Control optimization method and system for inverter

Abstract

The invention discloses a control optimization method and system for an inverter, and relates to the technical field of inverter control. The method comprises the following steps: acquiring current and voltage signals of an inverter on a time sequence, calculating instantaneous power based on a power observation model to obtain a harmonic power component, calculating fundamental power of the current and voltage signals to obtain a fundamental power component, and coupling the fundamental power component and the harmonic power component to obtain an overall coupling coefficient; and inputting the coefficient into an enhanced adaptive sliding mode controller to obtain an initial control signal, calculating a difference value through a damping compensation algorithm to obtain a damping compensation signal, and superposing the damping compensation signal to form a final control signal to control the action of the inverter. According to the method, through multi-dimensional improvement of an observation model, the observation precision of harmonic power is greatly improved; and then coupling analysis and a damping compensation mechanism are combined to perform dual optimization on the control signal, and closed-loop upgrading is formed from sensing to regulation, so that the accuracy of inverter control is remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of inverter control technology, specifically relating to a control optimization method and system for inverters. Background Technology

[0002] With the rapid development of new energy power generation, microgrids, and other fields, and the continuous improvement of power quality requirements, inverter control optimization has become a core element in ensuring the stable operation of the power system and improving energy utilization efficiency. However, current inverter control faces many complex challenges, such as diverse grid disturbances, dynamic and variable load conditions, and prominent harmonic interference in the inverter operating environment. These conditions not only cause high harmonic distortion rates and low power conversion efficiency in the inverter output current, but may also lead to decreased system stability and increased equipment losses. Therefore, constructing accurate and robust inverter control optimization methods and leveraging advanced technologies to achieve real-time perception and dynamic control of the inverter's operating status has become an urgent need to ensure the safe and efficient operation of the power system. Traditional inverter control harmonic power observation suffers from insufficient accuracy and real-time performance. Damping compensation often uses fixed parameters, which cannot be dynamically adjusted according to signal changes, easily leading to insufficient or over-compensation, making it difficult to meet the high-precision control requirements of inverters under complex operating conditions.

[0003] Patent CN120110200A discloses an inverter control method, system, and photovoltaic inverter, belonging to the field of inverter control technology. In this application, the sampled signal in the inverter is first analyzed to obtain several frequency domain signal points. Then, one or more target resonant points are determined from these frequency domain signal points. Filtering parameters are determined based on the frequency of the target resonant points to obtain a bandpass filter. The sampled signal is then filtered using the bandpass filter to obtain the signal component corresponding to the target resonant point. Finally, a virtual damping output is obtained based on the target resonant point and its corresponding signal component, and the inverter is controlled based on this virtual damping output. This achieves targeted resonance suppression for each target resonant point, thereby improving inverter performance. However, this invention does not consider the difficulty of balancing harmonic power observation accuracy and real-time performance, and its damping compensation mechanism does not adaptively adjust to dynamic disturbances and signal changes, resulting in low inverter control accuracy. Summary of the Invention

[0004] The purpose of this invention is to solve the problem of low control accuracy of inverters, and to propose a control optimization method and system for inverters.

[0005] In a first aspect of this invention, a control optimization method for an inverter is first proposed, the method comprising: Obtain the inverter's current and voltage signals in a time series; The instantaneous power of the current signal and the voltage signal is calculated using a power observation model to obtain the harmonic power components. The power observation model is an improvement on the LSTM model, specifically including: constructing a feature enhancement layer between the input layer and the hidden layer of the LSTM model; replacing the activation function used in the hidden layer of the LSTM model with an improved EIU activation function; and iteratively optimizing the parameters of the power observation model using a quantum particle swarm optimization algorithm. The fundamental power component is obtained by calculating the fundamental power of the current signal and the voltage signal. The fundamental power component and the harmonic power component are coupled to obtain the overall coupling coefficient; The overall coupling coefficient is input into the enhanced adaptive sliding mode controller to calculate the initial control signal; The damping compensation signal is obtained by calculating the difference between the initial control signal and the output current signal using a damping compensation algorithm. The damping compensation signal and the initial control signal are superimposed to obtain the final control signal; The final control signal controls the operation of the inverter.

[0006] Optionally, the current signal and the voltage signal are divided into sub-current signals and sub-voltage signals by a preset time interval; The sub-current signal and the sub-voltage signal are combined according to timestamps to obtain a sub-signal group; The sub-signal group is input into the input layer of the power observation model, and the sub-signal group is decomposed into sub-bands and dynamically weighted through the multi-scale wavelet attention mechanism of the feature enhancement layer to obtain a multi-scale feature vector. The multi-scale feature vector is input into the hidden layer of the power observation model to obtain the deep feature vector; Instantaneous power is obtained by integrating the deep feature vectors using the linear activation function of the output layer; The instantaneous power corresponding to the sub-signal group is subjected to spectral analysis and the harmonic power components are separated.

[0007] Optionally, the fundamental power component and the harmonic power component can be decomposed into fundamental power sub-component sets and harmonic power sub-component sets by wavelet packet transform at different frequency bands. The target fundamental power sub-component and the target harmonic power sub-component in the same frequency band are combined to obtain a power sub-component group; the target fundamental power sub-component is any one of the fundamental power sub-components in the set of fundamental power sub-components; the target harmonic power sub-component is any one of the harmonic power sub-components in the set of harmonic power sub-components; The fundamental power RMS value sequence and harmonic power RMS value sequence are obtained by aligning the power sub-components according to time stamps and calculating their effective values. A local coupling coefficient set is obtained by calculating the correlation coefficient of the fundamental power RMS value sequence and the harmonic power RMS value sequence through a first preset sliding window. The frequency band coupling coefficient is obtained by weighting the set of local coupling coefficients obtained for each power sub-component group according to the composite power energy within its corresponding time window. The overall coupling coefficient is obtained by weighting and summing the frequency band coupling coefficients of all frequency bands according to the proportion of power of each frequency band in the total power.

[0008] Optionally, the overall coupling coefficient is compared with a preset sliding surface function reference value to obtain the sliding surface function value; pass Calculate the initial control signal; where u is the initial control signal, k is the gain coefficient of the sliding mode reaching law (greater than 0), s is the sliding surface function value, and d is the total disturbance value estimated in real time by the disturbance observer. The weights are greater than 0, and sgn(s) is the sign function. Let be the rate of change of the disturbance.

[0009] Optionally, in step 1, the initial control signal and the output current signal are aligned according to a time window to obtain a signal group in a time sequence; Step 2: Calculate the instantaneous difference of the signal group to generate an instantaneous difference sequence; Step 3: Perform candidate sliding window averaging filtering on the instantaneous difference sequence to obtain a candidate filtered instantaneous difference sequence; Step 4: Calculate the average difference within each window of the candidate filter instantaneous difference sequence using a second preset sliding window to obtain the trend value of change; Step 5: If the trend value is greater than or equal to the preset trend value, adjust the candidate sliding window as a new candidate sliding window, and return to step 3 until the trend value is less than the preset trend value to finally obtain the filtered instantaneous difference sequence. Step 6: Analyze the dynamic changes of the filtered instantaneous difference sequence and adaptively adjust the damping compensation coefficient to obtain the initial damping compensation signal; Step 7: Limit the amplitude of the preliminary damping compensation signal to obtain the damping compensation signal.

[0010] In a second aspect of this invention, a control optimization system for an inverter is provided, comprising: The acquisition module is used to acquire the inverter's current and voltage signals over a time series. The harmonic power component generation module is used to calculate the instantaneous power of the current signal and the voltage signal to obtain the harmonic power components through a power observation model. The power observation model is an improvement based on the LSTM model, and the specific improvements include: constructing a feature enhancement layer between the input layer and the hidden layer of the LSTM model; replacing the activation function used in the hidden layer of the LSTM model with an improved EIU activation function; and iteratively optimizing the parameters of the power observation model through a quantum particle swarm optimization algorithm. The fundamental power component generation module is used to calculate the fundamental power component from the current signal and the voltage signal; The overall coupling coefficient generation module is used to couple the fundamental power component and the harmonic power component to obtain the overall coupling coefficient; The initial control signal calculation module is used to input the overall coupling coefficient into the enhanced adaptive sliding mode controller to calculate the initial control signal; The damping compensation signal calculation module is used to calculate the difference between the initial control signal and the output current signal using a damping compensation algorithm to obtain the damping compensation signal. The final control signal generation module is used to superimpose the damping compensation signal and the initial control signal to obtain the final control signal; A control module is used to control the operation of the inverter using the final control signal.

[0011] Optionally, the harmonic power component calculation module includes: A sub-signal generation module is used to divide the current signal and the voltage signal into sub-current signals and sub-voltage signals by a preset time interval; The sub-signal group generation module is used to combine the sub-current signal and the sub-voltage signal according to the timestamp to obtain the sub-signal group; A multi-scale feature vector generation module is used to input the sub-signal group into the input layer of the power observation model, and to obtain a multi-scale feature vector by performing sub-band decomposition and dynamic weight allocation on the sub-signal group through the multi-scale wavelet attention mechanism of the feature enhancement layer. A deep feature vector generation module is used to input the multi-scale feature vector into the hidden layer of the power observation model to obtain a deep feature vector; The instantaneous power calculation module is used to integrate and calculate the instantaneous power by using the linear activation function of the output layer to calculate the deep feature vector; The harmonic power component module is used to perform spectral analysis on the instantaneous power corresponding to the sub-signal group and separate the harmonic power components.

[0012] Optionally, the overall coupling coefficient generation module includes: The sub-component set generation module is used to decompose the fundamental power component and the harmonic power component into fundamental power sub-component sets and harmonic power sub-component sets by wavelet packet transform at different frequency bands. The power sub-component group module is used to combine the target fundamental power sub-component and the target harmonic power sub-component in the same frequency band to obtain a power sub-component group; the target fundamental power sub-component is any one of the fundamental power sub-components in the set of fundamental power sub-components; the target harmonic power sub-component is any one of the harmonic power sub-components in the set of harmonic power sub-components; An effective sequence generation module is used to align the power sub-component groups according to time markers and calculate the effective values ​​to obtain the fundamental power effective value sequence and the harmonic power effective value sequence; The local coupling coefficient calculation module is used to calculate the correlation coefficient of the fundamental power RMS value sequence and the harmonic power RMS value sequence through a first preset sliding window to obtain a local coupling coefficient set. The frequency band coupling coefficient calculation module is used to calculate the frequency band coupling coefficient of each power sub-component by weighting the set of local coupling coefficients obtained by each power sub-component according to the composite power energy within its corresponding time window. The overall coupling coefficient calculation module is used to calculate the overall coupling coefficient by weighting and summing the frequency band coupling coefficients of all frequency bands according to the proportion of power of each frequency band in the total power.

[0013] Optionally, the initial control signal calculation module includes: The sliding surface function value calculation module is used to compare the overall coupling coefficient with the preset sliding surface function reference value to calculate the sliding surface function value; The initial control signal generation module is used to generate the initial control signal through... Calculate the initial control signal; where u is the initial control signal, k is the gain coefficient of the sliding mode reaching law (greater than 0), s is the sliding surface function value, and d is the total disturbance value estimated in real time by the disturbance observer. The weights are greater than 0, and sgn(s) is the sign function. Let be the rate of change of the disturbance.

[0014] Optionally, the damping compensation signal calculation module includes: The signal group generation module is used to align the initial control signal and the output current signal with a time window to obtain a signal group in the time series. The instantaneous difference sequence generation module is used to calculate the instantaneous difference of the signal group and generate an instantaneous difference sequence. The candidate sequence generation module is used to perform candidate sliding window average filtering on the instantaneous difference sequence to obtain a candidate filtered instantaneous difference sequence; The trend value calculation module is used to calculate the average difference within each window of the candidate filtered instantaneous difference sequence through a second preset sliding window to obtain the trend value. The judgment module is used to adjust the candidate sliding window as a new candidate sliding window if the change trend value is greater than or equal to the preset change trend value, and return to the candidate sequence generation module until the change trend value is less than the preset change trend value to finally obtain the filtered instantaneous difference sequence. An initial damping compensation signal generation module is used to analyze the dynamic changes of the filtered instantaneous difference sequence and adaptively adjust the damping compensation coefficient to obtain an initial damping compensation signal. An amplitude limiting module is used to limit the amplitude of the preliminary damping compensation signal to obtain a damping compensation signal.

[0015] The beneficial effects of this invention are as follows: This invention proposes a control optimization method for inverters. First, the current and voltage signals of the inverter are acquired in a time series. Then, the instantaneous power is calculated based on a power observation model to obtain the harmonic power components. Simultaneously, the fundamental power is calculated from the current and voltage signals to obtain the fundamental power components. Next, the fundamental and harmonic power components are coupled to obtain the overall coupling coefficient. This coefficient is input into an enhanced adaptive sliding mode controller to obtain the initial control signal. The difference is calculated using a damping compensation algorithm to obtain the damping compensation signal. These are then superimposed to form the final control signal to control the inverter operation. This method significantly improves the observation accuracy of harmonic power through multi-dimensional improvements to the observation model. Furthermore, by combining coupling analysis and a damping compensation mechanism to perform dual optimization of the control signal, a closed-loop upgrade is formed from perception to regulation, thereby significantly improving the accuracy of inverter control. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 A flowchart of a control optimization method for an inverter provided in an embodiment of the present invention; Figure 2 This is a framework diagram of a control optimization system for an inverter provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention provides a control optimization method for inverters. See also... Figure 1 , Figure 1 A flowchart illustrating a control optimization method for an inverter, provided as an embodiment of the present invention. The method includes the following steps: S101 acquires the inverter's current and voltage signals in a time series. S102, the harmonic power components are obtained by calculating the instantaneous power of the current signal and voltage signal through the power observation model; S103 calculates the fundamental power component by performing fundamental power calculation on the current signal and voltage signal; S104, the fundamental power component and harmonic power component are coupled to obtain the overall coupling coefficient; S105, input the overall coupling coefficient to the enhanced adaptive sliding mode controller to calculate the initial control signal; S106, the damping compensation signal is obtained by calculating the difference between the initial control signal and the output current signal through the damping compensation algorithm; S107, the damping compensation signal and the initial control signal are superimposed to obtain the final control signal; S108, the final control signal controls the operation of the inverter; The power observation model is an improvement on the LSTM model. The specific improvements include: constructing a feature enhancement layer between the input layer and the hidden layer of the LSTM model; replacing the activation function used in the hidden layer of the LSTM model with an improved EIU activation function; and iteratively optimizing the parameters of the power observation model using the quantum particle swarm optimization algorithm.

[0021] The present invention provides a control optimization method for inverters. This method first acquires the inverter's current and voltage signals in a time series, then calculates the instantaneous power based on a power observation model to obtain harmonic power components. Simultaneously, it calculates the fundamental power from the current and voltage signals to obtain the fundamental power component. Next, it couples the fundamental and harmonic power components to obtain an overall coupling coefficient. This coefficient is input into an enhanced adaptive sliding mode controller to obtain an initial control signal. The difference is calculated using a damping compensation algorithm to obtain a damping compensation signal. These are then superimposed to form the final control signal to control the inverter's operation. This method significantly improves the accuracy of harmonic power observation through multi-dimensional improvements to the observation model. Furthermore, by combining coupling analysis and a damping compensation mechanism to perform dual optimization of the control signal, a closed-loop upgrade is formed from perception to regulation, thereby significantly improving the accuracy of inverter control.

[0022] In one implementation, the hidden layer activation function is replaced with an improved EIU activation function. This function is designed for the wide range and strong dynamic characteristics of power signals. The positive input uses a hybrid linear and exponential form, which retains the linear segment to avoid saturation with large signals, while introducing nonlinear fitting details for small signals through the exponential term. The negative input uses an exponential and decaying form to avoid gradient vanishing and enhance sensitivity to negative power fluctuations, replacing the tanh or sigmoid activation functions of traditional LSTM, thus improving the model's accuracy and dynamic response to power nonlinear mapping. The power observation model, iteratively optimized using the quantum particle swarm optimization algorithm, can learn parameters, specifically including the wavelet basis, decomposition scale, and weight matrix and bias vector of the attention mechanism in the wavelet decomposition in the feature enhancement layer; the weight matrix and bias vector of the forget gate, input gate, cell state update, and output gate in the LSTM layer; and the weight matrix and bias vector of the output layer used to integrate deep feature vectors to obtain instantaneous power. These parameters together determine the model's ability to process and predict power signals.

[0023] In one implementation, the inverter current and voltage signals obtained from the time series are first divided into sub-current and sub-voltage signals by a preset time interval. The fundamental components of the sub-current and sub-voltage signals are then separated by the Fourier transform method to obtain the fundamental current and fundamental voltage. Then, based on the active power calculation formula, the fundamental current and fundamental voltage are instantaneously multiplied to obtain the fundamental power component.

[0024] In one implementation, the amplitude of the damping compensation signal is detected to confirm that the signal amplitude meets the preset input requirements. Then, the waveform of the switch control signal is shaped to eliminate spikes and pulses in the signal. The shaped switch control signal and the detected qualified damping compensation signal are input to a signal superposition unit. In the signal superposition unit, the two signals are linearly superimposed according to a preset weight ratio to generate a superimposed initial modulation signal. The initial modulation signal is subjected to spectrum analysis to detect whether there are high-frequency components exceeding the preset range. If high-frequency components are present, the initial modulation signal is pre-filtered by a filtering preprocessing unit. The pre-filtered signal is then amplitude-calibrated again to ensure that the signal amplitude meets the control requirements of the full-bridge circuit. The calibrated signal is used as the final modulation signal and output to the switch control unit of the full-bridge circuit.

[0025] In one implementation, the final control signal is input to the inverter's signal adaptation unit. Through level conversion and timing calibration, it is converted into a pulse drive signal that meets the driving requirements of the switching devices in the inverter's full-bridge topology. This ensures that the voltage amplitude, rise or fall edge speed, and pulse width of the drive signal match the rated parameters of the switching devices. Subsequently, the adapted drive signal is transmitted to the drive circuit of the corresponding switching device. The drive circuit controls the switching device to turn on and off according to the high and low level states of the pulse signal. When the drive signal is high, the switching device is turned on, and the corresponding bridge arm forms a path. When the drive signal is low, the switching device is turned off, and the corresponding bridge arm is disconnected. By having each switching device alternately turn on and off according to a preset logic, the DC to AC inversion conversion is achieved. Simultaneously, during the control process, the on-state voltage drop, off-state leakage current, and other status parameters of the switching devices are collected in real time. Combined with the real-time current and voltage feedback at the inverter output, the timing and amplitude of the drive signal are dynamically fine-tuned to ensure the synchronization of the switching device operation and the stability of the inverter output, ultimately completing the closed-loop operation control of the inverter.

[0026] In one embodiment, the current signal and voltage signal are divided into sub-current signal and sub-voltage signal by a preset time interval; Sub-signal groups are obtained by combining sub-current signals and sub-voltage signals according to timestamps; The sub-signal group is input into the input layer of the power observation model, and the sub-signal group is decomposed into sub-bands and dynamically weighted through the multi-scale wavelet attention mechanism of the feature enhancement layer to obtain the multi-scale feature vector. The multi-scale feature vector is input into the hidden layer of the power observation model to obtain the deep feature vector; Instantaneous power is obtained by integrating the deep feature vectors through the linear activation function of the output layer; The instantaneous power of the sub-signal group is subjected to spectral analysis and the harmonic power components are separated.

[0027] In one implementation, the sub-signal group first enters the model input layer to complete data format standardization. Then, in the feature enhancement layer, the multi-scale wavelet attention mechanism first uses wavelet basis functions of different scales to perform multi-resolution sub-band decomposition on the current and voltage sub-signals to obtain sub-band features covering different frequency components. At the same time, the attention mechanism learns to dynamically allocate weights based on the correlation between each sub-band feature and power calculation, strengthens the feature expression of key frequency sub-bands, and weakens the influence of noise or irrelevant sub-bands. Finally, the weighted multi-scale sub-band features are fused to form a multi-scale feature vector containing multi-level and multi-frequency information.

[0028] In one implementation, multi-scale feature vectors are fed into an improved LSTM hidden layer. The hidden layer models the temporal dependencies and performs nonlinear transformations on the input multi-scale features using memory units and an improved EIU activation function. The EIU activation function enhances the nonlinear representation of features and the efficiency of gradient propagation. During processing, the hidden layer continuously updates the cell state and hidden state, capturing the deep correlation between current and voltage changes over time in the sub-signal group. Finally, it outputs a deep feature vector that has undergone multi-layer abstraction and temporal feature extraction. This vector contains key implicit features for calculating instantaneous power.

[0029] In one implementation, after the deep feature vector is input into the model output layer, the output layer uses a linear activation function to perform weighted summation and bias adjustment, mapping the deep features to the power numerical space. By integrating the current, voltage, and related feature information contained in the deep features, the instantaneous power value corresponding to the sub-signal group is directly calculated. This value can reflect the instantaneous power change of the inverter within that time period in real time. For the instantaneous power calculated for each sub-signal group, spectral analysis methods such as Fourier transform or wavelet transform are used to decompose it into power spectra of different frequency components. Then, based on the fundamental frequency, the power corresponding to all frequency components other than the fundamental frequency is separated from the power spectrum. The sum of the power of these non-fundamental frequency components is the harmonic power component, reflecting the power fluctuation part of the inverter output power caused by harmonics.

[0030] In one embodiment, the fundamental power component and harmonic power component are decomposed into fundamental power sub-component sets and harmonic power sub-component sets by wavelet packet transform at different frequency bands. The target fundamental power sub-component and the target harmonic power sub-component in the same frequency band are combined to obtain a power sub-component group; the target fundamental power sub-component is any one fundamental power sub-component in the fundamental power sub-component group; the target harmonic power sub-component is any one harmonic power sub-component in the harmonic power sub-component group. The fundamental power RMS value sequence and harmonic power RMS value sequence are obtained by aligning the power sub-components according to the time stamp and calculating the effective values. The local coupling coefficient set is obtained by calculating the correlation coefficient between the fundamental power RMS value sequence and the harmonic power RMS value sequence through the first preset sliding window. The frequency band coupling coefficient is obtained by weighting the set of local coupling coefficients obtained for each power sub-component group according to the composite power energy within its corresponding time window. The overall coupling coefficient is obtained by weighting and summing the frequency band coupling coefficients of all frequency bands according to the proportion of power of each frequency band in the total power.

[0031] In one implementation, the first preset sliding window is determined by a technician.

[0032] In one implementation, the detailed logical process of decomposing the fundamental power component and harmonic power component into fundamental power sub-component sets and harmonic power sub-component sets by wavelet packet transform in different frequency bands is as follows: using the wavelet packet transform algorithm, the fundamental power component and harmonic power component are decomposed into multiple fundamental power sub-components covering different frequency ranges according to the preset frequency resolution. By selecting appropriate wavelet basis functions and decomposition levels, the fundamental power component is decomposed into multiple fundamental power sub-components covering different frequency ranges to form a fundamental power sub-component set. At the same time, the harmonic power component is decomposed into harmonic power sub-components that correspond one-to-one with the frequency range of the fundamental power sub-component set to form a harmonic power sub-component set, ensuring that the decomposed sub-components can accurately reflect the power characteristics in each frequency band.

[0033] In one implementation, a fixed-length sliding window is used to sequentially slide along the time axis of the fundamental power RMS sequence and the harmonic power RMS sequence. After each slide, a subsequence of the fundamental power RMS and a subsequence of the harmonic power RMS within the window are extracted. The linear correlation between the two is calculated using the Pearson correlation coefficient to obtain the local coupling coefficient corresponding to that window. As the window continues to slide, a series of local coupling coefficients covering the entire time series are generated and combined to form a set of local coupling coefficients, reflecting the correlation strength between the fundamental and harmonic powers in different time segments. For the set of local coupling coefficients corresponding to each power sub-component, the composite power energy within the time window corresponding to each local coupling coefficient is calculated, i.e., the sum of the squares of the fundamental power RMS and the harmonic power RMS within that window, or other energy synthesis methods. Each local coupling coefficient is multiplied by its corresponding composite power energy and summed, then divided by the sum of the composite power energy of all windows. This weighted averaging process highlights the influence of time windows with high energy proportions on the coupling coefficient, ultimately obtaining a frequency band coupling coefficient that represents the overall correlation between the fundamental and harmonic powers in that frequency band. Calculate the total power of the fundamental power subcomponent and harmonic power subcomponent corresponding to each frequency band, and use their respective proportions to the total output power of the inverter as the weights of the coupling coefficients for each frequency band. Multiply the coupling coefficient of each frequency band by its corresponding weight and sum them up. Through this weighted summation process, the contribution of each frequency band to the overall power coupling relationship is integrated, and finally, the overall coupling coefficient reflecting the overall correlation characteristics of the inverter's fundamental power and harmonic power is obtained.

[0034] In one embodiment, the sliding surface function value is calculated by comparing the overall coupling coefficient with a preset sliding surface function reference value; pass Calculate the initial control signal; where u is the initial control signal, k is the gain coefficient of the sliding mode reaching law (greater than 0), s is the sliding surface function value, and d is the total disturbance value estimated in real time by the disturbance observer. The weights are greater than 0, and sgn(s) is the sign function. Let be the rate of change of the disturbance.

[0035] In one implementation, the preset sliding surface function reference value is determined by technicians.

[0036] In one implementation, the sliding surface function value is calculated by comparing the overall coupling coefficient with a preset sliding surface function reference value. This allows for the precise capture of the inverter system's current harmonic interference and operational deviations, using the harmonic voltage and current coupling strength reflected by the overall coupling coefficient as the core state variable. This provides an input basis that closely matches the actual operating state for sliding mode control. Furthermore, the initial control signal is calculated by introducing the sliding mode approach law gain coefficient, the total disturbance value estimated in real time by the disturbance observer, the weighting coefficient, and the control formula for the disturbance change rate. This approach leverages the inherent robustness of sliding mode control to address the system's nonlinear characteristics while simultaneously offsetting the effects of uncertainties such as load fluctuations and grid disturbances through real-time disturbance estimation and disturbance change rate compensation. The weighting coefficient allows for flexible adjustment of the disturbance change rate compensation, and the sign function ensures the control signal's rapid approach to the sliding surface. Ultimately, this ensures that the initial control signal balances response speed and disturbance rejection capability, laying a precise control foundation for stable inverter output.

[0037] In one embodiment, step 1 involves aligning the initial control signal and the output current signal according to a time window to obtain a signal group in a time sequence. Step 2: Calculate the instantaneous difference of the signal group to generate an instantaneous difference sequence; Step 3: Perform candidate sliding window averaging filtering on the instantaneous difference sequence to obtain the candidate filtered instantaneous difference sequence; Step 4: Calculate the average difference within each window of the candidate filter instantaneous difference sequence using a preset sliding window to obtain the trend value of change; Step 5: If the trend value is greater than or equal to the preset trend value, adjust the candidate sliding window as a new candidate sliding window, and return to step 3 until the trend value is less than the preset trend value to finally obtain the filtered instantaneous difference sequence. Step 6: Analyze the dynamic changes of the filtered instantaneous difference sequence and adaptively adjust the damping compensation coefficient to obtain the initial damping compensation signal; Step 7: Limit the amplitude of the preliminary damping compensation signal to obtain the damping compensation signal.

[0038] In one implementation, a second preset sliding window slides along the candidate filtered instantaneous difference sequence, calculates the average value of the filtered difference within the local interval covered by each window, obtains the average difference of that interval, and arranges these average differences in window order to form a trend value that reflects the overall change direction and magnitude of the filtered difference sequence, which is used to judge the stability of the difference sequence. The second preset sliding window will be determined by technical personnel.

[0039] In one implementation, the instantaneous difference sequence of the filter is dynamically analyzed. By monitoring the fluctuation amplitude and rate of change of the sequence over time, the dynamic characteristics of the deviation between the current control signal and the output current are determined. Based on the analysis results, the damping compensation coefficient is adaptively adjusted. When the difference fluctuation is large, the damping compensation coefficient is increased to enhance system damping and suppress oscillations; when the difference tends to stabilize, the damping compensation coefficient is decreased to avoid over-damping affecting the response speed, ultimately generating an initial damping compensation signal that matches the current deviation state.

[0040] In one implementation, the deviation between the control signal and the output is accurately captured through time alignment and difference calculation. The stability of the deviation sequence is improved by filtering through adaptive dynamic adjustment window. Combined with dynamic analysis, damping compensation is adaptively adjusted and the amplitude is limited. This not only ensures the accuracy of deviation perception and the adaptability of filtering, but also realizes the dynamic matching and safe controllability of damping compensation, effectively enhancing the accuracy, stability and reliability of system control.

[0041] Based on the same inventive concept, embodiments of the present invention also provide a control optimization system for an inverter. See also Figure 2 , Figure 2 A framework diagram of a control optimization system for an inverter provided in an embodiment of the present invention includes: The acquisition module is used to acquire the inverter's current and voltage signals over a time series. The harmonic power component generation module is used to calculate the instantaneous power of current and voltage signals to obtain harmonic power components through a power observation model. The power observation model is an improvement on the LSTM model, specifically including: constructing a feature enhancement layer between the input and hidden layers of the LSTM model; replacing the activation function used in the hidden layer of the LSTM model with an improved EIU activation function; and iteratively optimizing the parameters of the power observation model using a quantum particle swarm optimization algorithm. The fundamental power component generation module is used to calculate the fundamental power of current and voltage signals to obtain the fundamental power component. The overall coupling coefficient generation module is used to couple the fundamental power component and the harmonic power component to obtain the overall coupling coefficient; The initial control signal calculation module is used to input the overall coupling coefficient into the enhanced adaptive sliding mode controller to calculate the initial control signal; The damping compensation signal calculation module is used to calculate the damping compensation signal by performing a damping compensation algorithm on the difference between the initial control signal and the output current signal. The final control signal generation module is used to superimpose the damping compensation signal and the initial control signal to obtain the final control signal; The control module is used to control the inverter's operation using the final control signals.

[0042] The control optimization system for inverters provided in this invention first acquires the inverter's current and voltage signals in a time series, then calculates the instantaneous power based on a power observation model to obtain the harmonic power components. Simultaneously, it calculates the fundamental power from the current and voltage signals to obtain the fundamental power components. Next, it couples the fundamental and harmonic power components to obtain the overall coupling coefficient. This coefficient is input into an enhanced adaptive sliding mode controller to obtain the initial control signal. The difference is calculated using a damping compensation algorithm to obtain the damping compensation signal. These are then superimposed to form the final control signal to control the inverter's operation. This method significantly improves the accuracy of harmonic power observation through multi-dimensional improvements to the observation model. Furthermore, by combining coupling analysis and a damping compensation mechanism to perform dual optimization of the control signal, a closed-loop upgrade is formed from perception to regulation, thereby significantly improving the accuracy of inverter control.

[0043] In one embodiment, the sub-signal generation module is used to divide the current signal and voltage signal into sub-current signal and sub-voltage signal by a preset time interval; The sub-signal group generation module is used to combine the sub-current signal and the sub-voltage signal according to the timestamp to obtain the sub-signal group; The multi-scale feature vector generation module is used to input the sub-signal group into the input layer of the power observation model. The multi-scale feature vector is obtained by performing sub-band decomposition and dynamic weight allocation on the sub-signal group through the multi-scale wavelet attention mechanism of the feature enhancement layer. The deep feature vector generation module is used to input multi-scale feature vectors into the hidden layer of the power observation model to obtain deep feature vectors; The instantaneous power calculation module is used to calculate the instantaneous power by integrating the deep feature vectors through the linear activation function of the output layer; The harmonic power component module is used to perform spectral analysis on the instantaneous power corresponding to the sub-signal group and separate the harmonic power components.

[0044] In one embodiment, the overall coupling coefficient generation module includes: The sub-component set generation module is used to decompose the fundamental power component and harmonic power component into fundamental power sub-component sets and harmonic power sub-component sets by wavelet packet transform in different frequency bands. The power sub-component group module is used to combine the target fundamental power sub-component and the target harmonic power sub-component in the same frequency band to obtain a power sub-component group; the target fundamental power sub-component is any one of the fundamental power sub-components in the fundamental power sub-component group; the target harmonic power sub-component is any one of the harmonic power sub-components in the harmonic power sub-component group. The effective sequence generation module is used to align the power sub-components according to time stamps and calculate the effective values ​​to obtain the fundamental power effective value sequence and the harmonic power effective value sequence; The local coupling coefficient calculation module is used to calculate the correlation coefficient of the fundamental power effective value sequence and the harmonic power effective value sequence through a first preset sliding window to obtain a local coupling coefficient set. The frequency band coupling coefficient calculation module is used to calculate the frequency band coupling coefficient of each power sub-component by weighting the set of local coupling coefficients obtained by each power sub-component according to the composite power energy within its corresponding time window. The overall coupling coefficient calculation module is used to calculate the overall coupling coefficient by weighting and summing the frequency band coupling coefficients of all frequency bands according to the proportion of power of each frequency band in the total power.

[0045] In one embodiment, the initial control signal calculation module includes: The sliding surface function value calculation module is used to calculate the sliding surface function value by comparing the overall coupling coefficient with the preset sliding surface function reference value; The initial control signal generation module is used to generate the initial control signal through... Calculate the initial control signal; where u is the initial control signal, k is the gain coefficient of the sliding mode reaching law (greater than 0), s is the sliding surface function value, and d is the total disturbance value estimated in real time by the disturbance observer. The weights are greater than 0, and sgn(s) is the sign function. Let be the rate of change of the disturbance.

[0046] In one embodiment, the damping compensation signal calculation module includes: The signal group generation module is used to align the initial control signal and the output current signal with a time window to obtain a signal group in the time series. The instantaneous difference sequence generation module is used to calculate the instantaneous difference of the signal group and generate an instantaneous difference sequence. The candidate sequence generation module is used to perform candidate sliding window average filtering on the instantaneous difference sequence to obtain the candidate filtered instantaneous difference sequence; The trend value calculation module is used to calculate the average difference within each window of the candidate filtered instantaneous difference sequence through a second preset sliding window to obtain the trend value. The judgment module is used to adjust the candidate sliding window as a new candidate sliding window if the trend value is greater than or equal to the preset trend value, and return to the candidate sequence generation module until the trend value is less than the preset trend value to finally obtain the filtered instantaneous difference sequence. The initial damping compensation signal generation module is used to dynamically adjust the damping compensation coefficient based on the dynamic changes of the filtered instantaneous difference sequence to obtain the initial damping compensation signal. The amplitude limiting module is used to limit the amplitude of the initial damping compensation signal to obtain the damping compensation signal.

[0047] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A control optimization method for inverters, characterized in that, The method includes: Obtain the inverter's current and voltage signals in a time series; The instantaneous power of the current signal and the voltage signal is calculated using a power observation model to obtain the harmonic power components. The power observation model is an improvement on the LSTM model, specifically including: constructing a feature enhancement layer between the input layer and the hidden layer of the LSTM model; replacing the activation function used in the hidden layer of the LSTM model with an improved EIU activation function; and iteratively optimizing the parameters of the power observation model using a quantum particle swarm optimization algorithm. The fundamental power component is obtained by calculating the fundamental power of the current signal and the voltage signal. The fundamental power component and the harmonic power component are coupled to obtain the overall coupling coefficient; The overall coupling coefficient is input into the enhanced adaptive sliding mode controller to calculate the initial control signal; The damping compensation signal is obtained by calculating the difference between the initial control signal and the output current signal using a damping compensation algorithm. The damping compensation signal and the initial control signal are superimposed to obtain the final control signal; The final control signal controls the operation of the inverter.

2. The control optimization method for an inverter according to claim 1, characterized in that, The harmonic power components are obtained by calculating the instantaneous power of the current signal and the voltage signal using a power observation model, including: The current signal and the voltage signal are divided into sub-current signals and sub-voltage signals by a preset time interval; The sub-current signal and the sub-voltage signal are combined according to timestamps to obtain a sub-signal group; The sub-signal group is input into the input layer of the power observation model, and the sub-signal group is decomposed into sub-bands and dynamically weighted through the multi-scale wavelet attention mechanism of the feature enhancement layer to obtain a multi-scale feature vector. The multi-scale feature vector is input into the hidden layer of the power observation model to obtain the deep feature vector; Instantaneous power is obtained by integrating the deep feature vectors using the linear activation function of the output layer; The instantaneous power corresponding to the sub-signal group is subjected to spectral analysis and the harmonic power components are separated.

3. The control optimization method for an inverter according to claim 1, characterized in that, The overall coupling coefficient obtained by coupling the fundamental power component and the harmonic power component includes: The fundamental power component and the harmonic power component are decomposed into fundamental power component sets and harmonic power component sets by wavelet packet transform at different frequency bands. The target fundamental power sub-component and the target harmonic power sub-component in the same frequency band are combined to obtain a power sub-component group; the target fundamental power sub-component is any one of the fundamental power sub-components in the set of fundamental power sub-components; the target harmonic power sub-component is any one of the harmonic power sub-components in the set of harmonic power sub-components; The fundamental power RMS value sequence and harmonic power RMS value sequence are obtained by aligning the power sub-components according to time stamps and calculating their effective values. A local coupling coefficient set is obtained by calculating the correlation coefficient of the fundamental power RMS value sequence and the harmonic power RMS value sequence through a first preset sliding window. The frequency band coupling coefficient is obtained by weighting the set of local coupling coefficients obtained for each power sub-component group according to the composite power energy within its corresponding time window. The overall coupling coefficient is obtained by weighting and summing the frequency band coupling coefficients of all frequency bands according to the proportion of power of each frequency band in the total power.

4. The control optimization method for an inverter according to claim 1, characterized in that, The overall coupling coefficient is input into the enhanced adaptive sliding mode controller to calculate the initial control signal, which includes: The sliding surface function value is obtained by comparing the overall coupling coefficient with the preset sliding surface function reference value; pass Calculate the initial control signal; where u is the initial control signal, k is the gain coefficient of the sliding mode reaching law (greater than 0), s is the sliding surface function value, and d is the total disturbance value estimated in real time by the disturbance observer. The weights are greater than 0, and sgn(s) is the sign function. Let be the rate of change of the disturbance.

5. The control optimization method for an inverter according to claim 1, characterized in that, The damping compensation signal is obtained by calculating the difference between the initial control signal and the output current signal using a damping compensation algorithm. Step 1: Align the initial control signal and the output current signal according to a time window to obtain a signal group in the time sequence; Step 2: Calculate the instantaneous difference of the signal group to generate an instantaneous difference sequence; Step 3: Perform candidate sliding window averaging filtering on the instantaneous difference sequence to obtain a candidate filtered instantaneous difference sequence; Step 4: Calculate the average difference within each window of the candidate filter instantaneous difference sequence using a second preset sliding window to obtain the trend value of change; Step 5: If the trend value is greater than or equal to the preset trend value, adjust the candidate sliding window as a new candidate sliding window, and return to step 3 until the trend value is less than the preset trend value to finally obtain the filtered instantaneous difference sequence. Step 6: Analyze the dynamic changes of the filtered instantaneous difference sequence and adaptively adjust the damping compensation coefficient to obtain the initial damping compensation signal; Step 7: Limit the amplitude of the preliminary damping compensation signal to obtain the damping compensation signal.

6. A control optimization system for an inverter, characterized in that, The system includes: The acquisition module is used to acquire the inverter's current and voltage signals over a time series. The harmonic power component generation module is used to calculate the instantaneous power of the current signal and the voltage signal to obtain the harmonic power components through a power observation model. The power observation model is an improvement based on the LSTM model, and the specific improvements include: constructing a feature enhancement layer between the input layer and the hidden layer of the LSTM model; replacing the activation function used in the hidden layer of the LSTM model with an improved EIU activation function; and iteratively optimizing the parameters of the power observation model through a quantum particle swarm optimization algorithm. The fundamental power component generation module is used to calculate the fundamental power component from the current signal and the voltage signal; The overall coupling coefficient generation module is used to couple the fundamental power component and the harmonic power component to obtain the overall coupling coefficient; The initial control signal calculation module is used to input the overall coupling coefficient into the enhanced adaptive sliding mode controller to calculate the initial control signal; The damping compensation signal calculation module is used to calculate the difference between the initial control signal and the output current signal using a damping compensation algorithm to obtain the damping compensation signal. The final control signal generation module is used to superimpose the damping compensation signal and the initial control signal to obtain the final control signal; A control module is used to control the operation of the inverter using the final control signal.

7. A control optimization system for an inverter according to claim 6, characterized in that, The harmonic power component calculation module includes: A sub-signal generation module is used to divide the current signal and the voltage signal into sub-current signals and sub-voltage signals by a preset time interval; The sub-signal group generation module is used to combine the sub-current signal and the sub-voltage signal according to the timestamp to obtain the sub-signal group; A multi-scale feature vector generation module is used to input the sub-signal group into the input layer of the power observation model, and to obtain a multi-scale feature vector by performing sub-band decomposition and dynamic weight allocation on the sub-signal group through the multi-scale wavelet attention mechanism of the feature enhancement layer. A deep feature vector generation module is used to input the multi-scale feature vector into the hidden layer of the power observation model to obtain a deep feature vector; The instantaneous power calculation module is used to integrate and calculate the instantaneous power by using the linear activation function of the output layer to calculate the deep feature vector; The harmonic power component module is used to perform spectral analysis on the instantaneous power corresponding to the sub-signal group and separate the harmonic power components.

8. A control optimization system for an inverter according to claim 6, characterized in that, The overall coupling coefficient generation module includes: The sub-component set generation module is used to decompose the fundamental power component and the harmonic power component into fundamental power sub-component sets and harmonic power sub-component sets by wavelet packet transform in different frequency bands. The power sub-component group module is used to combine the target fundamental power sub-component and the target harmonic power sub-component in the same frequency band to obtain a power sub-component group; the target fundamental power sub-component is any one of the fundamental power sub-components in the set of fundamental power sub-components; the target harmonic power sub-component is any one of the harmonic power sub-components in the set of harmonic power sub-components; An effective sequence generation module is used to align the power sub-component groups according to time markers and calculate the effective values ​​to obtain the fundamental power effective value sequence and the harmonic power effective value sequence; The local coupling coefficient calculation module is used to calculate the correlation coefficient of the fundamental power RMS value sequence and the harmonic power RMS value sequence through a first preset sliding window to obtain a local coupling coefficient set. The frequency band coupling coefficient calculation module is used to calculate the frequency band coupling coefficient of each power sub-component by weighting the set of local coupling coefficients obtained by each power sub-component according to the composite power energy within its corresponding time window. The overall coupling coefficient calculation module is used to calculate the overall coupling coefficient by weighting and summing the frequency band coupling coefficients of all frequency bands according to the proportion of power of each frequency band in the total power.

9. A control optimization system for an inverter according to claim 6, characterized in that, The initial control signal calculation module includes: The sliding surface function value calculation module is used to compare the overall coupling coefficient with the preset sliding surface function reference value to calculate the sliding surface function value; The initial control signal generation module is used to generate the initial control signal through... Calculate the initial control signal; where u is the initial control signal, k is the gain coefficient of the sliding mode reaching law (greater than 0), s is the sliding surface function value, and d is the total disturbance value estimated in real time by the disturbance observer. The weights are greater than 0, and sgn(s) is the sign function. Let be the rate of change of the disturbance.

10. A control optimization system for an inverter according to claim 6, characterized in that, The damping compensation signal calculation module includes: The signal group generation module is used to align the initial control signal and the output current signal with a time window to obtain a signal group in the time series. The instantaneous difference sequence generation module is used to calculate the instantaneous difference of the signal group and generate an instantaneous difference sequence. The candidate sequence generation module is used to perform candidate sliding window average filtering on the instantaneous difference sequence to obtain a candidate filtered instantaneous difference sequence; The trend value calculation module is used to calculate the average difference within each window of the candidate filtered instantaneous difference sequence through a second preset sliding window to obtain the trend value. The judgment module is used to adjust the candidate sliding window as a new candidate sliding window if the change trend value is greater than or equal to the preset change trend value, and return to the candidate sequence generation module until the change trend value is less than the preset change trend value to finally obtain the filtered instantaneous difference sequence. An initial damping compensation signal generation module is used to analyze the dynamic changes of the filtered instantaneous difference sequence and adaptively adjust the damping compensation coefficient to obtain an initial damping compensation signal. An amplitude limiting module is used to limit the amplitude of the preliminary damping compensation signal to obtain a damping compensation signal.

Images