities, and can flexibly handle problems such as network learning coefficients. At the same time, it can be used in other fields such as: signal processing, Fault diagnosis, intelligent control, etc. have broad application prospects.

Through the selection rules of virtual parameters, in order to suppress the offset of frequency and voltage, realize stable double droop control, and realize stable grid connection, so the idea of neural network control is adopted to adjust virtual parameters in real time, and the control method is as follows:

Firstly, the angular frequency offset and the change slope of the angular frequency are obtained in real time through the neural network controller, namely:

Secondly, the neural network controller obtains the virtual parameters J and D according to the given virtual parameter selection rules, namely:

In the formula: J ₀ and D ₀ represent the inertia and damping parameters adjusted and output by the neural network; K _j and K _d are the adjustment coefficients of inertia and damping respectively; T _j and T _d are the upper and lower limits of parameter changes.

The virtual parameters output by the neural network can adjust the voltage in real time:

V＝(L _f C _f ) ^-1 [ω _i i _q L _f -r _f i _d -v _odi -i _odi L _f +v _id ] (11)

In the formula: v _odi is the d-axis voltage component output by the virtual synchronous machine, L _f , C _f , r _f represent the inductance, capacitance and resistance of the LC filter respectively, i _odi is the d-axis current component output by the VSG, i _d , i _q is the d and q axis current components of VSG input, and v _id is the d axis current component of VSG input. ω _i is the angular frequency of VSG output. In addition, the tracking error is:

It can be seen from the above formula that when the virtual synchronous machine uses the droop coefficient to allocate reactive power, it cannot accurately adjust the voltage. Therefore, the error function e _li is set in advance to achieve a compromise between voltage and reactive power. However, the defined error function requires The voltage is controlled twice, that is, e _1i reduces the deviation between v _odi -v _ref , and the error is reduced through the neural network controller, so the formula (12) can be transformed into:

In the formula: Another error variable e _2i can be expressed as:

e _2i ＝d _i v _odi -α _ij (14)

In the formula: α _ij is a virtual control quantity.

Using Lyapunov's theory to check its stability

Set the Lyape function as:

Differentiate it and put it into formula (13) to get:

Substituting formula (14) into formula (16) can get:

In the formula, α _i is expressed by the following formula:

Substituting formula (18) into formula (17) can get:

Another Lyape function obtained from the above formula is:

Combining equations (11), (19) and (20), the controller of this model can be obtained:

v _id ＝-e _1i -k _2i e _2i +F _i (x _i ) (21)

the above Can be expressed as:

Using a neural network to can be transformed into:

The neural network controller is used for real-time adjustment, first of all, it can be obtained from the droop control:

ω _i ＝ω _ni -k _p p _i (24)

In the formula: ω _i is the angular frequency output by the virtual synchronous machine, ω _ni is the per unit value of the angular frequency, and P _i is the output active power after filtering.

The adjusted error can be expressed as:

The chosen Lyapunov function is:

Differentiate it to get:

v _fi ＝e _ωi d _i (27)

Substituting formula (25) into formula (27) can get:

Substitute (24) into (28) to get:

Then use neural network control, namely:

So the angular velocity output is:

In the formula: is the weight of the neural network, and λ _i (xi ₎ is the approximate error controlled by the neural network.

The error is controlled within a reasonable range through the neural network controller, and the Lyapunov theory can be used to determine that the system stability has been greatly improved after the neural network control.

The frequency is adjusted by the controller through the neural network, so as to ensure that the active power can be distributed on demand and realize the stable control of pf. In order to keep the error between ω _i and ω _ref within a reasonable range, the neural network controller is used to adjust in real time, firstly, it can be obtained by droop control:

ω _i ＝ω _ni -k _p p _i (24)

In the formula: ω _i is the angular frequency output by the virtual synchronous machine, ω _ni is the per unit value of the angular frequency, and P _i is the output active power after filtering.

The adjusted error can be expressed as:

The chosen Lyapunov function is:

Differentiate it to get:

v _fi ＝e _ωi d _i (27)

Substituting formula (25) into formula (27) can get:

Substitute (24) into (28) to get:

Then use neural network control, namely:

So the angular velocity output is:

The above descriptions are only preferred embodiments of the present invention, and it should be understood that the present invention is not limited to the forms disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and Modifications can be made within the scope of the ideas described herein, by virtue of the above teachings or skill or knowledge in the relevant art. However, changes and changes made by those skilled in the art do not depart from the spirit and scope of the present invention, and should all be within the protection scope of the appended claims of the present invention.

Claims (4)

1. A micro-grid system VSG double droop control method based on an adaptive neural network, the micro-grid system architecture is mainly composed of a photovoltaic power generation unit, a storage battery energy storage unit, an inverter module, an LC filter module, and a VSG control module, wherein The front stage of the photovoltaic power generation unit uses a DC-DC boost converter to achieve maximum power point tracking, and the rear stage is a DC-AC grid-connected inverter, which is connected in parallel with the energy storage unit on the DC side, and VSG is introduced to provide inertia and damping for the system. Establish an adaptive neural network controller, obtain the angular frequency offset and the change slope of the angular frequency in real time through the neural network controller, combine the set virtual parameter selection rules to obtain virtual parameters J and D, and output through the adaptive neural network controller The virtual parameters adjust the frequency and voltage of the microgrid system to complete the double droop control. 2. the micro-grid system VSG double droop control method based on adaptive neural network according to claim 1, is characterized in that, described establishment adaptive neural network controller obtains angular frequency offset in real time by neural network controller and angular frequency change slope, combined with the set virtual parameter selection rules to obtain the virtual parameters J and D, including: Firstly, the angular frequency offset and the change slope of the angular frequency are obtained in real time through the neural network controller, namely: Secondly, the neural network controller obtains the virtual parameters J and D according to the given virtual parameter selection rules, namely: In the formula: J ₀ and D ₀ represent the inertia and damping parameters adjusted and output by the neural network; K _j and K _d are the adjustment coefficients of inertia and damping respectively; T _j and T _d are the upper and lower limits of parameter changes. 3. the micro-grid system VSG double droop control method based on the adaptive neural network according to claim 2, is characterized in that, the described virtual parameter regulation micro-grid system voltage by adaptive neural network controller output comprises: The voltage is adjusted by the virtual parameters output by the neural network: V＝(L _f C _f ) ^-1 [ω _i i _q L _f -r _f i _d -v _odi -i _odi L _f +v _id ] In the formula: v _odi is the d-axis voltage component of the VSG output, L _f , C _f , r _f represent the inductance, capacitance and resistance of the LC filter respectively, i _odi is the d-axis current component of the VSG output, i _d , i _q d and q-axis current components input by the VSG, v _id is the d-axis current component input by the VSG, and ω _i is the angular frequency output by the VSG; The tracking error is: In the formula: Another error variable e _2i is expressed as: e _2i ＝d _i v _0di -α _ij In the formula: α _ij is a virtual control quantity. 4. the micro-grid system VSG double droop control method based on the adaptive neural network according to claim 3, is characterized in that, the described angular frequency of regulating the micro-grid system by the virtual parameter output of the adaptive neural network controller comprises : From droop control: ω _i =ω _ni -k _p p _i In the formula: ω _i is the angular frequency output by the VSG, ω _ni is the per unit value of the angular frequency, and P _i is the output active power after filtering. The adjusted error is: The angular frequency output is: