A grid-side current determination method for an LCL filter

Abstract

The present application relates to a kind of grid-side current determination method of LCL filter, it is related to energy storage converter grid-connected control technical field, comprising: constructing including to be solved input gain and to be solved total disturbance LCL filter super-local model;The input gain characterizes the dynamic proportional relationship between filter capacitor voltage change rate and converter-side current, for eliminating the effect of filter capacitor drift uncertainty on grid-side current;The total disturbance characterizes the comprehensive effect of unknown factor that influences filter capacitor voltage change rate, unknown factor includes grid-side current;Based on the constraint that total disturbance of adjacent sampling time is approximately equal, according to the sampling value of converter-side current and filter capacitor voltage at multiple sampling times, input gain is solved;With the observation error of filter capacitor voltage tends to zero as target, based on super-local model, total disturbance is solved;According to total disturbance and input gain, grid-side current is determined.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of grid-connected control technology for energy storage converters, specifically to a method for determining the grid-side current of an LCL filter. Background Technology

[0002] To reduce hardware costs, energy storage converter systems connected to the grid via LCL filters typically only use a current sensor on the converter side to collect the converter-side filter current and a voltage sensor on the filter capacitor side to collect the filter capacitor voltage. The grid-side filter current is directly and equivalently replaced by the converter-side filter current. However, current actually flows in the filter capacitor branch, and the direct equivalent method ignores the capacitor current component, which introduces steady-state errors and high-frequency harmonic interference into the control system, reducing the quality of the grid-connected current.

[0003] In existing technologies, some solutions obtain the capacitor current by differentiating the filter capacitor voltage, and then reconstruct the grid-side filter current. However, this method is an open-loop observation, which is highly sensitive to voltage sampling noise and sensor interference, and relies entirely on an ideal capacitor model. When the capacitor parameters shift, the reconstructed grid-side filter current has phase and amplitude deviations, resulting in large observation errors, slow dynamic response, and insufficient robustness. Summary of the Invention

[0004] The purpose of the embodiments in this specification is to provide a method for determining the grid-side current of an LCL filter, so as to overcome the problems of large observation error, slow dynamic response, and insufficient robustness in existing methods.

[0005] To address the aforementioned technical problems, this specification provides, in one aspect, a method for determining the grid-side current of an LCL filter, comprising: A hyperlocal model of the LCL filter is constructed, comprising the input gain to be solved and the total disturbance to be solved. The hyperlocal model characterizes the correlation between the rate of change of the filter capacitor voltage and the converter-side current, the input gain, and the total disturbance. The input gain characterizes the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current, and is used to eliminate the effect of filter capacitor drift uncertainty on the grid-side current. The total disturbance characterizes the combined effect of unknown factors affecting the rate of change of the filter capacitor voltage, including the grid-side current. Based on the constraint that the total disturbance is approximately equal at adjacent sampling times, the input gain is solved according to the sampled values ​​of the converter side current and the filter capacitor voltage at multiple sampling times. With the goal of reducing the observation error of the filter capacitor voltage to zero, the total disturbance is solved based on the hyperlocal model. The grid-side current is determined based on the total disturbance and the input gain.

[0006] Furthermore, the construction of the LCL filter hyperlocal model, which includes the input gain to be solved and the total perturbation to be solved, includes: Based on the topology of the LCL filter, determine the physical relationship between the rate of change of the filter capacitor voltage and the converter-side current, the grid-side current, and the filter capacitor. The input gain to be solved is determined based on the filter capacitor; The total disturbance to be solved is determined based on the effect of grid-side current on the rate of change of filter capacitor voltage. Based on physical relationships, a hyperlocal model is constructed that takes the converter-side current as input, the change rate of the filter capacitor voltage as output, and includes the input gain and the total disturbance to be solved.

[0007] Furthermore, the hyperlocal model constructed based on physical relationships, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain to be solved and the total disturbance to be solved, includes: Based on physical relationships, a hyperlocal model is constructed using the following formula, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain and the total disturbance to be solved: ; In the formula, This is the voltage across the filter capacitor. The rate of change of the filter capacitor voltage. For converter-side current, The input gain to be solved, The total disturbance to be solved is denoted as .

[0008] Furthermore, the constraint that the total disturbance is approximately equal at adjacent sampling times, and the calculation of the input gain based on the sampled values ​​of the converter-side current and the filter capacitor voltage at multiple sampling times, includes: The hyperlocal model is discretized to establish discretized equations for the current sampling time and the previous sampling time, respectively. With the constraint that the total disturbance at adjacent sampling times is approximately equal, the discretization equations of the current sampling time and the previous sampling time are combined to eliminate the total disturbance; Based on the results of the simultaneous solution, the input gain at the current sampling moment is calculated using the sampled values ​​of the filter capacitor voltage and the converter side current at the current sampling moment and the previous sampling moment.

[0009] Furthermore, the discretization of the hyperlocal model to establish discretized equations for the current sampling time and the previous sampling time includes: The hyperlocal model is discretized using the following formula to establish the discretized equations for the current sampling time and the previous sampling time, respectively: ; In the formula, , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time The converter-side current and total disturbance, and Sampling time The converter-side current and total disturbance, This is the input gain.

[0010] Furthermore, the step of calculating the input gain at the current sampling moment based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling moments, includes: Based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling times, the input gain at the current sampling time is calculated using the following formula: ; In the formula, The input gain at the current sampling time. , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time and The converter-side current.

[0011] Furthermore, the step of aiming to minimize the observation error of the filter capacitor voltage, and solving for the total disturbance based on a hyperlocal model, includes: Input the input gain and converter-side current into the extended state observer; The deviation between the observed value and the sampled value of the filter capacitor voltage at the current moment is obtained as the observation error of the filter capacitor voltage. With the goal of reducing the observation error to zero, the total perturbation in the hyperlocal model is iteratively estimated using an extended state observer, and the total perturbation at the current sampling time is output.

[0012] Furthermore, the step of iteratively estimating the total perturbation in the hyperlocal model using an extended state observer with the goal of minimizing the observation error to zero, and outputting the total perturbation at the current sampling time, includes: With the goal of minimizing the observation error, the total perturbation in the hyperlocal model is iteratively estimated using the following formula, outputting the total perturbation at the current sampling time: ; In the formula, For observation error, and These are the filter capacitor voltages at the sampling time. The observed values ​​and sampled values, The voltage of the filter capacitor at the sampling time The observed values, The sampling period is Sampling time The converter-side current, For input gain, and Sampling time and The total disturbance, and These are the first and second gain coefficients of the extended state observer, respectively. To expand the bandwidth of the state observer.

[0013] Furthermore, determining the grid-side current based on the total disturbance and input gain includes: The grid-side current at the current sampling time is determined by comparing the total disturbance with the input gain at the current sampling time.

[0014] Further, determining the grid-side current at the current sampling time based on the comparison result of the total disturbance and the input gain at the current sampling time includes: Based on the comparison between the total disturbance and the input gain at the current sampling time, the grid-side current at the current sampling time is determined using the following formula: ; In the formula, , and These represent the grid-side current, total disturbance, and input gain at the sampling time, respectively.

[0015] Furthermore, embodiments of this specification provide a grid-side current determination device for an LCL filter, comprising: A construction module is used to construct a hyperlocal model of the LCL filter, including the input gain to be solved and the total disturbance to be solved. The hyperlocal model represents the correlation between the rate of change of the filter capacitor voltage and the converter-side current, the input gain, and the total disturbance. The input gain represents the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current, and is used to eliminate the effect of filter capacitor drift uncertainty on the grid-side current. The total disturbance represents the combined effect of unknown factors affecting the rate of change of the filter capacitor voltage, including the grid-side current. The first solution module is used to solve for the input gain based on the constraint that the total disturbance is approximately equal at adjacent sampling times, and the sampled values ​​of the converter side current and the filter capacitor voltage at multiple sampling times. The second solution module is used to solve the total disturbance based on the hyperlocal model, with the goal of making the observation error of the filter capacitor voltage approach zero. The determination module is used to determine the grid-side current based on the total disturbance and the input gain.

[0016] Furthermore, embodiments of this specification provide a computer device, including: Memory, used to store computer programs; A processor is used to execute the computer program to implement the above-described method for determining the grid-side current of the LCL filter.

[0017] In another aspect, embodiments of this specification provide a computer storage medium storing computer program instructions, which, when executed, implement the above-described method for determining the grid-side current of the LCL filter.

[0018] In another aspect, embodiments of this specification provide a computer program product, including a computer program that, when executed by a processor, implements the above-described method for determining the grid-side current of the LCL filter.

[0019] As can be seen from the technical solutions provided in the embodiments of this specification above, the input gain in the embodiments of this specification is adaptively adjusted according to the drift of the filter capacitor parameters, so that the determination result of the grid-side current is not affected by capacitor aging, temperature changes and batch differences, thus improving parameter robustness; the total disturbance is solved by the closed-loop feedback of the observation error tending to zero, rather than by open-loop differentiation or filtering approximation, which can suppress the interference of sampling noise and unmodeled dynamics on the observation results and improve the determination accuracy of the grid-side current; in addition, the grid-side current is determined by the total disturbance and the input gain together, without relying on the nominal parameters of the filter capacitor, thus avoiding steady-state errors introduced by parameter mismatch. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below.

[0021] Figure 1 This is a schematic diagram of the energy storage converter circuit topology connected to the grid via an LCL filter, as provided in the embodiments of this specification. Figure 2 This is a flowchart of a method for determining the grid-side current of an LCL filter provided in the embodiments of this specification; Figure 3 A schematic diagram of the control logic for a method for determining the grid-side current of an LCL filter provided in an embodiment of this specification; Figure 4 A schematic diagram comparing the observed current and actual current of the LCL filter provided in the embodiments of this specification; Figure 5 A schematic diagram illustrating the observation error of the LCL filter provided in the embodiments of this specification; Figure 6 A schematic diagram comparing the observed current and the actual current of an LCL filter current determination method provided in the embodiments of this specification. Figure 7 A schematic diagram illustrating the observation error of a grid-side current determination method for an LCL filter provided in the embodiments of this specification; Figure 8 This is a schematic diagram comparing the observed current and actual current of the LCL filter under power grid fault conditions provided in the embodiments of this specification; Figure 9 A schematic diagram illustrating the observation error of the LCL filter under power grid fault conditions provided in the embodiments of this specification; Figure 10 A schematic diagram comparing the observed current and the actual current under grid fault conditions for a grid-side current determination method for an LCL filter provided in the embodiments of this specification; Figure 11 A schematic diagram illustrating the observation error under grid fault conditions of a grid-side current determination method for an LCL filter provided in the embodiments of this specification; Figure 12 A schematic diagram comparing the observed current and the actual current under the attenuation of the filter capacitor value in a method for determining the grid-side current of an LCL filter provided in the embodiments of this specification. Figure 13 A schematic diagram illustrating the observation error under the attenuation of the filter capacitor value in a method for determining the grid-side current of an LCL filter provided in an embodiment of this specification; Figure 14 This is a schematic diagram of the structure of a grid-side current determination device for an LCL filter provided in the embodiments of this specification; Figure 15 This is a schematic diagram of the structural composition of a computer device provided in the embodiments of this specification. Detailed Implementation

[0022] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0023] It should be noted that the terms "first," "second," etc., used in this specification, claims, and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0024] In some embodiments, the energy storage converter circuit topology connected to the grid via an LCL filter is as follows: Figure 1 As shown, the direction of current flow towards the power grid is defined as positive.

[0025] The Power Conversion System (PCS) is a key power electronic device connecting the energy storage battery system to the AC power grid. Its core function is to achieve bidirectional energy conversion and flow control. In charging mode, the PCS converts AC power into DC power to charge the battery pack. In discharging mode, it inverts the DC power stored in the battery pack into AC power that matches the voltage amplitude, frequency, and phase of the grid, feeding it back to the AC power grid. The PCS typically uses Pulse Width Modulation (PWM) technology to control the switching of power switching devices (such as IGBTs and MOSFETs) to output an AC voltage with controllable amplitude and frequency. However, the PWM modulation process inevitably generates a large number of high-frequency harmonic components near the switching frequency. If directly fed into the grid, this will severely pollute the power quality and may cause equipment failure. Therefore, the AC output side of the PCS needs to be equipped with a filter to effectively suppress high-frequency switching harmonics and ensure that the grid-connected current meets power quality standards.

[0026] An LCL filter is a third-order passive filter network connected between the AC output of an energy storage converter and the grid's point of common coupling (PCC) to filter out high-frequency switching harmonics in the converter's output voltage. The LCL filter consists of three passive components connected in series and parallel: a first inductor (converter-side inductor), one end of which is connected to the AC output of the energy storage converter; a filter capacitor, one end of which is connected to the other end of the first inductor, and the other end of which is connected to the AC neutral point (or ground); and a second inductor (grid-side inductor), one end of which is connected to the PCC of the filter capacitor and the first inductor, and the other end of which is connected to the PCC. The first inductor, filter capacitor, and second inductor together form a "Π" or "T" type topology. Compared to a single-inductor filter, the LCL filter, with the same total inductance value, has superior attenuation characteristics for high-frequency harmonics (high-frequency harmonic attenuation rate can reach -60dB / dec), and is therefore widely used in high-power grid-connected converters. However, the LCL filter is a third-order system, and its amplitude-frequency response exhibits a significant resonance peak at the resonant frequency. If the control strategy is not properly designed, this resonance peak may cause system oscillation or even instability. To suppress resonance, damping measures need to be taken in the main circuit or control algorithm. In the embodiments described in this specification, the filter capacitor voltage refers to the voltage across the filter capacitor; the converter-side filter current is the current flowing through the first inductor; and the grid-side current is the current injected into the grid through the second inductor.

[0027] The power grid (also known as the power grid or the point of common coupling of the power system) is a unified power transmission and distribution network composed of power plants, transmission and transformation equipment, and distribution networks, providing stable AC power to various electrical loads. In grid-connected power generation systems, the power grid, as the receiving end of renewable energy, requires that the AC current injected by grid-connected equipment maintains the same frequency and phase as the grid voltage (i.e., unity power factor operation), and the harmonic distortion rate of the injected current must strictly comply with the limits of power quality standards such as IEEE 519 or GB / T 14549. The fundamental frequency of the power grid is the industrial frequency (50Hz or 60Hz), corresponding to a fundamental period of 20ms or 16.67ms. In actual operation, the voltage amplitude and frequency of the power grid may experience short-term deviations due to load fluctuations, fault disturbances, or large-scale integration of new energy sources. Such grid disturbances can be coupled to the converter system through LCL filters, affecting the control accuracy of the grid-connected current. Therefore, in the grid-connected control of energy storage converters, precise monitoring and control of the grid-side current are necessary to ensure power quality and stable system operation.

[0028] Existing technologies rely on open-loop observation, making them sensitive to voltage sampling noise and sensor interference. Switching ripple and white noise are superimposed on the capacitor voltage, and there is an inherent trade-off between noise suppression and observation bandwidth in high-pass filters, resulting in glitches in the reconstructed grid-side current. Traditional reconstruction methods depend on ideal capacitor models, neglecting parasitic parameters such as the capacitor's equivalent series resistance. When capacitor aging or temperature changes cause parameter shifts, phase and amplitude deviations are introduced into the observed capacitor current. Open-loop observation lacks feedback correction, and low-order components in the capacitor voltage are selectively amplified by the differentiating element, increasing the harmonic content of the reconstructed grid-side current. The phase of the ideal differential is a constant 90° lead across the entire frequency band, while the phase of the high-pass filter transitions from 90° to 0° near the cutoff frequency. In practical applications, the phase deviates from the pure differential phase, resulting in phase lag in the observed capacitor current.

[0029] To address the aforementioned issues, this specification provides a method for determining the grid-side current of an LCL filter, as illustrated in the embodiments below. Figure 2 and Figure 3 As shown, the specific implementation may include the following steps: S201: Construct a hyperlocal model of the LCL filter, including the input gain to be solved and the total disturbance to be solved; the hyperlocal model characterizes the correlation between the rate of change of the filter capacitor voltage and the converter-side current, the input gain, and the total disturbance; the input gain characterizes the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current, and is used to eliminate the effect of filter capacitor drift uncertainty on the grid-side current; the total disturbance characterizes the combined effect of unknown factors affecting the rate of change of the filter capacitor voltage, including the grid-side current.

[0030] To overcome the limitations of traditional grid-side current reconstruction methods, which rely on ideal capacitor models and lack parameter adaptation capabilities, a hyperlocal model of the LCL filter can be constructed, incorporating the input gain and the total disturbance to be solved. This hyperlocal model characterizes the relationship between the rate of change of the filter capacitor voltage and the converter-side current, input gain, and total disturbance. The input gain characterizes the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current, eliminating the impact of filter capacitor drift uncertainty on the accuracy of grid-side current determination. In other words, the input gain adaptively adjusts with changes in the filter capacitor parameters, ensuring that the grid-side current determination is independent of the nominal capacitor value. The total disturbance characterizes the combined effect of unknown factors influencing the rate of change of the filter capacitor voltage. These unknown factors include grid-side current information; the total disturbance, as a lumped term, integrates the influence of the grid-side current on the rate of change of the capacitor voltage, along with other unmodeled dynamics, simplifying the model structure and reducing dependence on precise physical parameters. By constructing a hyperlocal model, the problem of determining the grid-side current is transformed into a joint problem of solving the input gain and the total disturbance, laying the foundation for subsequent data-driven model-free observations and solving the observation bias problem caused by capacitor parameter mismatch in traditional methods.

[0031] In some embodiments, the physical relationship between the rate of change of the filter capacitor voltage and the converter-side current, the grid-side current, and the filter capacitor can be determined based on the topology of the LCL filter.

[0032] To establish the physical foundation upon which the hyperlocal model relies, the physical relationship between the rate of change of the filter capacitor voltage and the converter-side current, grid-side current, and filter capacitor parameters is determined based on the topology of the LCL filter. This physical relationship is based on Kirchhoff's current law, namely, at the capacitor node of the LCL filter, the current flowing through the converter-side inductor splits into two paths: one flows through the filter capacitor branch, and the other flows through the grid-side inductor and is injected into the grid. The relationship between the rate of change of the filter capacitor voltage and the current flowing through the filter capacitor branch is determined by the voltage-current characteristic equation of the capacitor element. This physical relationship serves as an initial reference for constructing the hyperlocal model, enabling the subsequent separation of the grid-side current term from the unknown combined effects. By establishing the above physical relationship, the causal dependencies between various electrical quantities are clarified, providing a physical basis for the definition of total disturbance and input gain in the hyperlocal model, thus giving the model construction clear engineering interpretability.

[0033] In some embodiments, the input gain to be solved can be determined based on the filter capacitor.

[0034] To eliminate the impact of filter capacitor parameter drift on observation accuracy, the input gain to be solved is determined based on the real-time response characteristics of the filter capacitor. During operation, the capacitance value of the filter capacitor drifts due to dielectric aging, temperature changes, or batch variations in manufacturing. This drift directly alters the sensitivity of the capacitor voltage change rate to the converter-side current; specifically, the same converter-side current at different capacitance values ​​results in different capacitor voltage change rates. The input gain, as a quantitative indicator of this sensitivity, adaptively determines its value based on the current actual state of the filter capacitor parameters. In one implementation, the input gain is calculated using the nominal parameters of the filter capacitor and updated in real-time during operation via an online identification algorithm. In another implementation, the input gain is calculated online directly using real-time sampled data of the converter-side current and the filter capacitor voltage through an algebraic solution, without relying on any prior information about the capacitor parameters. By introducing a dynamic adaptive mechanism for the input gain, the determination of the grid-side current does not depend on the nominal parameters of the filter capacitor, fundamentally eliminating the adverse effects of capacitor parameter uncertainty on observation accuracy.

[0035] In some embodiments, the total disturbance to be solved can be determined based on the effect of the grid-side current on the rate of change of the filter capacitor voltage.

[0036] To effectively extract grid-side current information from lumped disturbances, the total disturbance to be solved is determined based on the effect of grid-side current on the rate of change of the filter capacitor voltage. Grid-side current, as the current injected into the grid through the grid-side inductor, directly reflects the magnitude of grid-connected power and power quality in terms of amplitude and frequency. The total disturbance is constructed as a lumped variable containing grid-side current information; that is, the total disturbance includes at least a component of the rate of change of capacitor voltage caused by grid-side current, and may also include other influencing factors such as grid voltage fluctuations, LCL resonant modes, and unmodeled dynamics. In one embodiment, the total disturbance is defined as the negative of the ratio of grid-side current to filter capacitor parameters, i.e., the contribution term of grid-side current to the rate of change of capacitor voltage. In another embodiment, the total disturbance further includes a deviation term caused by parasitic parameters of the LCL filter; that is, the grid-side current contribution term and the parasitic parameter deviation term are superimposed to constitute the total disturbance. By embedding grid-side current information into the definition of total disturbance, the grid-side current is no longer solved directly as an independent unknown, but is obtained by inverse calculation after observing the total disturbance. This transforms the current observation problem into a disturbance observation problem, simplifying the observer design.

[0037] In some embodiments, a hyperlocal model can be constructed based on physical relationships, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain to be solved and the total disturbance to be solved.

[0038] To ensure that the hyperlocal model is mathematically solvable and physically equivalent to the original system, a hyperlocal model is constructed based on physical relationships. This model takes the converter-side current as input, the rate of change of the filter capacitor voltage as output, and includes the input gain and the total disturbance to be solved. This hyperlocal model reduces the third-order physical model of the original LCL filter to a first-order form, retaining only the filter capacitor voltage as a state variable. In one embodiment, the construction of the hyperlocal model includes, but is not limited to: multiplying the converter-side current by the input gain as the measurable driving term of the model, and using the total disturbance as the unknown compensation term; the sum of these two is equivalent to the rate of change of the filter capacitor voltage. In another embodiment, the construction of the hyperlocal model includes, but is not limited to: including all terms related to the grid-side current in the physical relationships as part of the total disturbance, and defining the coefficients related to the filter capacitor parameters in the physical relationships as the input gain, so that the hyperlocal model does not explicitly contain any capacitor parameters in form. By constructing a hyperlocal model, grid-side currents, parasitic parameters, and external disturbances that are difficult to model accurately in the original system are all included in the total disturbance. The model structure is simple and does not depend on precise parameters, providing a mathematical model that is easy to solve in real time for the design of subsequent observers.

[0039] S202: Based on the constraint that the total disturbance is approximately equal at adjacent sampling times, the input gain is solved according to the sampled values ​​of the converter side current and the filter capacitor voltage at multiple sampling times.

[0040] In some embodiments, the sampling period is much shorter than the fundamental frequency period. The sampling period is the sampling time interval, and the fundamental frequency period is the period corresponding to the fundamental frequency of the power grid.

[0041] To ensure the feasibility of solving the discrete domain problem, the sampling period is much smaller than the fundamental frequency period. The sampling period characterizes the time interval between two adjacent sampling operations, and its value determines the sampling resolution of the digital control system. The fundamental frequency period characterizes the length of a complete AC cycle corresponding to the fundamental frequency of the power grid. The ratio of the sampling period to the fundamental frequency period determines the number of sampling points for electrical quantities within one fundamental frequency cycle; the larger this number, the higher the reconstruction accuracy of the analog signal by the digital control system. The constraint that the sampling period is much smaller than the fundamental frequency period ensures that the change in the grid-side current between adjacent sampling moments is extremely small, thus making the premise that the total disturbance is approximately equal at adjacent sampling moments valid in engineering practice.

[0042] To utilize the slowly varying characteristics of total disturbance to achieve online calculation of input gain, the input gain is calculated based on the constraint that the total disturbance is approximately equal at adjacent sampling times, using the sampled values ​​of converter-side current and filter capacitor voltage at multiple sampling times. The premise that the total disturbance is approximately equal at adjacent sampling times is that the sampling period is sufficiently small relative to the grid fundamental period; under this premise, the difference in total disturbance between adjacent sampling times is negligible. In one implementation, the input gain at the current time is directly obtained by simultaneously solving the system of sampled values ​​from the current and previous sampling times. In another implementation, the input gain is obtained by combining sampled values ​​from multiple historical sampling times prior to the current time, and the solution is filtered to suppress measurement noise interference with the gain calculation. Since the change in input gain is only caused by capacitor parameter drift, and its rate of change is much lower than the sampling frequency, the input gain remains approximately constant at adjacent sampling times. By solving the equations at multiple sampling times under the constraint of approximately equal total perturbation, the input gain can be automatically updated to follow the time-varying characteristics of the capacitor parameters without offline calibration, thus solving the problem of mismatch in fixed gain after capacitor parameter drift.

[0043] In some embodiments, the hyperlocal model can be discretized to establish discretized equations for the current sampling time and the previous sampling time, respectively.

[0044] To convert the hyperlocal model in the continuous domain into a form executable by a digital signal processor, the hyperlocal model is discretized to establish discretized equations for the current sampling time and the previous sampling time. The discretization process employs a numerical differential approximation method, converting the continuous domain differential operations in the hyperlocal model into difference operations between adjacent sampling times. In one embodiment, a first-order forward Euler discretization method is used, approximating the rate of change of the filter capacitor voltage as the voltage difference between the current time and the previous sampling time divided by the sampling period. In another embodiment, a first-order backward Euler discretization method is used, approximating the rate of change of the filter capacitor voltage as the voltage difference between the current time and the next sampling time divided by the sampling period. After discretization, the input gain and total disturbance in the hyperlocal model are expressed as an algebraic combination of the sampled values ​​at each sampling time, facilitating real-time calculation by the digital signal processor within each sampling period. Through discretization, the hyperlocal model can be solved recursively step-by-step in the digital control system, providing an executable numerical calculation form for online identification of input gain and real-time observation of total disturbance.

[0045] In some embodiments, the total disturbance can be eliminated by simultaneously solving the discretization equations of the current sampling time and the previous sampling time, with the constraint that the total disturbance is approximately equal at adjacent sampling times.

[0046] To eliminate the interference of the unknown total disturbance on the input gain solution, the discretized equations of the current sampling time and the previous sampling time are simultaneously solved, constrained by the approximate equality of the total disturbance at adjacent sampling times. The basic principle of this elimination process is that the discretized equations of two adjacent sampling times contain the same unknown (i.e., the total disturbance), and the total disturbance in the two equations is approximately equal. Therefore, subtracting the two equations cancels out the total disturbance terms, resulting in a difference equation containing only the input gain as the unknown. Specific implementation methods for the elimination process include, but are not limited to: subtracting the discretized equation of the previous sampling time from the discretized equation of the current sampling time term by term. The total disturbance term is eliminated in the subtraction operation, and the remaining terms only contain the product of the input gain and the difference between the converter-side current, as well as related terms of the second-order difference of the filter capacitor voltage. After the elimination process, the input gain can be directly and explicitly expressed by the sampled values. By solving the simultaneous elimination process, the solution of the input gain does not depend on the specific value of the total disturbance, avoiding the difficulty caused by the unknown total disturbance and simplifying the gain calculation process.

[0047] In some embodiments, the input gain at the current sampling moment can be calculated based on the results of the simultaneous solution, using the sampled values ​​of the filter capacitor voltage at the current sampling moment and the sampled value of the converter side current at the previous sampling moment.

[0048] To obtain the numerical solution of the input gain at the current sampling moment, based on the simultaneous solution results, the input gain at the current sampling moment is calculated using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling moments. Specifically, the difference equation obtained after elimination is algebraically operated, i.e., a linear equation in one variable concerning the input gain is solved, yielding that the input gain is equal to the ratio of the second-order difference value of the filter capacitor voltage to the first-order difference value of the converter-side current. The calculated input gain is in complex form. In one embodiment, the real part of this complex number is taken as the input gain value at the current sampling moment. In another embodiment, the magnitude of this complex number is taken as the input gain value at the current sampling moment. After calculating the input gain, in one embodiment, the gain value is directly output to the extended state observer for total disturbance observation at the current moment. In another embodiment, the gain value is filtered by a low-pass filter before output to suppress the contamination of the gain value by high-frequency measurement noise. The closed-form solution of the input gain is obtained directly through algebraic operations, the calculation process does not require iteration, the computational load is small, and it is suitable for real-time execution within a microsecond-level sampling period.

[0049] S203: With the goal of reducing the observation error of the filter capacitor voltage to zero, solve for the total disturbance based on the hyperlocal model.

[0050] To separate the total disturbance from the lumped unknowns and achieve indirect observation of the grid-side current, a hyperlocal model is used to solve for the total disturbance, aiming to make the observation error of the filter capacitor voltage approach zero. The observation error of the filter capacitor voltage characterizes the deviation between the observed value and the actual sampled value of the filter capacitor voltage, and this deviation directly reflects the fitting accuracy of the hyperlocal model to the dynamic behavior of the system. When the observation error approaches zero, the observed voltage output by the hyperlocal model equals the actual sampled voltage, and at this time, the estimated total disturbance within the model must approximate its true physical value. The total disturbance implicitly contains grid-side current information; therefore, the grid-side current can be indirectly obtained by solving for the total disturbance. In one embodiment, an extended state observer is used to perform online recursive estimation of the total disturbance, using the observation error as a feedback signal to drive the observer state update until the observation error converges to an acceptable range. In another embodiment, an algebraic solution method is used to express the observation error of the filter capacitor voltage as a function of the total disturbance, and the numerical solution of the total disturbance is obtained by solving for the zeros of this function. By using the convergence criterion of observation error approaching zero to solve for the total disturbance, the estimation process of the total disturbance has a clear convergence target and a closed-loop correction mechanism, avoiding the cumulative error problem caused by the lack of feedback in the open-loop observation method, thereby improving the extraction accuracy of implicit grid-side current information.

[0051] In some embodiments, the input gain and converter-side current input can be used to expand the state observer.

[0052] To achieve closed-loop estimation of the total disturbance in the hyperlocal model, the input gain and converter-side current are input to an extended state observer. The extended state observer characterizes a class of state estimators based on state-space models, which can expand all unknown dynamics and external disturbances in the system into additional state variables, and achieve real-time estimation of the extended state through feedback correction of the output error. This extended state observer uses the hyperlocal model as its internal prediction model, the input gain and converter-side current as model inputs, and the deviation between the observed and sampled values ​​of the filter capacitor voltage as the correction signal. In one embodiment, the extended state observer adopts a linear extended state observer structure, with its gain coefficient set according to the pole placement method, achieving linear estimation of the total disturbance with a fixed bandwidth. In another embodiment, the extended state observer adopts a nonlinear extended state observer structure, with its gain coefficient changing nonlinearly with the magnitude of the observation error, to achieve a balance between fast tracking of abrupt disturbances and low-noise estimation of gradual disturbances. By extending the state observer by inputting the input gain and converter-side current, the observer can keep up with the dynamic changes of the system in a timely manner after the model parameters (input gain) are adaptively updated, ensuring the real-time performance and accuracy of the total disturbance estimation. This solves the problem of inaccurate estimation by fixed-parameter observers after changes in system parameters.

[0053] In some embodiments, the deviation between the observed value and the sampled value of the filter capacitor voltage at the current moment can be obtained as the observation error of the filter capacitor voltage.

[0054] To quantify the prediction bias of the hyperlocal model on the capacitor voltage and drive the correction mechanism of the extended state observer, the deviation between the observed value and the sampled value of the filter capacitor voltage at the current moment is obtained as the observation error of the filter capacitor voltage. The observed value of the filter capacitor voltage is obtained by the extended state observer recursively based on the hyperlocal model, and its physical meaning is the prediction result of the capacitor voltage at the current moment based on the system state, input gain, converter side current, and total disturbance at the previous moment. The sampled value of the filter capacitor voltage is acquired in real time by a voltage sensor and represents the true physical value of the capacitor voltage at the current moment. The difference between the observed value and the sampled value is the observation error. The sign and magnitude of this error reflect the direction and degree of deviation of the hyperlocal model and the total disturbance estimate from the true system state. In one embodiment, the observation error is directly taken as the difference between the observed value and the sampled value, and this difference is used as the error feedback term in the subsequent observer state update equation. In another embodiment, the observation error is used for feedback after being amplitude-limited to prevent sudden changes in the observer state caused by instantaneous spike noise from the sensor. By establishing the deviation between the observed value and the sampled value as an error signal, a quantitative correction basis is provided for the extended state observer, enabling the observer to adjust the estimated value of the total disturbance in real time according to the magnitude and direction of the error signal, thereby forming a closed-loop feedback mechanism.

[0055] In some embodiments, with the goal of minimizing the observation error, the total perturbation in the hyperlocal model can be iteratively estimated by an extended state observer, and the total perturbation at the current sampling time can be output.

[0056] To achieve a step-by-step recursive solution for the total disturbance in the discrete domain, aiming to minimize the observation error, an extended state observer iteratively estimates the total disturbance in the hyperlocal model, outputting the total disturbance at the current sampling moment. This iterative estimation process is executed once per sampling period. The extended state observer corrects the estimated value of the total disturbance according to its internal state update equation based on the observation error acquired at the current moment. In one embodiment, the extended state observer multiplies the observation error at the current moment by a first gain coefficient and feeds it back to the estimation update channel of the total disturbance, achieving step-by-step approximation of the total disturbance estimate through this feedback. In another embodiment, the extended state observer multiplies the combined value of the observation errors at the current moment and the previous moment by a second gain coefficient and feeds it back to the estimation update channel of the total disturbance, enhancing the observer's ability to suppress persistent disturbances through the combination of differentiation or integration of the error. After one or more iterative updates, the observation error is compressed to an acceptable range, and the total disturbance estimate output by the extended state observer at this point is the total disturbance at the current sampling moment. By using the iterative estimation mechanism of the extended state observer, the total disturbance can be corrected at each sampling time based on the latest voltage observation error, realizing real-time closed-loop tracking of the total disturbance. This solves the problem that open-loop observation cannot keep up with the total disturbance when it changes rapidly due to factors such as grid fluctuations and load changes.

[0057] S204: Determine the grid-side current based on the total disturbance and input gain.

[0058] In some embodiments, the grid-side current can be determined based on the total disturbance and the input gain.

[0059] To address the insufficient observability issue encountered when directly solving for grid-side current as an independent unknown using traditional methods, this paper determines the grid-side current based on the total disturbance and input gain. In one implementation, the total disturbance and input gain are related by division, meaning the grid-side current equals the negative of the ratio of the total disturbance to the input gain. In another implementation, they are related by multiplication, i.e., the total disturbance is first negatively evaluated, then multiplied by the reciprocal of the input gain. Physically, the input gain corresponds to the reciprocal of the filter capacitor parameter, and the component of the total disturbance related to the grid-side current corresponds to the negative of the ratio of the grid-side current to the filter capacitor parameter. Dividing these two components cancels out the filter capacitor parameters, thus directly obtaining the grid-side current. By determining the grid-side current based on the total disturbance and input gain, the acquisition of the grid-side current no longer depends on the nominal parameters of the filter capacitor or requires an additional grid-side current sensor. The grid-side current information can be extracted from the lumped disturbance simply through algebraic operations on the total disturbance and input gain, solving the problem of real-time acquisition when the grid-side current cannot be directly measured or the measurement cost is too high.

[0060] In some embodiments, the grid-side current at the current sampling time can be determined based on the comparison between the total disturbance at the current sampling time and the input gain.

[0061] To achieve real-time, step-by-step output of the grid-side current during the discrete control cycle, the grid-side current at the current sampling moment is determined based on the comparison between the total disturbance and the input gain at the current sampling moment. This comparison result represents the numerical relationship between the total disturbance and the input gain at the current sampling moment. In one embodiment, this comparison result is the quotient of the total disturbance divided by the input gain. In another embodiment, the comparison result is the product of the negative of the total disturbance and the reciprocal of the input gain. The input gain is adaptively updated according to the time-varying characteristics of the filter capacitor parameters, and the total disturbance is output in real-time by the extended state observer at the current sampling moment; both are instantaneous values ​​at the current sampling moment. In one embodiment, the comparison result is low-pass filtered before output to remove high-frequency measurement noise. In another embodiment, the comparison result is amplitude-limited before output to prevent the current observation value from exceeding a reasonable range due to numerical anomalies. By determining the grid-side current based on the total disturbance and the input gain at the current sampling moment, the observed grid-side current value can be updated in real-time within each sampling cycle following changes in the system state, providing step-by-step current feedback signals for the grid-connected control system.

[0062] In some embodiments, the step S201 above, which determines the physical relationship between the filter capacitor voltage change rate and the converter-side current, grid-side current, and filter capacitor based on the LCL filter topology, may further include: The physical relationship between the rate of change of the filter capacitor voltage and the converter-side current, the grid-side current, and the filter capacitor is determined using the following formula: ; In the formula, This is the voltage across the filter capacitor. The rate of change of the filter capacitor voltage. For converter-side current, For filtering capacitors, This represents the grid-side current.

[0063] The above formula characterizes the Kirchhoff's current law relationship for the current at the capacitor node of an LCL filter. The left side of the formula... The current flowing through the filter capacitor branch is represented by the product of the rate of change of the filter capacitor voltage and the capacitance value of the filter capacitor. This product is used to characterize the instantaneous value of the current in the capacitor branch, and its sign depends on the direction of change of the filter capacitor voltage. (The right side of the formula...) This formula represents the difference between the converter-side current and the grid-side current. The converter-side current is the current flowing through the converter-side inductor of the LCL filter, and the grid-side current is the current injected into the grid through the grid-side inductor of the LCL filter. Overall, this formula indicates that at the capacitor node of the LCL filter, the current flowing through the filter capacitor branch is equal to the difference between the converter-side current and the grid-side current; that is, the algebraic sum of the currents at the capacitor node is zero. Based on this formula, the portion of the converter-side current that does not flow into the capacitor branch is the current injected into the grid, thus establishing a correlation between the rate of change of the filter capacitor voltage and the grid-side current. In this formula, the direction of the grid-side current is positive when it flows towards the grid; a positive grid-side current indicates that power is being fed into the grid in the forward direction, while a negative value indicates that power is flowing in the reverse direction.

[0064] In some embodiments, the hyperlocal model constructed in step S201 above, based on physical relationships, with the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain to be solved and the total disturbance to be solved, may further include: Based on physical relationships, a hyperlocal model is constructed using the following formula, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain and the total disturbance to be solved: ; In the formula, This is the voltage across the filter capacitor. The rate of change of the filter capacitor voltage. For converter-side current, The input gain to be solved, The total disturbance to be solved is denoted as .

[0065] The above formula represents the basic mathematical form of the hyperlocal model. The left side of the formula... The rate of change of the filter capacitor voltage, i.e., the derivative of the filter capacitor voltage with respect to time, represents how quickly the capacitor voltage changes per unit time, and is used as the output of the model. The first term on the right-hand side of the formula... The measurable drive contribution characterizing the rate of change of the filter capacitor voltage to the converter-side current, where The input gain to be solved is used to characterize the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current; this product term represents the part of the model that can be explicitly explained by the input quantity. The second term on the right-hand side of the formula... The total disturbance to be solved is represented by the combined effect of all unknown factors influencing the rate of change of the filter capacitor voltage, excluding the contribution of the converter-side current. These unknown factors include at least grid-side current information, as well as unmodeled dynamics, external disturbances, and parameter perturbations. This formula comprehensively characterizes the mathematical relationship between the rate of change of the filter capacitor voltage, which is composed of a measurable driving term and an unknown combined disturbance term. In other words, the hyperlocal model incorporates all parts of the original LCL filter's physical model that are difficult to model precisely into the total disturbance term, while retaining only the product of the converter-side current and the input gain as the measurable driving term. Based on this formula, the grid-side current term in the physical relationship no longer appears explicitly but is implicitly included in the total disturbance. Therefore, the problem of determining the grid-side current can be transformed into a joint problem of solving for the input gain and the total disturbance. In this formula, both the input gain and the total disturbance are unknown variables to be solved, meaning both must be obtained through online calculation or observation, rather than relying on offline parameter identification or empirical settings. This allows the model to adapt to changes in system parameters in real time.

[0066] In some embodiments, the discretization of the hyperlocal model in step S202 above, to establish discretized equations for the current sampling time and the previous sampling time respectively, may further include: The hyperlocal model is discretized using the following formula to establish the discretized equations for the current sampling time and the previous sampling time, respectively: ; In the formula, , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time The converter-side current and total disturbance, and Sampling time The converter-side current and total disturbance, This is the input gain.

[0067] This discretization process employs a first-order forward Euler method, transforming the continuous-domain differentiation operations in the hyperlocal model into discrete-domain difference operations. In the first formula, Characterizing the voltage of the filter capacitor at the sampling time The first-order forward difference, which is the voltage difference between the current sampling time and the previous sampling time divided by the sampling period, is used to approximate the rate of change of voltage in the continuous domain. This difference value is derived from the converter-side current at the current sampling moment. Input gain and the total perturbation at the current sampling time Jointly decided. In the second formula, Characterizing the voltage of the filter capacitor at the sampling time The first-order forward differential, calculated by dividing the voltage difference between the previous sampling time and the sampling period, is used to approximate the rate of change of the filter capacitor voltage at the previous sampling time. This differential value is derived from the converter-side current at the previous sampling time. Input gain and the total perturbation at the previous sampling time The two formulas collectively characterize the discretized form of the hyperlocal model at two adjacent sampling times, transforming the continuous differential equation into an algebraic equation between adjacent sampling points. This facilitates real-time computation by the digital processor using sampled values ​​within each sampling period. Based on this discretization process, it is possible to perform calculations with the same input gain. The following establishes coupled equations for two sampling times, providing an algebraic basis for subsequent simultaneous elimination to solve for the input gain. Since the change in input gain is caused only by the drift of the capacitor parameter, and its time constant is much larger than the sampling period, the input gain in the two discretized equations is the same. In both formulas, the total perturbation is in discrete time series form, that is, the total perturbation at each sampling time is independently labeled to characterize its possible step-by-step changes over time.

[0068] In some embodiments, step S202 above, which calculates the input gain at the current sampling time based on the sampled values ​​of the filter capacitor voltage and the converter-side current at the current sampling time and the sampled values ​​of the current at the previous sampling time, may further include: Based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling times, the input gain at the current sampling time is calculated using the following formula: ; In the formula, The input gain at the current sampling time. , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time and The converter-side current.

[0069] The above formula is obtained by solving the simultaneous elimination problem of the discretized equations at two adjacent sampling times. (Molecular part) The second-order voltage difference characterizes the change in the first-order difference of the filter capacitor voltage within two adjacent sampling intervals. This second-order difference characterizes the rate of change of the filter capacitor voltage across each sampling interval. This characterizes the change in voltage during the previous sampling period. The denominator represents the change in voltage within the previous sampling period, and the difference between the two represents the difference in the rate of voltage change between two adjacent sampling periods. Characterized by the product of the sampling period and the first-order difference of the converter-side current, where This formula characterizes the change in converter-side current between two adjacent sampling times. This change, multiplied by the sampling period, is used to match the dimensions of the second-order voltage difference in the numerator. Overall, the formula represents the ratio of the second-order difference of the filter capacitor voltage to the first-order difference of the converter-side current (multiplied by the sampling period). This formula utilizes the constraint that the total disturbance is approximately equal between adjacent sampling times. By eliminating the total disturbance through division, an explicit expression for the input gain that depends only on the sampled value and the sampling period is obtained. Based on this formula, the input gain value automatically adjusts according to the time-varying characteristics of the filter capacitor parameters, without requiring offline calibration of the nominal capacitor value or relying on any prior capacitor parameters. In one embodiment, the calculated input gain is a complex number, and the real part of this complex number is taken as the input gain value at the current sampling time. In another embodiment, the calculated input gain is a complex number, and the magnitude of this complex number is taken as the input gain value at the current sampling time. In another implementation, after calculating the input gain, a low-pass filter is applied to suppress sampling noise from contaminating the gain value. The filtered gain value is then used in the hyperlocal model at the current sampling time. In this formula, the difference between the converter-side current at adjacent sampling times is not zero; that is, the input gain is calculated only when the current dynamically changes, thus avoiding division by zero anomalies. Based on this formula, the input gain can be updated in real time during each sampling period, ensuring that the hyperlocal model always uses the accurate response characteristics of the current capacitor parameters as the basis for subsequent total disturbance observations and grid-side current determination.

[0070] In some embodiments, step S203 above, which aims to make the observation error approach zero, iteratively estimates the total perturbation in the hyperlocal model using an extended state observer and outputs the total perturbation at the current sampling time, may further include: With the goal of minimizing the observation error, the total perturbation in the hyperlocal model is iteratively estimated using the following formula, outputting the total perturbation at the current sampling time: ; In the formula, For observation error, and These are the filter capacitor voltages at the sampling time. The observed values ​​and sampled values, The voltage of the filter capacitor at the sampling time The observed values, The sampling period is Sampling time The converter-side current, For input gain, and Sampling time and The total disturbance, and These are the first and second gain coefficients of the extended state observer, respectively. To expand the bandwidth of the state observer.

[0071] First formula This characterizes the definition of observation error, namely the deviation between the observed value of the filter capacitor voltage at the current sampling time and its actual sampled value. This deviation is used to quantify the degree of deviation of the hyperlocal model's prediction of the capacitor voltage, serving as a feedback signal to drive the extended state observer to perform correction. When the observed value is greater than the sampled value, A positive value indicates that the model predicts a higher capacitor voltage; when the observed value is lower than the sampled value, A negative value indicates that the model predicts a lower capacitor voltage. The absolute value of the value reflects the severity of the prediction bias.

[0072] Second formula This characterizes the recursive update relationship of the observed filter capacitor voltage from the current sampling time to the next sampling time. Among them, The characterization is based on the prediction of the change in capacitor voltage over a future sampling period using a hyperlocal model. Specifically, the rate of change of voltage at the current moment is multiplied by the sampling period to calculate the preliminary predicted value of the capacitor voltage at the next moment. Based on this prediction, This formula uses the current observation error to correct the voltage prediction value for the next time step. Specifically, it adds the feedback of the observation error to the voltage prediction channel through a first gain coefficient. The first gain coefficient characterizes the influence of the observation error on the voltage estimate update step size, and its value is related to the observer bandwidth and sampling period. A larger first gain coefficient results in a more sensitive response to the observation error and a faster tracking speed, but also a stronger amplification effect on measurement noise. A smaller first gain coefficient results in a smoother response to the observation error and better noise immunity, but a correspondingly slower tracking speed. This formula achieves the recursive progression of the capacitor voltage observation value from the current sampling time to the next sampling time through "model prediction and error feedback correction," ensuring that the voltage observation value converges to the actual sampled value at each step.

[0073] The third formula This characterizes the recursive update relationship of the total perturbation from the current sampling time to the next sampling time. Among them, This is the estimated total disturbance value at the current sampling time, which serves as the base value for the next recursive step. This formula uses the current observation error to correct the total disturbance estimate for the next time step. Specifically, it uses a second gain coefficient to superimpose the feedback of the observation error onto the total disturbance update channel. The second gain coefficient characterizes the influence of the observation error on the update step size of the total disturbance estimate, and its value is related to the observer bandwidth and sampling period. A larger second gain coefficient results in a more sensitive response to the observation error and stronger suppression of persistent disturbances, but also increases sensitivity to measurement noise. A smaller second gain coefficient results in a smoother change in the total disturbance estimate and better noise immunity, but reduces the tracking accuracy for low-frequency disturbances. This formula recursively advances the total disturbance from the current sampling time to the next sampling time through "current disturbance estimation and error feedback correction," ensuring that the total disturbance estimate converges to the true physical value at each time step.

[0074] The fourth formula And the fifth formula The quantitative configuration relationship between the first gain coefficient and the second gain coefficient and the bandwidth and sampling period of the extended state observer is characterized respectively. and They are linearly proportional. and The two components exhibit a quadratic relationship, jointly ensuring that the observer's convergence dynamics meet the preset bandwidth requirements. In one implementation, the bandwidth is set to one-tenth of the sampling frequency to achieve a balance between response speed and noise suppression. In another implementation, the bandwidth is adaptively adjusted according to the grid frequency, meaning that the bandwidth is adjusted accordingly when the grid frequency fluctuates to ensure observation accuracy. In yet another implementation, the bandwidth is dynamically adjusted according to the harmonic content of the converter-side current, meaning that the bandwidth is reduced when the harmonic content is high to enhance the filtering effect, and increased when the harmonic content is low to accelerate the response speed.

[0075] The above formula constitutes the complete recursive estimation process of the extended state observer from the current sampling time to the next sampling time: First, the observation error at the current time is calculated. Then, the observation error is fed back to the voltage observation channel and the total disturbance observation channel using the first gain coefficient and the second gain coefficient, respectively, to synchronously update the voltage observation value and the total disturbance estimate. This recursive process is executed once in each sampling period. After several iterations, the observation error tends to zero, and the output total disturbance estimate is the total disturbance at the current sampling time. Based on this recursive estimation mechanism, the total disturbance can be corrected in a closed loop at each sampling time according to the latest voltage observation error, realizing real-time tracking and step-by-step updating of the total disturbance, and solving the problem of accumulated bias caused by the lack of error feedback in the open-loop observation method. In this recursive estimation mechanism, the voltage observation value and the total disturbance estimate are updated simultaneously in the same sampling period to ensure the timing matching and physical logic consistency between the two.

[0076] In some embodiments, determining the grid-side current at the current sampling time based on the comparison result of the total disturbance and the input gain at the current sampling time in step S204 may further include: Based on the comparison between the total disturbance and the input gain at the current sampling time, the grid-side current at the current sampling time is determined using the following formula: ; In the formula, , and These represent the grid-side current, total disturbance, and input gain at the sampling time, respectively.

[0077] The above formula represents the quantitative conversion relationship between grid-side current, total disturbance, and input gain. The right side of the formula... The ratio representing the total disturbance to the input gain is inverted, where This represents the total disturbance estimate output by the extended state observer at the current sampling time. The current input gain calculated in step S202. (Molecular) It implicitly contains the grid-side current component (i.e. ), denominator Corresponding to the reciprocal of the filter capacitor parameter (i.e. After dividing the two, the filter capacitor parameters cancel each other out. Then, by taking the negative sign, the grid-side current can be directly obtained. This formula utilizes the equivalent correspondence between the total disturbance and the physical model in the hyperlocal model, extracting the grid-side current information implicit in the total disturbance through algebraic operations. Because... The formula adaptively adjusts to changes in the filter capacitor parameters in real time, thus the grid-side current determined by this formula does not depend on the nominal capacitor value and is not affected by capacitor parameter drift. In one implementation, the calculated grid-side current is low-pass filtered before output to remove observation noise. In another implementation, the calculated grid-side current is limited before output to prevent current observations from exceeding reasonable ranges due to numerical anomalies. In yet another implementation, when the absolute value of the input gain is less than a preset threshold, the grid-side current value at the previous sampling time remains unchanged to prevent numerical overflow due to an excessively small denominator. Based on this formula, the current grid-side current observation value can be directly calculated from the total disturbance and input gain in each sampling period, providing a real-time current feedback signal for the grid-connected control system. This eliminates the need for additional grid-side current sensors and overcomes the shortcomings of traditional methods where observation bias accumulates due to capacitor parameter mismatch.

[0078] In some embodiments, the above-described discretization of the hyperlocal model may further include: using the trapezoidal method or the Runge-Kutta method to discretize the hyperlocal model in order to establish discretized equations for the current sampling time and the previous sampling time.

[0079] To overcome the impact of truncation error introduced by the first-order forward Euler discretization method on the accuracy of input gain calculation when the sampling period is large, the above-mentioned discretization of the hyperlocal model also includes using the trapezoidal method or the Runge-Kutta method to discretize the hyperlocal model, in order to establish the discretized equations for the current sampling time and the previous sampling time. The trapezoidal method approximates the average slope of the continuous domain differential equation at the two endpoints of the sampling interval as the average slope within that interval, giving the discretized difference equation second-order accuracy. The Runge-Kutta method further improves the discretization accuracy by weighted averaging of the slopes at multiple points within the sampling interval. In one embodiment, the trapezoidal method is used to discretize the hyperlocal model, and the discretized equation for the current sampling time is: The discretization equation for the previous sampling time is then established. In another implementation, a third-order Runge-Kutta method is used to discretize the hyperlocal model, approximating the voltage change rate through a weighted combination of three slope estimates. After discretization, the original first-order forward Euler discretization equation is replaced with the aforementioned discretization equation, while the subsequent simultaneous elimination and gain calculation steps remain unchanged. By employing a higher-order discretization method, the accuracy of the input gain calculation results can be maintained even when the sampling period is relatively large or the system dynamics change rapidly.

[0080] In some embodiments, before discretizing the hyperlocal model as described above, the method may further include: obtaining the rate of change of the converter-side current or the filter capacitor voltage; shortening the sampling interval when the rate of change exceeds a preset threshold; and extending the sampling interval when the rate of change is lower than the preset threshold; and discretizing the hyperlocal model using the adjusted sampling interval.

[0081] To address the issues of unnecessary computational resource waste caused by a fixed sampling period during steady-state system operation, and delayed gain updates due to insufficient sampling points during dynamic system changes, the aforementioned method includes acquiring the rate of change of the converter-side current or filter capacitor voltage before discretizing the hyperlocal model. When the rate of change exceeds a preset threshold, the sampling interval is shortened; when the rate of change is below the preset threshold, the sampling interval is extended. The adjusted sampling interval is then used to discretize the hyperlocal model. The sampling interval characterizes the time interval between two adjacent sampling operations, and its value determines the time step parameter in the discretization equation. In one embodiment, the absolute value of the filter capacitor voltage change rate is used as the criterion. When this absolute value is greater than a first threshold, the sampling interval is shortened to half of the original interval; when this absolute value is less than a second threshold, the sampling interval is extended to twice the original interval. In another embodiment, the absolute value of the converter-side current change rate is used as the criterion. When this absolute value exceeds a preset threshold, the sampling interval is shortened; when this absolute value is below the preset threshold, the sampling interval is extended. After the sampling interval is adjusted, the sampling period parameter in the discretization equation is updated accordingly to the adjusted value. By introducing a dynamic sampling interval adjustment mechanism, the input gain calculation can adaptively adjust the sampling density according to the dynamic characteristics of the system, thereby optimizing the computational efficiency of the algorithm while ensuring the accuracy of gain calculation.

[0082] In some embodiments, the constraint that the total disturbance at adjacent sampling times is approximately equal may further include: obtaining the rate of change of the total disturbance at adjacent sampling times, constraining the rate of change to be less than a preset threshold, and simultaneously solving the discretization equations of the current sampling time and the previous sampling time; after simultaneously eliminating variables, introducing a correction term based on the rate of change of the total disturbance to compensate for the elimination result.

[0083] To overcome the model mismatch error introduced by the strict assumption that the total disturbance is completely equal at adjacent sampling times when the total disturbance changes rapidly, the above-mentioned method, constrained by the approximate equality of the total disturbance at adjacent sampling times, also includes obtaining the rate of change of the total disturbance at adjacent sampling times, constrained by the rate of change being less than a preset threshold, simultaneously solving the discretized equations of the current sampling time and the previous sampling time, and after simultaneous elimination, introducing a correction term based on the rate of change of the total disturbance to compensate for the elimination result. The rate of change of the total disturbance is used to characterize how fast the total disturbance changes per unit time, and its value reflects the degree to which the total disturbance deviates from the approximate equality assumption. In one embodiment, the rate of change of the total disturbance is approximately estimated by the total disturbance estimate output by the extended state observer at the previous time step. The input gain expression obtained after simultaneous elimination is: ;in This is a correction term based on the estimated total disturbance rate of change. In another implementation, a total disturbance rate of change less than a preset threshold is used as a prerequisite for simultaneous elimination. Elimination is performed only when this condition is met; otherwise, the input gain value from the previous sampling time is directly used. By introducing a constraint on the total disturbance rate of change and adding correction compensation, this method can maintain the accuracy of input gain calculation even under conditions where the total disturbance changes rapidly.

[0084] In some embodiments, the above-mentioned constraint that the total disturbance at adjacent sampling times is approximately equal, and the simultaneous discretization equations of the current sampling time and the previous sampling time, may further include: obtaining discretization equations including the current sampling time and the previous N sampling times, where N≥3; establishing a joint estimation equation including the input gain and the parameters of the total disturbance variation model, with the constraint that the total disturbance changes linearly or quadratically with time; and solving the joint estimation equations using the least squares method to obtain the input gain.

[0085] To address the issue of measurement noise significantly impacting gain calculations when solving equations using only the current and previous two sampling times, the above method, constrained by the approximate equality of total disturbance at adjacent sampling times, combines the discretized equations of the current and previous sampling times. It further includes obtaining discretized equations encompassing the current sampling time and the previous N sampling times, where N≥3. Constrained by the linear or quadratic change of the total disturbance over time, a joint estimation equation is established, incorporating the input gain and parameters of the total disturbance variation model. This joint estimation equation is solved using the least squares method to obtain the input gain. A linear change in total disturbance over time characterizes a monotonically changing total disturbance at a constant rate across multiple sampling times, while a quadratic change in total disturbance over time characterizes a changing total disturbance with a constant acceleration across multiple sampling times. In one implementation, it is assumed that the total disturbance changes linearly over time, i.e. ,in Let be the rate of change of the total disturbance. In this case, solving N equations simultaneously can estimate both the input gain and the rate of change of the total disturbance. In another implementation, it is assumed that the total disturbance changes quadratically with time, i.e. ,in and Let N be the first and second rates of change of the total disturbance, respectively. By simultaneously solving N equations, the input gain and the two rate of change parameters can be estimated at the same time. The larger the value of N, the more equations are involved in the joint estimation, and the stronger the suppression of measurement noise by the least squares estimation. By utilizing multiple historical sampling points for joint estimation, the interference of single-point measurement noise on the input gain calculation results is effectively suppressed, improving the numerical stability and noise resistance of the gain calculation in noisy environments.

[0086] In some embodiments, after calculating the input gain at the current sampling time based on the simultaneous solution results, the method may further include: filtering the calculated input gain using a Kalman filter, and outputting the filtered input gain.

[0087] To address the issues of phase lag introduced by low-pass filtering leading to gain update delay, and the inability of fixed filter coefficients to simultaneously achieve noise suppression and response speed, the method, after calculating the input gain at the current sampling time based on the simultaneous solution results, further includes using a Kalman filter to filter the calculated input gain and outputting the filtered input gain. The Kalman filter characterizes a class of optimal recursive filters based on state-space models. Through a joint prediction-correction mechanism of the system state transition model and the observation model, it outputs the optimal state estimate in the sense of minimum mean square error, given the statistical characteristics of system noise and observation noise. In one implementation, the input gain is used as the state variable, and its state transition model is... ,in For process noise, the observation model is: ,in To obtain the initial gain value from the simultaneous solution, To observe noise, the optimal estimate of the input gain at the current moment is obtained recursively through Kalman filtering. In another implementation, the state vector is constructed using the input gain and its rate of change, and a second-order Kalman filter is used to track the gain. By using a Kalman filter instead of a conventional low-pass filter, the phase lag introduced by the low-pass filter is avoided while ensuring noise suppression capability. This allows the input gain to respond quickly when the capacitance parameter changes abruptly, while maintaining a high-precision smooth estimate in steady state.

[0088] In some embodiments, the above-described filtering of the calculated input gain using a Kalman filter may further include: obtaining the innovation sequence of the Kalman filter, and adjusting the process noise covariance and observation noise covariance of the Kalman filter according to the statistical characteristics of the innovation sequence.

[0089] To address the issue of Kalman filter performance degradation or even divergence when noise statistics are inaccurate, the aforementioned method uses a Kalman filter to filter the calculated input gain. This also includes acquiring the Kalman filter's innovation sequence and adjusting the process noise covariance and observation noise covariance based on the statistical characteristics of the innovation sequence. The innovation sequence characterizes the difference sequence between the actual observed and predicted values ​​in each step of the Kalman filter's recursion. Its statistical characteristics (including but not limited to mean and covariance) reflect the degree of matching between the current filter noise model and the actual noise characteristics. When the covariance of the innovation sequence is consistently greater than the theoretical predicted value, it indicates that the actual intensity of the observed or process noise is higher than a preset value. In this case, the corresponding noise covariance matrix needs to be increased to avoid the filter over-relying on the model prediction. When the covariance of the innovation sequence is consistently less than the theoretical predicted value, it indicates that the actual noise intensity is lower than a preset value. In this case, the corresponding noise covariance matrix needs to be decreased to improve the filter's sensitivity to model changes. In one implementation, the amplification factor of the process noise covariance matrix is ​​dynamically adjusted based on the ratio of the actual covariance to the theoretical covariance of the innovation sequence. In another implementation, the observation noise covariance matrix is ​​dynamically adjusted based on the degree to which the mean of the innovation sequence deviates from zero. By introducing an adaptive covariance adjustment mechanism based on the innovation sequence, the Kalman filter can maintain optimal filtering performance even when the noise statistics are unknown or time-varying, thus improving the adaptability and robustness of the input gain filtering process.

[0090] In some embodiments, before calculating the input gain at the current sampling time based on the simultaneous solution results, the method may further include: obtaining the difference between the converter-side current at the current sampling time and the previous sampling time; and keeping the input gain at the previous sampling time unchanged when the absolute value of the difference is less than a preset threshold.

[0091] In some embodiments, to address the numerical anomaly problem caused by the denominator approaching zero and resulting in drastic fluctuations in the calculated input gain when the change in converter-side current between adjacent sampling times is too small, the method further includes obtaining the difference between the converter-side current at the current sampling time and the previous sampling time before calculating the input gain based on the simultaneous solution results. When the absolute value of this difference is less than a preset threshold, the input gain at the previous sampling time remains unchanged. The converter-side current difference characterizes the change in current sampling values ​​between the current sampling time and the previous sampling time, and this change is directly related to the denominator term in the input gain calculation formula. When the absolute value of this difference is too small, the absolute value of the denominator approaches zero, and even small sampling noise can cause drastic fluctuations in the calculated gain. The preset threshold is determined based on the sampling accuracy and the quantization error of the current sensor. In one embodiment, the threshold is taken as 0.5% to 1% of the rated value of the current sensor. In another embodiment, the threshold is adaptively set based on the fundamental amplitude of the converter-side current; that is, the threshold increases accordingly when the fundamental amplitude is larger. When the absolute value of the current difference is less than a preset threshold, simultaneous elimination calculation is not performed at the current sampling time, and the output gain remains at the value of the previous sampling time. By introducing a calculation condition monitoring and hold strategy, the numerical anomaly problem of drastic fluctuations in the calculated gain value due to excessively small current change rate is effectively avoided, thus improving the numerical robustness and reliability of the gain calculation process.

[0092] In some embodiments, when the absolute value of the difference is less than a preset threshold, the method may further include: obtaining multiple input gain calculation values ​​for the previous M sampling times, performing a weighted average on the multiple input gain calculation values, and using the weighted average result as the input gain for the current sampling time.

[0093] To provide a reasonable gain update value even when the current change rate is too small to effectively calculate the input gain through simultaneous elimination, the method further includes acquiring multiple input gain calculation values ​​from the previous M sampling times when the absolute value of the difference is less than a preset threshold. These multiple input gain calculation values ​​are then weighted and averaged, and the weighted average result is used as the input gain at the current sampling time. M is a preset positive integer used to determine the number of historical gain values ​​participating in the weighted average; in one embodiment, M is between 5 and 10. In the weighted average, gain calculation values ​​closer to the current time are assigned a larger weight, and gain calculation values ​​farther from the current time are assigned a smaller weight. The weighting coefficients are set according to an exponential decay law or a linear decay law. In one embodiment, the weighting coefficients follow an exponential decay law. Settings, in which As the attenuation factor, Where m is the sampling time number, and the sampling time closer to the current time corresponds to a larger weight. In another implementation, the weight coefficients follow a linear decay law. The weights are set such that the sum of all weights is 1. The calculated input gain values ​​are weighted and summed according to the above weights to obtain a weighted average. If there are fewer than M historical gain values, the average of all existing gain values ​​is taken. By using the weighted average of historical gain values ​​as a substitute output when the current change rate is too small, the input gain can still be smoothly updated according to historical trends even under poor computational conditions. This avoids model mismatch caused by prolonged stagnation of gain, while ensuring the continuity and smoothness of the output gain.

[0094] In some embodiments, the above-mentioned input gain and converter-side current input to the extended state observer may further include: using a nonlinear extended state observer as the extended state observer, wherein the error feedback term of the nonlinear extended state observer is a nonlinear function.

[0095] In some embodiments, to address the inherent contradiction between fast response and noise suppression in linear extended state observers, the input gain and converter-side current are input to the extended state observer. Furthermore, a nonlinear extended state observer is also used, with its error feedback term being a nonlinear function. The error feedback term of the linear extended state observer is a linear function of the observation error, and its gain coefficient is constant. When the gain is large, the observer tracks quickly but is sensitive to noise; when the gain is small, it has strong noise immunity but slow tracking speed. The nonlinear extended state observer transforms the observation error by introducing a nonlinear function before feeding it back to the state update channel. In one embodiment, aiming to make the observation error approach zero, the total disturbance in the hyperlocal model is iteratively estimated using the following formula, outputting the total disturbance at the current sampling time: , among which, Linear interval width threshold, The function exhibits higher gain when the observation error is large and lower gain when the observation error is small. In another implementation, the nonlinear function is an arctangent function, whose saturation characteristic causes the feedback quantity to tend to saturate when the error is large and to be approximately linear when the error is small. By employing a nonlinear extended state observer, the observer can respond quickly when the total disturbance changes abruptly, while maintaining low noise sensitivity during steady-state operation, thus resolving the inherent contradiction between fast response and noise suppression in linear observers.

[0096] In some embodiments, before inputting the input gain and converter-side current into the extended state observer, the method may further include: performing harmonic suppression preprocessing on the converter-side current, and inputting the preprocessed converter-side current into the extended state observer.

[0097] To address the issue of decreased observation accuracy caused by high-frequency switching ripple in the converter-side current coupling to the total disturbance estimation result through the extended state observer, a harmonic suppression preprocessing is performed on the converter-side current before inputting the input gain and converter-side current into the extended state observer. The preprocessed converter-side current is then input into the extended state observer. Harmonic suppression preprocessing characterizes the attenuation of high-frequency components near the switching frequency and its multiples in the converter-side current sampling signal. In one embodiment, a low-pass filter is used, with its cutoff frequency set to 1 / 10 to 1 / 5 of the switching frequency, to retain the fundamental frequency and lower harmonics while attenuating higher harmonics. In another embodiment, a band-stop filter is used, with its notch frequency set to the switching frequency and its major harmonics, to accurately attenuate switching ripple at specific frequencies. By preprocessing the input current, the input signal to the extended state observer is smoother, reducing the contamination of the total disturbance estimation result by high-frequency harmonics and improving the signal-to-noise ratio of the total disturbance observation.

[0098] In some embodiments, the above-mentioned harmonic suppression preprocessing of the converter-side current may further include: using a second-order generalized integrator to filter the converter-side current, extracting the fundamental component of the converter-side current, and using the fundamental component as the preprocessed converter-side current.

[0099] In some embodiments, to further improve the selectivity of harmonic suppression preprocessing and avoid fundamental amplitude attenuation while filtering out harmonics, harmonic suppression preprocessing of the converter-side current further includes using a second-order generalized integrator to filter the converter-side current, extracting the fundamental component of the converter-side current, and using this fundamental component as the preprocessed converter-side current input extended state observer. The second-order generalized integrator is used to characterize a class of resonant filters with frequency adaptive characteristics, which can achieve zero phase shift and unity gain at a specified frequency, while attenuating signals deviating from that frequency. In one embodiment, the center frequency of the second-order generalized integrator is set to the grid fundamental frequency, and its output signal is in phase with the input signal and has the same amplitude, thus achieving the extraction of the fundamental component. In another embodiment, two second-order generalized integrators are used to extract the fundamental components of the converter-side current on the α-axis and β-axis respectively, suitable for current signals in a two-phase stationary coordinate system. By employing a second-order generalized integrator to extract the fundamental component, high-order harmonics are effectively filtered out while ensuring that the fundamental component is input to the extended state observer without amplitude attenuation or phase shift, thus avoiding any impact on the overall disturbance observation accuracy.

[0100] In some embodiments, the above-mentioned method of obtaining the deviation between the observed value and the sampled value of the filter capacitor voltage at the current sampling time may further include: obtaining the observation error at the previous sampling time, predictively compensating the observation error at the current sampling time based on the difference between the observation error at the current sampling time and the observation error at the previous sampling time, and using the compensated deviation as the observation error of the filter capacitor voltage.

[0101] To improve the predictability of the extended state observer to system dynamic changes and shorten the convergence time of observation errors, the method obtains the deviation between the observed value and the sampled value of the filter capacitor voltage at the current sampling moment. It also includes obtaining the observation error at the previous sampling moment, and predictively compensating for the observation error at the current sampling moment based on the difference between the observation error at the current sampling moment and the observation error at the previous sampling moment. The compensated deviation is then used as the observation error of the filter capacitor voltage. This compensation method utilizes the trend information of the observation error. When the observation error shows a continuous increasing or decreasing trend, the compensation term allows the extended state observer to perform correction actions in advance before the error further expands. In one embodiment, the predictive compensation method is based on the cumulative change amount at the previous sampling moment. In another embodiment, the predictive compensation method is based on the cumulative change amount at the previous L sampling moments. By introducing a trend prediction mechanism into the observation error, the extended state observer has the predictability of the error evolution direction, improving the dynamic tracking performance of the observer in transient processes and shortening the number of iterations required for observation error convergence.

[0102] In some embodiments, the above-mentioned method of obtaining the deviation between the observed value and the sampled value of the filter capacitor voltage at the current sampling time may further include: obtaining multiple observation errors at the current sampling time and the previous N sampling times, performing a weighted combination of the multiple observation errors, and using the weighted combination result as the observation error of the filter capacitor voltage.

[0103] To suppress the instantaneous impact of sampling noise on the correction of the extended state observer, the deviation between the observed value and the sampled value of the filter capacitor voltage at the current sampling moment is obtained. This also includes acquiring multiple observation errors at the current sampling moment and the previous N sampling moments, weighting and combining these multiple observation errors, and using the weighted combination result as the observation error of the filter capacitor voltage. The weighted combination is used to characterize the summarization of observation errors at multiple sampling moments according to different weights, thereby reducing the impact of single-point measurement noise on the observation error. In one embodiment, the weighting coefficients are set according to the time decay law, that is, the closer the observation error is to the current moment, the greater the weight is assigned, and the farther the observation error is to the current moment, the smaller the weight is assigned, and the weighting coefficients adopt an exponential decay form. In another embodiment, the weighting coefficients are adaptively set according to the system dynamic characteristics at each sampling moment, that is, the observation error during periods of drastic system dynamic changes is assigned a smaller weight, and the observation error during periods of stable system dynamics is assigned a larger weight. By using weighted combination errors to replace single-point errors, the data redundancy of multiple sampling points effectively suppresses the interference of single-point measurement noise on the observer update, improving the stability and noise resistance of the observer in a strong noise environment.

[0104] In some embodiments, the iterative estimation of the total disturbance in the hyperlocal model using the extended state observer may further include: adjusting the bandwidth of the extended state observer according to the observation error of the filter capacitor voltage, wherein the bandwidth is positively correlated with the magnitude of the observation error.

[0105] To address the issue of fixed-bandwidth observers exhibiting insufficient response speed during transient processes and amplifying noise during steady-state processes, an extended state observer is used to iteratively estimate the total disturbance in the hyperlocal model. This includes adjusting the bandwidth of the extended state observer based on the observation error of the filter capacitor voltage. The bandwidth is positively correlated with the magnitude of the observation error; that is, the larger the magnitude of the observation error, the larger the bandwidth, and vice versa. This bandwidth adjustment mechanism uses the observation error magnitude as a real-time indicator of the system's dynamic characteristics. When the observation error is large, it indicates that the system is in a transient process or has undergone a sudden change, requiring an increase in bandwidth to improve tracking speed. When the observation error is small, it indicates that the system is approaching steady state, requiring a decrease in bandwidth to suppress measurement noise. In one implementation, the relationship between bandwidth and the observation error magnitude is linear. In another implementation, the relationship is a saturated nonlinear relationship. By adaptively adjusting the bandwidth according to the observation error magnitude, the extended state observer automatically optimizes its response characteristics under different operating conditions, balancing the dual requirements of rapid transient tracking and steady-state noise suppression.

[0106] In some embodiments, the above-mentioned iterative estimation of the total disturbance in the hyperlocal model by using an extended state observer may further include: using a cascaded extended state observer to iteratively estimate the total disturbance, wherein a first-level observer is used to observe the low-frequency component of the total disturbance, and a second-level observer uses the output residual of the first-level observer as input to observe the high-frequency component of the total disturbance, and the low-frequency component and the high-frequency component are superimposed to obtain the total disturbance.

[0107] To address the challenge of a single extended state observer simultaneously ensuring accurate steady-state estimation of low-frequency dominant disturbances and rapid tracking of high-frequency residual disturbances across the entire frequency band, this paper proposes an iterative estimation method for the total disturbance in the hyperlocal model using extended state observers. This method also includes employing cascaded extended state observers for iterative estimation. The first-stage observer observes the low-frequency components of the total disturbance, while the second-stage observer, using the output residual of the first-stage observer as input, observes the high-frequency components. The low-frequency and high-frequency components are then superimposed to form the total disturbance. The cascaded extended state observer characterizes multiple extended state observers connected in series, with the output residual of the previous stage serving as the input of the next stage. Each stage of the observer processes disturbance signals with different frequency components. In one implementation, the first-stage observer uses a narrow bandwidth (1 / 20 to 1 / 10 of the sampling frequency) to output the low-frequency dominant estimate of the total disturbance. The second-stage observer takes the observation residual of the first-stage observer as input and uses a wider bandwidth (1 / 5 to 1 / 2 of the sampling frequency) to output the high-frequency residual estimate of the total disturbance. The outputs of the two stages are superimposed to obtain the total disturbance. In another implementation, a three-stage cascaded structure is used to process the disturbances of low-frequency, mid-frequency, and high-frequency components separately. The cascaded structure enables frequency-band observation of the total disturbance, with low-frequency and high-frequency components estimated in observers with different bandwidths. This avoids high-frequency noise directly coupling to the estimation result of the low-frequency dominant component, ensuring the steady-state estimation accuracy of the low-frequency disturbance while maintaining the tracking capability of the high-frequency component, and broadening the effective bandwidth application range of the observer.

[0108] The following is a specific embodiment of this specification: Set the sampling period to 100 μs. Figures 4-7 Simulation results of two current observation methods were compared. The first channel represents the components of the actual grid-side filtered current and the observed current on the dq axis, while the second channel represents the observation error on the dq axis. As can be seen from the figures, the filter-based observation method exhibits DC error on the d-axis, with increased ripple in the observed current, and the observation error is around 1 A. In contrast, the proposed observation method shows no significant DC observation error, and the observed current ripple is smaller, with an observation error of approximately 0.2 A, demonstrating better steady-state performance.

[0109] Figures 8-11Simulation results of two observation methods under power grid fault conditions were compared. The fault occurred in the power grid at 0.8 s, and the grid-side filter current exhibited high-frequency oscillations. It can be seen that the filter-based observation method failed to track the high-frequency oscillating grid-side filter current during the dynamic process, with a maximum observation error exceeding 6 A. In contrast, the proposed observation method maintained good tracking performance for the high-frequency oscillating grid-side filter current, with a maximum observation error not exceeding 3 A, demonstrating stronger dynamic performance.

[0110] Figures 12-13 Simulation results of the proposed observation method under capacitor decay are presented, where the capacitor value suddenly decays to 50% of its rated value within 0.6 s. As shown in the figure, the observation error of the grid-side filter current increases abruptly at the moment of capacitance decay; however, because the input gain of the hyperlocal model can adaptively adjust according to the system's input and output, the model's gain coefficient is quickly updated, causing the observation error to converge rapidly. This demonstrates that the proposed observation method has strong parameter robustness.

[0111] Based on the above-described method for determining the grid-side current of an LCL filter, this specification also provides embodiments of a device for determining the grid-side current of an LCL filter. For example... Figure 14 As shown, the grid-side current determination device 1400 of the LCL filter may specifically include the following modules: Module 1401 is used to construct a hyperlocal model of the LCL filter, including the input gain to be solved and the total disturbance to be solved. The hyperlocal model represents the correlation between the rate of change of the filter capacitor voltage and the converter-side current, the input gain, and the total disturbance. The input gain represents the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current, and is used to eliminate the effect of filter capacitor drift uncertainty on the grid-side current. The total disturbance represents the combined effect of unknown factors affecting the rate of change of the filter capacitor voltage, including the grid-side current. The first solution module 1402 is used to solve the input gain based on the constraint that the total disturbance is approximately equal at adjacent sampling times, and according to the sampled values ​​of the converter side current and the filter capacitor voltage at multiple sampling times. The second solution module 1403 is used to solve the total disturbance based on the hyperlocal model with the goal of making the observation error of the filter capacitor voltage approach zero. The determination module 1404 is used to determine the grid-side current based on the total disturbance and the input gain.

[0112] In some embodiments, constructing the LCL filter hyperlocal model, which includes the input gain to be solved and the total perturbation to be solved, includes: Based on the topology of the LCL filter, determine the physical relationship between the rate of change of the filter capacitor voltage and the converter-side current, the grid-side current, and the filter capacitor. The input gain to be solved is determined based on the filter capacitor; The total disturbance to be solved is determined based on the effect of grid-side current on the rate of change of filter capacitor voltage. Based on physical relationships, a hyperlocal model is constructed that takes the converter-side current as input, the change rate of the filter capacitor voltage as output, and includes the input gain and the total disturbance to be solved.

[0113] In some embodiments, the construction of a hyperlocal model based on physical relationships, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain to be solved and the total disturbance to be solved, includes: Based on physical relationships, a hyperlocal model is constructed using the following formula, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain and the total disturbance to be solved: ; In the formula, This is the voltage across the filter capacitor. The rate of change of the filter capacitor voltage. For converter-side current, The input gain to be solved, The total disturbance to be solved is denoted as .

[0114] In some embodiments, the step of solving for the input gain based on the constraint that the total disturbance is approximately equal at adjacent sampling times, according to the sampled values ​​of the converter-side current and the filter capacitor voltage at multiple sampling times, includes: The hyperlocal model is discretized to establish discretized equations for the current sampling time and the previous sampling time, respectively. With the constraint that the total disturbance at adjacent sampling times is approximately equal, the discretization equations of the current sampling time and the previous sampling time are combined to eliminate the total disturbance; Based on the results of the simultaneous solution, the input gain at the current sampling moment is calculated using the sampled values ​​of the filter capacitor voltage and the converter side current at the current sampling moment and the previous sampling moment.

[0115] In some embodiments, the discretization of the hyperlocal model to establish discretized equations for the current sampling time and the previous sampling time includes: The hyperlocal model is discretized using the following formula to establish the discretized equations for the current sampling time and the previous sampling time, respectively: ; In the formula, , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time The converter-side current and total disturbance, and Sampling time The converter-side current and total disturbance, This is the input gain.

[0116] In some embodiments, the step of calculating the input gain at the current sampling moment based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current sampling moment and the previous sampling moment, includes: Based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling times, the input gain at the current sampling time is calculated using the following formula: ; In the formula, The input gain at the current sampling time. , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time and The converter-side current.

[0117] In some embodiments, the step of solving for the total disturbance based on a hyperlocal model with the goal of minimizing the observation error of the filter capacitor voltage includes: Input the input gain and converter-side current into the extended state observer; The deviation between the observed value and the sampled value of the filter capacitor voltage at the current moment is obtained as the observation error of the filter capacitor voltage. With the goal of reducing the observation error to zero, the total perturbation in the hyperlocal model is iteratively estimated using an extended state observer, and the total perturbation at the current sampling time is output.

[0118] In some embodiments, the step of iteratively estimating the total perturbation in the hyperlocal model using an extended state observer with the goal of minimizing the observation error to zero, and outputting the total perturbation at the current sampling time, includes: With the goal of minimizing the observation error, the total perturbation in the hyperlocal model is iteratively estimated using the following formula, outputting the total perturbation at the current sampling time: ; In the formula, For observation error, and These are the filter capacitor voltages at the sampling time. The observed values ​​and sampled values, The voltage of the filter capacitor at the sampling time The observed values, The sampling period is Sampling time The converter-side current, For input gain, and Sampling time and The total disturbance, and These are the first and second gain coefficients of the extended state observer, respectively. To expand the bandwidth of the state observer.

[0119] In some embodiments, determining the grid-side current based on the total disturbance and the input gain includes: The grid-side current at the current sampling time is determined by comparing the total disturbance with the input gain at the current sampling time.

[0120] In some embodiments, determining the grid-side current at the current sampling time based on the comparison result of the total disturbance and the input gain at the current sampling time includes: Based on the comparison between the total disturbance and the input gain at the current sampling time, the grid-side current at the current sampling time is determined using the following formula: ; In the formula, , and These represent the grid-side current, total disturbance, and input gain at the sampling time, respectively.

[0121] As can be seen from the technical solutions provided in the embodiments of this specification above, the input gain in the embodiments of this specification is adaptively adjusted according to the drift of the filter capacitor parameters, so that the determination result of the grid-side current is not affected by capacitor aging, temperature changes and batch differences, thus improving parameter robustness; the total disturbance is solved by the closed-loop feedback of the observation error tending to zero, rather than by open-loop differentiation or filtering approximation, which can suppress the interference of sampling noise and unmodeled dynamics on the observation results and improve the determination accuracy of the grid-side current; in addition, the grid-side current is determined by the total disturbance and the input gain together, without relying on the nominal parameters of the filter capacitor, thus avoiding steady-state errors introduced by parameter mismatch.

[0122] This specification also provides a computer device for determining the grid-side current of an LCL filter, including a processor and a memory for storing processor-executable instructions. In specific implementation, the processor can execute the above-described method for determining the grid-side current of an LCL filter according to the instructions.

[0123] To execute the above instructions more accurately, please refer to... Figure 15 As shown in the embodiments of this specification, another specific computer device 1500 is also provided, wherein the computer device 1500 includes a network communication port 1501, a processor 1502 and a memory 1503, and the above structures are connected by internal cables so that the various structures can perform specific data interaction.

[0124] The processor 1502 can be specifically used to execute the above-mentioned method for determining the grid-side current of the LCL filter.

[0125] The memory 1503 can be used to store the corresponding instruction program.

[0126] In this embodiment, the network communication port 1501 can be a virtual port bound to different communication protocols, thereby enabling the sending or receiving of different data. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. Furthermore, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM or CDMA; it can also be a Wi-Fi chip; or it can be a Bluetooth chip.

[0127] In this embodiment, the processor 1502 can be implemented in any suitable manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc. This specification is not limiting.

[0128] In this embodiment, the memory 1503 includes volatile memory and non-volatile memory. The memory 1503 can include multiple layers. In digital systems, anything that can store binary data can be a memory; in integrated circuits, a circuit with storage function but no physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0129] Furthermore, embodiments of this specification also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described... Figure 2 The instructions for determining the grid-side current of the LCL filter are shown.

[0130] Furthermore, embodiments of this specification provide a computer program product, including a computer program that, when executed by a processor, implements the above-described... Figure 2 The method for determining the grid-side current of the LCL filter is shown.

[0131] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.

[0132] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0133] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0134] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0135] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0136] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational tasks to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The task is a function specified in one or more boxes.

[0137] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for determining the grid-side current of an LCL filter, characterized in that, include: A hyperlocal model of the LCL filter is constructed, comprising the input gain to be solved and the total disturbance to be solved. The hyperlocal model characterizes the correlation between the rate of change of the filter capacitor voltage and the converter-side current, the input gain, and the total disturbance. The input gain characterizes the dynamic proportional relationship between the rate of change of the filter capacitor voltage and the converter-side current, and is used to eliminate the effect of filter capacitor drift uncertainty on the grid-side current. The total disturbance characterizes the combined effect of unknown factors affecting the rate of change of the filter capacitor voltage, including the grid-side current. Based on the constraint that the total disturbance is approximately equal at adjacent sampling times, the input gain is solved according to the sampled values ​​of the converter side current and the filter capacitor voltage at multiple sampling times. With the goal of reducing the observation error of the filter capacitor voltage to zero, the total disturbance is solved based on the hyperlocal model. The grid-side current is determined based on the total disturbance and the input gain.

2. The method according to claim 1, characterized in that, The construction of the LCL filter hyperlocal model, which includes the input gain to be solved and the total perturbation to be solved, includes: Based on the topology of the LCL filter, determine the physical relationship between the rate of change of the filter capacitor voltage and the converter-side current, the grid-side current, and the filter capacitor. The input gain to be solved is determined based on the filter capacitor; The total disturbance to be solved is determined based on the effect of grid-side current on the rate of change of filter capacitor voltage. Based on physical relationships, a hyperlocal model is constructed that takes the converter-side current as input, the change rate of the filter capacitor voltage as output, and includes the input gain and the total disturbance to be solved.

3. The method according to claim 2, characterized in that, The hyperlocal model, constructed based on physical relationships, takes the converter-side current as input, the filter capacitor voltage change rate as output, and includes the input gain and the total disturbance to be solved, and includes: Based on physical relationships, a hyperlocal model is constructed using the following formula, taking the converter-side current as input, the rate of change of the filter capacitor voltage as output, and including the input gain and the total disturbance to be solved: ； In the formula, This is the voltage across the filter capacitor. The rate of change of the filter capacitor voltage. For converter-side current, The input gain to be solved, The total disturbance to be solved is denoted as .

4. The method according to claim 1, characterized in that, The constraint that the total disturbance is approximately equal at adjacent sampling times is used to solve for the input gain based on the sampled values ​​of the converter-side current and the filter capacitor voltage at multiple sampling times, including: The hyperlocal model is discretized to establish discretized equations for the current sampling time and the previous sampling time, respectively. With the constraint that the total disturbance at adjacent sampling times is approximately equal, the discretization equations of the current sampling time and the previous sampling time are combined to eliminate the total disturbance; Based on the results of the simultaneous solution, the input gain at the current sampling moment is calculated using the sampled values ​​of the filter capacitor voltage and the converter side current at the current sampling moment and the previous sampling moment.

5. The method according to claim 4, characterized in that, The discretization of the hyperlocal model to establish discretized equations for the current sampling time and the previous sampling time includes: The hyperlocal model is discretized using the following formula to establish the discretized equations for the current sampling time and the previous sampling time, respectively: ； In the formula, , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time The converter-side current and total disturbance, and Sampling time The converter-side current and total disturbance, This is the input gain.

6. The method according to claim 4, characterized in that, The step of calculating the input gain at the current sampling moment based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling moments, includes: Based on the simultaneous solution results, using the sampled values ​​of the filter capacitor voltage and the converter-side current at the current and previous sampling times, the input gain at the current sampling time is calculated using the following formula: ； In the formula, The input gain at the current sampling time. , and Divided into sampling time , and The voltage of the filter capacitor, The sampling period is and Sampling time and The converter-side current.

7. The method according to claim 1, characterized in that, The goal is to minimize the observation error of the filter capacitor voltage. Based on a hyperlocal model, the total disturbance is solved, including: Input the input gain and converter-side current into the extended state observer; The deviation between the observed value and the sampled value of the filter capacitor voltage at the current moment is obtained as the observation error of the filter capacitor voltage. With the goal of reducing the observation error to zero, the total perturbation in the hyperlocal model is iteratively estimated using an extended state observer, and the total perturbation at the current sampling time is output.

8. The method according to claim 7, characterized in that, The goal is to minimize the observation error. This involves iteratively estimating the total perturbation in the hyperlocal model using an extended state observer, and outputting the total perturbation at the current sampling time. This includes: With the goal of minimizing the observation error, the total perturbation in the hyperlocal model is iteratively estimated using the following formula, outputting the total perturbation at the current sampling time: ； In the formula, For observation error, and These are the filter capacitor voltages at the sampling time. The observed values ​​and sampled values, The voltage of the filter capacitor at the sampling time The observed values, The sampling period is Sampling time The converter-side current, For input gain, and Sampling time and The total disturbance, and These are the first and second gain coefficients of the extended state observer, respectively. To expand the bandwidth of the state observer.

9. The method according to claim 1, characterized in that, The determination of the grid-side current based on the total disturbance and input gain includes: The grid-side current at the current sampling time is determined by comparing the total disturbance with the input gain at the current sampling time.

10. The method according to claim 9, characterized in that, The step of determining the grid-side current at the current sampling time based on the comparison result of the total disturbance and the input gain at the current sampling time includes: Based on the comparison between the total disturbance and the input gain at the current sampling time, the grid-side current at the current sampling time is determined using the following formula: ； In the formula, , and These represent the grid-side current, total disturbance, and input gain at the sampling time, respectively.

Images