An adaptive control method and system for an inverter

Abstract

The application discloses an adaptive control method and system of an inverter, and the method comprises the following steps: acquiring current operation sampling data, working condition parameters and device rated parameters of an arc load in real time; performing fast Fourier transform to obtain frequency spectrum characteristic data; based on an arc load model library, performing optimization parameter adjustment to obtain preliminary control parameters; inputting the current operation sampling data into a pre-trained support vector machine model to obtain target control parameters; performing load adjustment based on a fuzzy logic control algorithm to obtain adjusted current operation sampling data; performing distortion analysis to obtain an adjustment result, and adjusting the preliminary control parameters to obtain initial control parameters; and performing working condition analysis and adjustment according to the environmental temperature, the device rated parameters and the initial control parameters to obtain final control parameters. The method realizes adaptive control of the inverter on the arc load, and realizes sensitive and accurate control of arc stable combustion.

Description

Technical Field

[0001] This invention relates to the field of inverter control technology, and in particular to an adaptive control method and system for inverters. Background Technology

[0002] Currently, arc welding inverter power supplies are widely used in various industrial applications, especially in scenarios requiring matching with arc loads, such as welding and arc heating systems. However, the dynamic characteristics of arc loads are extremely complex and influenced by various factors, such as electrode materials, electrode spacing, gas composition, and gas flow rate. This causes the voltage and current characteristics of the arc to change continuously, posing a significant challenge to inverter control.

[0003] Existing arc welding inverters often employ fixed-parameter control, such as PID gain and proportional gain, which are optimized only based on specific operating conditions. However, the voltage-current characteristics of arc loads exhibit significant nonlinearity, time-varying nature, and randomness. The negative resistance characteristic of the arc (voltage decreases as current increases) conflicts with the linear assumptions of fixed-parameter control, easily leading to positive feedback instability. Furthermore, the dynamic behavior of the arc is affected by various factors. Once the arc characteristics change, such as electrode ablation leading to increased spacing, the original fixed parameters, such as proportional gain, cannot compensate for the new error characteristics. Excessively high proportional gain may cause oscillations, while excessively low proportional gain leads to response lag.

[0004] Therefore, traditional inverter control methods cannot adapt to the dynamic changes of arc load, resulting in insufficient real-time control accuracy and insufficient stability of arc combustion. Summary of the Invention

[0005] This invention provides an adaptive control method and system for inverters to achieve adaptive control of arc loads, solving the problems of insufficient sensitivity of precise control and unstable arc combustion.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides an adaptive control method for an inverter, comprising:

[0007] S1. Obtain the current operating sampling data, operating condition parameters, and device rated parameters of the arc load; wherein, the current operating sampling data includes real-time voltage and real-time current; the operating condition parameters include ambient temperature, solder supply rate, and solder parameters;

[0008] S2. Perform a Fast Fourier Transform on the currently running sampled data to obtain spectral characteristic data;

[0009] S3. Based on the spectral characteristic data, the current arc load type is matched using a preset arc load model library, and a preliminary set of control parameters is extracted from the matched model.

[0010] S4. Input the preliminary control parameter set, the operating condition parameters, and the solder supply acceleration parameters into the pre-trained support vector machine model to obtain the target control parameters;

[0011] S5. Input the target control parameters into a preset fuzzy logic controller for fuzzification processing to obtain fuzzy control quantities, and defuzzify the fuzzy control quantities to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions;

[0012] S6. Monitor the output voltage and output current waveforms of the inverter in real time, use wavelet transform to detect the distortion components in the waveforms, and judge the stability of arc combustion based on the distortion components.

[0013] S7. When unstable arc combustion is detected, the fluctuation amplitude is calculated based on the rated parameters of the device to determine the current operating condition type.

[0014] S8. Based on the current operating condition type, search the preset operating condition parameter mapping table to obtain the optimal parameter combination, and correct the current control parameters according to the optimal parameter combination to ensure that the output characteristics of the inverter remain stable under different operating conditions.

[0015] In one optional implementation, the step of matching the current arc load type using a preset arc load model library based on the spectral characteristic data, and extracting a preliminary control parameter set from the matched model, includes:

[0016] Spectral feature extraction is performed on the spectral characteristic data to obtain a spectral feature vector;

[0017] The similarity between the spectral feature vector and the model feature vector in the preset arc load model library is calculated to obtain the similarity; wherein, each arc load model includes a model feature vector and preliminary control parameters corresponding to the arc load type;

[0018] The model with the highest similarity is selected as the current arc load model;

[0019] Extract the control parameters corresponding to the current arc load model to obtain a preliminary control parameter set.

[0020] In one optional implementation, the step of calculating the similarity between the spectral feature vector and the model feature vector in a preset arc load model library to obtain the similarity includes:

[0021] The similarity between the spectral feature vector and the model feature vector is calculated using cosine similarity to obtain the similarity score.

[0022] The formula for calculating the similarity is as follows:

[0023]

[0024] In the formula, This represents the similarity, where n represents the dimension of the feature vector. This represents the i-th component of the spectral eigenvector. This represents the i-th component of the model's feature vector.

[0025] In one optional implementation, the training process of the support vector machine model includes:

[0026] Record historical control parameter sets, historical operating condition parameters, and corresponding historical target control parameters under different electric arc conditions;

[0027] The historical control parameter set and the historical ambient temperature are standardized and used as inputs, and the historical target control parameters are used as outputs to construct a support vector regression model and train the kernel function of the support vector regression model.

[0028] When the loss function value of the support vector regression model is less than the preset loss threshold, the trained support vector machine model is obtained.

[0029] Furthermore, the operating parameters include solder supply acceleration parameters, which correspond to acceleration duration and acceleration magnitude.

[0030] In one optional implementation, the step of inputting the target control parameters into a preset fuzzy logic controller for fuzzification processing to obtain fuzzy control quantities includes:

[0031] Determine the fuzzy set of target control parameters;

[0032] Calculate the membership degree of the target control parameter to the fuzzy set; wherein the membership degree calculation formula is as follows:

[0033]

[0034] in, Indicates membership degree; This represents the normalized target control parameters; , and All of these are pre-defined constants.

[0035] Based on the membership degree and combined with preset fuzzy rules, the fuzzy control quantity is obtained.

[0036] In one optional implementation, the step of using wavelet transform to detect distortion components in the waveform and determining the stability of arc combustion includes:

[0037] Wavelet transform is used to decompose the output voltage and output current waveforms and extract the distortion component;

[0038] The amplitude-frequency characteristic data are obtained from the distortion component to determine the state of electric arc combustion.

[0039] When the amplitude-frequency characteristic data exceeds the preset amplitude-frequency threshold, the arc combustion is determined to be unstable.

[0040] In one optional implementation, the step of calculating the fluctuation amplitude based on the device's rated parameters and determining the current operating condition type includes:

[0041] Calculate the current voltage fluctuation range and current fluctuation range based on the device's rated parameters;

[0042] The formulas for calculating the voltage fluctuation amplitude and the current fluctuation amplitude are as follows:

[0043]

[0044]

[0045] in, Indicates the voltage fluctuation range; Indicates the current voltage waveform peak value; Indicates the current voltage waveform valley value; Indicates the rated voltage of the device; Indicates the amplitude of current fluctuation; Indicates the current current waveform peak value; Indicates the current voltage waveform valley value; Indicates the rated voltage of the device;

[0046] The voltage fluctuation amplitude and current fluctuation amplitude are compared with the preset fluctuation amplitude range to determine the current operating condition type.

[0047] In one optional implementation, the step of correcting the current control parameters according to the optimal parameter combination to ensure that the inverter's output characteristics remain stable under different operating conditions includes:

[0048] The control parameters are corrected based on the optimal parameter combination to obtain the corrected control parameters;

[0049] The inverter settings are updated using the corrected control parameters to determine whether the output characteristics remain stable.

[0050] If the output characteristics do not remain stable, the optimal parameter combination is re-obtained according to the operating condition type, and the control parameters are adjusted again until the output characteristics remain stable.

[0051] In a second aspect, the present invention provides an adaptive control system for an inverter, comprising:

[0052] The data acquisition module is used to acquire the current operating sampling data, operating parameters, and device rated parameters of the arc load;

[0053] The data processing module is used to perform a fast Fourier transform on the currently running sampled data to obtain spectral characteristic data;

[0054] The initial parameter acquisition module includes a preset arc load model library and corresponding control parameters, and extracts the corresponding preliminary control parameter set based on the spectrum characteristic data to match the current arc load type.

[0055] The target parameter acquisition module includes a pre-trained support vector machine model, and takes the preliminary control parameter set, operating condition parameters and solder supply acceleration parameters input into the support vector machine model, and outputs the target control parameters.

[0056] The adjustment module includes a fuzzy logic controller with preset adjustment logic, and performs fuzzification processing on the target control parameters input to the fuzzy logic controller to obtain fuzzy control quantities, and defuzzifies the fuzzy control quantities to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions;

[0057] The stability judgment module is used to monitor the output voltage waveform and output current waveform of the inverter in real time, use wavelet transform to detect the distortion component in the waveform, and judge the stability and operating condition type of the arc combustion based on the distortion component.

[0058] The parameter correction module is used to search for the optimal parameter combination from a preset working condition parameter mapping table according to the current working condition type, and to correct the current control parameters according to the optimal parameter combination.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] This invention acquires data in real time and incorporates operating parameters into the training model. In particular, by inputting disturbances to the acceleration parameters of the solder supply, it can not only stably reproduce the original probabilistic welding dynamics, but also seek optimized control information under corresponding conditions after instantaneous changes. This provides accurate input-output information for subsequent analysis and control, ensuring that the control strategy can be adjusted based on the latest operating state. At the same time, it simplifies the amount of calculation in the actual control process and reduces the dependence on computing resources and cost investment.

[0061] This invention extracts a preliminary set of control parameters from the matched model and obtains more accurate target control parameters through a support vector machine model, laying the foundation for achieving precise and sensitive control.

[0062] This invention enables the inverter to control the output voltage and current more precisely through fuzzy logic control and defuzzification processing, allowing the inverter to flexibly adapt to changes in arc load and improve control accuracy and response speed.

[0063] This invention monitors the inverter's output waveform in real time to detect instability in arc combustion, providing a basis for subsequent adjustments and ensuring the stability of arc combustion.

[0064] In summary, this invention can effectively ensure that the inverter can cope with changes in arc load under different operating conditions, improve the accuracy of arc load control, and stabilize arc combustion. Attached Figure Description

[0065] Figure 1 This is a schematic flowchart of the adaptive control method for an inverter provided in the first embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram of the adaptive control system structure of the inverter provided in the second embodiment of the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0068] Reference Figure 1 The first embodiment of the present invention provides an adaptive control method for an inverter, comprising the following steps:

[0069] S1. Obtain the current operating sampling data, operating parameters, and device rated parameters of the arc load; wherein, the current operating sampling data includes real-time voltage and real-time current; the operating parameters include ambient temperature, solder supply speed, and solder parameters; and also include solder supply acceleration parameters, which correspond to acceleration duration and acceleration magnitude. This can both disturb the environment and create changes in the welding arc through solder supply acceleration, and obtain the optimal control matching parameters through statistical analysis and training of control parameters after acceleration disturbance.

[0070] S2. Perform a Fast Fourier Transform on the currently running sampled data to obtain spectral characteristic data;

[0071] S3. Based on the spectral characteristic data, the current arc load type is matched using a preset arc load model library, and a preliminary set of control parameters is extracted from the matched model.

[0072] S4. Input the preliminary control parameter set and the operating condition parameters into the pre-trained support vector machine model to obtain the target control parameters;

[0073] S5. Input the target control parameters into a preset fuzzy logic controller for fuzzification processing to obtain fuzzy control quantities, and defuzzify the fuzzy control quantities to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions;

[0074] S6. Monitor the output voltage and current waveforms of the inverter in real time, and use wavelet transform to detect the distortion component in the waveform to determine the stability of the arc combustion.

[0075] S7. When unstable arc combustion is detected, the fluctuation amplitude is calculated based on the rated parameters of the device to determine the current operating condition type.

[0076] S8. Based on the current operating condition type, search the preset operating condition parameter mapping table to obtain the optimal parameter combination, and correct the current control parameters according to the optimal parameter combination to ensure that the output characteristics of the inverter remain stable under different operating conditions.

[0077] In step S1, when acquiring the current operating sampling data of the arc load, voltage and current values ​​are first collected in real time using sensors. For example, a voltage sensor collects a real-time voltage of 220 volts, and a current sensor collects a real-time current of 10 amps. Next, the ambient temperature is acquired using a temperature sensor, for example, the current ambient temperature is 25 degrees Celsius. Simultaneously, the rated parameters of the devices are retrieved from the database or entered into the interactive system, including IGBT withstand voltage and heat dissipation coefficient, for example, the rated voltage of the IGBT is 1200V and the rated current is 100A. The synthesis of the above data can provide a basis for subsequent optimization to evaluate the stability and efficiency of the arc load.

[0078] In step S2, the currently running sampled data is subjected to a fast Fourier transform to obtain spectral characteristic data.

[0079] In the process of performing a Fast Fourier Transform (FFT) on the current operating sampling data of the arc load, the acquired real-time voltage and current signals are first digitized, assuming a sampling frequency of 1000 Hz and 1024 sampling points. The FFT converts the time-domain signal into a frequency-domain signal, obtaining the spectral characteristics of the voltage and current. For example, the voltage signal exhibits a fundamental frequency peak at 50 Hz with an amplitude of 220 volts, while harmonic components with amplitudes of 10 volts and 5 volts are detected at higher harmonic frequencies such as 150 Hz and 250 Hz, respectively. The current signal has a fundamental frequency amplitude of 10 amperes at 50 Hz, and harmonic amplitudes of 5 amperes and 3 amperes at 150 Hz and 250 Hz, respectively. These spectral characteristics allow for the identification of harmonic sources and their impact on the load system, providing data support for subsequent harmonic suppression and load optimization.

[0080] In step S3, based on the spectral characteristic data, a preset arc load model library is used to match the current arc load type, and a preliminary set of control parameters is extracted from the matched model, including:

[0081] S31, extract spectral features from the spectral characteristic data to obtain a spectral feature vector;

[0082] S32, calculate the similarity between the spectral feature vector and the model feature vector in the preset arc load model library to obtain the similarity; wherein, each arc load model includes a model feature vector and preliminary control parameters corresponding to the arc load type;

[0083] S33, Select the model with the highest similarity as the current arc load model;

[0084] S34, extract the control parameters corresponding to the current arc load model to obtain a preliminary control parameter set.

[0085] In step S31, spectral features are extracted from the spectral characteristic data to obtain a spectral feature vector.

[0086] It's worth noting that spectral feature extraction from spectral characteristic data is the process of transforming complex spectral data into a quantifiable feature vector. For example, assuming features are extracted from the spectral characteristic data of an arc load, we can focus on the fundamental frequency amplitude, harmonic amplitude, and the statistical characteristics of the frequency distribution. One possible implementation involves first analyzing the voltage spectrum data to extract the fundamental frequency amplitude of 220 volts at 50 Hz, the harmonic amplitudes of 10 volts at 150 Hz, and the harmonic amplitudes of 5 volts at 250 Hz. Then, combining this with the mean and variance of the frequency distribution, a feature vector containing five elements is formed. The advantage of this method is that it condenses the key information in the spectrum, facilitating subsequent processing.

[0087] In step S32, the similarity between the spectral feature vector and the model feature vector in the preset arc load model library is calculated to obtain the similarity; wherein, each arc load model includes a model feature vector and preliminary control parameters corresponding to the arc load type; including:

[0088] The similarity between the spectral feature vector and the model feature vector is calculated using cosine similarity to obtain the similarity score.

[0089] The formula for calculating the similarity is as follows:

[0090]

[0091] In the formula, This represents the similarity, where n represents the dimension of the feature vector. This represents the i-th component of the spectral eigenvector. This represents the i-th component of the model's feature vector.

[0092] The similarity calculation between the spectral feature vector and the feature vectors of models in the arc load model library aims to find the best-matching model. Specifically, cosine similarity can be used as the calculation method. For example, if the extracted feature vector is [220, 10, 5, 50, 2], and a model in the library has a feature vector of [230, 12, 4, 48, 1.8], the similarity between the two can be calculated to obtain a value such as 0.99.

[0093] In step S33, the model with the highest similarity is selected as the current arc load model.

[0094] It should be noted that each model in the model library pre-stores its corresponding feature vectors and control parameters. For example, this model might correspond to the "TIG welding model" type, and the control parameter set includes static characteristic parameters, dynamic response parameters, and stability protection parameters. This method can quickly identify the load type.

[0095] Selecting the model with the highest similarity as the current arc load model is the core of the entire process. In one embodiment, assuming the model library contains three models: TIG welding model, MIG welding model, and manual welding model, calculations show that the "TIG welding model" has the highest similarity, reaching 0.95, while the other models are only 0.85 and 0.70. Preferably, the model corresponding to 0.95 is selected as the current model. The advantage of this selection is that it is data-driven, ensuring the accuracy of the matching and providing a reliable foundation for subsequent control.

[0096] In step S34, the control parameters corresponding to the current arc load model are extracted to obtain a preliminary control parameter set.

[0097] Extracting the control parameters corresponding to the current arc load model to form a preliminary control parameter set is a crucial step in applying the analysis results to practice. For example, if the current arc load model is a TIG welding model, the included control parameter set might have a base current of 5-20A, an open-circuit voltage of 60-80V, and a peak current of 50-300A. In one embodiment, if the ambient temperature is 30 degrees Celsius and the solder supply speed level is 1, the parameter set may need further fine-tuning to adapt to actual operating conditions. The advantage of this method is that it utilizes the experience of the preset model while combining real-time data to improve applicability. During similarity calculation, the feature vectors in the model library will also contain different dimensions to ensure comprehensive matching. The final extracted control parameter set can effectively guide the equipment to adjust its operating state and improve system stability.

[0098] In step S4, the preliminary control parameter set and the operating condition parameters are input into the pre-trained support vector machine model to obtain the target control parameters.

[0099] The training process of the support vector machine model includes:

[0100] Record historical control parameter sets, historical operating condition parameters, and corresponding historical target control parameters under different electric arc conditions;

[0101] The historical control parameter set and the historical operating condition parameters are standardized and used as input, and the historical target control parameters are used as output. A support vector regression model is constructed and the kernel function of the support vector regression model is trained.

[0102] When the loss function value of the support vector regression model is less than the preset loss threshold, the trained support vector machine model is obtained.

[0103] It should be noted that the training process of the support vector machine model begins with data collection, selecting 1000 sets of historical data. Each set of data includes a set of control parameters, operating condition parameters, and corresponding target control parameters.

[0104] These data undergo preprocessing, including cleaning and standardization. For example, the switching frequency is normalized from 20 kHz to a range of 0 to 1, and the ambient temperature is mapped from 30 degrees Celsius to a range of 0 to 1. Next, a Gaussian kernel function is used for data mapping, with the kernel parameter γ set to 1 and the penalty coefficient C set to 10. During training, the model uses this data to find the optimal hyperplane in high-dimensional space, optimizing the parameters by minimizing the objective function.

[0105] For example, during training, the model found that when the ambient temperature exceeds 35 degrees Celsius, the switching frequency needs to be reduced to 18 kHz to maintain system stability. After adding positive or negative acceleration level 1 for 0.2 seconds, the model training found that the switching frequency also needs to be increased / decrease by 10 kHz to maintain system stability. Furthermore, the model optimized the PID parameters in the stability protection parameters from [8,2,1] to [85,22,12] using cross-validation, while also adjusting the harmonic suppression coefficient from 9 to 8. The entire training process is performed iteratively using the gradient descent algorithm. After each iteration, the loss function is calculated and the model parameters are updated, ultimately resulting in a model that can dynamically adjust control parameters based on input characteristics. This process ensures that the model can output the optimal control strategy based on real-time data in practical applications, improving the system's operating efficiency and stability.

[0106] It is worth noting that the initial control parameter set [8, 2, 5, 9] and the ambient temperature of 25 degrees Celsius were input into a pre-trained support vector machine (SVM) model. The initial control parameter set could include the output voltage setpoint, output current setpoint, switching frequency, and PID parameters. The model optimized the parameters based on a radial basis function (RBF) kernel, with the kernel parameter γ set to 1 and the penalty coefficient C set to 100. The SVM model learned the nonlinear relationship between the parameter set and the ambient temperature through the training data, outputting the target control parameters [85, 15, 55, 92]. During model training, a dataset containing 1000 samples was used, with 80% used for training and 20% for validation. The model achieved an accuracy of 95% on the validation set. The SVM model ensured the stability and reliability of the target control parameters by maximizing the classification margin. When inputting the initial control parameter set, the model first normalized the parameters, mapping the parameter values ​​to the [0, 1] interval, and then used a kernel function to map the data to a high-dimensional space for linear classification. Finally, the target control parameters output by the model are denormalized to restore them to the actual parameter range.

[0107] In this way, the support vector machine model can dynamically adjust control parameters according to changes in ambient temperature, ensuring stable operation of the system under different conditions.

[0108] In step S5, the target control parameters are input into a preset fuzzy logic controller for fuzzification to obtain fuzzy control quantities, and the fuzzy control quantities are defuzzified to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions.

[0109] The step of inputting the target control parameters into a preset fuzzy logic controller for fuzzification processing to obtain fuzzy control quantities includes:

[0110] Determine the fuzzy set of target control parameters;

[0111] Calculate the membership degree of the target control parameter to the fuzzy set; where the membership degree calculation formula is as follows:

[0112]

[0113] where, represents the membership degree; represents the normalized target control parameter; , and are all preset constants.

[0114] Obtain the fuzzy control quantity according to the membership degree and in combination with the preset fuzzy rules.

[0115] It should be noted that determining the fuzzy set of the target control parameter is the basis of fuzzy logic control.

[0116] For example, for the output voltage of an inverter, three fuzzy sets of "low", "medium", and "high" can be defined. Specifically, if the output voltage range is 0 - 400V, then 0 - 150V can be classified as "low", 100 - 300V as "medium", and 250 - 400V as "high". Setting sets with overlapping regions is also one of the characteristics of fuzzy logic.

[0117] Exemplarily, for the fuzzy set of "medium", the three vertex coordinates of the triangular function can be set as (100, 0), (200, 1), (300, 0). When the input voltage is 180V, its membership degree to the "medium" set can be calculated to be approximately 0.8 through the formula. This method is simple and intuitive, with high calculation efficiency and is suitable for real - time control systems.

[0118] Obtaining the fuzzy control quantity according to the membership degree and preset rules is a key step. When performing the fuzzification process, the control rule base plays a key role. The rule base consists of a series of fuzzy control rules based on the working conditions. For example, "If the current is high and the temperature is high, then the output arc voltage increases"; "If the current is low and the temperature is low, then the output arc current decreases". These rules define how to adjust the output control quantity according to the input data. In actual operation, the system processes the fuzzified input data according to these rule bases to obtain the fuzzy control quantity, and this process provides a basis for the generation of subsequent control instructions.

[0119] It can be understood that this fuzzy control method can effectively handle the uncertainty and non - linear characteristics of the system, improve the control accuracy and response speed of the inverter. By reasonably setting the fuzzy sets, selecting appropriate membership degree functions, and formulating appropriate fuzzy rules, the stable regulation of the inverter output can be achieved, reducing harmonic interference and improving the power quality.

[0120] Defuzzification is a crucial step in the fuzzy control process, aiming to convert fuzzy control quantities into specific numerical control commands for the inverter to execute. A commonly used method for defuzzification is the center-average method, which calculates the output value by taking the weighted average of all control quantities.

[0121] For example, the output of a fuzzy control rule is also a fuzzy set (e.g., "increase arc voltage"), which may correspond to... Each rule has its own activation strength, which is determined by the minimum or maximum membership degree of the input target control parameter to the fuzzy set. Assuming the above... With a corresponding activation strength of 0.6, the membership degree of 3V is 0.6 and that of 4V is 0.2 after MIN truncation. The final control command obtained through weighted calculation is 2.6V. This value is the inverter adjustment command obtained after defuzzification. The numerical control command obtained through defuzzification will be transmitted to the inverter, which will dynamically adjust the arc load accordingly, thereby controlling key parameters such as arc current and voltage.

[0122] In one specific embodiment, after receiving the defuzzified control command, the inverter will perform corresponding dynamic adjustment operations on the arc load. Specifically, the inverter will adjust various parameters of the arc according to the output current, voltage, or temperature control commands to ensure that the arc process operates under optimal conditions. For example, when the arc current needs to be increased, the inverter will increase the output current by adjusting the current source or current regulator; if the arc voltage needs to be reduced, the inverter will reduce the voltage output accordingly. In addition, the inverter further optimizes the stability of the arc load by adjusting the gas flow rate or other relevant control parameters.

[0123] It should be noted that the dynamic adjustment of the arc load is a real-time process. The inverter gradually adjusts various parameters of the arc according to the input commands to maintain the arc in its optimal operating state. For example, when the arc load undergoes a sudden change, causing a sharp increase in current, the inverter will quickly adjust according to the pre-set control strategy to reduce the current output, thereby preventing the arc from overheating or becoming unstable. The system also provides feedback on the current operating sampling data after each adjustment and further adjusts the control parameters based on the feedback information.

[0124] In step S6, the output voltage waveform and output current waveform of the inverter are monitored in real time, and wavelet transform is used to detect the distortion component in the waveform to determine the stability of the electric arc combustion.

[0125] The method of using wavelet transform to detect distortion components in waveforms and determine the stability of arc combustion includes:

[0126] Wavelet transform is used to decompose the output voltage and output current waveforms and extract the distortion component;

[0127] The amplitude-frequency characteristic data are obtained from the distortion component to determine the state of arc combustion.

[0128] When the amplitude-frequency characteristic data exceeds the preset amplitude-frequency threshold, the arc combustion is determined to be unstable.

[0129] It should be noted that, specifically, by acquiring the inverter's output voltage and current in real time through sensors, waveform data reflecting the system's operating status can be obtained. For example, in an inverter operating at 50 Hz, the sensor acquires data 1000 times per second, obtaining continuous voltage and current waveforms. These waveforms often contain noise or distortion, requiring further processing to extract key information.

[0130] In one possible implementation, wavelet transform is used to decompose the waveform data. Wavelet transform can divide the signal into different frequency components, preserving both time-domain and frequency-domain characteristics, which facilitates distortion analysis. For example, the Daubechies wavelet basis is used to decompose the voltage waveform into 6-8 layers, with the low-frequency portion reflecting the fundamental frequency and the high-frequency portion containing distortion or transient interference. For instance, a sudden increase in amplitude detected in the high-frequency layer might originate from irregular jumps during arcing.

[0131] Specifically, extracting amplitude-frequency characteristic data from the distortion component is a crucial step in determining the state of the electric arc. This amplitude-frequency characteristic data includes amplitude, frequency distribution, and duration. For example, in a certain decomposition, the amplitude of the high-frequency distortion component reaches 2 volts, the duration is 5 milliseconds, and the frequency is concentrated above 200 Hz. These data, compared with the baseline during normal operation, reflect abnormalities in arc combustion. Preferably, the amplitude can be analyzed from multiple perspectives, such as peak value, rate of change, and periodicity. In one embodiment, if the peak value is 50% higher than the historical average, the rate of change exceeds 0.5 volts per millisecond, and there is no obvious periodicity, it indicates arc instability. This multi-faceted analysis ensures sufficient basis for judgment and avoids misjudgment based on a single indicator.

[0132] It should be noted that the setting of the preset threshold directly affects the accuracy of the judgment. In one embodiment, the threshold is set based on historical data statistics. For example, the upper limit of the high-frequency distortion amplitude is 1.5 volts, and the duration does not exceed 3 milliseconds. If the amplitude-frequency characteristic data exceeds this range, such as an amplitude of 2 volts lasting for 5 milliseconds, the arc combustion is determined to be unstable; if the amplitude-frequency characteristic data does not exceed this range, the arc combustion is determined to be stable.

[0133] For example, during one operation, the distorted waveform captured by the sensor showed that the amplitude frequently exceeded the limit and the duration was prolonged. Combined with frequency analysis, it was confirmed that the proportion of high-frequency components was increased, which is consistent with the physical characteristics of arc runaway. From another perspective, if only the amplitude exceeds the limit but the duration is short, it may only be a momentary interference. Multi-dimensional verification enhances reliability.

[0134] In one possible implementation, the wavelet transform can be implemented using a digital signal processor, offering strong real-time performance. For example, the processor completes the decomposition every 10 milliseconds, outputs the distortion features, compares them with a threshold, and uses the result directly for subsequent control. This method is easy to integrate into inverter systems, enabling timely detection of arcing issues. For instance, if arcing instability is detected, the system can trigger a protection mechanism to adjust operating parameters to restore stability. From a business perspective, arcing often affects inverter output quality; real-time monitoring and assessment can effectively reduce fault risks and improve equipment lifespan and safety.

[0135] In step S7, when unstable arc combustion is detected, the fluctuation amplitude is calculated based on the rated parameters of the device to determine the current operating condition type, including:

[0136] Calculate the current voltage fluctuation range and current fluctuation range based on the device's rated parameters;

[0137] The formulas for calculating the voltage fluctuation amplitude and the current fluctuation amplitude are as follows:

[0138]

[0139]

[0140] in, Indicates the magnitude of voltage fluctuation; Indicates the peak value of the real-time voltage waveform; Indicates the valley value of the real-time voltage waveform; Indicates the rated voltage of the device; Indicates the amplitude of current fluctuation; Indicates the peak value of the real-time current waveform; Indicates the valley value of the real-time current waveform; Indicates the rated current of the device;

[0141] The voltage fluctuation amplitude and current fluctuation amplitude are compared with the preset fluctuation amplitude range to determine the current operating condition type.

[0142] In inverter operation monitoring, calculating voltage and current fluctuation amplitudes is a crucial step in assessing system status. Understandably, voltage fluctuation amplitude typically reflects the degree of change in output voltage over a period of time, while current fluctuation amplitude reflects the stability of the load or system operation.

[0143] For example, in an inverter operating at 50 Hz, the voltage data collected by the sensor may vary around a reference value of 220 volts, reaching a maximum of 230 volts and a minimum of 215 volts. Simple analysis shows a fluctuation of approximately 6.8%, which is a percentage change relative to the reference value. Similarly, the current fluctuation can be calculated based on measured data; for example, with a reference current of 10 amps, a fluctuation range between 9.5 amps and 10.8 amps yields a fluctuation of 13%. This calculation method intuitively reflects the relative magnitude of the fluctuation, facilitating subsequent comparative analysis.

[0144] Specifically, comparing the fluctuation range with a preset range is the core step in determining the type of operating condition. The preset fluctuation range is typically set based on equipment design or historical operating data. For example, the voltage fluctuation range might be set to ±5%, and the current fluctuation range to ±10%. In one possible implementation, if the voltage fluctuation is 6.8% exceeding 5% and the current fluctuation is 13% exceeding 10%, the system might determine the current operating condition as "slightly abnormal."

[0145] It should be noted that this judgment relies on the joint verification of multi-dimensional data to avoid misjudgment due to a single exceedance. For example, if voltage fluctuation exceeds the limit but current fluctuation is within the range, it may only be an external power grid disturbance, rather than a problem with the equipment itself. Preferably, the operating condition type can be divided into several states such as normal, slightly abnormal, and severely abnormal. In one embodiment, a slightly abnormal condition may correspond to a voltage fluctuation of 6.8% and a current fluctuation of 13%. At this time, the system can still operate, but attention is required; if the voltage fluctuation rises to 10% and the current fluctuation reaches 20%, it may be judged as a severely abnormal condition, indicating that immediate intervention is required. For example, during a certain operation, the voltage fluctuation increased from 5% to 8%, and the current fluctuation increased from 8% to 15%. Combined with the analysis of the operation log, it was found that a sudden change in load was the main cause. This multi-faceted consideration ensures the reliability of the operating condition judgment and avoids misjudging a serious problem due to transient interference.

[0146] In one possible implementation, fluctuation amplitude calculation can be performed by smoothing the data with a digital filter, for example, taking the average value within a 10-millisecond window as a benchmark, and then calculating the fluctuation percentage. This method can reduce noise interference and improve data reliability. For example, in a certain data acquisition, the voltage instantaneously jumps to 235 volts, but after smoothing, the benchmark value stabilizes at 222 volts, and the fluctuation amplitude drops to 5.4%, which is still within the normal range. This method is easy to integrate into the inverter control system in real time, providing timely feedback on the operating condition type and providing a basis for subsequent adjustments. It is understandable that the setting of the preset range needs to be adjusted according to the specific application scenario. For example, in high-precision equipment, the voltage fluctuation range may be tightened to ±3%, while in ordinary industrial scenarios it can be relaxed to ±7%. In one embodiment, during the operation of an inverter, the voltage fluctuation is 6% and the current fluctuation is 12%. Compared with the relaxed ranges of ±7% and ±15%, it is judged as a normal operating condition; however, if evaluated according to the strict ranges of ±3% and ±10%, it is considered a slight anomaly.

[0147] This flexibility ensures the applicability of the monitoring methods, while the accurate classification of operating conditions helps to optimize equipment operation strategies and extend service life.

[0148] In step S8, the optimal parameter combination is obtained by searching a preset operating condition parameter mapping table according to the current operating condition type, and the current control parameters are corrected according to the optimal parameter combination to ensure that the output characteristics of the inverter remain stable under different operating conditions, including:

[0149] The control parameters are corrected based on the optimal parameter combination to obtain the corrected control parameters;

[0150] The inverter settings are updated using the corrected control parameters to determine whether the output characteristics remain stable.

[0151] If the output characteristics do not remain stable, the optimal parameter combination is re-obtained according to the operating condition type, and the control parameters are adjusted again until the output characteristics remain stable.

[0152] Among these steps, retrieving the corresponding parameter combination from a preset table based on the current operating condition is a fundamental part of inverter operation adjustment. It's understandable that preset tables are typically designed based on equipment characteristics, including mappings between operating condition types and parameters such as voltage and frequency. For example, under a "slightly abnormal" operating condition, the preset table might provide a parameter combination for fine-tuning the frequency to 49.8 Hz and raising the voltage reference to 225 volts. This method quickly matches optimal parameters using existing data, avoiding blind adjustments.

[0153] In one possible implementation, the control parameters are corrected based on the optimal parameters to determine the adjusted control value, which needs to be combined with actual operating data. For example, the original control value of the inverter is 220 volts and 50 Hz, while the optimal parameters suggest 225 volts and 49.8 Hz. The corrected control value is then set according to the new value. This adjustment is usually executed directly by the control module to ensure real-time performance. It should be noted that the correction process depends on the rationality of the parameter combination. If the preset table is poorly designed, the adjustment may deviate from the expected result.

[0154] Specifically, after updating the inverter settings with adjusted control values, the goal is to achieve stable output characteristics. Maintaining stable output characteristics means that fluctuations in key electrical parameters are within acceptable limits. For example, in a certain adjustment, if voltage fluctuation decreases from 6% to 3% and current fluctuation decreases from 12% to 8%, while the acceptable range for both voltage and current fluctuations is 5%, the output characteristics will tend to stabilize. The advantage of this method is that it quickly restores system stability and improves operating efficiency.

[0155] In one embodiment, if the updated output remains unstable, for example, if voltage fluctuations rise to 7%, the operating condition type needs to be re-examined, possibly changing from "mild anomaly" to "severe anomaly," thereby obtaining a more stringent parameter combination, such as reducing the frequency to 49.5 Hz. Preferably, if the output characteristics do not remain stable, the parameter combination is re-acquired and it is determined whether to correct it again; if the output characteristics remain stable, it indicates that the parameter correction is correct.

[0156] For example, if the preset range is voltage fluctuation ±5%, and during a certain operation the fluctuation reaches 6%, the system will re-match parameters according to the operating condition type "minor anomaly," such as when the voltage rises to 228 volts. This cyclic adjustment effectively copes with dynamic changes, ensuring that the output remains controllable. In one possible implementation, the inverter's operating state is determined by calculating the characteristic adjustment range through the corrected parameter combination, reflecting attention to detail.

[0157] For example, when the voltage is adjusted from 220 volts to 225 volts, the characteristic adjustment range is about 2.3%. Combined with the current fluctuation decreasing from 10% to 7%, the operating status can be determined as "approaching normal".

[0158] Specifically, during a particular operation, after adjustment, voltage fluctuations decreased to 4% and current fluctuations to 6%. Analysis of load changes and environmental factors confirmed stability. This multi-dimensional verification enhances the reliability of the judgment and helps to optimize operating strategies in a timely manner. It is understandable that the above method, through a combination of preset tables and dynamic correction, flexibly adapts to different operating conditions. For example, when an inverter experiences a sudden increase in load, the initial adjustment does not meet expectations, but by correcting the parameters again, the output characteristics return to the normal range. This approach not only improves equipment adaptability but also effectively extends its service life, ensuring continuous operation.

[0159] Reference Figure 2 The second embodiment of the present invention provides an adaptive control system for an inverter, comprising:

[0160] The data acquisition module is used to acquire the current operating sampling data, operating parameters, and device rated parameters of the arc load;

[0161] The data processing module is used to perform a fast Fourier transform on the currently running sampled data to obtain spectral characteristic data;

[0162] The initial parameter acquisition module includes a preset arc load model library and corresponding control parameters, and extracts the corresponding preliminary control parameter set based on the spectrum characteristic data to match the current arc load type.

[0163] The target parameter acquisition module includes a pre-trained support vector machine model, and takes the preliminary control parameter set and operating condition parameters input into the support vector machine model, and outputs the target control parameters.

[0164] The adjustment module includes a fuzzy logic controller with preset adjustment logic, and performs fuzzification processing on the target control parameters input to the fuzzy logic controller to obtain fuzzy control quantities, and defuzzifies the fuzzy control quantities to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions;

[0165] The stability assessment module is used to monitor the output voltage and output current waveforms of the inverter in real time. It uses wavelet transform to detect the distortion component in the waveform and to determine the stability and operating condition of the arc combustion.

[0166] The parameter correction module is used to find the optimal parameter combination from a preset operating condition parameter mapping table according to the current operating condition type, and to correct the current control parameters according to the optimal parameter combination to ensure that the output characteristics of the inverter remain stable under different operating conditions.

[0167] The data acquisition module may include:

[0168] Voltage sensor: A Hall effect sensor or a voltage divider resistor network is used to monitor the voltage signal of the arc load in real time.

[0169] Current sensor: Uses Rogowski coil or Hall current sensor to collect arc current.

[0170] Temperature sensor: Integrated thermocouple or digital temperature sensor (such as DS18B20) to monitor ambient temperature.

[0171] Communication modules: such as wired communication modules like CAN, RS485, or Ethernet, or wireless communication modules.

[0172] Interactive systems: such as touchscreens.

[0173] Protection design: TVS diodes and filter circuits are added to the front end of the sensor to suppress high-frequency noise from electric arcs and overvoltage surges.

[0174] The data processing module may include:

[0175] ADC converter: Selects a high-speed, high-precision ADC (such as 16-bit resolution, 1MSPS sampling rate) to convert analog signals into digital signals.

[0176] Signal conditioning circuit: includes amplification, filtering (low-pass filter cutoff frequency 500Hz~1kHz) and isolation circuit (optical isolation) to ensure signal purity.

[0177] The main control unit can be a high-performance DSP or ARM Cortex-M7, supporting floating-point operations to meet the real-time requirements of FFT, wavelet transform and SVM algorithms. It can also be configured with multiple chips for different modules.

[0178] The system architecture also includes the inverter's drive control section, such as IGBT drive circuits and overcurrent / overvoltage protection circuits.

[0179] It should be noted that the adaptive control system for an inverter provided in this embodiment of the invention is used to execute all the process steps of the adaptive control method for an inverter in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0180] 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. In particular, it should be noted that 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 for those skilled in the art.

Claims

1. An adaptive control method for an inverter, characterized in that, include: S1. Obtain the current operating sampling data, operating condition parameters, and device rated parameters of the arc load; wherein, the current operating sampling data includes real-time voltage and real-time current; the operating condition parameters include ambient temperature, solder supply rate, solder parameters, and solder supply acceleration parameters, and the acceleration parameters correspond to acceleration duration and acceleration magnitude; S2. Perform a Fast Fourier Transform on the currently running sampled data to obtain spectral characteristic data; S3. Based on the spectral characteristic data, the current arc load type is matched using a preset arc load model library, and a preliminary set of control parameters is extracted from the matched model. S4. Input the preliminary control parameter set, the operating condition parameters, and the solder supply acceleration parameters into the pre-trained support vector machine model to obtain the target control parameters; the training process of the support vector machine model includes: Record historical control parameter sets, historical operating condition parameters, and corresponding historical target control parameters under different electric arc conditions; The historical control parameter set and the historical operating condition parameters are standardized and used as input, and the historical target control parameters are used as output. A support vector regression model is constructed and the kernel function of the support vector regression model is trained. When the loss function value of the support vector regression model is less than the preset loss threshold, the trained support vector machine model is obtained. S5. Input the target control parameters into a preset fuzzy logic controller for fuzzification processing to obtain fuzzy control quantities, and defuzzify the fuzzy control quantities to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions; S6. Monitor the output voltage and output current waveforms of the inverter in real time, use wavelet transform to detect the distortion components in the waveforms, and judge the stability of arc combustion based on the distortion components. S7. When unstable arc combustion is detected, the fluctuation amplitude is calculated based on the rated parameters of the device to determine the current operating condition type. S8. Based on the current operating condition type, search the preset operating condition parameter mapping table to obtain the optimal parameter combination, and correct the current control parameters according to the optimal parameter combination to ensure that the output characteristics of the inverter remain stable under different operating conditions.

2. The adaptive control method for an inverter according to claim 1, characterized in that, Step S3 includes: Spectral feature extraction is performed on the spectral characteristic data to obtain a spectral feature vector; The similarity between the spectral feature vector and the model feature vector in the preset arc load model library is calculated to obtain the similarity; wherein, each arc load model includes a model feature vector and preliminary control parameters corresponding to the arc load type; The model with the highest similarity is selected as the current arc load model; Extract the control parameters corresponding to the current arc load model to obtain a preliminary control parameter set.

3. The adaptive control method for the inverter according to claim 2, characterized in that, The similarity between the spectral feature vector and the model feature vector is calculated using cosine similarity to obtain the similarity score. The formula for calculating the similarity is as follows: In the formula, This represents the similarity, where n represents the dimension of the feature vector. This represents the i-th component of the spectral eigenvector. This represents the i-th component of the model's feature vector.

4. The adaptive control method for an inverter according to claim 1, characterized in that, In step S5, the deriving of the fuzzy control quantity includes the following steps: Determine the fuzzy set of target control parameters; Calculate the membership degree of the target control parameter to the fuzzy set; wherein the membership degree calculation formula is as follows: in, Indicates membership degree; This represents the normalized target control parameters; , and All of these are pre-defined constants; Based on the membership degree and combined with preset fuzzy rules, the fuzzy control quantity is obtained.

5. The adaptive control method for an inverter according to claim 1, characterized in that, The method of using wavelet transform to detect distortion components in waveforms and determine the stability of electric arc combustion includes: Wavelet transform is used to decompose the output voltage and output current waveforms and extract the distortion component; The amplitude-frequency characteristic data are obtained from the distortion component to determine the state of arc combustion. When the amplitude-frequency characteristic data exceeds the preset amplitude-frequency threshold, the arc combustion is determined to be unstable.

6. The adaptive control method for an inverter according to claim 1, characterized in that, The step of calculating the fluctuation amplitude based on the rated parameters of the device and determining the current operating condition type includes: Calculate the current voltage fluctuation amplitude and current fluctuation amplitude based on the device's rated parameters; The formulas for calculating the voltage fluctuation amplitude and the current fluctuation amplitude are as follows: in, Indicates the voltage fluctuation range; Indicates the current voltage waveform peak value; Indicates the current voltage waveform valley value; Indicates the rated voltage of the device; Indicates the amplitude of current fluctuation; Indicates the current current waveform peak value; Indicates the current current waveform valley value; Indicates the rated current of the device; The voltage fluctuation amplitude and current fluctuation amplitude are compared with the preset fluctuation amplitude range to determine the current operating condition type.

7. The adaptive control method for an inverter according to claim 1, characterized in that, The step of correcting the current control parameters based on the optimal parameter combination to ensure that the inverter's output characteristics remain stable under different operating conditions includes: The control parameters are corrected based on the optimal parameter combination to obtain the corrected control parameters; The inverter settings are updated using the corrected control parameters to determine whether the output characteristics remain stable. If the output characteristics do not remain stable, the optimal parameter combination is re-obtained according to the operating condition type, and the control parameters are adjusted again until the output characteristics remain stable.

8. An adaptive control system for an inverter, characterized in that, Configured to implement the adaptive control method for the inverter as described in claim 1, comprising: The data acquisition module is used to acquire the current operating sampling data, operating parameters, and device rated parameters of the arc load; The data processing module is used to perform a fast Fourier transform on the currently running sampled data to obtain spectral characteristic data; The initial parameter acquisition module includes a preset arc load model library and corresponding control parameters, and extracts the corresponding preliminary control parameter set based on the spectrum characteristic data to match the current arc load type. The target parameter acquisition module includes a pre-trained support vector machine model, and takes the preliminary control parameter set, operating condition parameters and solder supply acceleration parameters input into the support vector machine model, and outputs the target control parameters. The adjustment module includes a fuzzy logic controller with preset adjustment logic, and performs fuzzification processing on the target control parameters input to the fuzzy logic controller to obtain fuzzy control quantities, and defuzzifies the fuzzy control quantities to generate adjustment instructions so that the inverter can adjust according to the adjustment instructions; The stability judgment module is used to monitor the output voltage waveform and output current waveform of the inverter in real time, use wavelet transform to detect the distortion component in the waveform, and judge the stability and operating condition type of the arc combustion based on the distortion component. The parameter correction module is used to search for the optimal parameter combination from a preset working condition parameter mapping table according to the current working condition type, and to correct the current control parameters according to the optimal parameter combination.

Images