Power dispatching system based on multi-level direct current chopping technology

Abstract

The invention relates to the technical field of converters, in particular to a power dispatching system based on a multi-level direct current chopping technology, which comprises a flux linkage trend analysis module, a duty ratio constraint module, a standby unit evaluation module and a path synchronization coordination module. According to the method, the power integral curve is constructed in real time, the flux linkage increment trend is analyzed, the relevance between the flux linkage direction and the current change is dynamically judged, the duty ratio amplitude limiting range is adjusted in combination with the exponential smoothing prediction model, and phase compensation correction is executed, so that the dynamic response capability of the system is remarkably improved; a sliding standard deviation algorithm is adopted to screen a standby unit group and analyze a current fluctuation coefficient, linear regression is combined to process a temperature change trend, the anti-interference capability of the system is enhanced, a dynamic time warping algorithm is adopted to calculate a phase offset and construct a voltage difference vector, accurate synchronous control of a multi-level architecture is realized, and the stability of the system is improved. Harmonic content and energy loss are effectively reduced, electric energy conversion efficiency and equipment operation stability are improved, and the service life of a power device is prolonged.

Description

Technical Field

[0001] This invention relates to the field of converter technology, and more particularly to a power dispatching system based on multi-level DC chopper technology. Background Technology

[0002] The field of converter technology involves the conversion of AC to DC power. Its core lies in the efficient conversion of electrical energy through power semiconductor devices, and it is mainly applied in new energy power generation systems, power transmission networks, and industrial control equipment. The technology system in this field includes three core aspects: topology design, switching device selection, and control strategy optimization. Among these, multi-level converter technology, by reducing harmonic content through stepped voltage output, has become the mainstream solution for medium- and high-voltage power scenarios. Traditional power dispatching systems refer to devices that use multi-level DC chopper circuits to achieve energy distribution between power network nodes. Their core technology lies in constructing a stepped DC voltage through multiple sets of series-connected power switching devices, and maintaining level balance using capacitor voltage equalization and inductor energy storage methods. Traditional solutions typically use IGBTs or MOSFETs as basic switching units, building a multi-level architecture through a cascaded H-bridge structure, adjusting the switching timing with pulse width modulation control strategies, and using LC filter circuits to eliminate high-frequency ripple components.

[0003] Traditional power dispatching systems rely on fixed capacitor voltage equalization and inductor energy storage to maintain level balance. Under sudden load changes or high-frequency switching scenarios, they are prone to mismatch between flux direction and current changes, resulting in uneven voltage step distribution. Cascaded H-bridge structures use fixed pulse width modulation strategies to coordinate the timing of switching transistors, which is difficult to adapt to phase shifts under complex operating conditions, causing harmonic superposition and ripple residue. IGBT or MOSFET switching units lack dynamic evaluation mechanisms when temperature fluctuates, and the lag in standby unit switching causes the risk of local overload. LC filter circuits are limited in suppressing high-frequency ripples by fixed parameter configurations and cannot match current fluctuation characteristics in real time. Long-term operation can easily lead to power quality degradation and accumulation of thermal stress in devices. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a power dispatching system based on multi-level DC chopper technology.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a power dispatching system based on multi-level DC chopping technology includes: The flux linkage trend analysis module is used to construct a power integral curve by collecting the voltage and current sampling values ​​at the inductor terminals, calculate the flux linkage increment difference using the sliding window difference method, determine the correlation between the flux linkage direction vector and the current rate of change vector using the Pearson correlation coefficient method, generate a flux linkage imbalance trigger signal and transmit it to the duty cycle constraint module. The duty cycle constraint module is used to receive the flux imbalance trigger signal, call the flux direction vector, adjust the duty cycle limit range using an exponential smoothing prediction model, perform linear interpolation compensation correction based on the phase angle difference, and output the duty cycle correction parameters to the standby unit evaluation module. The standby unit evaluation module is used to select standby unit groups after obtaining the duty cycle correction parameters, call the sliding standard deviation algorithm to calculate the current fluctuation coefficient, use the linear regression analysis method to process the temperature slope, generate a standby unit activation command and transmit it to the path synchronization coordination module. The path synchronization coordination module is used to receive the backup unit activation command, call the duty cycle correction parameters, construct the voltage difference vector and monitor the current zero-crossing time difference, calculate the phase offset through the dynamic time warping algorithm and the weighted fusion algorithm, and output the phase compensation coefficient for converter control.

[0006] As a further embodiment of the present invention, the flux imbalance trigger signal specifically includes a correlation threshold, direction vector polarity, and time window length; the duty cycle correction parameters include a limiting range, phase offset, and smoothing coefficient; the backup unit activation command specifically refers to a current fluctuation threshold temperature-current fluctuation correlation threshold and a unit group number; and the phase compensation coefficient includes voltage difference amplitude, time difference value, and carrier phase difference.

[0007] As a further aspect of the present invention, the calculation formula of the weighted fusion algorithm is as follows: ; in, This is the phase offset. Represents the weighting coefficient of voltage difference amplitude. The time difference is the weighting coefficient. This represents the time difference between the zero-crossing points of adjacent channels.

[0008] As a further aspect of the present invention, the magnetic flux linkage trend analysis module includes: The signal acquisition submodule acquires the voltage and current sampling values ​​at the inductor terminal, performs sliding mean filtering on the voltage signal to smooth high-frequency noise, converts the current sampling value into an instantaneous power value through multiplication, and performs time-domain integration to construct a power integral curve. Based on the power integral curve, the difference calculation submodule uses a sliding window to extract adjacent data points, calculates the first-order difference value of continuous data points within the window, performs second-order difference operation on the beginning and end endpoints of the window, and accumulates the absolute values ​​of all differences within the window to generate the flux linkage increment difference. The correlation determination submodule calls the flux linkage increment difference sequence and the current change rate sequence to calculate the covariance matrix of the two sequences and obtain the standard deviation parameter. At the same time, it calculates the quotient of the product of the covariance of the two sequences and the standard deviation of the two sequences to obtain the Pearson coefficient. The coefficient is compared with the dynamic threshold through a sliding window. When the number of consecutive cycles reaches the set condition, a flux linkage imbalance trigger signal is generated. The dynamic threshold is obtained by optimizing historical operating data using the gradient descent method, and its determination condition is that the Pearson coefficient is less than 0.7 for three consecutive sampling periods.

[0009] As a further aspect of the present invention, the duty cycle constraint module includes: The signal parsing submodule receives the flux imbalance trigger signal, calls the flux direction vector, performs sliding window mean filtering on the vector, calculates the standard deviation parameter of the vector in the time domain, and generates a direction feature sequence. The amplitude limiting adjustment submodule uses the exponential smoothing prediction model to perform a weighted average calculation on the time-series duty cycle data based on the directional feature sequence, calculates the absolute deviation between the current duty cycle and the predicted value, and adjusts the amplitude range of the upper and lower limits of the duty cycle according to the deviation value to generate a dynamic amplitude limiting interval. The phase correction submodule calls the dynamic limiting interval, calculates the phase angle difference between the current duty cycle and the limiting boundary, uses linear interpolation to calculate the compensation coefficient for the difference, and outputs the duty cycle correction parameters.

[0010] As a further aspect of the present invention, the weighting coefficients of the exponential smoothing prediction model... The value range is 0.2-0.5, and its specific value satisfies ,in Let | be the Pearson correlation coefficient. | represents the absolute value of the Pearson correlation coefficient.

[0011] As a further aspect of the present invention, the backup unit evaluation module includes: The unit group filtering submodule obtains the duty cycle correction parameter, establishes a mapping relationship between the correction parameter and the unit group number, and adopts a dual threshold filtering mechanism to mark unit groups with parameter values ​​below the lower threshold as failure status and unit groups with parameter values ​​above the upper threshold as overload status, retains unit groups with parameter values ​​in the dual threshold range, and generates valid unit group identifiers. The fluctuation coefficient calculation submodule extracts the current sampling sequence corresponding to the effective unit group identifier, sets an adaptive window width adjustment mechanism, calculates the standard deviation of each window in time order, performs an arithmetic mean operation on the standard deviations of continuous windows, and generates fluctuation characteristic values. The instruction generation submodule collects time-series data from the temperature sensor, constructs a linear regression equation with the fluctuation characteristic value as the dependent variable, calculates the slope parameter of the temperature change rate and the fluctuation characteristic, and when the absolute value of the slope exceeds the temperature-fluctuation correlation benchmark, it encapsulates the unit group number, activation timestamp, and slope parameter into a data packet to generate a backup unit activation instruction.

[0012] As a further aspect of the present invention, the adaptive window width adjustment mechanism satisfies ; in, To adjust the coefficient, .

[0013] As a further aspect of the present invention, the path synchronization coordination module includes: The instruction parsing submodule parses the binary data stream in the backup unit enable instruction, separates the duty cycle correction parameter from the unit group address code, matches the converter control channel according to the address code, converts the correction parameter into three-phase voltage difference vector components, and generates a channel correction vector. The phase calculation submodule collects sampling points of current waveforms from multiple channels, detects the time stamp difference of zero crossing points of adjacent channels, uses a dynamic time warping algorithm to align the current sequences of differentiated channels, calculates the cumulative distance of the optimal path, and generates the phase offset. The compensation control submodule inputs the phase offset and the channel correction vector into the compensation coefficient calculation model, establishes a linear compensation relationship between the phase difference and the voltage vector, updates the converter PWM modulation parameters according to the control cycle, and generates the phase compensation coefficient.

[0014] As a further aspect of the present invention, the weight allocation coefficients of the linear compensation relationship satisfy... ; in, These are the voltage vector weighting coefficients. The inverter's operating frequency, It is a natural constant.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by constructing the power integral curve in real time and analyzing the flux linkage increment trend, the correlation between flux linkage direction and current change is dynamically determined. The duty cycle limiting range is adjusted and phase compensation correction is performed by combining the exponential smoothing prediction model, which significantly improves the dynamic response capability of the system. The sliding standard deviation algorithm is used to screen the backup unit group and analyze the current fluctuation coefficient. The temperature change trend is processed by linear regression to enhance the anti-interference capability of the system. The phase offset is calculated by the dynamic time warping algorithm and the voltage difference vector is constructed to achieve precise synchronous control of the multi-level architecture, effectively reducing harmonic content and energy loss, improving power conversion efficiency and equipment operation stability, and extending the service life of power devices. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the power dispatching system of the present invention; Figure 2 This is an internal flowchart of the magnetic flux linkage trend analysis module of the present invention; Figure 3 This is an internal flowchart of the duty cycle constraint module of the present invention; Figure 4 This is an internal flowchart of the backup unit evaluation module of the present invention; Figure 5 This is an internal flowchart of the path synchronization and coordination module of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the software-based technical solution is described in detail below with reference to system architecture diagrams and embodiments. It should be understood that the specific embodiments described herein are only for explaining the technical solutions of this invention and do not constitute a limitation on the scope of protection.

[0018] In the description of this invention, the system architecture relationships or data processing flows indicated by terms such as "layer," "module," "interface," "data flow," "client," and "server" are all defined based on the architecture diagram or flowchart corresponding to the embodiments. This way of describing is only used to clearly illustrate the logical relationships between the elements in the technical solution, and not to limit the physical deployment form. The term "multiple" includes two or more technical units, including but not limited to multiple data nodes, processing threads, service instances, or functional components and other scalable elements. The specific number is determined according to the actual business scenario and needs to be specifically specified.

[0019] Please see Figure 1 and Figure 2 This invention provides a technical solution: a power dispatching system based on multi-level DC chopping technology includes: The flux linkage trend analysis module is used to construct a power integral curve by collecting the voltage and current sampling values ​​at the inductor terminals, calculate the flux linkage increment difference using the sliding window difference method, determine the correlation between the flux linkage direction vector and the current rate of change vector using the Pearson correlation coefficient method, generate a flux linkage imbalance trigger signal and transmit it to the duty cycle constraint module. The specific magnetic flux imbalance trigger signal includes the correlation threshold, the polarity of the direction vector, and the length of the time window. The magnetic flux linkage trend analysis module includes: The signal acquisition submodule acquires the voltage and current sampling values ​​at the inductor terminal, performs sliding mean filtering on the voltage signal to smooth high-frequency noise, converts the current sampling value into an instantaneous power value through multiplication, and performs time-domain integration to construct a power integral curve. The signal acquisition submodule uses a high-speed data acquisition card connected in parallel with the three-phase bridge arm inductor to synchronously and continuously acquire data on the voltage across the inductor and the current measured by the Hall current sensor connected in series in the circuit, at a sampling frequency of 20kHz. Within a monitoring period of 5 milliseconds, the submodule acquires 100 discrete voltage sampling points and 100 discrete current sampling points, forming the original voltage sequence and the original current sequence.

[0020] For the acquired voltage signal sequence, this submodule performs a moving average filtering operation. A fixed-width sliding window of 5 data points is set, and this window moves point by point across the voltage sequence. At each window position, the arithmetic mean of the 5 voltage data points within the window is calculated, and the calculated mean is assigned to the center point of the window as the filtered voltage value for that point. This process smooths out high-frequency noise components in the original voltage sequence caused by switching actions and electromagnetic interference. Specifically, if in to The sequence of raw voltage values ​​collected at five consecutive sampling times is [100.2V, 101.1V, 100.5V, 99.8V, 100.9V]. Then, the voltage value located at... Filtered voltage value at time 1 The calculation process is as follows: This operation is performed rolling across the entire voltage sequence, generating a smoother voltage sequence with the same number of data points.

[0021] Simultaneously, the signal acquisition submodule combines the acquired raw current sample value sequence with the filtered voltage sequence to obtain the instantaneous power sequence. At each sampling moment, the filtered voltage value at that moment is multiplied by the corresponding raw current value. Within a specific small time interval... (This value is determined by a sampling rate of 20kHz, i.e.) ),like The filtered voltage value at time t is The current sampling value at this moment is Then the instantaneous power value at that moment The calculation is as follows This calculation is repeated at each sampling point to generate an instantaneous power sequence of the same length as the original sequence. Then, a time-domain integration operation is performed on this instantaneous power sequence to construct a power integral curve. This integration operation is achieved through accumulation: starting from the initial time... Begin by taking the instantaneous power value at each sampling time. Multiply by the sampling time interval Then the result is added to the integral value of the previous time step. Above, that is If in The instantaneous power at times [1000.00W, 1013.03W, 1025.50W] are respectively, then the values ​​on the power integral curve at the corresponding times are respectively , , .

[0022] Based on the power integral curve, the difference calculation submodule uses a sliding window to extract adjacent data points, calculates the first-order difference value of continuous data points within the window, performs second-order difference operation on the beginning and end endpoints of the window, and accumulates the absolute values ​​of all differences within the window to generate the flux linkage increment difference. After receiving the power integral curve data sequence constructed in the previous steps, the difference calculation submodule processes the sequence using the sliding window difference method. A sliding window with a width of 10 data points is set. Within this window, the submodule extracts all adjacent data points and calculates their first-order difference values ​​pairwise. This aims to quantify the local rate of change of the power integral curve. Let the first three values ​​of the power integral sequence within the window be [0.1019265J, 0.1025300J, 0.1032150J], then the calculated first two first-order difference values ​​are respectively... and After calculating all nine first-order differences within the window, the submodule then performs second-order difference operations on the first and last endpoints of the sliding window. Specifically, it calculates the difference between the first first-order difference value within the window and the immediately preceding first-order difference value outside the window, and the difference between the last first-order difference value within the window and the immediately following first-order difference value outside the window. Finally, it sums the absolute values ​​of all nine first-order differences within the window, and the absolute values ​​of the two second-order differences calculated at the first and last endpoints. If the sum of the absolute values ​​of the first-order differences within the window is 0.00582, and the absolute values ​​of the first and last second-order differences are 0.000082 and 0.000095 respectively, then the final flux linkage increment difference generated by the window is... This process is repeated as the sliding window moves across the entire power integral curve, thereby generating a continuous sequence of flux linkage increment differences.

[0023] The correlation determination submodule calls the flux linkage increment difference sequence and the current change rate sequence to calculate the covariance matrix of the two sequences and obtain the standard deviation parameter. At the same time, it calculates the quotient of the product of the covariance of the two sequences and the standard deviation of the two sequences to obtain the Pearson coefficient. The coefficient is compared with the dynamic threshold through a sliding window. When the number of consecutive cycles reaches the set condition, a flux linkage imbalance trigger signal is generated. The dynamic threshold is obtained by optimizing historical operating data using the gradient descent method, and its determination condition is that the Pearson coefficient is less than 0.7 for three consecutive sampling periods.

[0024] The correlation determination submodule calls the flux linkage increment difference sequence generated in the previous step, and a method that performs a point-by-point first-order difference operation on the original current sampling value sequence to obtain the absolute difference between adjacent sampling points, i.e., the current change, and divides this change by the sampling time interval. The current change rate sequence is obtained. Let the flux linkage increment difference sequence be... The current change rate sequence is The submodule first calculates the lengths of these two units. The covariance matrix of the sequences is obtained by calculating the arithmetic mean of each sequence (let's say it's a set of sequences). and Then calculate the covariance. At the same time, the standard deviations of the two sequences are calculated separately. and After obtaining these statistical parameters, through... Divide by and The product of these factors is used to calculate the Pearson correlation coefficient. .

[0025] The dynamic threshold was set at 0.7, a value determined based on optimized analysis of historical operating data. The specific experimental verification process was as follows: Historical data covering 100 sets of normal power grid operating conditions and 100 sets of known flux imbalance operating conditions were selected, with each set containing sampled values ​​from 10 consecutive power frequency cycles. The corresponding Pearson correlation coefficients were calculated for each of these 200 sets of data. A set of candidate thresholds was set, ranging from 0.50 to 0.90, with a step size of 0.05. Each candidate threshold was applied to all 200 calculated Pearson coefficients, and the number of times normal operating conditions were misclassified as imbalance (false positives) and imbalanced operating conditions were misclassified as normal (false negatives) was counted. Experimental data showed that when the threshold was set at 0.65 or below, the false positive rate exceeded 5%, causing unnecessary system adjustments; when the threshold was set at 0.75 or above, the false negative rate exceeded 4%, posing a risk of missed detections. When the threshold was set at 0.7, the overall misclassification rate reached its lowest point of 2.5%, achieving a balance between reliability and sensitivity. Therefore, the final decision threshold was set at 0.7. The overall false positive rate was defined as the ratio of the sum of false positives and false negatives to the total number of test samples.

[0026] The correlation determination submodule continuously compares the latest calculated Pearson coefficient value with this dynamic threshold of 0.7. This comparison is performed within a sliding window set to 5 sampling periods. If, within this window, the Pearson coefficient is detected to be below 0.7 for 3 or more consecutive sampling periods, the flux imbalance condition is determined to be triggered. Assuming the Pearson coefficient values ​​calculated for 5 consecutive periods are [0.82, 0.68, 0.65, 0.69, 0.75], since the coefficient values ​​in the 2nd, 3rd, and 4th periods are all less than 0.7, the system generates a flux imbalance trigger signal at the completion of the calculation in the 4th period (with a value of 0.69). This signal is a data packet containing three items: the correlation threshold of 0.7, the polarity of the direction vector determined by the sign of the algebraic sum of the original first-order difference values ​​(not taken as absolute values) within the window (if the algebraic sum is positive, the polarity is +1), and the length of the time window used for the current analysis (10 data points).

[0027] Please see Figure 1 and Figure 3 The duty cycle constraint module is used to receive the flux imbalance trigger signal, call the flux direction vector, adjust the duty cycle limit range using an exponential smoothing prediction model, perform linear interpolation compensation correction based on phase angle difference, and output the duty cycle correction parameters to the standby unit evaluation module. The duty cycle correction parameters include the limiting range, phase offset, and smoothing coefficient. The duty cycle constraint module includes: The signal parsing submodule receives the flux imbalance trigger signal, calls the flux direction vector, performs sliding window mean filtering on the vector, calculates the standard deviation parameter of the vector in the time domain, and generates a direction feature sequence. Upon receiving the flux linkage imbalance trigger signal, the signal parsing submodule immediately parses and extracts three key parameters from the signal data packet: correlation threshold, direction vector polarity, and time window length. Simultaneously, it retrieves the flux linkage direction vector directly constructed from the flux linkage increment difference sequence. To eliminate potential instantaneous disturbances in this vector, the submodule applies a moving average filter with a window width of 5. The processing method is entirely consistent with the aforementioned voltage signal filtering, aiming to obtain a smoother vector that better reflects the trend. After processing, the filtered value of the second point in an original vector sequence [+0.8, -0.2, +1.0, +0.9, -0.1] will become... After filtering, the submodule calculates the standard deviation of the smoothed vector over a time window comprised of the most recent 50 sampling points. This standard deviation numerically represents the dispersion and uncertainty of the flux linkage direction change. If the mean of the smoothed vector sequence within this time window is 0.4, then its standard deviation is... The calculation yields a specific value, set at 0.15. This standard deviation is appended to the end of the directional feature sequence, and this submodule generates a continuously updated directional feature sequence over time.

[0028] The amplitude limiting adjustment submodule uses the exponential smoothing prediction model to perform a weighted average calculation on the time-series duty cycle data based on the directional feature sequence, calculates the absolute deviation between the current duty cycle and the predicted value, and adjusts the amplitude range of the upper and lower limits of the duty cycle according to the deviation value to generate a dynamic amplitude limiting interval. The weighting coefficients of the exponential smoothing prediction model The value range is 0.2-0.5, and its specific value satisfies ,in The Pearson correlation coefficient generated by the flux linkage trend analysis module, | | represents the absolute value of the Pearson correlation coefficient; The amplitude limiting adjustment submodule calls the generated direction feature sequence and applies an exponential smoothing prediction model to dynamically adjust the allowable range of the PWM duty cycle. The core parameter of this model is the weighting coefficient. Through formula Perform the calculation. The formula contains... The parameter is the Pearson correlation coefficient, obtained directly from the trigger signal generated in the previous module (magnetic flux trend analysis module). This formula is based on the fact that when the system is stable, the magnetic flux is highly correlated with the rate of change of current, and the Pearson coefficient... absolute value Approaching 1, at this point Approaching 0, making When the value is small, the model relies more on the smoothing trend of historical data; when the system shows signs of imbalance... The decline caused Increasing the value gives the model higher weight to the current actual observation, thus achieving a faster response. A coefficient of 0.5 is used to... The value is scaled to a reasonable range.

[0029] Weighting coefficients The value of is limited to between 0.2 and 0.5. This range was determined through extensive simulation experiments. In the experiments, disturbances of different intensities and types were injected into the power system model, and the effects of different... The performance of the duty cycle constraint module under different values. The results show that when... When the value is below 0.2, the adjustment of the limiting range is too slow to respond to sudden changes in the duty cycle, with an average delay exceeding 5 switching cycles, which cannot effectively constrain potential overcurrent. When the value is higher than 0.5, the limiting adjustment becomes overly sensitive to measurement noise, and the output limiting range produces useless oscillations exceeding the duty cycle resolution, thus reducing system stability. Therefore, the interval [0.2, 0.5] is determined to be the optimal operating range that balances response speed and stability.

[0030] In this embodiment, the Pearson coefficient obtained from the trigger signal Its absolute value is 0.65. The value is 0.65. Substitute this value into the formula to calculate: Since the calculated result of 0.175 is lower than the preset effective range lower limit of 0.2, the amplitude limiting adjustment submodule will... The final value is forcibly set to 0.2.

[0031] The submodule retrieves the current real-time duty cycle. and the duty cycle prediction value maintained by the model for the previous time step. Using a defined Perform exponential smoothing to update the predicted values: Then, calculate the current actual duty cycle. Compared with the new forecast value The absolute deviation between: The system limits the duty cycle range of the reference based on this deviation value. Perform a contraction adjustment. The adjustment rule is: subtract twice the deviation value from the upper limit, and add twice the deviation value to the lower limit. The new duty cycle upper limit is... The new lower limit is Therefore, a dynamic and tightening amplitude limit range is established. It was generated.

[0032] The phase correction submodule calls the dynamic limiting interval, calculates the phase angle difference between the current duty cycle and the limiting boundary, uses linear interpolation to calculate the compensation coefficient for the difference, and outputs the duty cycle correction parameters.

[0033] The phase correction submodule receives this dynamic limiting range. It maps the numerical range of the duty cycle to the phase angle range, establishing a value from... arrive A linear mapping relationship. Based on this relationship, the current duty cycle... The corresponding phase angle is rad. Similarly, the lower limit of the amplitude limit range. The corresponding phase angle is rad, upper limit The corresponding phase angle is rad. This submodule calculates the difference between the current duty cycle phase angle and the phase at the center point of the dynamic limiting interval. The phase at the center point of the interval is... rad. The absolute value of the difference between the current phase and the center is... rad. Next, a compensation coefficient is calculated using linear interpolation, which is proportional to the aforementioned phase difference. A reference compensation coefficient is set. This value is obtained based on system calibration, so the final compensation amount is... This compensation amount will be used for subsequent fine-tuning of the duty cycle phase. Finally, the module integrates and outputs a set of duty cycle correction parameters, which includes: dynamic limiting range. The calculated phase offset is 0.027, and the smoothing coefficient used is... .

[0034] Please see Figure 1 and Figure 4 The standby unit evaluation module is used to select standby unit groups after obtaining the duty cycle correction parameters, call the sliding standard deviation algorithm to calculate the current fluctuation coefficient, use the linear regression analysis method to process the temperature slope, generate a standby unit activation command and transmit it to the path synchronization coordination module. The specific instructions for enabling the backup unit refer to the current fluctuation threshold, the temperature-current fluctuation correlation threshold, and the unit group number. The backup unit evaluation module includes: The unit group filtering submodule obtains the duty cycle correction parameter, establishes a mapping relationship between the correction parameter and the unit group number, and adopts a dual threshold filtering mechanism to mark unit groups with parameter values ​​below the lower threshold as failure status and unit groups with parameter values ​​above the upper threshold as overload status, retains unit groups with parameter values ​​in the dual threshold range, and generates valid unit group identifiers. After acquiring the duty cycle correction parameters, the unit group screening submodule first performs the screening. The system is configured with 10 independent spare unit groups, numbered from Unit-01 to Unit-10. Each unit group has a pre-defined safe and stable duty cycle operating range based on the specifications and heat dissipation design of its internal power devices. The screening process employs a dual-threshold mechanism: the lower threshold is the minimum value of the unit group's designed duty cycle range, and the upper threshold is its maximum value. The duty cycle dynamic limiting range received by the module is... The submodule compares the upper and lower limits of this range with the preset range for each spare unit group.

[0035] For Unit-05, the design scope is as follows: Due to the received dynamic lower limit Below the permissible lower limit of Unit-05 The unit group cannot guarantee stable startup under this operating condition, and is therefore marked as "failed".

[0036] For Unit-07, the design scope is as follows: Due to the received dynamic limiting upper limit. Higher than the allowed limit of Unit-07 Enabling this unit group poses an overload risk and is therefore marked as "overloaded".

[0037] For Unit-06, the design scope is as follows: Received dynamic limiting range If a unit group is fully included within its scope of work, it is deemed applicable and marked as "valid".

[0038] After comparing all 10 unit groups one by one, the submodule filters out all unit groups with a status of "valid" to form a list of valid unit group identifiers, such as [Unit-01, Unit-03, Unit-06].

[0039] The fluctuation coefficient calculation submodule extracts the current sampling sequence corresponding to the effective unit group identifier, sets an adaptive window width adjustment mechanism, calculates the standard deviation of each window in time order, performs an arithmetic mean operation on the standard deviations of continuous windows, and generates fluctuation characteristic values. The adaptive window width adjustment mechanism satisfies ; in, These are the adjusted and baseline window widths, respectively. To adjust the coefficient, This is the absolute value of the current difference between the current sampling point and the previous sampling point; The fluctuation coefficient calculation submodule extracts the real-time current sampling sequences corresponding to the above-mentioned valid unit group identifiers (Unit-01, Unit-03, Unit-06). During the calculation process, the submodule employs an adaptive window width adjustment mechanism that follows the calculation formula. . In the formula, This is the adjusted window width used to calculate the standard deviation, expressed in units of the number of sampling points. It is the baseline window width; It is an adjustment factor; It is the absolute value of the difference between the current value at the current sampling point and the current value at the previous sampling point, expressed in amperes (A). The design logic of this formula is: when the grid current is stable, The value is very small, and the window width is small. Approximately the base width To obtain statistically stable eigenvalues; when the current fluctuates drastically... The value increases, through The coefficient affects the window width. This reduces the computational sensitivity and improves the ability to capture transient processes.

[0040] All parameter settings in the formula are based on experimental evidence. (Base window width) The sampling point was set to 128. This value was selected after power spectrum analysis of historical stable operating data, resulting in a window length that covers the main low-frequency fluctuation components and has a moderate computational cost. Adjustment coefficient. The setup was optimized through a series of impulse response experiments. In the experiments, a standard current impulse signal (stepping from 10A to 20A within 1ms) was applied to the test system, and the response was recorded at different... The response time and overshoot of the calculated fluctuation characteristic value under the given value setting.

[0041] Table 1 Value Optimization Experimental Data Table As shown in Table 1, the experimental data, when the adjustment coefficient... When the value is 15.0 (points / A), the system achieves the optimal balance between the two key performance indicators: response time (2.5ms) and overshoot (4.1%). Therefore, in this embodiment, The value was determined to be 15.0.

[0042] Perform a calculation example: At a certain sampling moment, the module detects the current value of Unit-03 as follows: The value of the immediately preceding sampling time is The absolute value of the current difference was calculated. Substitute the values ​​into the formula to calculate the adaptive window width: Since the window width must be an integer, the calculation result is rounded down to obtain the actual window width used. One sampling point.

[0043] The submodule then uses this adaptive window with a width of 123 to capture the latest 123 current sampling points from Unit-03 and calculates their standard deviation to obtain an instantaneous fluctuation value, set to 0.58. This process is performed continuously on the time axis, calculated every fixed step (e.g., 10 sampling points). To further smooth the results, the submodule performs an arithmetic mean operation on the standard deviations of the most recent 5 consecutive calculations (set as the sequence [0.58, 0.61, 0.59, 0.62, 0.60]) to generate the final fluctuation characteristic value: .

[0044] The instruction generation submodule collects time-series data from the temperature sensor, constructs a linear regression equation with the fluctuation characteristic value as the dependent variable, calculates the slope parameter of the temperature change rate and the fluctuation characteristic, and when the absolute value of the slope exceeds the temperature-fluctuation correlation benchmark, it encapsulates the unit group number, activation timestamp, and slope parameter into a data packet to generate a backup unit activation instruction.

[0045] The instruction generation submodule collects real-time temperature data from an NTC thermistor mounted on the heatsink of the Unit-03 power module. Simultaneously, it uses the fluctuation characteristic value (0.60) calculated in the previous step as the dependent variable to construct a linear regression analysis model. This model aims to explore the correlation between the rate of temperature change and the current fluctuation characteristics. The submodule extracts temperature data from the most recent minute and calculates its linear rate of change. Within this minute, the temperature changes from... linearly rising to Its rate of change is Meanwhile, the module recorded that the fluctuation characteristic value increased linearly from 0.45 to 0.60 during this period. The submodule used the rate of temperature change as the independent variable and the fluctuation characteristic value as the dependent variable to calculate the slope parameter of the regression equation using the least squares method.

[0046] The "Temperature-Fluctuation Correlation Baseline" is preset to 0.2. This baseline value is set based on a safety assessment of the thermal characteristics of power devices. When the absolute value of this slope parameter exceeds 0.2, it indicates a strong positive correlation between the increase in temperature and the aggravation of current fluctuations, suggesting a risk of thermal runaway due to positive feedback, requiring external intervention. If the currently calculated slope parameter is 0.25, due to its absolute value... If the value exceeds the associated baseline value by 0.2, the system determines that a new backup unit must be activated to share the electrical and thermal stress of the current unit. At this time, the submodule encapsulates the current unit group number Unit-03, the current Coordinated Universal Time (UTC) timestamp, and the calculated slope parameter 0.25 into a standard format data packet, generates a backup unit activation instruction, and passes it to the next module.

[0047] Please see Figure 1 and Figure 5 The path synchronization coordination module is used to receive the backup unit activation command, call the duty cycle correction parameters, construct the voltage difference vector and monitor the current zero-crossing time difference, calculate the phase offset through the dynamic time warping algorithm and the weighted fusion algorithm, and output the phase compensation coefficient for converter control. The phase compensation coefficient includes the voltage difference amplitude, time difference value, and carrier phase difference; The calculation formula for the weighted fusion algorithm is as follows: ; in, This is the phase offset. Represents the weighting coefficient of voltage difference amplitude. The absolute value of the magnitude of the three-phase voltage difference vector. The time difference is the weighting coefficient. This represents the time difference between the zero-crossing points of adjacent channels. The path synchronization coordination module includes: The instruction parsing submodule parses the binary data stream in the backup unit enable instruction, separates the duty cycle correction parameter from the unit group address code, matches the converter control channel according to the address code, converts the correction parameter into three-phase voltage difference vector components, and generates a channel correction vector. Upon receiving the standby unit activation command, the instruction parsing submodule immediately parses the binary data stream within the command. It precisely separates three parts: the address code Unit-03 of the unit group requiring attention, the timestamp of the activation operation, and the slope parameter 0.25, characterizing its overheating risk. Based on the address code Unit-03, the submodule queries and matches the specific hardware control channel of the three-phase bridge arm converter corresponding to that unit in the preset system topology table. Simultaneously, the submodule invokes the duty cycle correction parameters passed from the preceding module, particularly the dynamic limiting range. And a phase offset of 0.027. It converts these abstract correction parameters into concrete three-phase voltage difference vector components. This conversion process will affect the dynamic range of the duty cycle. As a reference, and in conjunction with phase offset information, a target voltage correction vector is generated for this channel via table lookup or a preset function. The components of this vector specify the exact adjustment values ​​required for the three-phase voltage. Let's assume... This is the channel correction vector.

[0048] The phase calculation submodule collects sampling points of current waveforms from multiple channels, detects the time stamp difference of zero crossing points of adjacent channels, uses a dynamic time warping algorithm to align the current sequences of differentiated channels, calculates the cumulative distance of the optimal path, and generates the phase offset. The phase calculation submodule synchronously acquires current waveform sampling points from multiple related channels, including backup unit channels soon to be put into use and one or more adjacent working channels already in stable operation. One of the core tasks of this submodule is to accurately detect the instant the AC current waveform of each channel crosses zero from negative to positive or from positive to negative, and to assign high-precision timestamps to these "zero-crossing points." Let the timestamp for detecting a zero-crossing point of working channel A be... Seconds, while the corresponding zero-crossing timestamp of the adjacent channel B to be synchronized is... The submodule calculates the difference between these two timestamps to obtain the direct time difference between them. Considering that the transmission delay and response characteristics of different channels may introduce nonlinear time distortion, the submodule further performs a dynamic time warping operation to align the current waveform sequences of the two channels. This operation establishes a cost matrix, in which the values ​​located at... The element value at position A is the first element of channel A. The amplitude of the current sampling point and the first current sampling point of channel B The algorithm then finds the absolute value of the difference in amplitude between the current sampling points. The algorithm then searches this matrix for a path from the bottom left to the top right corner where the cumulative sum of all elements is minimized. The cumulative distance corresponding to this path (designed value of 85.4) is used as a metric to quantify the overall phase difference between the two waveforms, and is compared with the aforementioned direct timestamp difference. Together, they are used to generate a comprehensive, more accurate phase offset.

[0049] The compensation control submodule inputs the phase offset and the channel correction vector into the compensation coefficient calculation model, establishes a linear compensation relationship between the phase difference and the voltage vector, updates the converter PWM modulation parameters according to the control cycle, and generates the phase compensation coefficient. The weight allocation coefficients of the linear compensation relationship satisfy... ; in, These are the voltage vector weighting coefficients. The inverter's operating frequency, It is a natural constant.

[0050] The compensation control submodule receives the integrated phase offset calculated in the previous step and the channel correction vector generated by the command parsing submodule. These are then fed as inputs into a compensation coefficient calculation model. In this model, a weighted fusion algorithm is first used to calculate the final phase offset that needs to be compensated. The calculation formula is as follows: .

[0051] In the formula, It is the final output phase offset, in radians (rad); It is the magnitude of the three-phase voltage difference vector (i.e., the channel correction vector), and its unit is volts (V); It is the zero-crossing time difference between adjacent channels, and the unit is seconds (s); and These are two dimensionless weighting coefficients. The advantage of this formula is that it does not consider only the time difference or voltage difference in isolation, but rather weights and fuses the voltage amplitude difference reflecting steady-state imbalance and the timestamp difference reflecting dynamic phase drift. This allows the phase compensation decision to take into account both sources of error simultaneously, improving the accuracy of synchronization and adaptability to complex operating conditions.

[0052] Weighting coefficient and The values ​​were set to 0.6 and 0.4. This set of values ​​was determined based on sensitivity analysis of the system model. In the simulation environment, voltage imbalance and phase difference were independently introduced into the model, and the critical values ​​leading to system out-of-synchronization were observed. Analysis showed that in this system, voltage imbalance has a slightly greater impact on synchronization stability than time drift, therefore it was assigned a slightly higher weight of 0.6. The sum of the two weights is 1.0, constituting the normalization process.

[0053] Before performing calculations, it is necessary to distinguish between units. and The data is processed to allow for a weighted summation within the same framework. The processing rule here is to convert both to unitless per-unit values. The voltage conversion reference is the system's rated line voltage, set at 400V. The time conversion reference is one complete cycle of the AC power grid, which is 20ms for a 50Hz system.

[0054] The calculation process for the example is as follows: 1. Obtain the channel correction vector from the instruction parsing submodule. Calculate its magnitude (amplitude): 2. Perform per-unit processing: 3. Obtain the time difference from the phase calculation submodule. 4. [The sentence is incomplete and requires more context to be translated accurately.] Perform per-unit processing: 5. Substitute the standardized values ​​and weighting coefficients into the formula: 6. Convert the unitless calculation result back to the physical unit radian, and multiply by... : rad.

[0055] The result Radius is the amount of phase compensation that ultimately needs to be applied to the PWM carrier of the backup unit.

[0056] Next, when establishing the linear compensation relationship between the phase difference and the voltage vector, the submodule introduces another dynamic weight allocation coefficient determined by the operating frequency. The calculation formula is as follows: . In the formula, These are the weighting coefficients of the voltage vector; This is the current operating frequency of the converter, measured in kilohertz (kHz). It is a natural constant. The innovation of this formula lies in its ability to adaptively adjust the voltage compensation level according to the converter's operating frequency. As the operating frequency... As the voltage increases, factors such as nonlinear effects and dead time during the switching process have a more significant impact on voltage output accuracy. This is further explained by the exponential term. The weight of the voltage vector in the final compensation decision can be increased nonlinearly, thereby achieving stronger voltage correction at high frequencies and improving the dynamic performance and output accuracy of the system.

[0057] The above embodiments illustrate preferred embodiments of the present invention. Any equivalent adjustments to the technical solution based on software engineering methods are within the scope of protection, including but not limited to: implementing algorithm logic using different programming languages, refactoring functional modules into services, adjusting data interaction protocols, and optimizing resource scheduling strategies. Any implementation scheme derived from reasonable modifications to the data processing flow, service call chain, or system architecture layer without departing from the core technology of the present invention should be considered within the scope of protection defined by the claims of the present invention.

Claims

1. A power dispatching system based on multi-level DC chopping technology, characterized in that, The system includes: The flux linkage trend analysis module is used to construct a power integral curve by collecting the voltage and current sampling values ​​at the inductor terminals, calculate the flux linkage increment difference using the sliding window difference method, determine the correlation between the flux linkage direction vector and the current rate of change vector using the Pearson correlation coefficient method, generate a flux linkage imbalance trigger signal and transmit it to the duty cycle constraint module. The duty cycle constraint module is used to receive the flux imbalance trigger signal, call the flux direction vector, adjust the duty cycle limit range using an exponential smoothing prediction model, perform linear interpolation compensation correction based on the phase angle difference, and output the duty cycle correction parameters to the standby unit evaluation module. The standby unit evaluation module is used to select standby unit groups after obtaining the duty cycle correction parameters, call the sliding standard deviation algorithm to calculate the current fluctuation coefficient, use the linear regression analysis method to process the temperature slope, generate a standby unit activation command and transmit it to the path synchronization coordination module. The path synchronization coordination module is used to receive the backup unit activation command, call the duty cycle correction parameters, construct the voltage difference vector and monitor the current zero-crossing time difference, calculate the phase offset through the dynamic time warping algorithm and the weighted fusion algorithm, and output the phase compensation coefficient for converter control.

2. The power dispatching system based on multi-level DC chopper technology according to claim 1, characterized in that, The flux imbalance trigger signal specifically includes the correlation threshold, direction vector polarity, and time window length. The duty cycle correction parameters include the amplitude limiting range, phase offset, and smoothing coefficient. The backup unit activation command specifically refers to the current fluctuation threshold temperature-current fluctuation correlation threshold and the unit group number. The phase compensation coefficient includes the voltage difference amplitude, time difference value, and carrier phase difference.

3. The power dispatching system based on multi-level DC chopper technology according to claim 2, characterized in that, The calculation formula for the weighted fusion algorithm is as follows: ； in, This is the phase offset. Represents the weighting coefficient of voltage difference amplitude. The time difference is the weighting coefficient. This represents the time difference between the zero-crossing points of adjacent channels.

4. The power dispatching system based on multi-level DC chopper technology according to claim 3, characterized in that, The magnetic flux linkage trend analysis module includes: The signal acquisition submodule acquires the voltage and current sampling values ​​at the inductor terminal, performs sliding mean filtering on the voltage signal to smooth high-frequency noise, converts the current sampling value into an instantaneous power value through multiplication, and performs time-domain integration to construct a power integral curve. Based on the power integral curve, the difference calculation submodule uses a sliding window to extract adjacent data points, calculates the first-order difference value of continuous data points within the window, performs second-order difference operation on the beginning and end endpoints of the window, and accumulates the absolute values ​​of all differences within the window to generate the flux linkage increment difference. The correlation determination submodule calls the flux linkage increment difference sequence and the current change rate sequence to calculate the covariance matrix of the two sequences and obtain the standard deviation parameter. At the same time, it calculates the quotient of the product of the covariance of the two sequences and the standard deviation of the two sequences to obtain the Pearson coefficient. The coefficient is compared with the dynamic threshold through a sliding window. When the number of consecutive cycles reaches the set condition, a flux linkage imbalance trigger signal is generated. The dynamic threshold is obtained by optimizing historical operating data using the gradient descent method, and its determination condition is that the Pearson coefficient is less than 0.7 for three consecutive sampling periods.

5. The power dispatching system based on multi-level DC chopper technology according to claim 4, characterized in that, The duty cycle constraint module includes: The signal parsing submodule receives the flux imbalance trigger signal, calls the flux direction vector, performs sliding window mean filtering on the vector, calculates the standard deviation parameter of the vector in the time domain, and generates a direction feature sequence. The amplitude limiting adjustment submodule uses the exponential smoothing prediction model to perform a weighted average calculation on the time-series duty cycle data based on the directional feature sequence, calculates the absolute deviation between the current duty cycle and the predicted value, and adjusts the amplitude range of the upper and lower limits of the duty cycle according to the deviation value to generate a dynamic amplitude limiting interval. The phase correction submodule calls the dynamic limiting interval, calculates the phase angle difference between the current duty cycle and the limiting boundary, uses linear interpolation to calculate the compensation coefficient for the difference, and outputs the duty cycle correction parameters.

6. The power dispatching system based on multi-level DC chopper technology according to claim 5, characterized in that, The weighting coefficients of the exponential smoothing prediction model The value range is 0.2-0.5, and its specific value satisfies ,in Let | be the Pearson correlation coefficient. | represents the absolute value of the Pearson correlation coefficient.

7. The power dispatching system based on multi-level DC chopper technology according to claim 6, characterized in that, The backup unit evaluation module includes: The unit group filtering submodule obtains the duty cycle correction parameter, establishes a mapping relationship between the correction parameter and the unit group number, and adopts a dual threshold filtering mechanism to mark unit groups with parameter values ​​below the lower threshold as failure status and unit groups with parameter values ​​above the upper threshold as overload status, retains unit groups with parameter values ​​in the dual threshold range, and generates valid unit group identifiers. The fluctuation coefficient calculation submodule extracts the current sampling sequence corresponding to the effective unit group identifier, sets an adaptive window width adjustment mechanism, calculates the standard deviation of each window in time order, performs an arithmetic mean operation on the standard deviations of continuous windows, and generates fluctuation characteristic values. The instruction generation submodule collects time-series data from the temperature sensor, constructs a linear regression equation with the fluctuation characteristic value as the dependent variable, calculates the slope parameter of the temperature change rate and the fluctuation characteristic, and when the absolute value of the slope exceeds the temperature-fluctuation correlation benchmark, it encapsulates the unit group number, activation timestamp, and slope parameter into a data packet to generate a backup unit activation instruction.

8. The power dispatching system based on multi-level DC chopper technology according to claim 7, characterized in that, The adaptive window width adjustment mechanism satisfies ; in, To adjust the coefficient, .

9. The power dispatching system based on multi-level DC chopper technology according to claim 8, characterized in that, The path synchronization coordination module includes: The instruction parsing submodule parses the binary data stream in the backup unit enable instruction, separates the duty cycle correction parameter from the unit group address code, matches the converter control channel according to the address code, converts the correction parameter into three-phase voltage difference vector components, and generates a channel correction vector. The phase calculation submodule collects sampling points of current waveforms from multiple channels, detects the time stamp difference of zero crossing points of adjacent channels, uses a dynamic time warping algorithm to align the current sequences of differentiated channels, calculates the cumulative distance of the optimal path, and generates the phase offset. The compensation control submodule inputs the phase offset and the channel correction vector into the compensation coefficient calculation model, establishes a linear compensation relationship between the phase difference and the voltage vector, updates the converter PWM modulation parameters according to the control cycle, and generates the phase compensation coefficient.

10. The power dispatching system based on multi-level DC chopper technology according to claim 9, characterized in that, The weight allocation coefficients of the linear compensation relationship satisfy... ; in, These are the voltage vector weighting coefficients. For the inverter's operating frequency, It is a natural constant.

Images