Frequency modulation adaptive neural network control method of T-type three-level converter for energy storage

Abstract

The invention discloses a frequency modulation adaptive neural network control method for a T-type three-level converter for energy storage, and relates to the technical field of converters, S100, constructing the T-type three-level converter for energy storage, and taking the converter as primary frequency modulation equipment; s200, making a T-type three-level modulation signal strategy according to the circuit structure of the selected T-type three-level converter; and S300, combining the T-type three-level modulation signal strategy with an upper neural network module to learn and optimize a droop coefficient so as to carry out conversion on a signal of the converter. According to the method adopted by the invention, the T-type three-level converter for energy storage is controlled to increase inertia through the primary frequency modulation adaptive neural network, the stability of the frequency and the voltage of the grid-connected point is kept, and rapid active and reactive power adjustment is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of converters, and in particular to a frequency modulation adaptive neural network control method and system for a T-type three-level converter for energy storage. Background Art

[0002] Current energy storage batteries suffer from non-convergence issues with droop control, which prevents them from allocating power according to their capacity. This results in poor system frequency stability and a lack of adaptability. Traditional capacity allocation methods rely solely on fixed-parameter droop control strategies, which fail to accurately reflect the remaining capacity of energy storage units during frequency modulation. This results in units with high remaining capacity being discharged less and charged more, exacerbating the aging process of energy storage units. Summary of the Invention

[0003] In view of one or more of the above-mentioned existing problems, the present invention proposes a frequency modulation adaptive neural network control method for a T-type three-level converter for energy storage.

[0004] According to a first aspect of the present application, a frequency modulation adaptive neural network control method for a T-type three-level converter for energy storage includes the following:

[0005] S100. Constructing a T-type three-level converter for energy storage and using the converter as a primary frequency modulation device;

[0006] S200. Formulate a T-type three-level modulation signal strategy based on the circuit structure of the selected T-type three-level converter;

[0007] S300. The T-type three-level modulation signal strategy is combined with the neural network module to learn and optimize the droop coefficient of the converter signal.

[0008] As an achievable preferred method, S100. Construct a T-type three-level converter for energy storage, and use the converter as a primary frequency modulation device; the topology of the converter includes a power supply end and a three-phase bridge arm in parallel, the power supply end is connected to two DC capacitors C1 and C2 in series, each phase bridge arm includes two IGBT tubes in series, one side of the midpoint of each phase bridge arm is connected in series with two IGBT tubes in different directions, and the other side is connected to the midpoint of the DC capacitors C1 and C2, the midpoints of the two capacitors are connected to one end of two IGBT tubes in different directions in each phase bridge arm, and each IGBT tube is driven by a control circuit; the output end of the three-phase bridge arm is connected to an LCL filter.

[0009] As an achievable preferred method, S200 formulates a T-type three-level modulation signal strategy and designs filter parameters based on the circuit structure of the selected T-type three-level converter; specifically includes:

[0010] The ratio of the two inductors La1 / La2 is between 3 and 7. Since the size of the filter capacitor C is fixed, the inverter side inductance can be calculated, that is, the grid side inductance can be obtained. The maximum value of the grid side inductance is as follows (3):

[0011]

[0012] The minimum value of the grid-side inductance is given by the following formula (4):

[0013]

[0014] Among them, f s is the switching frequency, I is the rated current effective value, v aN′ is the grid voltage, ω0 is the grid voltage angular frequency, and the filter capacitor Where P is the grid voltage and ω0 is the grid voltage angular frequency.

[0015] As an achievable preferred embodiment, S300. The T-type three-level modulation signal strategy is combined with the neural network module to learn and optimize the droop coefficient to adjust the signal of the converter, including the following steps:

[0016] S301. The primary frequency regulation energy storage power station is adjusted according to the Pf droop characteristic, so that the grid connection point has the ability to actively support and participate in the rapid adjustment of the grid frequency. The obtained grid frequency is shown in formula (5):

[0017] P=P0-k p (ff d )(5)

[0018] In formula (5), P0 is the initial power (MW), f d is the primary frequency modulation dead zone (±0.05Hz), f is the grid connection point frequency, k P is the active power droop coefficient;

[0019] S302. Reactive power regulation is achieved by real-time monitoring of the voltage at the grid connection point. When the voltage exceeds the dead zone, the voltage is drooped by QU, as shown in formula (6):

[0020] Q=Q0-k q (UU d )(6)

[0021] In formula (5), Q0 is the initial power (MW), U d is the voltage regulation dead zone, U is the grid connection point voltage;

[0022] S303. Adaptive droop coefficient k P and k Q The signal of the converter is regulated by introducing an adaptive neural network approximation function.

[0023] The beneficial effects of the present invention are:

[0024] The method adopted by the present invention is to increase the inertia of a T-type three-level converter for energy storage through primary frequency modulation adaptive neural network control, maintain the stability of the frequency and voltage at the grid connection point, and realize rapid active and reactive power regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic diagram of the application of the adaptive neural network control method for frequency modulation of a T-type three-level converter for energy storage;

[0026] Figure 2 This is a waveform diagram of the frequency modulation adaptive neural network control method of the T-type three-level converter for energy storage using the present application;

[0027] Figure 3 This is a waveform diagram of another embodiment of the frequency modulation adaptive neural network control method of a T-type three-level converter for energy storage according to the present application;

[0028] Figure 4 This is a waveform diagram of another embodiment of the frequency modulation adaptive neural network control method of a T-type three-level converter for energy storage according to the present application;

[0029] The above drawings illustrate specific embodiments of the present disclosure, which will be described in more detail below. These drawings and textual descriptions are not intended to limit the scope of the present disclosure in any way, but rather to illustrate the concepts of the present disclosure to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0030] The technical solution of the application is further described in detail below with reference to the accompanying drawings.

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0033] The following is combined with Figure 1-3, some embodiments of the present invention are described in detail. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0034] The present invention provides a frequency modulation adaptive neural network control method for a T-type three-level converter for energy storage, and the frequency modulation adaptive neural network control method for a T-type three-level converter for energy storage includes the following contents:

[0035] S100. Construct a T-type three-level converter for energy storage and use the converter as a primary frequency modulation device; S100. Construct a T-type three-level converter for energy storage and use the converter as a primary frequency modulation device; the topology of the converter includes a power supply end and a three-phase bridge arm in parallel, the power supply end is connected to two DC capacitors C1 and C2 in series, each phase bridge arm includes two IGBT tubes in series, one side of the midpoint of each phase bridge arm is connected in series with two IGBT tubes in different directions, and the other side is connected to the midpoint of the DC capacitors C1 and C2, the midpoints of the two capacitors are connected to one end of two IGBT tubes in different directions of each phase bridge arm, and each IGBT tube is driven by a control circuit; the output end of the three-phase bridge arm is connected to an LCL filter.

[0036] like Figure 1 As shown in the figure, a T-type three-level converter for energy storage is used as a key device for primary frequency modulation. This topology uses an LCL filter, as shown in the figure. Figure 1 As shown. ao 、v bo 、v co is the potential difference between the three-phase bridge arm connection point and the DC capacitor midpoint O, i a1 、i b1 、i c1 is the three-phase inductor current value on the inverter side, v aN 、v bN 、v cN is the three-phase filter capacitor voltage, i Ca 、i Cb 、i Cc is the current flowing through the three-phase filter capacitor, and i a2 、i b2 、i c2 is the current in the three-phase inductance on the grid side. NO is the potential difference between point N and point O, v NN′ is the potential difference between points N and N′. When the three-phase grid is balanced, the voltage and current at each reference point satisfy the equilibrium equation. Calculate the power and PQ-based adaptive droop control equations in the dq coordinate system.

[0037] S200. According to the circuit structure of the selected T-type three-level converter to develop T-type three-level modulation signal strategy; S200. According to the circuit structure of the selected T-type three-level converter to develop T-type three-level modulation signal strategy and design filter parameters;

[0038] Performing switching mode analysis, the topology of the T-type three-level converter can output the same output as the conventional three-level converter at three levels, namely +U dc 、-U dc , 0, set the above three states to P, N, and O states respectively, set point o to the midpoint of the two capacitors C1 and C2, and take the current flowing out of the bridge arm as the positive reference direction. When the current is positive or negative, there are three switch switching states.

[0039] Output P state: Set S a1 In the on state, S a2 、S a3 、S a4 In the off state, when the current is positive / negative, the current only flows through S a1 , does not flow through S a2 、S a3 、S a4 At this time, the inverter voltage relative to the reference point o is U dc / 2.

[0040] Output N state: Set S a4 In the on state, S a1 、S a2 、S a3 In the off state, the current only flows through S a4 , does not flow through S a1 、S a2 、S a3 At this time, the inverter voltage relative to the reference point o is -U dc / 2.

[0041] Output O state: Set S a2 、S a3 In the on state, S a1 、S a4 In the off state, the current flows through S a2 、S a3 , does not flow through S a1 、S a4 , at this time the inverter voltage relative to the reference point o is 0.

[0042] In a T-type three-level inverter, each power switch withstands a voltage equal to half the bus voltage. When the output is in the P or N state, only one power switch is required to conduct, and the positive or negative output voltage is independent of the current reference direction. Table 2 shows the relationship between the T-type three-level output voltage and the switch state.

[0043] Table 2

[0044] Level <![CDATA[S a1 ]]> <![CDATA[S a2 ]]> <![CDATA[S a3 ]]> <![CDATA[S a4 ]]> state <![CDATA[+U dc ]]> 1 0 0 0 P 0 0 1 1 0 O <![CDATA[-U dc ]]> 0 0 0 1 N

[0045] When the above switch state switching method is used, there is only the conversion between O state and P state; O state and N state, but it is impossible to convert between P state and N state. That is, the switching must be carried out in the PON sequence, otherwise it may cause the DC bus to be short-circuited. At the same time, during the process of switching from P state to O state, S a1 Complete shutdown, S a2 、S a3 During this process, three switching transistors switch on and off, complicating the commutation process. Table 2 uses a dual-carrier modulation strategy, driving only one transistor at high voltage and only four transistors at low voltage. This modulation strategy has the advantage of simple driving, but the disadvantage is that the switching frequencies of transistors 1-4 are inconsistent, which is not conducive to heat dissipation design.

[0046] To solve this problem, a new commutation method is chosen to change the switch state switching. Table 3 uses a dual-carrier modulation strategy to drive tubes 1 and 2 at a high level and tubes 3 and 4 at a low level. The relationship between the improved output voltage and the switch state is shown in Table 3.

[0047] Table 3

[0048] Level <![CDATA[S a1 ]]> <![CDATA[S a2 ]]> <![CDATA[S a3 ]]> <![CDATA[S a4 ]]> state <![CDATA[+U dc ]]> 1 1 0 0 P 0 0 1 1 0 O <![CDATA[-U dc ]]> 0 0 1 1 N

[0049] As shown in the table above, in the forward-backward T-shaped three-level commutation mode, the switching states of the two switching elements change only when the switching states are switched. From the perspective of the magnitude of switching losses, the improved commutation mode is more suitable.

[0050] It should be noted that the inductance on the inverter side determines the current ripple. The smaller the inductance, the greater the current ripple. The inductance on the grid side depends on the attenuation of high frequencies. Therefore, the inductance on the grid side cannot be too small. Under the same filtering conditions, the filter capacitor is limited by its output. Under the same filtering effect, increasing the filter capacitor will reduce the sum of the inductance value on the inverter side and the inductance on the grid side, while increasing the ratio of the inductance on the inverter side to the inductance on the grid side. To ensure unity power factor, the smaller the filter capacitor, the better. However, to ensure attenuation performance, the larger the filter capacitor, the better. Otherwise, a larger inductance will be required. Therefore, it is best to start selecting at 1 / 2 of the maximum capacitance. If the restriction conditions are not met, the size of the filter capacitor can be further increased to the maximum value.

[0051] Specifically include:

[0052] The ratio of the two inductors La1 / La2 is between 3 and 7. Since the size of the filter capacitor C is fixed, the inverter side inductance can be calculated, that is, the grid side inductance can be obtained. The maximum value of the grid side inductance is as follows (3):

[0053]

[0054] The minimum value of the grid-side inductance is given by the following formula (4):

[0055]

[0056] Among them, f s is the switching frequency, I is the rated current effective value, v aN′ is the grid voltage, ω0 is the grid voltage angular frequency, and the filter capacitor Where P is the grid voltage and ω0 is the grid voltage angular frequency. The parameters of the grid-connected LCL filter designed in this application are shown in Table 1.

[0057] Table 1 1MW energy storage frequency regulation TPC parameters

[0058]

[0059]

[0060] S300. The T-type three-level modulation signal strategy is combined with the neural network module to learn and optimize the droop coefficient of the converter signal.

[0061] S301. The primary frequency regulation energy storage power station is adjusted according to the Pf droop characteristic, so that the grid connection point has the ability to actively support and participate in the rapid adjustment of the grid frequency. The obtained grid frequency is shown in formula (5):

[0062] P=P0-k p (ff d )(5)

[0063] In formula (5), P0 is the initial power (MW), f d is the primary frequency modulation dead zone (±0.05Hz), f is the grid connection point frequency, k P is the active power droop coefficient;

[0064] S302. Reactive power regulation is achieved by real-time monitoring of the voltage at the grid connection point. When the voltage exceeds the dead zone, the voltage is drooped by QU, as shown in formula (6):

[0065] Q=Q0-k q (UU d )(6)

[0066] In formula (5), Q0 is the initial power (MW), U d is the voltage regulation dead zone, U is the grid connection point voltage;

[0067] S303. Adaptive droop coefficient k P and k Q The signal of the converter is regulated by introducing an adaptive neural network approximation function.

[0068] Introducing adaptive neural network approximation function f(x)=W *T h(x)+ε(x), to optimize k P As an example, we introduce an adaptive neural network to approximate the function

[0069] Where, input variable X = [f(k), P(k), P(k-1)] T , W is the RBF neural network weight vector, and H is the radial basis vector.

[0070] The Gaussian function can be written as:

[0071]

[0072] Where x is the 3D input vector; c i is the center of the ith radial basis function, δ i is the i-th perceptual variable, which determines the width of the basis function around the center point; ||xc i || is the norm of the vector, indicating the difference between x and c i The distance between them.

[0073] Performance indicator function:

[0074]

[0075] In order to minimize the error objective function between the expected output and the actual output of the RBF network model, the output weight, node center and node base width parameters are iteratively adjusted, with learning efficiency η∈[0,1] and momentum factor α∈[0,1].

[0076]

[0077] After optimization, the RBF network basis function half σ = 0.15, the initial value of the weighted coefficient w1 = 0.02, w2 = 0.01, w3 = 0.01. After multiple training experiments, the final parameters were selected as the learning efficiency η = 0.22 and the momentum factor α = 0.05. Q The optimization process and k P After the carrier in-phase pulse width modulation technology is processed, the harmonic distribution of the output phase voltage and the harmonics of the current under the resistance inductive load are as low as 0.73%.

[0078] like Figure 2As shown in the figure, in the range of 0-1000Hz, 6 cycles were taken from 2 seconds, and the measured waveform distortion rate was 0.73%. The frequency dropped from 50.2Hz to 49.8Hz, increasing the active output to reach a new balance point. The dynamic response can be completed within one power frequency cycle, and the voltage and power factor remain stable. Figure 3 As shown in the figure, the frequency increases from 49.8Hz to 50.2Hz, reducing the active output to reach a new balance point. The dynamic response can be completed within one power frequency cycle, and the voltage and power factor remain stable. Figure 4 As shown in the figure, the frequency rises from 49.8Hz to 50.2Hz to reach a steady state. The adaptive neural network control of the primary frequency modulation of the T-type three-level converter for energy storage can quickly reach a new equilibrium point when the frequency suddenly changes. The voltage and frequency remain stable and the dynamic performance is good.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

[0080] The above are only some embodiments of the present invention. For those skilled in the art, several modifications and improvements can be made without departing from the inventive concept of the present invention, which all fall within the scope of protection of the present invention.

Claims

1. A frequency modulation adaptive neural network control method for a T-type three-level converter for energy storage, characterized in that: The method comprises the following: S100. Constructing a T-type three-level converter for energy storage and using the converter as a primary frequency modulation device; S200. Formulate a T-type three-level modulation signal strategy based on the circuit structure of the selected T-type three-level converter; S300. The T-type three-level modulation signal strategy is combined with the neural network module to learn and optimize the droop coefficient of the converter signal.

2. The frequency modulation adaptive neural network control method of a T-type three-level converter for energy storage according to claim 1 is characterized in that: S100. Construct a T-type three-level converter for energy storage, and use the converter as a primary frequency modulation device; the topological structure of the converter includes a power supply end and a three-phase bridge arm in parallel, the power supply end is connected to two DC capacitors C1 and C2 in series, each phase bridge arm includes two IGBT tubes in series, one side of the midpoint of each phase bridge arm is connected in series with two IGBT tubes in different directions, and the other side is connected to the midpoint of the DC capacitors C1 and C2, the midpoints of the two capacitors are connected to one end of two IGBT tubes in different directions of each phase bridge arm, and each IGBT tube is driven by a control circuit; the output end of the three-phase bridge arm is connected to an LCL filter.

3. The frequency modulation adaptive neural network control method of a T-type three-level converter for energy storage according to claim 2 is characterized in that: S200. Formulate a T-type three-level modulation signal strategy and design filter parameters according to the circuit structure of the selected T-type three-level converter; specifically including: The ratio of the two inductors La1 / La2 is between 3 and 7. Since the size of the filter capacitor C is constant, the inverter side inductance is calculated, that is, the grid side inductance is obtained; the maximum value of the grid side inductance is as follows: (3) The minimum value of the grid-side inductance is given by the following formula (4): (4) Among them, f s is the switching frequency, I is the rated current effective value, is the grid voltage, is the grid voltage angular frequency, the filter capacitor Where P is the grid voltage, is the grid voltage angular frequency.

4. The frequency modulation adaptive neural network control method of a T-type three-level converter for energy storage according to claim 3 is characterized in that: S300. The T-type three-level modulation signal strategy is combined with the neural network module to learn and optimize the droop coefficient to adjust the signal of the converter, comprising the following steps: S301. The primary frequency regulation energy storage power station is adjusted according to the Pf droop characteristics, so that the grid connection point has the ability to actively support and participate in the rapid adjustment of the grid frequency. The obtained grid frequency is shown in formula (5): （5） In formula (5), P0 is the initial power (MW), f d is the primary frequency modulation dead zone (±0.05Hz), f is the grid connection point frequency, k P is the active power droop coefficient; S302. Reactive power regulation is achieved by real-time monitoring of the voltage at the grid connection point. When the voltage exceeds the dead zone, the voltage is drooped by QU, as shown in equation (6): （6） In formula (5), Q0 is the initial power (MW), U d is the voltage regulation dead zone, U is the grid connection point voltage; S303. Adaptive droop coefficient k P and k Q The signal of the converter is regulated by introducing an adaptive neural network approximation function.

Images