Control method of high-power isolated power supply

Abstract

The invention relates to the technical field of power supply control, and particularly discloses a control method of a high-power isolated power supply, which realizes full-life-cycle intelligent management of an isolated power supply oil paper insulation system by constructing a complete technical chain from dielectric loss factor monitoring to PDIV drift inversion and then to insulation life prediction and operation control. The insulation state on-line monitoring unit collects the key insulation performance index of the dielectric loss factor in real time, the aging degree and the performance change of the oil paper insulation medium in the operation process can be accurately reflected, and compared with a traditional off-line detection method, the real-time performance and the accuracy of monitoring are greatly improved. The operation parameters of the equipment can be dynamically adjusted according to the real-time evaluation result of the insulation state, sudden faults caused by insulation failure are effectively avoided, and the safety and reliability of operation of the equipment are remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of power supply control and relates to a control method for a high-power isolated power supply. Background Art

[0002] In plateaus above 3,000 meters, the significantly lower atmospheric pressure causes the partial discharge inception voltage (PDIV) of oil-paper insulation systems to drop by 8%-15% compared to plains. This phenomenon cannot be detected under standard atmospheric pressure testing conditions at the factory, posing a potential insulation safety hazard during field operation. Due to the harsh plateau environment, field monitoring data is difficult to obtain, sparse, and with a low signal-to-noise ratio. Operations and maintenance personnel can only rely on secondary voltage, current, temperature, and a small amount of partial discharge pulse data for condition assessment. When the PDIV gradually approaches or even falls below the system operating voltage, the insulation system enters a critical discharge state, which can easily cause short-circuit failures, resulting in equipment damage and power outages. However, traditional monitoring methods based on fixed thresholds cannot effectively distinguish between weak partial discharge signals during normal operation and abnormal discharge signals in the critical discharge state, resulting in high false alarm and missed alarm rates. Furthermore, due to the lack of in-depth analysis of long-term operating data, existing technologies struggle to develop effective predictive models to proactively identify PDIV drift trends and insulation aging. In addition, the incompleteness of field data and the complexity of environmental interference factors make it impossible for the monitoring method based on a single electrical parameter to accurately reflect the actual status of the insulation system, and unable to ensure the safe and stable operation of high-power isolated power supplies in plateau areas. Summary of the Invention

[0003] In view of the above problems in the prior art, the present invention provides a control method for a high-power isolated power supply to solve the above technical problems.

[0004] In order to achieve the above-mentioned and other purposes, the technical solutions adopted by the present invention are as follows: The present invention provides a method for controlling a high-power isolated power supply, the method comprising: Step S1: Real-time acquisition of dielectric loss factor of the oil-paper insulation medium inside the isolated power supply is performed through an insulation status online monitoring unit to obtain dielectric loss online monitoring data; based on the dielectric loss online monitoring data, combined with a dielectric loss factor and partial discharge inception voltage drift correlation curve pre-calibrated through accelerated aging experiments, PDIV dynamic inversion calculation is performed to obtain current insulation PDIV drift data; Step S2: Based on the PDIV drift data, a time series coupling process is performed with historical operating environment parameters and a power load curve to obtain insulation performance time series degradation data; an insulation life remaining prediction model is established based on the insulation performance time series degradation data to obtain an insulation remaining life prediction result; Step S3: Controlling the operation of the isolated power supply based on the insulation remaining life prediction result.

[0005] As described above, the present invention provides a method for controlling a high-power isolated power supply, which has at least the following beneficial effects: The present invention achieves intelligent management of the entire life cycle of the oil-paper insulation system of the isolated power supply by constructing a complete technical chain from dielectric loss factor monitoring to PDIV drift inversion, and then to insulation life prediction and operation control. The dielectric loss factor, a key insulation performance indicator, is collected in real time through the online insulation status monitoring unit, which can accurately reflect the aging degree and performance changes of the oil-paper insulation medium during operation. Compared with traditional offline detection methods, the real-time and accuracy of monitoring are greatly improved. Secondly, based on the dielectric loss factor and PDIV drift correlation curve pre-calibrated through accelerated aging experiments, PDIV dynamic inversion calculation is performed, which effectively solves the technical problem that PDIV degradation in plateau environments is difficult to detect during factory testing, allowing the insulation performance of the equipment to be accurately predicted before it is put into field operation. Thirdly, by coupling PDIV drift data with historical operating environment parameters and power load curves in time series, a multi-dimensional analysis model for insulation performance degradation is established, which fully considers the impact of multiple environmental factors such as temperature, humidity, and load fluctuations on insulation aging, making the insulation life prediction results closer to actual operating conditions. Finally, the isolated power supply's operational control, based on the predicted insulation remaining life, shifts from passive maintenance to predictive maintenance. This allows for dynamic adjustment of equipment operating parameters based on real-time insulation status assessments, effectively avoiding sudden failures caused by insulation failure and significantly improving equipment safety and reliability. This closed-loop control strategy, driven by online monitoring data, not only extends equipment life and reduces maintenance costs, but also provides crucial technical support for the safe and stable operation of high-power electrical equipment in plateau regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0007] Figure 1 It is a schematic diagram of the connection of each step of the method of the present invention. DETAILED DESCRIPTION

[0008] The above contents described below in conjunction with the implementation of the present invention are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they shall fall within the scope of protection of the present invention.

[0009] Example 1, please refer to Figure 1 As shown, a control method for a high-power isolated power supply includes: Step S1: Real-time acquisition of dielectric loss factor of the oil-paper insulation medium inside the isolated power supply is performed through an insulation status online monitoring unit to obtain dielectric loss online monitoring data; based on the dielectric loss online monitoring data, combined with a dielectric loss factor and partial discharge inception voltage drift correlation curve pre-calibrated through accelerated aging experiments, PDIV dynamic inversion calculation is performed to obtain current insulation PDIV drift data; Step S1 includes: Step S11: performing millisecond-level current phase angle tracking on the oil-paper insulation medium through the insulation state online monitoring unit to obtain dielectric polarization current amplitude data within the power frequency cycle, and generating dielectric loss online monitoring data after Fourier baseband filtering; Step S12: performing environmental error compensation on the dielectric loss online monitoring data, using a preset temperature loss gradient matrix and humidity compensation factor to eliminate dielectric loss measurement deviation, and outputting corrected dielectric loss characteristic data; Step S13: Calling the dielectric loss factor and PDIV drift correlation curve library calibrated by the accelerated aging experiment, performing loss offset mapping matching on the corrected dielectric loss characteristic data, and establishing a dynamic quantization matrix of loss factor and voltage drift through bilinear interpolation operation of adjacent aging nodes; Step S14: Combined with the historical data of the aging cumulative rate of oil-paper insulation, the dynamic quantization matrix is ​​input into the preset aging effect time integral model, and a time deconvolution operation is performed on the partial discharge inception voltage based on the dielectric loss change gradient to output the current insulation PDIV drift data.

[0010] In an embodiment of the present invention, a high-precision current phase angle sensor mounted on an online insulation status monitoring unit is used to capture the polarization current waveform of the oil-paper insulation dielectric at a 50Hz power frequency at a fixed sampling frequency. Synchronous phase-locked amplification is then used to separate the dielectric polarization current and capacitive current components. The original current signal is truncated using a Hanning window function to eliminate spectral leakage. The fundamental frequency current amplitude phase difference is extracted using a fast Fourier transform. Combined with the reactive power reference value of a standard capacitor, online monitoring data for the dielectric loss factor is calculated. The monitoring data is then subjected to multi-dimensional environmental parameter correction based on the temperature-dissipation gradient matrix in the environmental compensation database. The oil temperature data on the top layer of the transformer tank is collected in real time using an embedded temperature sensor. The gradient matrix is ​​indexed at 0.1°C intervals to obtain the temperature compensation coefficient. A humidity compensation factor lookup table is used to perform secondary compensation using the relative humidity value measured by the dew point sensor. The original dielectric loss value is then subtracted from the temperature compensation offset and multiplied by the humidity correction factor to generate dielectric loss characteristic data that eliminates environmental interference. Then, in the accelerated aging correlation curve matching stage, the corrected dielectric loss characteristic value is imported into a correlation database containing 200 sets of aging samples. The spatiotemporal position of the dielectric loss value in the aging curve cluster is determined using a three-dimensional coordinate projection algorithm. Four adjacent aging nodes are selected to form a bilinear interpolation unit. The dielectric loss gradient slope is calculated along the time-domain aging axis, and the voltage attenuation corresponding to each 10% dielectric loss increment along the voltage drift axis. A dynamic quantization matrix of dielectric loss and voltage drift is generated through a weighted average of the two axes. Finally, when performing the time-domain deconvolution calculation, the historical aging cumulative rate spectrum from the equipment operation log is loaded. In the time integration model, every 6 minutes is set as the basic integration unit. The dielectric loss-voltage relationship parameters in the quantization matrix are time-series convolved with the aging rate. The reverse iterative method is used to decompose the contribution weight of the aging effect at different historical moments to the current voltage drift. The early aging effect is exponentially weighted attenuated by establishing an attenuation memory function, and the PDIV drift data at the current moment is finally output.

[0011] Step S13 includes: Calling the dielectric loss factor and PDIV drift correlation curve library established by the accelerated aging experiment, dynamically matching the corrected dielectric loss characteristic data during the aging stage, and determining the coordinate projection interval of the current dielectric loss value on the correlation curve; Two adjacent aging stage nodes are selected within the coordinate projection interval, and the corresponding dielectric loss factor threshold and voltage drift reference value of the nodes are extracted to generate aging stage node matching data. Based on the aging stage node matching data, a bidirectional interpolation operation is performed, in which the dielectric loss change gradient is interpolated along the time aging axis and the drift amount is linearly compensated along the voltage drift axis, and the quantitative relationship parameters after interpolation compensation are output. Cross-dimensional weighted fusion is performed on the quantization relationship parameters after interpolation compensation to generate a dynamic quantization matrix of dielectric loss factor and voltage drift.

[0012] In an embodiment of the present invention, a cluster of three-dimensional correlation curves containing the corresponding relationship between dielectric loss and PDIV drift under different aging cycles is first retrieved from the accelerated aging experiment database. The corrected dielectric loss characteristic value is projected into three-dimensional space along the time aging axis, the dielectric loss change axis, and the voltage drift axis. The projection area of ​​the dielectric loss value on the aging surface is determined using the nearest neighbor density estimation algorithm, and a cubic search space containing eight adjacent aging sample points is constructed with the projection point as the center. Subsequently, based on the thin plate spline interpolation principle, the time attenuation gradient of the current dielectric loss value and the adjacent aging node is calculated along the time aging axis, and the correlation coefficient between the dielectric loss increment and the voltage drift is calculated along the dielectric loss change axis. The two nearest aging stage nodes are screened out using dual constraints, and their dielectric loss threshold boundary values ​​and corresponding voltage drift reference values ​​are extracted respectively. During bidirectional interpolation, a nonlinear interpolation method is used for the time-aging axis. Based on the second-order derivative of the aging curve, a logarithmic decay function is used to calculate the dielectric loss gradient interpolation coefficient between two aging nodes. For the voltage drift axis, a piecewise linear interpolation strategy is used. Based on the linear regression slope of the dielectric loss increment and voltage drift, drift compensation is calculated based on the current dielectric loss value's relative position relative to the adjacent node. After completing bidirectional interpolation, a weighting model for dielectric loss and voltage drift is established, assigning a 60% weight to the interpolated results for the time-aging axis and a 40% weight to the interpolated results for the voltage drift axis. A weighted average algorithm is then used to generate a cross-dimensional quantized relational parameter matrix. A parameter anomaly check mechanism is implemented. If the dispersion of the bidirectional interpolation results exceeds 15%, the calculation automatically switches to the cubic spline interpolation algorithm for recalculation. A dielectric loss change rate correction factor is introduced in the weighting stage, assigning an additional 20% gain to voltage drift during the rapid degradation phase to ensure parameter accuracy during the accelerated aging period. The final generated dynamic quantization matrix contains dielectric loss change step size, corresponding voltage drift and confidence interval data.

[0013] Step S14 includes: Extracting a historical data set of the aging cumulative rate of the oil-paper insulation material, loading the dynamic quantitative relationship matrix generated in step S13, and establishing an input parameter set for the time integral of the aging effect; Import the input parameter group into the preset aging effect time integral model, perform iterative calculation of the cumulative dielectric loss gradient in hourly time units, and output the cumulative dielectric loss drift sequence on the time axis; Based on the cumulative dielectric loss drift sequence, the partial discharge inception voltage is subjected to reverse time decoupling. The exponential decay weighting algorithm is used to separate the voltage drift contributions of different aging stages and generate time-based voltage drift components. The time domain normalization processing is performed on the periodic voltage drift component, and the current insulation PDIV drift data is output through component amplitude superposition and phase compensation calculation.

[0014] In an embodiment of the present invention, first, a historical data set of the aging rate of oil-paper insulation materials over the past five years is called from the insulation life management system. The data set contains the dielectric loss change gradient, load rate, and ambient temperature and humidity parameters recorded every hour. These historical data are aligned and matched with the dynamic quantization relationship matrix in time and space, and the sliding time window technology is used to divide the data blocks into 72-hour units. An input parameter group containing the initial value of dielectric loss, aging gradient, and voltage drift rate is established in each data block, and the numerical jump caused by inconsistent data acquisition intervals is eliminated by cubic spline interpolation. When the parameter group is imported into the aging effect time integral model, the adaptive step-size Runge-Kutta method is used for iterative calculation: in each hour time unit, according to the corresponding dielectric loss-voltage drift coefficient in the dynamic quantization matrix, the contribution value of the dielectric loss increment to the voltage drift in this period is calculated, and the cumulative drift is updated in real time through the integral accumulator. The convergence condition is set to that the relative error of the results of two adjacent iterations is less than 0.05%. When the model iterates to the current time node, the cumulative dielectric loss drift sequence on the continuous time axis is output. When performing reverse time decoupling, a deconvolution model based on an exponential decay kernel function is constructed, tracing back along the time axis to determine the weight of each historical period's impact on the current voltage drift. A 90% weight coefficient is retained for the aging effect of the last 30 days, and the weight distribution decreases according to the natural logarithm as the time span increases. The voltage recovery component caused by the material relaxation effect is simultaneously considered in the decoupling calculation, and an improved Wiener filter algorithm is used to separate the net voltage drift contribution value of each time period. When completing the time domain normalization processing, the voltage drift components of each time period are first normalized to the maximum value, and the drift amounts of different magnitudes are uniformly mapped to the range of 0-1. Then, based on the phase spectrum characteristics of the cumulative dielectric loss sequence, the Hilbert transform is used to extract the phase delay parameters of each component. The time domain misalignment error is eliminated by constructing a phase compensation vector. Finally, the standardized amplitude and phase correction values ​​are vector-superimposed to output the current insulation PDIV drift data.

[0015] Step S2: Based on the PDIV drift data, a time series coupling process is performed with historical operating environment parameters and a power load curve to obtain insulation performance time series degradation data; an insulation life remaining prediction model is established based on the insulation performance time series degradation data to obtain an insulation remaining life prediction result; Step S2 includes: Performing time stamp alignment processing on the PDIV drift data, historical ambient temperature data, and power load curve to establish a time-scale unified data set based on the equipment operation cycle; Based on a unified time-scale data set, coupled environmental and electrical stress analysis is performed. A composite aging stress parameter sequence is generated through a temperature load factor weighting algorithm and power fluctuation gradient compensation calculation. The time-varying response of the insulation medium is modeled based on the composite aging stress parameter sequence, and a dielectric polarization reconstruction method based on current density distribution is used to output time-series insulation performance degradation data. The accelerated aging sample library of insulating materials is called, and the time series degradation data is input into the LSTM neural network training model. The insulation remaining life prediction model is established through the dynamic adjustment mechanism of the forgetting gate threshold. The dielectric response spectrum data of the current oil-paper medium is used as the model input parameter, and the recursive forward calculation of the insulation remaining life prediction model is performed to output the insulation remaining life prediction result.

[0016] In an embodiment of the present invention, the collected PDIV drift data, historical ambient temperature data, and power load curve are first aligned according to a unified timestamp. Through interpolation algorithms and data resampling technology, a time-scale unified data set based on the equipment operation cycle is established to ensure that data from different sources have an accurate correspondence in the time dimension. On this basis, an environmental and electrical stress coupling analysis is performed, and the ambient temperature data is quantified using a temperature load factor weighting algorithm. This algorithm converts different temperature values ​​into corresponding aging weight coefficients based on the thermal aging characteristic curve of the oil-paper insulation material, and performs compensation calculations based on the fluctuation gradient of the power load curve. Specifically, by performing sliding window averaging on the power change rate, time periods with sharp power changes are identified, and the temperature load factor is dynamically adjusted within these time periods, thereby generating a composite aging stress parameter sequence that can accurately reflect the actual operating conditions of the equipment. Subsequently, the time-varying response of the insulation dielectric was modeled for the composite aging stress parameter sequence. A dielectric polarization reconstruction method based on current density distribution was employed. This method analyzes the current response characteristics of the oil-paper insulation system under different aging stresses to establish a spatiotemporal current density distribution model. Combined with dielectric polarization theory, this method reconstructs the polarization state changes of the insulation material during long-term operation, thereby outputting time-series degradation data that accurately describes the degradation of insulation performance over time. To establish the remaining insulation life prediction model, a pre-built accelerated aging sample library of insulation materials was first used. This library contains long-term aging experimental data of oil-paper insulation materials under different temperature, humidity, and electric field strength conditions. This acquired insulation performance time-series degradation data was used as input parameters to a pre-trained LSTM neural network training model. This model uses a dynamic forget gate threshold adjustment mechanism. By monitoring changes in the network's internal state in real time, the control threshold of the forget gate is dynamically adjusted, allowing the model to better adapt to the nonlinear characteristics and long-term dependencies of the insulation aging process. During model training, historical aging data is used to iteratively optimize the network parameters to ensure good generalization capabilities. Finally, the dielectric response spectrum data of the current oil-paper medium is used as the real-time input parameter of the model. This data is collected in real time by the online monitoring system and contains the current frequency response characteristics of the insulation system. By executing the recursive forward calculation process of the insulation life remaining prediction model, that is, according to the timing processing mechanism of the LSTM network, the input data is processed time step by time, and finally the insulation remaining life prediction result is output.

[0017] The sub-steps in step S2 include: The ambient temperature series and power load gradient series are extracted from the unified time-scale data set. The equivalent aging amount of the Arrhenius equation is calculated for the temperature series using the temperature-aging rate conversion coefficient to obtain the temperature load factor weighted data. Perform peak hold analysis and valley recovery compensation on the power load gradient sequence, and generate power gradient compensation data according to preset load fluctuation sensitivity parameters; The temperature load factor weighted data and the power gradient compensation data are fused in time domain convolution, and a sliding time window integration algorithm is used to generate a composite aging stress parameter sequence. The spatial charge distribution model of the insulating medium is reconstructed based on the composite aging stress parameter sequence, and the dielectric polarization response characteristic matrix is ​​established by fitting the Lorentz distribution of the current density vector. A time-varying degradation simulation is performed on the dielectric polarization response characteristic matrix, and the depolarization current is iteratively inverted using the preset dielectric relaxation time spectrum to output the insulation performance time-series degradation data.

[0018] In an embodiment of the present invention, an ambient temperature sequence and a power load gradient sequence are first extracted from a time-scaled unified data set. For the ambient temperature sequence, the temperature values ​​at each moment are converted to corresponding aging rate equivalents using a pre-calibrated temperature-aging rate conversion coefficient. This conversion process is based on the physical principle of the Arrhenius equation, which states that chemical reaction rates increase exponentially with each temperature increase. This calculation results in a cumulative equivalent aging value corresponding to the entire temperature sequence, generating temperature load factor-weighted data that accurately reflects the cumulative impact of temperature on the aging process of insulation materials. Simultaneously, a peak hold analysis is performed on the power load gradient sequence. By statistically analyzing the duration and amplitude characteristics of periods of sharp power increases, the device's ability to withstand electrical stress shocks is assessed. Recovery compensation is then calculated for valley values ​​during periods of power decreases. This compensation process takes into account the insulation material's self-recovery properties during low load periods and combines it with a preset load fluctuation sensitivity parameter, which is pre-determined based on the oil-paper insulation's response to electrical stress changes. This generates power gradient compensation data that accurately describes the impact of power fluctuations on aging. Subsequently, the temperature load factor weighted data and the power gradient compensation data are convolved and fused in the time domain. This fusion process uses a sliding time window integration algorithm. By setting a sliding window of fixed time length, the two data are weightedly integrated within each time window. The weight coefficient is predetermined based on the relative importance of the effects of temperature and electrical stress on aging, thereby generating a composite aging stress parameter sequence that can comprehensively reflect the combined effects of environmental and electrical stress. On this basis, the spatial charge distribution model of the insulating medium is reconstructed based on the composite aging stress parameter sequence. This reconstruction process establishes a three-dimensional distribution model of spatial charge density by analyzing the migration and accumulation of charge within the insulating material under the action of composite stress. In combination with the Lorentz distribution fitting method of the current density vector, that is, using statistical methods to fit the distribution characteristics of the current density in various directions in space to the form of a Lorentz function, a dielectric polarization response characteristic matrix is ​​established that can describe the electrical response characteristics of the insulating material under the action of composite stress. Finally, a time-varying degradation simulation is performed on the dielectric polarization response characteristic matrix. This simulation process utilizes a preset dielectric relaxation time spectrum, which describes the time distribution characteristics required for the insulating material to recover from the polarization state to the equilibrium state. Through the depolarization current iterative inversion algorithm, that is, repeatedly calculating and adjusting the correspondence between the polarization state and the depolarization current, the actual degradation process is gradually approached, thereby outputting the insulation performance time-series degradation data.

[0019] The sub-steps in step S2 further include: Call multiple sets of dielectric loss-PDIV drift time series data sets from the insulation material accelerated aging sample library, perform sample balancing on the time series degradation data, and generate an LSTM neural network training data set; Build an initial LSTM network model with forget gate control, input the training data set into the model for forward propagation calculation, and dynamically adjust the forget gate activation threshold through output layer error backpropagation; The forget gate threshold interval was optimized based on the dielectric response characteristic deviation of the validation set, and the threshold dynamic adjustment interval was set to [0.2, 0.8] to establish an insulation life remaining prediction model. The broadband dielectric response spectrum data of the current oil-paper medium is collected, and the loss factor spectrum eigenvector in the 10Hz-1MHz frequency band is extracted as the model input parameter. The dielectric spectrum eigenvector is input into the trained prediction model, and a forward propagation calculation based on time step recursion is performed to output the remaining insulation life prediction result in years.

[0020] In an embodiment of the present invention, a pre-built accelerated aging sample library of insulating materials is first called. The sample library contains multiple sets of dielectric loss-PDIV drift time series data sets obtained under multiple stress conditions such as different temperatures, humidity, and electric field strengths. These data sets cover the performance characteristics of oil-paper insulation materials in various aging stages. A matching degree analysis is performed on the acquired time series degradation data and the data in the sample library. By calculating the similarity characteristics of the data distribution, the sample data is balanced. This processing process includes interpolation and expansion of specific aging stage data with insufficient sample number, and downsampling of data with excessive sample number, to ensure the uniform distribution of training data in each aging stage, thereby generating a standardized data set suitable for LSTM neural network training. Subsequently, an initial LSTM network model with forget gate control was constructed. This model adopts a multi-layer recursive neural network structure, in which the forget gate control mechanism is used to determine the amount of historical information that the network needs to retain or forget in each time step. The balanced training data set is input into the initial model for forward propagation calculation. The output prediction result is calculated through the weight connection between the network layers, and the output result is compared with the actual aging status label to calculate the prediction error. This error information is transmitted layer by layer in the network through the backpropagation algorithm, and the internal parameters of the network are dynamically adjusted, especially the activation threshold of the forget gate. This threshold controls the degree to which the network forgets historical information and is automatically adjusted according to the error changes during training. During the model training process, validation set data is introduced to monitor the model performance in real time. The validation set contains dielectric response feature data independent of the training set. The generalization ability of the model is evaluated by calculating the feature deviation between the model prediction results and the actual labels of the validation set. Based on the evaluation results of the deviation, the threshold range of the forget gate is optimized and adjusted, and the dynamic adjustment range of the threshold is set to the range of 0.2 to 0.8. This range can balance the model's ability to remember historical information and adapt to new information, avoiding the loss of important historical information due to an excessively large threshold, or the inability of the model to adapt to new aging characteristics due to an excessively small threshold, thereby establishing an insulation life remaining prediction model with good prediction accuracy and stability. During the model application stage, the broadband dielectric response spectrum data of the current oil-paper medium is collected through the online monitoring system, which can comprehensively reflect the changes in the dielectric properties of the insulating material at different frequencies. During the data preprocessing process, the loss factor spectrum eigenvector within the frequency band is extracted. This eigenvector contains the key parameters that can characterize the current insulation state. These eigenvectors are used as real-time input parameters of the model and input into the trained prediction model. The forward propagation calculation process based on time step recursion is executed. That is, according to the timing processing mechanism of the LSTM network, the input features are processed time step by time step, and the historical information and current input information remembered by the network are fully utilized to finally output the remaining insulation life prediction result in years.

[0021] Step S3: Controlling the operation of the isolated power supply based on the insulation remaining life prediction result.

[0022] Step S3 includes: Obtain the insulation remaining life prediction results, extract the remaining life value and credibility index corresponding to the current moment, and obtain insulation status assessment data; Comparing the insulation status assessment data with a preset life threshold interval to determine the operating level of the current insulation status, wherein the operating level includes normal operation, warning operation, and emergency derating operation, and obtaining an insulation operating level determination result; According to the insulation operation level determination result, a corresponding power output control strategy is matched from a pre-configured operation strategy library, a power adjustment instruction adapted to the current insulation state is generated, and a power control strategy instruction set is obtained; The power control strategy instruction set is sent to the power regulation module of the isolated power supply to control it to adjust the output power level and operation timing, and to collect the adjusted power supply output parameters and insulation status feedback data in real time to obtain power regulation execution data; Based on the power regulation execution data, the isolated power supply is controlled.

[0023] In an embodiment of the present invention, the prediction result output by the aforementioned insulation life remaining prediction model is first obtained, and the result includes the remaining life value corresponding to the current moment and the corresponding credibility index, wherein the remaining life value represents the expected time that the equipment insulation system can still operate normally in units of years, and the credibility index reflects the reliability of the prediction result, which is quantified by analyzing the convergence of the internal state parameters of the prediction model and the historical prediction accuracy statistics, thereby obtaining comprehensive insulation state assessment data. Subsequently, the insulation state assessment data is compared and analyzed with a pre-set life threshold range, which is determined based on the design life of the equipment, safety margin requirements, and historical fault statistics. It usually includes multiple classification standards. When the remaining life value is greater than a first threshold, it is determined to be a normal operating state; when the remaining life value is between the first threshold and the second threshold, it is determined to be a warning operating state; when the remaining life value is less than the second threshold, it is determined to be an emergency derated operating state. At the same time, the judgment result is confidence-weighted in combination with the credibility index to ensure the accuracy of the judgment result, thereby determining the specific operating level of the current insulation state and obtaining the insulation operating level judgment result. Based on the judgment result, the corresponding power output control strategy is automatically matched from the pre-configured operation strategy library. The strategy library contains detailed control schemes for different operation levels. The standard power output mode is adopted in the normal operation state, the restrictive power output mode is adopted and the monitoring frequency is increased in the warning operation state, and the minimum safe power output mode is adopted in the emergency derating operation state and the backup protection mechanism is activated. The control strategy most suitable for the current state is retrieved from the strategy library through the table lookup method or rule matching algorithm, and a power adjustment instruction set suitable for the current insulation state is generated to form a complete power control strategy instruction set. The power control strategy instruction set is then sent to the power regulation module of the isolated power supply through a communication interface. This module includes core components such as power switching devices, voltage regulators, and current controllers. This module controls the module to adjust the output power level and operating sequence according to the instructions. For example, in a warning state, the output power is reduced to 80% of the rated value and the shutdown interval is increased. In an emergency state, the output power is reduced to a safe limit and a cyclic start-stop mechanism is activated. At the same time, the adjusted power supply output parameters, including key indicators such as output voltage, current, and power factor, are collected in real time. Insulation status feedback data, including real-time dielectric loss factor and partial discharge level, is obtained through the online monitoring system to obtain comprehensive power regulation execution data. Finally, based on this power regulation execution data, closed-loop control is implemented on the isolated power supply. By continuously monitoring the deviation between the regulation effect and the expected target, the control parameters are dynamically adjusted to ensure the safe and stable operation of the equipment under the current insulation state.

[0024] It should be noted that the intervals and thresholds are set for ease of comparison. The threshold size depends on the amount of sample data and the cardinality set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, derived from software simulations of the most recent real-world conditions using large amounts of data. The preset parameters in the formulas are set by those skilled in the art based on actual conditions.

[0025] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0026] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0027] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling a high-power isolated power supply, characterized in that: include: Step S1: Real-time acquisition of dielectric loss factor of the oil-paper insulation medium inside the isolated power supply is performed through an insulation status online monitoring unit to obtain dielectric loss online monitoring data; based on the dielectric loss online monitoring data, combined with a dielectric loss factor and partial discharge inception voltage drift correlation curve pre-calibrated through accelerated aging experiments, PDIV dynamic inversion calculation is performed to obtain current insulation PDIV drift data; Step S2: Based on the PDIV drift data, a time series coupling process is performed with historical operating environment parameters and a power load curve to obtain insulation performance time series degradation data; an insulation life remaining prediction model is established based on the insulation performance time series degradation data to obtain an insulation remaining life prediction result; Step S3: Controlling the operation of the isolated power supply based on the insulation remaining life prediction result.

2. The control method of a high-power isolated power supply according to claim 1, characterized in that: Step S1 includes: Step S11: performing millisecond-level current phase angle tracking on the oil-paper insulation medium through the insulation state online monitoring unit to obtain dielectric polarization current amplitude data within the power frequency cycle, and generating dielectric loss online monitoring data after Fourier baseband filtering; Step S12: performing environmental error compensation on the dielectric loss online monitoring data, using a preset temperature loss gradient matrix and humidity compensation factor to eliminate dielectric loss measurement deviation, and outputting corrected dielectric loss characteristic data; Step S13: Calling the dielectric loss factor and PDIV drift correlation curve library calibrated by the accelerated aging experiment, performing loss offset mapping matching on the corrected dielectric loss characteristic data, and establishing a dynamic quantization matrix of loss factor and voltage drift through bilinear interpolation operation of adjacent aging nodes; Step S14: Combined with the historical data of the aging cumulative rate of oil-paper insulation, the dynamic quantization matrix is ​​input into the preset aging effect time integral model, and a time deconvolution operation is performed on the partial discharge inception voltage based on the dielectric loss change gradient to output the current insulation PDIV drift data.

3. The control method of a high-power isolated power supply according to claim 2, characterized in that: Step S13 includes: Calling the dielectric loss factor and PDIV drift correlation curve library established by the accelerated aging experiment, dynamically matching the corrected dielectric loss characteristic data during the aging stage, and determining the coordinate projection interval of the current dielectric loss value on the correlation curve; Two adjacent aging stage nodes are selected within the coordinate projection interval, and the corresponding dielectric loss factor threshold and voltage drift reference value of the nodes are extracted to generate aging stage node matching data. Based on the aging stage node matching data, a bidirectional interpolation operation is performed, in which the dielectric loss change gradient is interpolated along the time aging axis and the drift amount is linearly compensated along the voltage drift axis, and the quantitative relationship parameters after interpolation compensation are output. Cross-dimensional weighted fusion is performed on the quantization relationship parameters after interpolation compensation to generate a dynamic quantization matrix of dielectric loss factor and voltage drift.

4. The control method of a high-power isolated power supply according to claim 2, characterized in that: Step S14 includes: Extracting a historical data set of the aging cumulative rate of the oil-paper insulation material, loading the dynamic quantitative relationship matrix generated in step S13, and establishing an input parameter set for the time integral of the aging effect; Import the input parameter group into the preset aging effect time integral model, perform iterative calculation of the cumulative dielectric loss gradient in hourly time units, and output the cumulative dielectric loss drift sequence on the time axis; Based on the cumulative dielectric loss drift sequence, the partial discharge inception voltage is subjected to reverse time decoupling. The exponential decay weighting algorithm is used to separate the voltage drift contributions of different aging stages and generate time-based voltage drift components. The time domain normalization processing is performed on the periodic voltage drift component, and the current insulation PDIV drift data is output through component amplitude superposition and phase compensation calculation.

5. The method for controlling a high-power isolated power supply according to claim 1, wherein: Step S2 includes: Performing time stamp alignment processing on the PDIV drift data, historical ambient temperature data, and power load curve to establish a time-scale unified data set based on the equipment operation cycle; Based on a unified time-scale data set, coupled environmental and electrical stress analysis is performed. A composite aging stress parameter sequence is generated through a temperature load factor weighting algorithm and power fluctuation gradient compensation calculation. The time-varying response of the insulation medium is modeled based on the composite aging stress parameter sequence, and a dielectric polarization reconstruction method based on current density distribution is used to output time-series insulation performance degradation data. The accelerated aging sample library of insulating materials is called, and the time series degradation data is input into the LSTM neural network training model. The insulation remaining life prediction model is established through the dynamic adjustment mechanism of the forgetting gate threshold. The dielectric response spectrum data of the current oil-paper medium is used as the model input parameter, and the recursive forward calculation of the insulation remaining life prediction model is performed to output the insulation remaining life prediction result.

6. The method for controlling a high-power isolated power supply according to claim 5, characterized in that: The sub-steps in step S2 include: The ambient temperature series and power load gradient series are extracted from the unified time-scale data set. The equivalent aging amount of the Arrhenius equation is calculated for the temperature series using the temperature-aging rate conversion coefficient to obtain the temperature load factor weighted data. Perform peak hold analysis and valley recovery compensation on the power load gradient sequence, and generate power gradient compensation data according to preset load fluctuation sensitivity parameters; The temperature load factor weighted data and the power gradient compensation data are fused in time domain convolution, and a sliding time window integration algorithm is used to generate a composite aging stress parameter sequence. The spatial charge distribution model of the insulating medium is reconstructed based on the composite aging stress parameter sequence, and the dielectric polarization response characteristic matrix is ​​established by fitting the Lorentz distribution of the current density vector. A time-varying degradation simulation is performed on the dielectric polarization response characteristic matrix, and the depolarization current is iteratively inverted using the preset dielectric relaxation time spectrum to output the insulation performance time-series degradation data.

7. The method for controlling a high-power isolated power supply according to claim 5, wherein: The sub-steps in step S2 further include: Call multiple sets of dielectric loss-PDIV drift time series data sets from the insulation material accelerated aging sample library, perform sample balancing on the time series degradation data, and generate an LSTM neural network training data set; Build an initial LSTM network model with forget gate control, input the training data set into the model for forward propagation calculation, and dynamically adjust the forget gate activation threshold through output layer error backpropagation; The forget gate threshold interval was optimized based on the dielectric response characteristic deviation of the validation set, and the threshold dynamic adjustment interval was set to [0.2, 0.8] to establish an insulation life remaining prediction model. The broadband dielectric response spectrum data of the current oil-paper medium is collected, and the loss factor spectrum eigenvector in the 10Hz-1MHz frequency band is extracted as the model input parameter. The dielectric spectrum eigenvector is input into the trained prediction model, and a forward propagation calculation based on time step recursion is performed to output the remaining insulation life prediction result in years.

8. The method for controlling a high-power isolated power supply according to claim 1, wherein: Step S3 includes: Obtain the insulation remaining life prediction results, extract the remaining life value and credibility index corresponding to the current moment, and obtain insulation status assessment data; Comparing the insulation status assessment data with a preset life threshold interval to determine the operating level of the current insulation status, wherein the operating level includes normal operation, warning operation, and emergency derating operation, and obtaining an insulation operating level determination result; According to the insulation operation level determination result, a corresponding power output control strategy is matched from a pre-configured operation strategy library, a power adjustment instruction adapted to the current insulation state is generated, and a power control strategy instruction set is obtained; The power control strategy instruction set is sent to the power regulation module of the isolated power supply to control it to adjust the output power level and operation timing, and to collect the adjusted power supply output parameters and insulation status feedback data in real time to obtain power regulation execution data; Based on the power regulation execution data, the isolated power supply is controlled.

Images