A power regulation method and system for parallel modules of flexible equipment

Abstract

The present invention relates to the field of information processing technology, and discloses a power regulation method and system for parallel modules of flexible equipment. The method and system analyze and process the real-time power data of each parallel module in the flexible equipment to obtain the power regulation target value and power curve of each parallel module. The power state of the parallel module to which it belongs is evaluated based on the power curve to determine the first parallel module, and the first parallel module is controlled to generate a power regulation instruction to send to the second parallel module adjacent to it, so that each second parallel module performs power regulation according to the power regulation instruction and in combination with its own corresponding power regulation target value to obtain a power regulation amount. A power regulation strategy based on predictive control is introduced according to the historical power data of each parallel module to predict the future power change trend of each parallel module, and the power regulation amount is corrected in advance to suppress oscillation in the power climbing process, so as to realize accurate power dynamic regulation of each parallel module in the flexible equipment.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a power regulation method and system for a flexible device parallel module. Background Art

[0002] Currently, precise control algorithms, such as PID control, fuzzy control or neural network control, are mostly used to adjust the output power of each parallel module in the flexible device in real time to achieve dynamic current sharing. However, the control algorithm still has differences in physical characteristics and inconsistent control timing between each parallel module, resulting in large power output fluctuations of the parallel module under non-ideal switching conditions. Therefore, it is difficult to achieve accurate dynamic power regulation of the parallel modules, which in turn affects the system efficiency and stability.

[0003] It can be seen that how to solve the problem that the existing control algorithm is difficult to accurately adjust the power of each parallel module in the flexible device has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0004] The present invention provides a power regulation method and system for parallel modules of flexible devices, which solves the problem that the existing control algorithm is difficult to accurately regulate the power of each parallel module in the flexible device.

[0005] In order to solve the above technical problems, the first aspect of the present invention provides a power regulation method for a flexible device parallel module, comprising:

[0006] Acquire real-time power data of each parallel module in the flexible device, and analyze and process the real-time power data to obtain a power regulation target value and a power curve of each parallel module;

[0007] Based on each of the power curves, the power state of each of the parallel modules is evaluated to determine a first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module, so that each of the second parallel modules performs a power adjustment according to the power adjustment instruction and the corresponding power adjustment target value, thereby obtaining a power adjustment amount;

[0008] Acquire historical power data of each of the parallel modules, and construct a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model;

[0009] The power adjustment amount of each parallel module is corrected through the power prediction sequence of each parallel module to obtain the power adjustment correction amount and control each parallel module to perform secondary power adjustment based on its own power adjustment correction amount, so as to realize power adjustment of each parallel module in the flexible device.

[0010] As one of the preferred solutions, the real-time power data is analyzed and processed to obtain the power regulation target value and power curve of each parallel module, including:

[0011] Performing noise reduction and filtering processing on the real-time power data by wavelet transform to obtain a first power data set, and performing data standardization processing on the first power data set by using a Kalman filter to obtain a second power data set;

[0012] The second power data set is input into the feedforward neural network for state calculation to obtain the operating state of each of the parallel modules for comparison with the preset state range, and the power mutual feedback numerical calculation is performed on the corresponding parallel modules according to the comparison result to obtain the power regulation target value of each of the parallel modules.

[0013] As one of the preferred solutions, the power mutual feedback numerical calculation of the parallel modules is performed according to the comparison result to obtain the power regulation target value of each parallel module, including:

[0014] The parallel modules whose operating states exceed the preset state range are regarded as over-limit operating parallel modules, and the remaining parallel modules are regarded as normal operating parallel modules;

[0015] Determine the power value to be adjusted of each of the over-limit running parallel modules, and prioritize each of the over-limit running parallel modules according to the relationship between the power value to be adjusted and the preset adjustment power range;

[0016] According to the distance between each of the normally operating parallel modules and each of the over-limit operating parallel modules, the adjustment target corresponding to each of the over-limit operating parallel modules is selected from the normally operating parallel modules based on the priority order to perform power mutual feedback numerical calculation to obtain the power adjustment target value of each of the parallel modules.

[0017] As one of the preferred solutions, the analyzing and processing of the real-time power data to obtain the power regulation target value and power curve of each parallel module also includes:

[0018] Performing noise filtering on the real-time power data by using a band-stop filter to obtain a third power data set, and performing weighted processing on the third power data set by using a Hanning window function to obtain a fourth power data set;

[0019] Correcting the fluctuation points in the fourth power data set according to an adaptive Kalman filter, and fitting the corrected fourth power data set using a feedforward neural network to obtain a primary power curve of each of the parallel modules;

[0020] The key features in each of the primary power curves are extracted, and the fundamental component and harmonic component of the key features are calculated by Fourier transform to compare with a preset harmonic distortion rate threshold. The curve interval exceeding the preset harmonic distortion rate threshold is marked as a switching distortion interval and eliminated to obtain the power curve of each of the parallel modules; the key features are used to characterize the high-frequency power fluctuation components of the corresponding parallel modules under non-ideal switching conditions.

[0021] As one of the preferred solutions, the power state of each of the parallel modules is evaluated based on each of the power curves to determine a first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module, including:

[0022] Decomposing each of the power curves into four layers by using discrete wavelet transform, and reconstructing the decomposition results by wavelet coefficients to obtain a characteristic data matrix, so as to extract a power characteristic vector sequence from the characteristic data matrix;

[0023] The power feature vector sequence is input into a mapping relationship model established by a radial basis kernel function support vector regressor for processing, so as to obtain the current power state of the corresponding parallel module, calculate the Euclidean distance between the current power state and a preset average sharing threshold, and obtain the power state evaluation value of the corresponding parallel module;

[0024] The parallel module whose power state evaluation value exceeds the preset performance index is used as the first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module.

[0025] As one of the preferred solutions, the step of causing each of the second parallel modules to perform power regulation once according to the power regulation instruction and the power regulation target value corresponding to each of the second parallel modules to obtain a power regulation amount includes:

[0026] According to the power adjustment instruction, each of the second parallel modules performs quantitative constraints on its corresponding power adjustment target value through a gradient algorithm and takes minimization of the power climbing rate variance as an objective function, and generates a power adjustment sequence including an adjustment target value, an adjustment time and an adjustment direction;

[0027] Based on the power adjustment sequence of each of the second parallel modules, the climbing rate value of each of the second parallel modules is calculated by Gaussian elimination method, the climbing rate value exceeding the preset climbing rate is limited, and the power adjustment increment of the corresponding second parallel module is allocated according to the preset attenuation ratio;

[0028] The power regulation increment and its corresponding climbing rate value are respectively used as the state quantity and the control quantity, and the power regulation quantity of each of the second parallel modules is calculated by Lagrange multiplier method.

[0029] As one of the preferred solutions, the power adjustment amount of each second parallel module is calculated by Lagrange multiplier method, including:

[0030] The power regulation rate constraint and the load power constraint of each of the second parallel modules are used as constraint conditions, and the power ramp rate of each parallel module in the flexible device is made consistent as a target model;

[0031] The target model is solved by the Lagrange multiplier method. When the iteration stop criterion is not met, the climbing rate value is updated according to the steepest descent method, and the target model is iteratively solved by the Lagrange multiplier method based on the updated climbing rate value until the iteration stop criterion is met, and the power adjustment amount of each parallel module in the flexible device when the iteration round is stopped is output; the iteration stop criterion is that the change rate of the power adjustment increment between adjacent rounds is less than a preset power adjustment increment change threshold.

[0032] As one of the preferred solutions, the acquiring of historical power data of each of the parallel modules and constructing a power trend prediction model of each of the parallel modules according to the historical power data to obtain a power prediction sequence according to the power trend prediction model includes:

[0033] The historical power data of each parallel module is collected through a rolling window, and the historical power data in the rolling window is normalized based on the rated power value, so as to train the long short-term memory neural network as sample data, and the power trend prediction model of the corresponding parallel module is obtained to generate a power trend prediction sequence;

[0034] Extracting the oscillation characteristics of the historical power data based on fast Fourier transform, and performing weighted processing on the power trend prediction sequence of the parallel module to which the historical power data belongs according to the inverse of the oscillation period to obtain a power prediction compensation sequence;

[0035] The power prediction compensation sequence is corrected according to the Kalman filter to obtain the power prediction sequence of each parallel module.

[0036] As one of the preferred solutions, the power adjustment amount of each parallel module is corrected through the power prediction sequence of each parallel module, the power adjustment correction amount is obtained, and each parallel module is controlled to perform secondary power adjustment based on its own power adjustment correction amount, including:

[0037] Collecting real-time power data, real-time temperature data, real-time voltage data, and real-time current data of each parallel module in the flexible device after secondary power regulation to train the deep neural network as training samples and output the load power change rate;

[0038] When the load power change rate exceeds a preset power change rate threshold, triggering parameter correction of a controller in the flexible device to obtain a parameter correction value;

[0039] Based on the parameter correction value, a genetic algorithm is used to optimize the power regulation correction amount of each parallel module to obtain a power regulation correction optimization sequence, and the ratio of the load power change rate of each parallel module to its own rated power is used as a compensation coefficient;

[0040] The power regulation correction optimization sequence is weighted by the compensation coefficient to obtain the power regulation compensation amount, and each corresponding parallel module is controlled to perform three power adjustments based on its own power regulation compensation amount to achieve power regulation of each parallel module in the flexible device.

[0041] A second aspect of the present invention provides a power regulation system for a flexible device parallel module, comprising:

[0042] A data processing module is used to obtain real-time power data of each parallel module in the flexible device, and analyze and process the real-time power data to obtain a power regulation target value and a power curve of each parallel module;

[0043] A primary regulation module, used for evaluating the power state of each of the parallel modules based on each of the power curves to determine a first parallel module, controlling the first parallel module to generate a power regulation instruction and sending it to a second parallel module adjacent to the first parallel module, so that each of the second parallel modules performs a primary power regulation according to the power regulation instruction and the corresponding power regulation target value, to obtain a power regulation amount;

[0044] A power prediction module, used for acquiring historical power data of each of the parallel modules, and constructing a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model;

[0045] The secondary regulation module is used to correct the power regulation amount of each parallel module through the power prediction sequence of each parallel module, obtain the power regulation correction amount and control each parallel module to perform secondary power regulation based on its own power regulation correction amount, so as to realize power regulation of each parallel module in the flexible device.

[0046] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0047] (1) By acquiring the power data of each parallel module in the flexible device in real time, the working status of each parallel module can be quickly understood; the analysis and processing of real-time power data can accurately calculate the power adjustment target value and power curve of each parallel module, providing an accurate basis for subsequent adjustment; by sending power adjustment instructions to adjacent parallel modules so that they can adjust the power in combination with their corresponding power adjustment target values, the accuracy and timeliness of the adjustment can be ensured, and information sharing and coordinated adjustment between various parallel modules can be realized;

[0048] (2) According to the historical power data of each parallel module, a power regulation strategy based on predictive control is introduced to predict the future power change trend of each parallel module, and the power regulation amount is corrected in advance, which suppresses the oscillation during the power ramp process and realizes accurate power dynamic regulation of each parallel module in the flexible device;

[0049] (3) The present invention integrates multiple intelligent links such as data collection, analysis and processing, prediction and optimization. Through intelligent control strategies, it realizes precise control and optimal adjustment of the parallel modules of flexible equipment to reduce manual intervention and improve production efficiency and automation level. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the implementation mode will be briefly introduced below. Obviously, the drawings described below are only some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 It is a flow chart of a power regulation method for parallel modules of flexible devices provided by a certain embodiment of the present invention;

[0052] Figure 2 It is a structural diagram of a power regulation system for a flexible device parallel module provided by a certain embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0055] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two components. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are for illustrative purposes only, and do not indicate or imply that the system or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0056] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0057] In one embodiment, if Figure 1 As shown, the first aspect of the present invention provides a power regulation method for a flexible device parallel module, comprising:

[0058] S1. Acquire real-time power data of each parallel module in the flexible device, and analyze and process the real-time power data to obtain a power regulation target value and a power curve of each parallel module;

[0059] Specifically, the present invention uses a voltage transformer and a current transformer to obtain real-time data from an embedded parallel module in a flexible device at a fixed data acquisition frequency; wherein the sampling frequency is set to collect 100 data points per second, and the collected real-time power data includes a voltage fundamental wave value, a current fundamental wave value, and a power phase value. The data is collected for voltage data ranging between positive and negative 380 volts, current data ranging between positive and negative 200 amperes, and phase angles ranging between positive and negative 180 degrees to avoid sampling distortion caused by data fluctuations.

[0060] In one embodiment, the analyzing and processing of the real-time power data to obtain the power regulation target value and power curve of each of the parallel modules includes:

[0061] Performing noise reduction and filtering processing on the real-time power data by wavelet transform to obtain a first power data set, and performing data standardization processing on the first power data set by using a Kalman filter to obtain a second power data set;

[0062] The second power data set is input into the feedforward neural network for state calculation to obtain the operating state of each of the parallel modules for comparison with the preset state range, and the power mutual feedback numerical calculation is performed on the corresponding parallel modules according to the comparison result to obtain the power regulation target value of each of the parallel modules.

[0063] Specifically, in the process of wavelet transform filtering and noise reduction, the present invention uses the db4 wavelet basis function to perform four-layer decomposition on the collected voltage fundamental wave value, current fundamental wave value, and power phase value, removes high-frequency noise and random disturbance, and retains the main frequency components. The first power data set obtained after filtering has good data continuity, and then the data is standardized through the Kalman filter, and the abnormal mutation points are corrected to obtain a stable and reliable second power data set.

[0064] Then, according to the voltage fundamental wave value, current fundamental wave value, and power phase value contained in the second power data set, the current operating state of each parallel module is numerically calculated in combination with a feedforward neural network, and the preset power threshold interval, voltage threshold interval, and current threshold interval obtained based on historical data statistics are obtained from the controller of the flexible device as a reference, and compared with the calculated operating state according to each data collection cycle, and the power mutual feedback numerical calculation is performed on the corresponding parallel modules according to the comparison result to obtain the power regulation target value of each parallel module;

[0065] Among them, when the feedforward neural network calculates the operating status, the input layer contains three neurons: power value, voltage value, and current value. The hidden layer uses ten neurons. The output layer corresponds to three operating states: normal operating state, critical operating state, and over-limit operating state. The threshold interval corresponding to each state is obtained by statistically analyzing historical operating data. For example, the normal operating power threshold interval is within the range of plus or minus 10% of the rated power, the critical operating power threshold interval is within the range of plus or minus 20% of the rated power, and the over-limit operating power threshold interval exceeds the rated power by plus or minus 20%.

[0066] The present invention uses wavelet transform to perform noise reduction filtering on real-time power data, which can effectively remove noise and interference in the data and improve the accuracy and reliability of the data; performs data standardization processing through Kalman filter, which can further reduce errors and uncertainties in the data and improve the stability and consistency of the data; uses the powerful nonlinear mapping ability of the feedforward neural network to learn and simulate complex system behaviors, thereby achieving accurate prediction and evaluation of the operating status of the parallel module; adopts an intelligent adjustment mechanism to ensure that the parallel module always operates in the optimal state, thereby improving the overall performance and stability of the system.

[0067] In one embodiment, the power mutual feedback value calculation is performed on the parallel modules according to the comparison result to obtain the power regulation target value of each parallel module, including:

[0068] The parallel modules whose operating states exceed the preset state range are regarded as over-limit operating parallel modules, and the remaining parallel modules are regarded as normal operating parallel modules;

[0069] Determine the power value to be adjusted of each of the over-limit running parallel modules, and prioritize each of the over-limit running parallel modules according to the relationship between the power value to be adjusted and the preset adjustment power range;

[0070] According to the distance between each of the normally operating parallel modules and each of the over-limit operating parallel modules, the adjustment target corresponding to each of the over-limit operating parallel modules is selected from the normally operating parallel modules based on the priority order to perform power mutual feedback numerical calculation to obtain the power adjustment target value of each of the parallel modules.

[0071] Specifically, when calculating the power mutual feedback value (which can also be considered as the power adjustment amount that a single parallel module needs to feed back to another parallel module, which can be positive, negative or zero), the present invention divides each parallel module into an over-limit operating parallel module and a normally operating parallel module based on the relationship between the operating state of each parallel module and the preset state interval, and determines the power value to be adjusted of the over-limit operating parallel module, that is, the value of the rated power that exceeds the preset state interval, and performs power adjustment on the parallel module in the over-limit operating state.

[0072] For scenarios with large power regulation amounts, the smaller the priority value, the higher the regulation priority. Adjustment starts from the parallel modules with low power regulation amounts and gradually transitions to the parallel modules with high power regulation amounts to avoid large power mutations. For example, when multiple parallel modules require power regulation at the same time, the parallel modules with a power regulation amount less than 10% of the rated power have a priority of 1, the parallel modules with a power regulation amount between 10% and 20% of the rated power have a priority of 2, the parallel modules with a power regulation amount greater than 20% of the rated power have a priority of 3, and so on.

[0073] The regulation object is selected based on the priority order and the power mutual feedback value is calculated. The regulation object is selected from the normal operating parallel module according to the distance from the over-limit parallel module. It is generally provided by the adjacent (i.e., within the preset distance interval) normal operating state parallel module. For example, when the power of a parallel module exceeds the rated value by 30%, the feedback amount required as the target parallel module is 30% of its rated value (the feedback amount is the power regulation target value of the target parallel module). Since it is the receiver, the value is negative; and the two adjacent parallel modules each provide 15% of the power regulation amount, that is, the parallel module needs to feedback 15% of the rated value of the target parallel module to the target parallel module. Since it is the feedback party, the value is positive. During the power mutual feedback process, the stability is ensured by the preset attenuation ratio. The preset attenuation ratio is set to 5%, and the power regulation target value of each parallel module can be obtained.

[0074] The present invention compares the operating status of the parallel modules with a preset status interval to accurately identify the parallel modules that are operating beyond the limit and the parallel modules that are operating normally, which helps to quickly locate the problem parallel modules; the priority sorting of the parallel modules that are operating beyond the limit ensures that the parallel modules with high priorities can be adjusted first, so as to more effectively control the operating status of the overall system and reduce the negative impact caused by the parallel modules operating beyond the limit; through the numerical calculation of power mutual feedback, the power regulation target value of each parallel module can be obtained, ensuring the scientificity and rationality of the regulation measures; the scheme realizes global optimization of the entire system by comprehensively considering factors such as the operating status, priority, and distance of the parallel modules.

[0075] In one embodiment, the analyzing and processing the real-time power data to obtain the power regulation target value and power curve of each of the parallel modules further includes:

[0076] Performing noise filtering on the real-time power data by using a band-stop filter to obtain a third power data set, and performing weighted processing on the third power data set by using a Hanning window function to obtain a fourth power data set;

[0077] Correcting the fluctuation points in the fourth power data set according to an adaptive Kalman filter, and fitting the corrected fourth power data set using a feedforward neural network to obtain a primary power curve of each of the parallel modules;

[0078] The key features in each of the primary power curves are extracted, and the fundamental component and harmonic component of the key features are calculated by Fourier transform to compare with a preset harmonic distortion rate threshold. The curve interval exceeding the preset harmonic distortion rate threshold is marked as a switching distortion interval and eliminated to obtain the power curve of each of the parallel modules; the key features are used to characterize the high-frequency power fluctuation components of the corresponding parallel modules under non-ideal switching conditions.

[0079] Specifically, the present invention filters out the high-frequency components of the real-time power data near the switching frequency through a band-stop filter with a center frequency set at 2 kHz and a bandwidth set at 400 Hz, thereby obtaining a third power data set; according to the fluctuation characteristics of the third power data set, a Hanning window function is used to perform weighted processing on the power data at different times, the window length is set to five times the switching period, and the data points in the window are given an attenuation weight with a center weight value of one and an edge weight value of zero point two, and a fourth power data set is obtained through weighted average calculation;

[0080] For the mutation points in the fourth power data set, the adaptive Kalman filter is used to set the observation quantity as the power data sequence and the state quantity as the power mean and variance, and the fluctuation points that deviate from the power mean three times the variance are corrected. The feedforward neural network input layer is set with twelve time series neurons, the hidden layer is set with eight neurons, and the output layer is set with four neurons corresponding to the amplitude, phase, frequency, and waveform characteristics of the power curve to fit the corrected power data, and the primary power curve of each parallel module is obtained;

[0081] Key features such as zero crossover frequency, fluctuation amplitude ratio, waveform distortion rate, power factor, etc. are extracted from the primary power curve. The fundamental component and harmonic component of the feature are calculated by Fourier transform, and compared with the preset harmonic distortion rate threshold. The interval exceeding the threshold is marked as a switching distortion interval and eliminated, and the power curve of each parallel module can be obtained; among them, when extracting the power curve features, the zero crossover frequency reflects the basic period of power fluctuation, which is usually around 50 Hz. The fluctuation amplitude ratio indicates the relative size between the peaks and troughs, which is less than 0.1 during normal operation. The waveform distortion rate describes the degree of deviation of the curve from the ideal sine wave. The power factor represents the ratio of active power to apparent power. The fundamental component and the first 40 harmonic components are calculated by 1024-point fast Fourier transform. The total harmonic distortion rate should be less than 0.05 during normal operation. The time interval exceeding this threshold is marked as a switching distortion interval.

[0082] The DC converter with a switching frequency of 2 kHz will produce oscillations during the switching process due to the influence of parasitic inductance and parasitic capacitance, resulting in high-frequency fluctuations in the power curve. After band-stop filtering and weighted averaging, these high-frequency components are effectively suppressed, and the remaining low-frequency components can reflect the load change characteristics. When the load suddenly changes, a step response will be formed on the power curve. Through Kalman filtering and neural network fitting, a smooth transition curve can be obtained, which is convenient for subsequent feature extraction and analysis. The present invention accurately corrects the fluctuation points in the power data set through an adaptive Kalman filter, thereby improving the stability and consistency of the data; the powerful nonlinear mapping ability of the feedforward neural network is used to fit the corrected data well with the characteristics of the complex power change relationship to generate the primary power curve of each parallel module; and the key features in the primary power curve can characterize the high-frequency power fluctuation components of the parallel module under non-ideal switching conditions, providing an important basis for subsequent harmonic analysis; accurately identifying and eliminating the curve interval exceeding the threshold value helps to eliminate power distortion caused by improper switch operation or equipment failure, and improves the accuracy and reliability of the power curve.

[0083] S2, evaluating the power state of each of the parallel modules based on each of the power curves to determine a first parallel module, controlling the first parallel module to generate a power adjustment instruction and sending it to a second parallel module adjacent to the first parallel module, so that each of the second parallel modules performs a power adjustment according to the power adjustment instruction and the corresponding power adjustment target value, and obtains a power adjustment amount;

[0084] In one embodiment, the power state of each of the parallel modules is evaluated based on each of the power curves to determine a first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module, including:

[0085] Decomposing each of the power curves into four layers by using discrete wavelet transform, and reconstructing the decomposition results by wavelet coefficients to obtain a characteristic data matrix, so as to extract a power characteristic vector sequence from the characteristic data matrix;

[0086] The power feature vector sequence is input into a mapping relationship model established by a radial basis kernel function support vector regressor for processing, so as to obtain the current power state of the corresponding parallel module, calculate the Euclidean distance between the current power state and a preset average sharing threshold, and obtain the power state evaluation value of the corresponding parallel module;

[0087] The parallel module whose power state evaluation value exceeds the preset performance index is used as the first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module.

[0088] Specifically, the present invention uses discrete wavelet transform to perform four-layer decomposition on the power curve, selects the second-layer high-frequency coefficients to calculate the power fluctuation amplitude, the third-layer high-frequency coefficients to calculate the power fluctuation period, the fourth-layer low-frequency coefficients to calculate the power change rate, and the adjacent-layer coefficients to calculate the power phase difference according to the energy distribution law of the wavelet coefficients, and reconstructs the wavelet coefficients to obtain a standardized characteristic data matrix to extract a power characteristic vector sequence from the data matrix; wherein, the second-layer high-frequency coefficients reflect the short-term change characteristics of the power fluctuation, and the corresponding time scale is millisecond level, and the power fluctuation amplitude is obtained by calculating the sum of the squares of the coefficients, the third-layer high-frequency coefficients reflect the medium-term change characteristics of the power fluctuation, and the corresponding time scale is second level, and the power fluctuation period is obtained by calculating the zero-crossing interval of the coefficients, and the fourth-layer low-frequency coefficients reflect the long-term trend of the power change, and the corresponding time scale is minute level, and the power change rate is obtained by calculating the difference of the coefficients;

[0089] Then, the power state mapping relationship model is established by training the radial basis kernel function support vector regressor with historical data, and the standardized power feature vector sequence is used as input to obtain the current power state of the corresponding parallel module, and the state data with a power averaging rate exceeding 95% in the historical data is used as the preset averaging threshold, and the Euclidean distance between the current power state and the preset averaging threshold is calculated to obtain the power state evaluation value; wherein, when the radial basis kernel function support vector regressor establishes the power state mapping relationship, the width parameter of the kernel function is set to 0.5, the penalty factor is set to 100, and the optimal parameter combination is selected through cross-validation, and the preset averaging threshold is selected based on the state with the best power averaging effect in the historical data, under which the power deviation of each parallel module is less than 5%, the power fluctuation amplitude is less than 10% of the rated value, and the standard deviation of the power fluctuation period is less than 20% of the average value;

[0090] Finally, an equal-sharing performance judgment rule is set to control the parallel module whose power state evaluation value exceeds 20% of the preset performance index as the first parallel module and generate a power adjustment instruction to be sent to the second parallel module adjacent to it through the optical fiber ring network. Therefore, each parallel module in the flexible device can be a first parallel module or a second parallel module of other first parallel modules. Adjacent can be understood as within a preset distance range, but no specific distance value is limited here. The power adjustment instruction adopts hexadecimal format, including the parallel module address, power adjustment target value, and adjustment priority. For example, the power adjustment instruction format of a parallel module is: start bit 0xAA, parallel module address 0x01, power value 0x0F, priority 0x02, and end bit 0xFF. The parallel modules exchange data through the serial communication interface at a baud rate of 115200. The data frame interval is 10 milliseconds, and the feedback checker ensures the correctness of data transmission through cyclic redundancy check. In addition, when making an averaging performance judgment, the calculation of the Euclidean distance takes into account the three dimensions of power amplitude difference, phase difference, and fluctuation period difference. When the difference in a certain dimension exceeds the corresponding threshold, the power averaging request is triggered, and the optical fiber ring network adopts a bidirectional redundant architecture.

[0091] In the power sharing control process of the parallel modules, when the load mutation causes the power of a parallel module to exceed the rated value by 30%, the power characteristics extracted by wavelet transform show that the power fluctuation amplitude increases and the fluctuation period shortens, and the state evaluation value output by the support vector regressor significantly deviates from the preset sharing threshold. After the power sharing request is triggered, the adjacent parallel modules respond one by one in the token transfer order, so that the power deviation of each parallel module after power adjustment is reduced to within 5% of the rated value. The present invention enhances the expression ability and robustness of feature data through multi-level decomposition and reconstruction; extracts the power feature vector sequence from the feature data matrix to convert the complex power signal into a concise and efficient feature vector, which provides convenience for subsequent processing and analysis; performs Euclidean distance calculation and power state evaluation through a high-precision mapping relationship model to quantify the degree of deviation of the power state of the parallel module, which provides an important basis for subsequent control and adjustment; through the coordinated adjustment between adjacent parallel modules, the power fluctuation and loss can be further reduced, which helps to improve the stability and reliability of the system.

[0092] In one embodiment, the step of causing each of the second parallel modules to perform power regulation once according to the power regulation instruction and the power regulation target value corresponding to each of the second parallel modules to obtain a power regulation amount includes:

[0093] According to the power adjustment instruction, each of the second parallel modules performs quantitative constraints on its corresponding power adjustment target value through a gradient algorithm and takes minimization of the power climbing rate variance as an objective function, and generates a power adjustment sequence including an adjustment target value, an adjustment time and an adjustment direction;

[0094] Based on the power adjustment sequence of each of the second parallel modules, the climbing rate value of each of the second parallel modules is calculated by Gaussian elimination method, the climbing rate value exceeding the preset climbing rate is limited, and the power adjustment increment of the corresponding second parallel module is allocated according to the preset attenuation ratio;

[0095] The power regulation increment and its corresponding climbing rate value are respectively used as the state quantity and the control quantity, and the power regulation quantity of each of the second parallel modules is calculated by Lagrange multiplier method.

[0096] Specifically, the present invention supports the maximum power regulation target value of the parallel module to be plus or minus thirty percent of the rated power, and uses the gradient algorithm to minimize the variance of the power climbing rate as the optimization goal, quantizes the power adjustment amount of each parallel module, and generates a power regulation sequence including the regulation target value, regulation time, and regulation direction. The gradient algorithm calculates the deviation between the current climbing rate and the average rate in each round of iteration, and corrects the power regulation amount of the parallel module with a large deviation. Taking into account the physical constraints of the parallel module, the power regulation upper limit does not exceed twenty percent of the rated power, the regulation time is not less than 10 milliseconds, and the regulation direction is consistent with the load change trend. The generated power regulation sequence includes the target power value at each time point;

[0097] Then, based on the power regulation sequence of each second parallel module, the maximum climbing rate that satisfies the regulation capacity of all parallel modules is calculated through the consistency constraint operator, and the Gaussian elimination method is used to solve the climbing rate distribution problem to obtain the climbing rate value of each parallel module, and the rate value exceeding the regulation range is limited, and the power regulation increment is allocated according to the preset attenuation ratio; wherein, when calculating the maximum climbing rate, the consistency constraint operator comprehensively considers the difference in regulation capacity of each parallel module, for example, the maximum regulation rate of a parallel module with a power of 1000 watts is 100 watts per millisecond, and the maximum regulation rate of a parallel module with a power of 500 watts is 50 watts per millisecond, and the consistency of the regulation capacity is used as a consistency constraint. Conditions, combined with the working characteristics of each parallel module to establish a mathematical model, the model is used to describe the interaction between the modules and the overall output characteristics of the parallel module, using optimization algorithms (such as genetic algorithms, particle swarm algorithms, etc.) or numerical solution methods (such as Newton iteration method, gradient descent method, etc.), under the premise of meeting the consistency constraint conditions, solve the maximum rate of change of the output power of the parallel module over time, that is, the maximum climbing rate; based on the consistency of the regulation capabilities of each parallel module and its corresponding regulation capability constraints to construct a set of equations, and solve the climbing rate value through Gaussian elimination method to ensure that all parallel modules can achieve it, and limit the rate value beyond the range to 80%;

[0098] Finally, the power regulation increment is selected as the state quantity and its corresponding climbing rate is selected as the control quantity, and the power regulation quantity of each corresponding parallel module is obtained by solving the Lagrange multiplier method. Among them, in the state space expression, the power value as the state quantity describes the operating state of the parallel module, and the climbing rate as the control quantity reflects the dynamic characteristics of power change. The Lagrange multiplier method also introduces power balance constraints and rate consistency constraints when solving the optimal control problem.

[0099] The present invention adopts a gradient algorithm, takes minimizing the variance of the power climbing rate as the objective function, and performs quantitative constraints on the power regulation target values ​​of each adjacent parallel module, which can ensure the stability and consistency of the power regulation process and avoid the impact of excessive power fluctuations on the system; the Gaussian elimination method is used to calculate the climbing rate value of each corresponding parallel module to ensure the accuracy of the calculation result.

[0100] In one embodiment, the calculating of the power regulation amount of each of the second parallel modules by the Lagrange multiplier method includes:

[0101] The power regulation rate constraint and the load power constraint of each of the second parallel modules are used as constraint conditions, and the power ramp rate of each parallel module in the flexible device is made consistent as a target model;

[0102] The target model is solved by the Lagrange multiplier method. When the iteration stop criterion is not met, the climbing rate value is updated according to the steepest descent method, and the target model is iteratively solved by the Lagrange multiplier method based on the updated climbing rate value until the iteration stop criterion is met, and the power adjustment amount of each parallel module in the flexible device when the iteration round is stopped is output; the iteration stop criterion is that the change rate of the power adjustment increment between adjacent rounds is less than a preset power adjustment increment change threshold.

[0103] Specifically, when solving the target model by the Lagrange multiplier method, the present invention also introduces power balance constraints and rate consistency constraints, and stops iteration when the power adjustment increment change rate of two adjacent iterations is less than five percent; wherein, the step size factor of the steepest descent method is set to 0.5, and the iterative solution process is continued according to the optimized climbing rate value until the iteration stop criterion is met, and the power adjustment amount of each parallel module when the iteration round is stopped is output. The optimal climbing rate distribution scheme can be obtained through multiple rounds of iterative calculations, and all parallel modules synchronously adjust the power at a uniform rate, avoiding power conflicts and oscillations between parallel modules. The response time of the entire averaging process is less than 500 milliseconds, and the power averaging accuracy can be better than five percent. The present invention takes the power climbing rate of each parallel module in the flexible device as the target model, which helps to ensure the coordination and stability of the parallel module during operation. By optimizing the power climbing rate, the power difference between the parallel modules can be reduced, and overload or insufficient power can be avoided, thereby improving the operation efficiency and reliability of the entire flexible device.

[0104] S3, acquiring historical power data of each of the parallel modules, and constructing a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model;

[0105] In one embodiment, step S3 includes:

[0106] The historical power data of each parallel module is collected through a rolling window, and the historical power data in the rolling window is normalized based on the rated power value, so as to train the long short-term memory neural network as sample data, and the power trend prediction model of the corresponding parallel module is obtained to generate a power trend prediction sequence;

[0107] Extracting the oscillation characteristics of the historical power data based on fast Fourier transform, and performing weighted processing on the power trend prediction sequence of the parallel module to which the historical power data belongs according to the inverse of the oscillation period to obtain a power prediction compensation sequence;

[0108] The power prediction compensation sequence is corrected according to the Kalman filter to obtain the power prediction sequence of each parallel module.

[0109] Specifically, the present invention selects a time window with a fixed length of 1 second through a rolling window calculator, normalizes the power data in the window based on the rated power value, uses a long short-term memory neural network input layer to set twelve time series neurons to capture the time series characteristics of power changes, sets eight neurons in the hidden layer to extract the deep characteristics of power changes, and sets four neurons in the output layer to correspond to four prediction points within the next 100 milliseconds, respectively, and performs model training based on the pre-collected standard power curve to obtain a power trend prediction model to generate a power trend prediction sequence; wherein, the training data selects the climbing section in the standard power curve, including three typical scenarios of steady climbing, oscillating climbing, and step climbing;

[0110] Then, the prediction time domain is set to 100 milliseconds using the variable predictor, and the historical power state values ​​sampled at a 1 millisecond period are obtained from the real-time database. The spectrum components in the range of 0 to 500 Hz are calculated by fast Fourier transform, and the main oscillation frequency and amplitude in the power oscillation characteristics are extracted. The power trend prediction sequence is weighted according to the inverse of the oscillation period as the weighting coefficient to generate a power prediction compensation sequence.

[0111] Finally, the prediction error minimization target is established through the sliding time domain controller. The Kalman filter is used to set the state quantity as the power mean and variance, and the observation quantity as the real-time power value. The power prediction compensation value is adaptively corrected to obtain the power prediction sequence of each parallel module. The prediction compensation value is corrected by the adaptive gain. When the oscillation amplitude is large, the introduction of the damping compensation term reduces the response speed of the control loop and suppresses power overshoot.

[0112] The present invention combines a variety of advanced algorithms and technologies, which can efficiently process and predict the power data of parallel modules, providing strong support for the real-time control and optimization of the parallel module system; through the synergy of multiple steps such as normalization processing, FFT extraction of oscillation characteristics, weighted processing and Kalman filter correction, it can generate a highly accurate power prediction sequence, providing a strong guarantee for the stable operation of the system; by introducing a power regulation strategy based on predictive control, the future power change trend of each parallel module is predicted, and the power regulation amount is corrected in advance to suppress oscillations during the power climbing process, thereby realizing accurate power dynamic regulation of each parallel module in the flexible device.

[0113] S4, correcting the power adjustment amount of each parallel module through the power prediction sequence of each parallel module, obtaining the power adjustment correction amount and controlling each parallel module to perform secondary power adjustment based on its own power adjustment correction amount, so as to realize power adjustment of each parallel module in the flexible device;

[0114] Specifically, based on the power prediction sequence, a piecewise linear interpolation calculator is used to decompose the power regulation amount according to a fixed slope and divide it into several time periods. A fixed regulation slope is maintained in each period, and the typical slope is set to 10 percent of the rated power per millisecond. Then, based on the token ring network, the power feedback value of each parallel module is read from the serial bus, and the power deviation exceeding the prediction interval is proportionally corrected by a feedforward compensator to obtain the power regulation correction amount, and the power of the corresponding parallel module is regulated. The regulation priority is allocated according to the size of the power regulation increment to achieve real-time suppression of power oscillation.

[0115] In addition, when one of the parallel modules has an overtemperature fault, the status monitor detects the temperature overlimit within 3 milliseconds, immediately reduces the priority of the parallel module to the lowest, and limits the power to less than 300 watts. After the other parallel modules are regrouped, the first group of parallel modules completes power sharing within 5 milliseconds, with a sharing accuracy better than 2%. The second group of parallel modules follows and adjusts in place within 10 milliseconds. The last group of parallel modules is used as a backup and gradually adjusts to the sharing state within 15 milliseconds. When the load power suddenly changes from 50% to 80% of the rated power, the hierarchical optimization calculator prioritizes the output power of the first group of parallel modules to ensure that the core parallel module responds quickly to the load change. The other groups of parallel modules gradually follow according to the preset response sequence. The entire sharing process is completed within 20 milliseconds. Even if the power of individual parallel modules overshoots during this period, the fluctuation can be controlled within 10% of the rated value through feedforward compensation.

[0116] In one embodiment, the power adjustment amount of each parallel module is corrected through the power prediction sequence of each parallel module to obtain the power adjustment correction amount and control each parallel module to perform secondary power adjustment based on its own power adjustment correction amount, including:

[0117] Collecting real-time power data, real-time temperature data, real-time voltage data, and real-time current data of each parallel module in the flexible device after secondary power regulation to train the deep neural network as training samples and output the load power change rate;

[0118] When the load power change rate exceeds a preset power change rate threshold, triggering parameter correction of a controller in the flexible device to obtain a parameter correction value;

[0119] Based on the parameter correction value, a genetic algorithm is used to optimize the power regulation correction amount of each parallel module to obtain a power regulation correction optimization sequence, and the ratio of the load power change rate of each parallel module to its own rated power is used as a compensation coefficient;

[0120] The power regulation correction optimization sequence is weighted by the compensation coefficient to obtain the power regulation compensation amount, and each corresponding parallel module is controlled to perform three power adjustments based on its own power regulation compensation amount to achieve power regulation of each parallel module in the flexible device.

[0121] Specifically, the present invention collects real-time power data, real-time temperature data, real-time voltage data, and real-time current data of each parallel module in the flexible device after secondary power regulation to train the deep neural network as training samples. The deep neural network is set to have an input layer including twelve feature vectors, a hidden layer including three layers with eight neurons each, and an output layer including four optimization parameters. Mapping relationships are established for temperature changes, voltage fluctuations, current fluctuations, ambient temperature, and ambient humidity data under different load rates, respectively, so as to output the load power change rate through the trained model;

[0122] When the load power change rate exceeds 20% of the rated value or the environmental parameters exceed the preset range, the correction is triggered, and the parameters of the controller of the flexible device are automatically corrected online. According to the corrected parameter correction value, the power balancing parameter is optimized and calculated using a genetic algorithm with a chromosome length of 32 bits. The controller parameter value is mapped through binary coding, and the crossover probability is set to 80% and the mutation probability is set to 20%. The parameter values ​​of the proportional gain between 0.1 and 10, the integral time constant between 1 and 100 milliseconds, and the response dead zone value between 1% and 5% of the rated power are read from the controller parameter library. The fitness function value is calculated using the power balancing deviation and the adjustment time weighted to generate a power adjustment correction optimization sequence;

[0123] The ratio of the load power change rate of each parallel module to its own rated power is used as the compensation coefficient. For example, when the load power changes from 500 watts to 800 watts within 10 milliseconds, the power change rate is 30 watts per millisecond, and the compensation coefficient is 0.03. The sliding mean calculator smoothes the compensated control parameters within a 50-millisecond window to avoid the impact of parameter jumps on control performance. Then, the power regulation correction optimization sequence is weighted by the compensation coefficient to obtain the power regulation compensation amount, and each corresponding parallel module is controlled to perform power regulation again based on its own power regulation compensation amount to achieve power regulation of each parallel module in the flexible device.

[0124] The coordinated control process of parallel modules under different working conditions, the change of ambient temperature directly affects the heat dissipation capacity of the parallel modules, and then affects the power regulation characteristics. Through the parameter mapping relationship established by the deep neural network, the response speed of the controller is automatically reduced when the temperature rises to avoid overshoot, and the response speed is appropriately increased when the temperature drops to improve the dynamic performance. The change of load power reflects the dynamic characteristics of the electrical equipment. The control parameters optimized by the genetic algorithm can adapt to different load change rates, increase the response dead zone value to reduce the control sensitivity when the power changes greatly, and reduce the response dead zone value to improve the control accuracy when the power changes little, so as to realize the dynamic adaptation of power sharing control.

[0125] The present invention can search for the optimal solution in a complex parameter space by triggering parameter correction and optimization of genetic algorithms, thereby obtaining a power regulation correction optimization sequence; the ratio of the load power change rate of each parallel module to its own rated power is used as a compensation coefficient to perform weighted processing on the power regulation correction optimization sequence, taking into account the differences and actual working conditions between the parallel modules, making power regulation more flexible and accurate. The established process optimization mechanism of the autonomous collaborative control algorithm adaptively adjusts the parameters of the collaborative control algorithm, autonomously optimizes the power sharing strategy according to the actual working conditions of the flexible equipment and changes in the external environment, and makes the power regulation between parallel modules dynamically adaptive, effectively solving the problem of unbalanced power regulation between parallel modules in the flexible equipment, and improving the stability and reliability of the system.

[0126] In order to solve the problems of inconsistent power regulation states and mismatched power ramp rates of each parallel module in the flexible device, the present invention obtains the power data of each parallel module in real time, and determines the power regulation target value of each parallel module through an autonomous collaborative control algorithm. By filtering the power data, a parallel module power state evaluation model is constructed to judge in real time whether the power regulation state of each parallel module meets the equal distribution performance index; when the judgment result is not satisfied, the power equal distribution adjustment between the parallel modules is started, and the distributed optimization algorithm is used to collaboratively determine the power regulation amount of each parallel module, so that the power ramp rate of each parallel module tends to be consistent. At the same time, a power regulation strategy based on predictive control is introduced to predict the future power change trend of the parallel module, correct the power regulation amount in advance, and suppress oscillation during the power ramp process.

[0127] In the embodiment of the present application, the existing control algorithm is difficult to accurately regulate the power of each parallel module in the flexible device. A power regulation method for the parallel modules of the flexible device is designed, which realizes the acquisition of real-time power data of each parallel module in the flexible device, and analyzes and processes the real-time power data to obtain the power regulation target value and power curve of each parallel module; based on each power curve, the power state of each parallel module is evaluated to determine the first parallel module, and the first parallel module is controlled to generate a power regulation instruction and send it to the second parallel module adjacent to the first parallel module, so that each of the second parallel modules The module performs a power regulation according to the power regulation instruction and the corresponding power regulation target value to obtain the power regulation amount; obtains the historical power data of each of the parallel modules, and constructs a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model; corrects the power regulation amount of each of the parallel modules through the power prediction sequence of each of the parallel modules to obtain the power regulation correction amount and controls each of the parallel modules to perform secondary power regulation based on its own power regulation correction amount, so as to realize accurate power dynamic regulation of each parallel module in the flexible device.

[0128] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0129] In another embodiment, if Figure 2 As shown, the second aspect of the present invention provides a power regulation system for a flexible device parallel module, comprising:

[0130] The data processing module 10 is used to obtain the real-time power data of each parallel module in the flexible device, and analyze and process the real-time power data to obtain the power regulation target value and power curve of each parallel module;

[0131] A primary regulation module 20 is used to evaluate the power state of each of the parallel modules based on each of the power curves to determine a first parallel module, control the first parallel module to generate a power regulation instruction and send it to a second parallel module adjacent to the first parallel module, so that each of the second parallel modules performs a primary power regulation according to the power regulation instruction and the corresponding power regulation target value, and obtains a power regulation amount;

[0132] A power prediction module 30, used for acquiring historical power data of each of the parallel modules, and constructing a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model;

[0133] The secondary regulation module 40 is used to correct the power regulation amount of each parallel module through the power prediction sequence of each parallel module, obtain the power regulation correction amount and control each parallel module to perform secondary power regulation based on its own power regulation correction amount, so as to realize power regulation of each parallel module in the flexible device.

[0134] It should be noted that each module in the above-mentioned power regulation system for parallel modules of flexible devices can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. For the specific definition of a power regulation system for parallel modules of flexible devices, please refer to the definition of a power regulation method for parallel modules of flexible devices above. The two have the same functions and effects, which will not be repeated here.

[0135] In summary, the present invention relates to the field of information processing technology, and discloses a power regulation method and system for parallel modules of flexible devices. The power regulation target value and power curve of each parallel module are obtained by analyzing and processing the real-time power data of each parallel module in the flexible device; the power state of the parallel module to which it belongs is evaluated based on the power curve to determine the first parallel module, and the first parallel module is controlled to generate a power regulation instruction and send it to the second parallel module adjacent to it, so that each second parallel module performs power regulation according to the power regulation instruction and in combination with its own corresponding power regulation target value to obtain a power regulation amount; a power regulation strategy based on predictive control is introduced according to the historical power data of each parallel module, the future power change trend of each parallel module is predicted, and the power regulation amount is corrected in advance to suppress oscillation during power climbing, so as to realize accurate power dynamic regulation of each parallel module in the flexible device.

[0136] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A power regulation method for a flexible device parallel module, characterized in that: include: Acquire real-time power data of each parallel module in the flexible device, and analyze and process the real-time power data to obtain a power regulation target value and a power curve of each parallel module; Based on each of the power curves, the power state of each of the parallel modules is evaluated to determine a first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module, so that each of the second parallel modules performs a power adjustment according to the power adjustment instruction and the corresponding power adjustment target value, thereby obtaining a power adjustment amount; Acquire historical power data of each of the parallel modules, and construct a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model; The power adjustment amount of each parallel module is corrected through the power prediction sequence of each parallel module to obtain the power adjustment correction amount and control each parallel module to perform secondary power adjustment based on its own power adjustment correction amount, so as to realize power adjustment of each parallel module in the flexible device; The analyzing and processing the real-time power data to obtain the power regulation target value and power curve of each parallel module includes: Performing noise reduction and filtering processing on the real-time power data by wavelet transform to obtain a first power data set, and performing data standardization processing on the first power data set according to a Kalman filter to obtain a second power data set; Inputting the second power data set into a feedforward neural network for state calculation to obtain the operating state of each of the parallel modules for comparison with a preset state interval, and performing power mutual feedback numerical calculation on the corresponding parallel modules according to the comparison result to obtain the power regulation target value of each of the parallel modules; The power mutual feedback numerical calculation is performed on the parallel modules according to the comparison result to obtain the power regulation target value of each parallel module, including: The parallel modules whose operating states exceed the preset state range are regarded as over-limit operating parallel modules, and the remaining parallel modules are regarded as normal operating parallel modules; Determine the power value to be adjusted of each of the over-limit running parallel modules, and prioritize each of the over-limit running parallel modules according to the relationship between the power value to be adjusted and the preset adjustment power range; According to the distance between each of the normally operating parallel modules and each of the over-limit operating parallel modules, the adjustment target corresponding to each of the over-limit operating parallel modules is selected from each of the normally operating parallel modules based on the priority order to perform power mutual feedback numerical calculation to obtain the power adjustment target value of each of the parallel modules; The analyzing and processing the real-time power data to obtain the power regulation target value and power curve of each parallel module also includes: Performing noise filtering on the real-time power data by using a band-stop filter to obtain a third power data set, and performing weighted processing on the third power data set by using a Hanning window function to obtain a fourth power data set; Correcting the fluctuation points in the fourth power data set according to an adaptive Kalman filter, and fitting the corrected fourth power data set using a feedforward neural network to obtain a primary power curve of each of the parallel modules; The key features in each of the primary power curves are extracted, and the fundamental component and harmonic component of the key features are calculated by Fourier transform to compare with a preset harmonic distortion rate threshold. The curve interval exceeding the preset harmonic distortion rate threshold is marked as a switching distortion interval and eliminated to obtain the power curve of each of the parallel modules; the key features are used to characterize the high-frequency power fluctuation components of the corresponding parallel modules under non-ideal switching conditions.

2. A power regulation method for a flexible device parallel module according to claim 1, characterized in that: The power states of the parallel modules are evaluated based on the power curves to determine a first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and sent to a second parallel module adjacent to the first parallel module, including: Decomposing each of the power curves into four layers by using discrete wavelet transform, and reconstructing the decomposition results by wavelet coefficients to obtain a characteristic data matrix, so as to extract a power characteristic vector sequence from the characteristic data matrix; The power feature vector sequence is input into a mapping relationship model established by a radial basis kernel function support vector regressor for processing, so as to obtain the current power state of the corresponding parallel module, calculate the Euclidean distance between the current power state and a preset average sharing threshold, and obtain the power state evaluation value of the corresponding parallel module; The parallel module whose power state evaluation value exceeds the preset performance index is used as the first parallel module, and the first parallel module is controlled to generate a power adjustment instruction and send it to a second parallel module adjacent to the first parallel module.

3. A power regulation method for a flexible device parallel module according to claim 1, characterized in that: The step of causing each of the second parallel modules to perform power regulation once according to the power regulation instruction and the power regulation target value corresponding to each of the second parallel modules to obtain a power regulation amount includes: According to the power adjustment instruction, each of the second parallel modules performs quantitative constraints on its corresponding power adjustment target value through a gradient algorithm and takes minimization of the power ramp rate variance as an objective function, and generates a power adjustment sequence including an adjustment target value, an adjustment time and an adjustment direction; Based on the power adjustment sequence of each of the second parallel modules, the climbing rate value of each of the second parallel modules is calculated by Gaussian elimination method, the climbing rate value exceeding the preset climbing rate is limited, and the power adjustment increment of the corresponding second parallel module is allocated according to the preset attenuation ratio; The power regulation increment and its corresponding climbing rate value are respectively used as the state quantity and the control quantity, and the power regulation quantity of each of the second parallel modules is calculated by Lagrange multiplier method.

4. A power regulation method for a flexible device parallel module according to claim 3, characterized in that: The calculating the power regulation amount of each of the second parallel modules by the Lagrange multiplier method includes: The power regulation rate constraint and the load power constraint of each of the second parallel modules are used as constraint conditions, and the power ramp rate of each parallel module in the flexible device is made consistent as a target model; The target model is solved by the Lagrange multiplier method. When the iteration stop criterion is not met, the climbing rate value is updated according to the steepest descent method, and the target model is iteratively solved by the Lagrange multiplier method based on the updated climbing rate value until the iteration stop criterion is met, and the power adjustment amount of each parallel module in the flexible device when the iteration round is stopped is output; the iteration stop criterion is that the change rate of the power adjustment increment between adjacent rounds is less than a preset power adjustment increment change threshold.

5. A power regulation method for a flexible device parallel module according to claim 1, characterized in that: The acquiring of historical power data of each of the parallel modules and constructing a power trend prediction model of each of the parallel modules according to the historical power data to obtain a power prediction sequence according to the power trend prediction model includes: The historical power data of each parallel module is collected through a rolling window, and the historical power data in the rolling window is normalized based on the rated power value, so as to train the long short-term memory neural network as sample data, and the power trend prediction model of the corresponding parallel module is obtained to generate a power trend prediction sequence; Extracting the oscillation characteristics of the historical power data based on fast Fourier transform, and performing weighted processing on the power trend prediction sequence of the parallel module to which the historical power data belongs according to the inverse of the oscillation period to obtain a power prediction compensation sequence; The power prediction compensation sequence is corrected according to the Kalman filter to obtain the power prediction sequence of each parallel module.

6. A power regulation method for a flexible device parallel module according to claim 1, characterized in that: After the power adjustment amount of each parallel module is corrected by the power prediction sequence of each parallel module, the power adjustment correction amount is obtained and each parallel module is controlled to perform secondary power adjustment based on its own power adjustment correction amount, the method includes: Collecting real-time power data, real-time temperature data, real-time voltage data, and real-time current data of each parallel module in the flexible device after secondary power regulation to train the deep neural network as training samples and output the load power change rate; When the load power change rate exceeds a preset power change rate threshold, triggering parameter correction of a controller in the flexible device to obtain a parameter correction value; Based on the parameter correction value, a genetic algorithm is used to optimize the power regulation correction amount of each parallel module to obtain a power regulation correction optimization sequence, and the ratio of the load power change rate of each parallel module to its own rated power is used as a compensation coefficient; The power regulation correction optimization sequence is weighted by the compensation coefficient to obtain the power regulation compensation amount, and each corresponding parallel module is controlled to perform three power adjustments based on its own power regulation compensation amount to achieve power regulation of each parallel module in the flexible device.

7. A power regulation system for parallel modules of flexible devices, characterized in that: include: A data processing module is used to obtain real-time power data of each parallel module in the flexible device, and analyze and process the real-time power data to obtain a power regulation target value and a power curve of each parallel module; A primary regulation module, used for evaluating the power state of each of the parallel modules based on each of the power curves to determine a first parallel module, controlling the first parallel module to generate a power regulation instruction and sending it to a second parallel module adjacent to the first parallel module, so that each of the second parallel modules performs a primary power regulation according to the power regulation instruction and the corresponding power regulation target value, to obtain a power regulation amount; A power prediction module, used for acquiring historical power data of each of the parallel modules, and constructing a power trend prediction model of each of the parallel modules according to the historical power data, so as to obtain a power prediction sequence according to the power trend prediction model; A secondary regulation module, used for correcting the power regulation amount of each parallel module through the power prediction sequence of each parallel module, obtaining the power regulation correction amount and controlling each parallel module to perform secondary power regulation based on its own power regulation correction amount, so as to realize power regulation of each parallel module in the flexible device; The analyzing and processing the real-time power data to obtain the power regulation target value and power curve of each parallel module includes: Performing noise reduction and filtering processing on the real-time power data by wavelet transform to obtain a first power data set, and performing data standardization processing on the first power data set according to a Kalman filter to obtain a second power data set; Inputting the second power data set into a feedforward neural network for state calculation to obtain the operating state of each of the parallel modules for comparison with a preset state interval, and performing power mutual feedback numerical calculation on the corresponding parallel modules according to the comparison result to obtain the power regulation target value of each of the parallel modules; The power mutual feedback numerical calculation is performed on the parallel modules according to the comparison result to obtain the power regulation target value of each parallel module, including: The parallel modules whose operating states exceed the preset state range are regarded as over-limit operating parallel modules, and the remaining parallel modules are regarded as normal operating parallel modules; Determine the power value to be adjusted of each of the over-limit running parallel modules, and prioritize each of the over-limit running parallel modules according to the relationship between the power value to be adjusted and the preset adjustment power range; According to the distance between each of the normally operating parallel modules and each of the over-limit operating parallel modules, the adjustment target corresponding to each of the over-limit operating parallel modules is selected from each of the normally operating parallel modules based on the priority order to perform power mutual feedback numerical calculation to obtain the power adjustment target value of each of the parallel modules; The analyzing and processing the real-time power data to obtain the power regulation target value and power curve of each parallel module also includes: Performing noise filtering on the real-time power data by using a band-stop filter to obtain a third power data set, and performing weighted processing on the third power data set by using a Hanning window function to obtain a fourth power data set; Correcting the fluctuation points in the fourth power data set according to an adaptive Kalman filter, and fitting the corrected fourth power data set using a feedforward neural network to obtain a primary power curve of each of the parallel modules; The key features in each of the primary power curves are extracted, and the fundamental component and harmonic component of the key features are calculated by Fourier transform to compare with a preset harmonic distortion rate threshold. The curve interval exceeding the preset harmonic distortion rate threshold is marked as a switching distortion interval and eliminated to obtain the power curve of each of the parallel modules; the key features are used to characterize the high-frequency power fluctuation components of the corresponding parallel modules under non-ideal switching conditions.

Images