A Power System Equalization Scheduling Method and System Based on Fuzzy Health Determination
By collecting electrical and thermal parameters of the power supply modules, calculating the equivalent aging time and health score, and generating a dynamic power-on/off sequence, the problem of uneven module aging is solved, and intelligent balanced scheduling of high-power microwave power supply systems is realized, extending service life and reducing failure risk.
Patent Information
- Application Number
- CN202611569478.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-12
- Publication Date
- 2026-11-10
AI Technical Summary
In traditional high-power microwave power supply systems, module aging is polarized, health assessment is incomplete, and module priority cannot be dynamically adjusted, resulting in high operation and maintenance costs and a high risk of sudden shutdown.
By collecting the electrical and thermal parameters of the module, the equivalent aging time is calculated using an adaptive exponential smoothing algorithm, a fuzzy evaluation matrix is constructed, and the health score is calculated using the AHP and TOPSIS algorithms. A dynamic power-on/off sequence is generated, and power closed-loop processing is performed during the module switching process.
It achieves a synchronous balance between module wear and aging, extends the service life of the whole machine, reduces the probability of sudden failures, simplifies operation and maintenance, and saves costs.
Smart Images

Figure CN123074072A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of high-power power supply technology, and more specifically to a power system equalization scheduling method and system based on fuzzy health determination. Background Technology
[0002] Currently, high-power microwave power supply systems (such as megawatt-level power supplies for industrial microwave heating and radar transmitters) commonly employ multi-module parallel expansion to increase overall output power and achieve redundancy backup. However, traditional parallel power supply systems have three core defects: First, they use fixed timing logic to set the order of module power-on and power-off. The modules that start first are subjected to power-on surges and standby losses for a long time, while the modules that follow are idle for a long time, leading to polarized aging of modules and a significant reduction in the mean time between failures (MTBF) of the entire system. Second, they only achieve even distribution of steady-state output current and only collect single parameters such as output voltage and load current, without considering the individual power supplies. The system comprehensively evaluates multiple performance indicators such as internal temperature rise of the source module, switching transistor loss, output linearity, and protection threshold offset; it cannot quantify the degree of module health degradation, and only triggers protection after a complete module failure, lacking early prediction and load transfer mechanisms; thirdly, the existing solution cannot quantify the cumulative working time and performance health degradation of each module, and cannot dynamically adjust the priority of module commissioning / exit; the system load distribution and start-up / shutdown sequence are fixed, and it is impossible to dynamically balance losses based on the real-time status of modules, and it is impossible to achieve synchronous aging and replacement of all modules in the entire unit, resulting in high operation and maintenance costs and a high risk of sudden shutdown.
[0003] Therefore, there is an urgent need for an intelligent balanced scheduling method and system that can comprehensively quantify the equivalent aging time and multi-dimensional health of each parallel power module, and dynamically generate start-stop sequences through multi-objective decision-making, in order to solve the pain points of uneven module losses and lack of comprehensive health assessment. Summary of the Invention
[0004] To overcome the aforementioned technical problems, this invention provides a balanced scheduling method and system for power systems based on fuzzy health-related decision-making. This balanced scheduling method is applicable to high-power microwave power systems containing multiple parallel power modules. By collecting the physical start-up and shutdown times of each module and multi-dimensional original parameters such as electrical and thermal parameters, an adaptive exponential smoothing algorithm incorporating a power-on impact loss correction coefficient is used to calculate the equivalent aging time and equivalent time deviation. An evaluation index set including output linear margin and IGBT temperature rise margin is constructed, and a fuzzy evaluation matrix is generated through membership function mapping. The Saaty 1-9 scaling method and consistency verification are used to obtain the AHP index weights, and the health score of each module is calculated. A decision matrix is constructed using the health score and time deviation, and the comprehensive proximity is calculated through TOPSIS multi-objective decision-making. A dynamic power-on sequence is generated in descending order, and a power-off sequence is generated in reverse order. A PID algorithm is used for power closed-loop processing during module switching. This achieves synchronous balance between the losses and aging degrees of parallel modules, extends the overall service life of the system, and reduces the probability of sudden failures.
[0005] To achieve the above objectives, the present invention provides a power system equalization scheduling method based on fuzzy health determination, wherein the power system includes multiple power modules connected in parallel, and the equalization scheduling method includes: Collect the electrical operating parameters of each power module; An adaptive duration estimation algorithm is used to obtain the equivalent aging duration of each power module based on electrical operating parameters, and the equivalent duration deviation of each power module is calculated accordingly. The health score of each power module was obtained by using the fuzzy evaluation matrix and the AHP (Analytic Hierarchy Process). The TOPSIS algorithm is used to construct a decision matrix based on the equivalent duration deviation and health score. The normalized decision matrix is used to obtain the positive ideal solution and the negative ideal solution. The comprehensive proximity of each power module is obtained based on the positive ideal solution and the negative ideal solution. The comprehensive proximity is the ratio of the Euclidean distance from the power module to the negative ideal solution to the sum of the Euclidean distances to the positive ideal solution and the negative ideal solution. The current power-on sequence is generated by sorting each power module in descending order based on their overall proximity, and the current power-off sequence is generated in reverse order. Based on the current power-on and power-off sequences, a PID algorithm is used to perform closed-loop power processing for each power module.
[0006] Preferably, the electrical operating parameters include at least one of the following: the physical duration of operation during each start-up and shutdown, the voltage output under different load conditions at rated input, the junction temperature of the IGBT device, the voltage and current of the input DC bus, and the actual OVP trigger voltage.
[0007] Preferably, an adaptive exponential smoothing cumulative duration estimation algorithm is used to obtain the equivalent aging duration of each power module based on electrical operating parameters, and the equivalent duration deviation of each power module is calculated accordingly, including: Obtain the physical duration of operation for each power module during each start-up and shutdown; Based on the physical duration of operation during each start-up and shutdown, an adaptive duration estimation algorithm is used to obtain the equivalent aging duration of the power module using formula (1). (1) in, For the power module index, Index for the number of runs, For the first The power module is in the first Equivalent aging time for each run For the first The power module is in the first Equivalent aging time for each run For exponentially smoothed forgetting factor, This is the correction factor for power-on surge loss. For the first The power module is the first Physical duration of each run; The equivalent duration deviation of each power module is obtained using formulas (2)-(3). (2) (3) in, This serves as the baseline for the overall average equivalent aging of the power supply system. This represents the total number of power modules. For the first Equivalent aging time of each power module For the first Equivalent duration deviation of each power module.
[0008] Preferably, the balanced scheduling method further includes: An evaluation index set is constructed based on electrical operating parameters. The evaluation index set includes the output linear margin of benefit-type indicators and the temperature rise margin of IGBT devices, and the no-load standby loss and protection threshold offset of cost-type indicators. The load regulation rate is obtained using formula (4) based on the voltage output and nominal rated output voltage of the power module under different load conditions at rated input. The output linear margin of the power module is obtained using formula (5). (4) (5) in, For the load regulation rate of the power module, This refers to the maximum voltage output of the power module under rated input conditions. This is the minimum voltage output of the power module under rated input. The nominal rated output voltage, This is the factory-permitted maximum load regulation rate; To output linear margin; The temperature rise margin of the IGBT device is obtained based on the junction temperature of the IGBT device in the power module using formula (6). (6) in, This refers to the junction temperature limit of IGBT devices. This refers to the junction temperature of the IGBT device. The no-load standby loss is obtained using formula (7) based on the voltage and current of the input DC bus of the power supply module. (7) in, This refers to the power loss during idle standby. The input voltage is the DC bus input voltage. For the input DC bus current; The protection threshold offset is obtained using formula (8) based on the actual OVP trigger voltage of the power module. (8) in, To protect the threshold offset, This is the actual OVP trigger voltage of the power module. This refers to the nominal OVP threshold of the power module. The benefit-type indicators and cost-type indicators of each power module are mapped according to formulas (9)-(10) to obtain the fuzzy membership degree of the indicators. (9) (10) in, For fuzzy membership functions of benefit-type indicators, These are the original values for the current benefit-type indicators. This represents the historical minimum value of the current benefit-related indicators. This represents the historical maximum value of the current benefit-related indicators. For cost-type indicators, fuzzy membership function. These are the original values for the current cost-related indicators. This is the historical minimum value of the current cost-related indicator. This represents the historical maximum value of the current cost-related indicator. The fuzzy evaluation matrix of the power system is constructed based on the fuzzy membership degrees of all indicators of all power modules.
[0009] Preferably, the health score of each power module is obtained using the AHP (Analytic Hierarchy Process) method based on the fuzzy evaluation matrix, including: The Saaty1–9 scaling method was used to construct pairwise comparison judgment matrices between the four indicators of each power module; Based on pairwise comparisons, the maximum eigenvalue is obtained using formula (11) from the judgment matrix. (11) in, The largest eigenvalue, For the number of indicators, For indexing indicators, This is a pairwise comparison judgment matrix. The initial weight vector, The first step is to compare the product of the pairwise comparison judgment matrix and the initial weight vector. One portion, For the first The current weight of each indicator; The consistency ratio is obtained using formula (12) based on the largest eigenvalue; (12) in, To determine the consistency index of the matrix by comparing each pair of matrices, To determine the consistency ratio of the matrices by comparing them pairwise, This is the average random consistency index value based on the number of indicators; Determine if the consistency ratio is greater than the preset verification threshold. If the consistency ratio is not greater than the preset verification threshold, reconstruct the pairwise comparison judgment matrix; If the consistency ratio is greater than the preset verification threshold, the pairwise comparison judgment matrix is normalized using formula (13) to obtain the corresponding index weight vector. (13) in, For the indicator weight vector, To output the index weight vector components corresponding to the linear residual, These are the weighted vector components of the index corresponding to the temperature rise margin of IGBT devices. These are the weighted vector components of the index corresponding to idle standby power loss. To protect the index weight vector components corresponding to the threshold offset, For the first The indicator weight vector of each indicator; The health score of each power module is obtained using formula (14) based on the fuzzy evaluation matrix and the index weight vector. (14) in, For the first Health score of each power module For fuzzy membership functions, For the first The power module's first Fuzzy membership degree of each indicator.
[0010] Preferably, the TOPSIS algorithm is used to construct a decision matrix based on the equivalent duration deviation and health score. The normalized decision matrix is used to obtain positive and negative ideal solutions. Based on the positive and negative ideal solutions, the overall proximity of each power module is calculated, including: Based on the equivalent duration deviation and health score, a decision matrix is constructed using formula (15). (15) in, For the decision matrix, For the first The health score of the power module is the first one. The first attribute of a power module For the first The equivalent duration deviation of the power module is the first... The second attribute of a power module, This represents the total number of power modules; The decision matrix is vector normalized using formula (16). (16) in, The elements of the normalized decision matrix are the _th _th_ element. The power module's first One attribute, The original element of the decision matrix is the i-th element. The power module's first Attributes The positive and negative ideal solutions in the power supply system are determined using formula (17). (17) in, This is the ideal solution in the power supply system. For the normalized first Health score of each power module For the normalized first Equivalent duration deviation of each power module To find the maximum value function, The function is for finding the minimum value; The Euclidean distances from each power module to the positive and negative ideal solutions are obtained using formula (18). (18) in, For the first The Euclidean distance from each power module to the ideal solution. For the first The Euclidean distance from each power module to the negative ideal solution. The first component of the positive ideal solution is the maximum health score among all normalized power modules. The second component of the positive ideal solution is the minimum equivalent duration deviation among all power modules after normalization. The first component of the negative ideal solution is the minimum health score among all power modules after normalization. The second component of the negative ideal solution is the maximum value of the equivalent duration deviation among all power modules after normalization. The overall proximity of each power module is obtained using formula (19) based on the Euclidean distance of each power module. (19) in, For the first Overall proximity of the power modules.
[0011] Preferably, a PID algorithm is used to perform power closed-loop processing for each power module based on the current power-on and power-off sequences, including: Obtain the target power and determine whether the current power system is in power output ramp-up mode or power output ramp-down mode; Once the power system is confirmed to be in power output ramp-up mode, the corresponding power modules are turned on sequentially according to the current power-on sequence, and the power output is increased synchronously with a preset fixed ramp-up step size, and the current sampling power is obtained in real time. Determine whether the current sampling power is not less than the difference between the target power and the preset first threshold; When the current sampling power is not less than the difference between the target power and the preset first threshold, exit the power output gradual increase mode and switch to the fine segmented closed-loop adjustment stage to adjust the power until the output power reaches the target power. The power system is determined to be in power output easing mode. The power output is reduced by a preset fixed easing step size, and the corresponding power modules are shut down synchronously in the current shutdown sequence. The current sampled power is obtained in real time. Determine whether the current sampling power is not greater than the sum of the target power and the preset first threshold; When the current sampling power is not greater than the sum of the target power and the preset first threshold, the power output slow-down mode is exited and the fine segmented closed-loop adjustment stage is entered to adjust the power until the output power reaches the target power.
[0012] Preferably, power regulation is performed by entering a fine-segmented closed-loop regulation stage until the output power reaches the target power, including: The sampling power is obtained according to the preset frequency; The power deviation between the sampled power and the target power is obtained according to formula (20). (20) in, For the first Power deviation per sampling period For the first The absolute value of the power deviation over each sampling period. For the target power, For the first Sampling power per sampling period; The power adjustment step size is determined using formula (21) based on the absolute value of the power deviation. ,(twenty one) in, For the first The adjustment step size for each sampling period, For symbolic functions, This is an adjustment function based on the absolute value of the power deviation; Adjust the output power according to formula (22) until the output power reaches the target power. ,(twenty two) in, For the first Output power per sampling period For the first The output power of each sampling period. Preferably, the equalization scheduling method includes: Determine whether the current power system is experiencing one of three conditions: power-on initialization, the health score of any of the power modules is lower than a preset warning score threshold, or a change in the number of power modules. If any of the following three conditions occur in the current power system: power-on initialization, health score of any power module falling below the preset warning score threshold, or change in the number of power modules, the step of collecting electrical operating parameters of each power module shall be re-executed.
[0013] A second aspect of the present invention provides a power system balancing scheduling system based on fuzzy health determination, wherein the power system includes multiple power modules connected in parallel, and the scheduling system includes: The acquisition unit is used to collect the electrical operating parameters of each power module; The equivalent aging calculation unit is used to obtain the equivalent aging time of each power module based on the electrical operating parameters using an adaptive duration estimation algorithm, and to calculate the equivalent duration deviation of each power module accordingly. The fuzzy evaluation matrix construction unit is used to construct an evaluation index set based on electrical operating parameters and generate a fuzzy evaluation matrix based on the evaluation index set. The health score calculation unit is used to obtain the health score of each power module based on the fuzzy evaluation matrix using the AHP (Analytic Hierarchy Process). The TOPSIS decision unit is used to construct a decision matrix based on the equivalent duration deviation and health score using the TOPSIS algorithm, normalize the decision matrix to obtain positive and negative ideal solutions, calculate the comprehensive proximity of each power module based on the positive and negative ideal solutions, sort the modules in descending order based on the comprehensive proximity to generate the current power-on sequence, and generate the current power-off sequence in reverse order. The comprehensive proximity is the ratio of the Euclidean distance from the power module to the negative ideal solution to the sum of the Euclidean distances to the positive and negative ideal solutions. The PID power closed-loop unit is used to perform power closed-loop processing for each power module based on the current power-on and power-off sequences using a PID algorithm, so as to achieve intelligent balanced scheduling of the power system.
[0014] The above technical solution involves first collecting the physical runtime of each module during start-up and shutdown, as well as raw parameters such as output voltage, IGBT junction temperature, and ambient temperature under different load conditions under rated input. An adaptive exponential smoothing algorithm is used to calculate the equivalent aging time of each module and determine the duration deviation, quantifying the aging deviation between modules. Based on the raw parameters, a four-dimensional evaluation index set is constructed, including output linear margin, IGBT temperature rise margin, no-load standby loss, and protection threshold offset. Each index is mapped to an interval using benefit-type and cost-type membership functions, generating a fuzzy evaluation matrix. A pairwise comparison judgment matrix is constructed using the Saaty 1-9 scaling method. After consistency verification, the index weight vector is obtained through eigenvector normalization. The fuzzy evaluation matrix and the weight vector are then fused to calculate the health score of each module. A two-dimensional decision matrix is constructed, with the health score as a benefit attribute and the duration deviation as a cost attribute. After vector normalization, positive and negative ideal solutions are determined. The Euclidean distance from each module to the positive and negative ideal solutions is calculated, and the comprehensive proximity of each module is obtained through a comprehensive proximity formula. Dynamic power-on sequences are generated in descending order, and power-off sequences are generated in reverse order. During module switching, a PID algorithm is used for power closed-loop processing. Through the synergy of gradual increase / decrease mode and fine segmented variable step size closed-loop regulation, a smooth transition of output power is achieved. Three types of trigger conditions are monitored: power-on initialization, module health score below the warning threshold, and changes in the number of modules. If any one of these conditions is met, the entire process is re-executed for reordering. This method achieves synchronous balancing of the working losses and aging levels of all parallel power modules, significantly extends the mean time between failures (MTBF) of the entire system, reduces output power fluctuations during module switching, reduces the probability of sudden module failures, simplifies operation and maintenance, and saves costs. Attached Figure Description
[0015] Figure 1 This is a connection block diagram of a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention. Figure 2 This is a connection block diagram for obtaining the equivalent duration deviation of a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention. Figure 3 This is a connection block diagram for constructing a fuzzy evaluation matrix in a power system balancing scheduling method based on fuzzy health degree decision according to an embodiment of the present invention. Figure 4 This is a connection block diagram illustrating the acquisition of health scores in a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention. Figure 5 This is a connection block diagram for obtaining the comprehensive proximity degree of a power system balancing scheduling method based on fuzzy health degree decision according to an embodiment of the present invention. Figure 6This is a connection block diagram of a power system balancing scheduling method based on fuzzy health decision according to an embodiment of the present invention, which performs power closed-loop processing. Figure 7 This is a connection block diagram illustrating the implementation of an output power regulation method for a power system based on fuzzy health determination, according to an embodiment of the present invention. Figure 8 This is a connection block diagram of the module dynamic reordering of a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention. Detailed Implementation
[0016] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.
[0017] like Figure 1 The diagram shown is a connection block diagram of a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention; Figure 1 The balanced scheduling method includes: In step S10, the electrical operating parameters of each power module are collected; In step S11, an adaptive duration estimation algorithm is used to obtain the equivalent aging duration of each power module based on the electrical operating parameters, and the equivalent duration deviation of each power module is calculated accordingly. In step S12, the health score of each power module is obtained by using the AHP (Analytic Hierarchy Process) method based on the fuzzy evaluation matrix. In step S13, the TOPSIS algorithm is used to construct a decision matrix based on the equivalent duration deviation and health score. The decision matrix is normalized to obtain the positive ideal solution and the negative ideal solution. The comprehensive proximity of each power module is obtained based on the positive ideal solution and the negative ideal solution. The comprehensive proximity is the ratio of the Euclidean distance from the power module to the negative ideal solution to the sum of the Euclidean distances to the positive ideal solution and the negative ideal solution. In step S14, the current power-on sequence is generated by sorting each power module in descending order based on the overall proximity, and the current power-off sequence is generated in reverse order. In step S15, a PID algorithm is used to perform closed-loop power processing for each power module based on the current power-on sequence and power-off sequence, so as to realize intelligent balanced scheduling of the power system.
[0018] Step S10 collects the physical running time of each parallel power module during start-up and shutdown, as well as the output voltage, IGBT junction temperature, input DC bus voltage and current, and actual overvoltage protection trigger voltage under different load conditions under rated input, as the data basis for subsequent equivalent aging time calculation and health assessment.
[0019] Step S11 calculates the equivalent aging time of each module by using an adaptive exponential smoothing algorithm based on the physical runtime of each start-up and shutdown, and calculates the equivalent time deviation of each module relative to the average equivalent aging benchmark of the whole machine, thereby quantifying the aging deviation between modules.
[0020] Step S12 constructs pairwise comparison judgment matrices for the four indicators using the Saaty1-9 scaling method. After consistency verification, the indicator weight vectors are obtained by normalizing the feature vectors. The fuzzy evaluation matrix and the weight vectors are then fused to calculate the health score of each power module.
[0021] Step S13 constructs a two-dimensional decision matrix using health score as benefit attribute and equivalent duration deviation as cost attribute. After vector normalization, positive and negative ideal solutions are determined. The Euclidean distance from each module to the positive and negative ideal solutions is calculated, and the overall proximity is obtained. The startup sequence is generated in descending order of overall proximity, and the shutdown sequence is generated in reverse order.
[0022] Step S14 generates the power-on sequence in descending order of overall proximity and the power-off sequence in reverse order.
[0023] Step S15 uses a PID algorithm for power closed-loop processing during module switching based on the dynamic power-on and power-off sequences. Through the synergy of gradual increase / decrease mode and fine segmented closed-loop adjustment, the output power is smoothly transitioned and accurately tracked, avoiding power oscillations caused by module switching.
[0024] Through the above technical solution, the balanced scheduling method is applicable to high-power microwave power supply systems containing multiple parallel power modules. By collecting the physical duration of each module's start-up and shutdown, as well as multi-dimensional raw parameters such as electrical and thermal parameters, an adaptive exponential smoothing algorithm with a power-on impact loss correction coefficient is used to calculate the equivalent aging time and equivalent time deviation. An evaluation index set including output linear margin and IGBT temperature rise margin is constructed, and a fuzzy evaluation matrix is generated through membership function mapping. The Saaty 1-9 scaling method and consistency verification are used to obtain the AHP index weights and calculate the health score of each module. A decision matrix is constructed using the health score and time deviation, and the comprehensive proximity is calculated through TOPSIS multi-objective decision-making. A dynamic start-up sequence is generated in descending order, and a shutdown sequence is generated in reverse order. A PID algorithm is used for power closed-loop processing during module switching. This achieves synchronous balance between the losses and aging degree of parallel modules, extends the service life of the entire unit, and reduces the probability of sudden failures.
[0025] To clarify the scope and types of electrical operating parameters to ensure the algorithm is feasible, in one embodiment of the present invention, the electrical operating parameters include at least one of the following: physical duration of operation during each start-up and shutdown, voltage output under different load conditions at rated input, junction temperature of IGBT devices, voltage and current of the input DC bus, and actual OVP trigger voltage.
[0026] Collect various historical operating data, including the physical duration of each power module's start-up and shutdown, the output voltage sampled at multiple points within the 10%~100% load range under rated input, the junction temperature of IGBT devices, the voltage and current of the input DC bus, and the actual overvoltage protection trigger voltage. These parameters are all monitoring signals that high-power microwave power modules are conventionally available and can be obtained without adding extra hardware. By comprehensively collecting these raw parameters, a complete data foundation is provided for the subsequent calculation of equivalent aging time, ensuring that the entire intelligent balancing scheduling method can be directly implemented in engineering, and that the calculation results can truly reflect the actual operating status and aging degree of each module.
[0027] like Figure 2 The diagram shown is a connection block diagram for obtaining the equivalent duration deviation in a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention; Figure 2 In order to quantify start-stop impact losses and overcome the polarization of fixed-sequence aging, in one embodiment of the present invention, an adaptive exponential smoothing cumulative duration estimation algorithm is used to obtain the equivalent aging duration of each power module based on electrical operating parameters, and the equivalent duration deviation of each power module is calculated accordingly, which may include the following steps: In step S20, the physical duration of operation for each power module during each start-up and shutdown is obtained; In step S21, the equivalent aging time of the power module is obtained using formula (1) based on the physical duration of operation during each start-up and shutdown. (1) in, For the power module index, Index for the number of runs, For the first The power module is in the first Equivalent aging time for each run For the first The power module is in the first Equivalent aging time for each run For exponentially smoothed forgetting factor, This is the correction factor for power-on surge loss. For the first The power module is the first Physical duration of each run; In step S32, the equivalent duration deviation of each power module is obtained using formulas (2)-(3). (2) (3) in, This serves as the baseline for the overall average equivalent aging of the power supply system. This represents the total number of power modules. For the first Equivalent aging time of each power module For the first Equivalent duration deviation of each power module.
[0028] The system obtains the physical duration of each module's start-up and shutdown, introduces an exponential smoothing forgetting factor and a power-on surge loss correction coefficient, and calculates the equivalent aging duration of each module using an iterative formula. This leads to the determination of the overall average equivalent aging baseline and the equivalent deviation duration of each module. The power-on surge loss correction coefficient equates a single power-on surge to several times the aging loss of normal operation, overcoming the shortcomings of traditional methods that simply accumulate physical durations without quantifying start-up and shutdown impacts. The exponential smoothing forgetting factor preserves historical aging data and mitigates short-term interference. The resulting equivalent duration deviation accurately reflects the aging deviation between modules.
[0029] like Figure 3 The diagram shown is a connection block diagram for constructing a fuzzy evaluation matrix in a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention; Figure 3 In one embodiment of the present invention, the balanced scheduling method includes: In step S30, an evaluation index set is constructed based on electrical operating parameters. The evaluation index set includes the output linear margin of benefit-type indicators and the temperature rise margin of IGBT devices, and the no-load standby loss and protection threshold offset of cost-type indicators. In step S31, the load regulation rate is obtained using formula (4) based on the voltage output and nominal rated output voltage of the power module under different load conditions at rated input, and the output linearity margin of the power module is obtained using formula (5). (4) (5) in, For the load regulation rate of the power module, This refers to the maximum voltage output of the power module under rated input conditions. This is the minimum voltage output of the power module under rated input. The nominal rated output voltage, This is the factory-permitted maximum load regulation rate; To output linear margin; In step S32, the temperature rise margin of the IGBT device is obtained based on the junction temperature of the IGBT device in the power module using formula (6). (6) in, This refers to the junction temperature limit of IGBT devices. This refers to the junction temperature of the IGBT device. In step S33, the no-load standby loss is obtained using formula (7) based on the voltage and current of the input DC bus of the power module. (7) in, This refers to the power loss during idle standby. The input voltage is the DC bus input voltage. For the input DC bus current; In step S34, the protection threshold offset is obtained based on the actual OVP trigger voltage of the power module using formula (8). (8) in, To protect the threshold offset, This is the actual OVP trigger voltage of the power module. This refers to the nominal OVP threshold of the power module. In step S35, the benefit-type indicators and cost-type indicators of each power module are mapped according to formulas (9)-(10) to obtain the fuzzy membership degree of the indicators. (9) (10) in, For fuzzy membership functions of benefit-type indicators, These are the original values for the current benefit-type indicators. This represents the historical minimum value of the current benefit-related indicators. This represents the historical maximum value of the current benefit-related indicators. For cost-type indicators, fuzzy membership function. These are the original values for the current cost-related indicators. This is the historical minimum value of the current cost-related indicator. This represents the historical maximum value of the current cost-related indicator. In step S36, a fuzzy evaluation matrix of the power system is constructed based on the fuzzy membership degrees of all indicators of all power modules.
[0030] Based on the collected electrical operating parameters, a set of four evaluation indicators is constructed, including benefit-based indicators (output linear margin, IGBT temperature rise margin) and cost-based indicators (no-load standby loss, protection threshold offset). Output linear margin is obtained by comparing the load regulation rate with the factory-allowed maximum load regulation rate, reflecting the module's margin from the performance failure boundary. IGBT temperature rise margin is obtained by the difference between the junction temperature limit and the current junction temperature, reflecting the device's lifespan reserve. No-load standby loss is obtained by the product of the input DC bus voltage and current, reflecting the aging degree of the switching transistor drive and core losses. Protection threshold offset is obtained by the absolute value of the difference between the actual overvoltage protection trigger voltage and the nominal threshold, reflecting the temperature drift aging of the reference source and voltage divider resistors. The four indicators are mapped to the [0, 1] interval using benefit-based and cost-based membership functions, constructing a fuzzy evaluation matrix to provide standardized input for subsequent AHP health assessment.
[0031] like Figure 4 The diagram shown is a connection block diagram illustrating the acquisition of health scores in a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention; Figure 4 In order to obtain the objective weights of the indicators and facilitate the subsequent calculation of health scores, in one embodiment of the present invention, the health score of each power module is obtained by using the AHP (Analytic Hierarchy Process) based on the fuzzy evaluation matrix, which may include the following steps: In step S40, the Saaty1–9 scaling method is used to construct a pairwise comparison judgment matrix between the four indicators of each power module; In step S41, the maximum eigenvalue is obtained using formula (11) based on the pairwise comparison judgment matrix. (11) in, The largest eigenvalue, For the number of indicators, For indexing indicators, This is a pairwise comparison judgment matrix. The initial weight vector, The first step is to compare the product of the pairwise comparison judgment matrix and the initial weight vector. One portion, For the first The current weight of each indicator; In step S42, the consistency ratio is obtained using formula (12) based on the largest eigenvalue; (12) in, To determine the consistency index of the matrix by comparing each pair of matrices, To determine the consistency ratio of the matrices by comparing them pairwise, This is the average random consistency index value based on the number of indicators; In step S43, it is determined whether the consistency ratio is greater than a preset verification threshold. In step S44, if the consistency ratio is not greater than the preset verification threshold, the pairwise comparison judgment matrix is reconstructed. In step S45, if the consistency ratio is greater than the preset verification threshold, the pairwise comparison judgment matrix is normalized using formula (13) to obtain the corresponding index weight vector. (13) in, For the indicator weight vector, To output the index weight vector components corresponding to the linear residual, These are the weighted vector components of the index corresponding to the temperature rise margin of IGBT devices. These are the weighted vector components of the index corresponding to idle standby power loss. To protect the index weight vector components corresponding to the threshold offset, For the first The indicator weight vector of each indicator; In step S46, the health score of each power module is obtained using formula (14) based on the fuzzy evaluation matrix and the index weight vector. (14) in, For the first Health score of each power module For fuzzy membership functions, For the first The power module's first Fuzzy membership degree of each indicator.
[0032] The Saaty 1-9 scaling method is used to construct pairwise comparison judgment matrices for four indicators. The maximum eigenvalue is calculated based on the judgment matrices and the initial weight vectors to obtain the consistency index and consistency ratio. The consistency ratio is then checked against a preset verification threshold (e.g., 0.1). If the ratio is not less than this threshold, the judgment matrix is reconstructed; otherwise, the eigenvectors of the judgment matrix are normalized to obtain the indicator weight vectors. Based on the fuzzy evaluation matrix and the indicator weight vectors, the health score of each module is calculated through weighted summation. This scheme eliminates the subjective contradictions in expert scoring through consistency verification, ensuring the rationality of the weight vectors. By integrating the fuzzy evaluation matrix (objective data) with the AHP weight vector (subjective experience), the objectivity of the actual performance of each module is preserved, while incorporating the subjective judgment of the importance of various indicators to power supply lifespan, enabling the health score to scientifically and comprehensively reflect the true health status of the modules.
[0033] like Figure 5 The diagram shown is a connection block diagram for obtaining the comprehensive proximity score in a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention; Figure 5 In order to comprehensively quantify the quality of power modules and facilitate the generation of dynamic power-on sequences for subsequent power modules, in one embodiment of the present invention, the TOPSIS algorithm is used to construct a decision matrix based on the equivalent duration deviation and health score. The decision matrix is normalized to obtain positive and negative ideal solutions. The comprehensive proximity of each power module is obtained based on the positive and negative ideal solutions, which may include the following steps: In step S50, a decision matrix is constructed using formula (15) based on the equivalent duration deviation and health score. (15) in, For the decision matrix, For the first The health score of the power module is the first one. The first attribute of a power module For the first The equivalent duration deviation of the power module is the first... The second attribute of a power module, This represents the total number of power modules; In step S51, the decision matrix is vector normalized using formula (16). (16) in, The elements of the normalized decision matrix are the _th _th_ element. The power module's first One attribute, The original element of the decision matrix is the i-th element. The power module's first Attributes In step S52, the positive and negative ideal solutions in the power supply system are determined using formula (17). (17) in, This is the ideal solution in the power supply system. For the normalized first Health score of each power module For the normalized first Equivalent duration deviation of each power module To find the maximum value function, The function is for finding the minimum value; In step S53, the Euclidean distance from each power module to the positive ideal solution and the negative ideal solution is obtained using formula (18). (18) in, For the first The Euclidean distance from each power module to the ideal solution. For the first The Euclidean distance from each power module to the negative ideal solution. The first component of the positive ideal solution is the maximum health score among all normalized power modules. The second component of the positive ideal solution is the minimum equivalent duration deviation among all power modules after normalization. The first component of the negative ideal solution is the minimum health score among all power modules after normalization. The second component of the negative ideal solution is the maximum value of the equivalent duration deviation among all power modules after normalization. In step S54, the overall proximity of each power module is obtained using formula (19) based on the Euclidean distance of each power module. (19) in, For the first Overall proximity of the power modules.
[0034] Using health score as the benefit attribute and equivalent duration deviation as the cost attribute, a two-dimensional decision matrix is constructed. The decision matrix is then vector normalized. Positive ideal solution (maximum health score, minimum duration deviation) and negative ideal solution (minimum health score, maximum duration deviation) are determined. The Euclidean distance from each module to the positive and negative ideal solutions is calculated. Based on the ratio of the distance to the negative ideal solution to the sum of the distances to the positive and negative ideal solutions, the comprehensive proximity of each module is calculated, with a value closer to 1 indicating a better module state. Dynamic startup sequences are generated by sorting modules in descending order of comprehensive proximity, and shutdown sequences are generated in reverse order. The TOPSIS algorithm, by simultaneously considering the geometric distances of "approaching the optimal" and "far from the worst," eliminates the need for manually setting the weights of the dual objectives. The scheduling priority is objectively determined by the multi-dimensional data of the modules themselves, truly achieving the optimal approximation of the two conflicting objectives of "maximizing health" and "minimizing duration deviation," overcoming the shortcomings of the traditional linear weighted model's extreme bias in single objectives.
[0035] In a specific embodiment of the present invention, the startup sequence of the power modules of a power system is determined. The specific process can be as follows: the power system is configured to include four power modules connected in parallel, wherein the junction temperature limit of the IGBT device is 125°C and the factory-permitted maximum load regulation is [not specified]. The percentage is 0.8%, and the upper limit of idle standby power loss is... 450W, nominal OVP threshold The voltage is 19.0 kV, with a maximum permissible offset of ±0.8 kV. Electrical operating parameters of the four power modules were collected, as shown in Table 1 below. Table 1
[0036] As shown in Table 1, the equivalent aging time of power module 1 is 153 hours, the load regulation rate is 0.32%, the IGBT junction temperature is 87℃, the no-load standby power loss is 162W, and the actual OVP trigger voltage is 19.42kV. The equivalent aging time of power module 2 is 113 hours, the load regulation rate is 0.239%, the IGBT junction temperature is 80.58℃, the no-load standby power loss is 113.36W, and the actual OVP trigger voltage is 19.3558kV. The equivalent aging time of power module No. 1 is 164h, the load regulation rate is 0.2858%, the junction temperature of IGBT devices is 84.29℃, the no-load standby loss is 141.48W, and the actual OVP trigger voltage is 19.393kV. The equivalent aging time of power module No. 4 is 157h, the load regulation rate is 0.1523%, the junction temperature of IGBT devices is 73.73℃, the no-load standby loss is 61.38W, and the actual OVP trigger voltage is 19.2873kV. According to formulas (4)-(8), the electrical operating parameters of each power module in Table 1 above are processed to obtain the output linear margin, IGBT device temperature rise margin, no-load standby loss, and protection threshold offset of each power module, as shown in Table 2 below. Table 2
[0037] As shown in Table 2, the output linearity margin of power module 1 is 0.4800%, the temperature rise margin of the IGBT device is 38.00℃, the no-load standby power loss is 162.00W, and the protection threshold offset is 0.4200kV; the output linearity margin of power module 2 is 0.5610%, the temperature rise margin of the IGBT device is 44.42℃, the no-load standby power loss is 113.36W, and the protection threshold offset is 0.3558kV; the output linearity margin of power module 3 is 0.5142%, the temperature rise margin of the IGBT device is 40.71℃, the no-load standby power loss is 141.48W, and the protection threshold offset is 0. .3930kV; The output linear margin of power module No. 4 is 0.6477%, the temperature rise margin of IGBT devices is 51.27℃, the no-load standby loss is 61.38W, and the protection threshold offset is 0.2873kV; Among them, X1 (output linear margin) and X2 (IGBT temperature rise margin) are benefit-type indicators, and the larger the value, the better the performance; X3 (no-load standby loss) and X4 (protection threshold offset) are cost-type indicators, and the smaller the value, the better the performance; Power module No. 4 performs best in all four indicators (X1 and X2 are the highest, X3 and X4 are the lowest), while the indicators of power module No. 1 are relatively weak. According to formula (9)-(10), the electrical operating parameters of each power module in Table 2 above are processed to obtain the fuzzy membership degree of each indicator of each power module, and the indicator weight vector is obtained according to the AHP hierarchical analysis method. The health score of each power module is calculated, and the equivalent duration deviation is calculated based on the equivalent duration of each power module in Table 1. The resulting efficiency and cost indicators of the power system are shown in Table 3 below. Table 3
[0038] As shown in Table 3, the health score of power module 1 is 0.521, and its equivalent duration deviation is 6 hours; the health score of power module 2 is 0.609, and its equivalent duration deviation is 34 hours; the health score of power module 3 is 0.5581, and its equivalent duration deviation is 17 hours; and the health score of power module 4 is 0.703, and its equivalent duration deviation is 10 hours. Based on the above data, the normalization matrix based on the TOPSIS algorithm is obtained by normalizing using formula (16).
[0039] in, This is a normalized matrix; The positive and negative ideal solutions of the power supply system are obtained according to formulas (17)-(18). Positive Ideal Solution
[0040] Negative ideal solution .
[0041] The overall proximity of each power module in the power system is calculated according to formulas (18)-(19), as shown in Table 4 below. Table 4
[0042] As shown in Table 4, the power supply modules in the power system are ordered from largest to smallest as follows: Power Module 4, Power Module 1, Power Module 3, and Power Module 2. Therefore, the scheduling startup sequence is also ordered accordingly. Power Module 4 has the highest health score, the smallest equivalent duration deviation, and the best overall condition, so it starts first. Power Module 1 has a relatively low health score but a very small equivalent duration deviation, and good aging balance. Power Module 3 has a moderate health score and a medium equivalent duration deviation. Power Module 2 has a very large equivalent duration deviation, excessive accumulated aging, and the lowest priority, so it starts last, in order to achieve dynamic adjustment of start and stop.
[0043] like Figure 6 The diagram shown is a connection block diagram for power closed-loop processing in a power system balancing scheduling method based on fuzzy health decision according to an embodiment of the present invention; Figure 6 In order to facilitate the switching of power modules in the power system and achieve steady-state power regulation, in one embodiment of the present invention, a PID algorithm is used to perform closed-loop power processing for each power module based on the current power-on sequence and power-off sequence, which may include the following steps: In step S60, the target power is obtained, and it is determined whether the current power system is in power output ramp-up mode or power output ramp-down mode. In step S61, it is determined that the power system is in the power output ramp-up mode. The corresponding power modules are turned on sequentially according to the current power-on sequence, and the power output is increased synchronously with a preset fixed ramp-up step size. The current sampling power is obtained in real time. In step S62, it is determined whether the current sampling power is not less than the difference between the target power and the preset first threshold. In step S63, when the current sampling power is not less than the difference between the target power and the preset first threshold, the power output gradual increase mode is exited and the fine segmented closed-loop adjustment stage is entered to adjust the power until the output power reaches the target power. In step S64, it is determined that the power system is in power output easing mode, the power output is reduced by a preset fixed easing step size, and the corresponding power modules are shut down synchronously in the current shutdown sequence, and the current sampling power is obtained in real time. In step S65, it is determined whether the current sampling power is not greater than the sum of the target power and the preset first threshold. In step S66, when the current sampled power is not greater than the sum of the target power and the preset first threshold, the power output slow-down mode is exited and the fine segmented closed-loop adjustment stage is entered to adjust the power until the output power reaches the target power.
[0044] By acquiring the target power, the system determines whether it is in a power output ramp-up or ramp-down mode. In ramp-up mode, the corresponding modules are turned on sequentially according to the power-on sequence, increasing the power output with a preset fixed ramp-up step size, and the sampled power is acquired in real time. When the sampled power is not less than the difference between the target power and a preset first threshold, the system exits ramp-up mode and enters a fine segmented closed-loop adjustment stage until the output power reaches the target power. In ramp-down mode, the power output is reduced with a preset fixed ramp-down step size, and the corresponding modules are turned off sequentially according to the power-off sequence, acquiring the sampled power in real time. When the sampled power is not greater than the sum of the target power and a preset first threshold, the system exits ramp-down mode and enters a fine segmented closed-loop adjustment stage. Through the synergy of PID closed-loop and TOPSIS dynamic sorting, a smooth transition of output power is ensured at the moment of module switching, avoiding bus voltage drops and circulating current impacts caused by the switching on of new modules, as well as output oscillations caused by the need for the remaining modules to quickly replenish power when a module is switched off, thereby protecting the microwave power supply load (magnetron / klystron) from power surge damage.
[0045] like Figure 7 The diagram shown is a connection block diagram illustrating the implementation of an output power regulation method for a power system based on fuzzy health determination, according to an embodiment of the present invention. Figure 7 In order to achieve fine power adjustment, rapid convergence and overshoot suppression, and avoid power oscillation, in one embodiment of the present invention, power adjustment is performed by entering a fine segmented closed-loop adjustment stage until the output power reaches the target power. This may include the following steps: In step S70, the sampling power is acquired according to a preset frequency; In step S71, the power deviation between the sampled power and the target power is obtained according to formula (20). (20) in, For the first Power deviation per sampling period For the first The absolute value of the power deviation over each sampling period. For the target power, For the first Sampling power per sampling period; In step S72, the power adjustment step size is determined using formula (21) based on the absolute value of the power deviation. ,(twenty one) in, For the first The adjustment step size for each sampling period, For symbolic functions, This is an adjustment function based on the absolute value of the power deviation; In step S73, the output power is adjusted according to formula (22) until the output power reaches the target power. ,(twenty two) in, For the first Output power per sampling period For the first Output power per sampling period.
[0046] The sampling power is acquired according to a preset frequency, and the power deviation and its absolute value between the sampling power and the target power are calculated. Based on the absolute value of the power deviation, a specific adjustment function is used to determine the adjustment step size, where the adjustment step size is the product of the sign function and the adjustment function based on the absolute value of the deviation. The output power is adjusted according to the determined adjustment step size until the output power reaches the target power. When the absolute value of the power deviation is large, a large step size is used to achieve fast convergence. When the deviation decreases, the step size is gradually reduced to achieve smooth fine adjustment, thus taking into account both dynamic response speed and steady-state accuracy. This effectively avoids the drawbacks of traditional fixed-step PID control, such as slow convergence at large deviations and overshoot oscillation at small deviations, ensuring that the output power fluctuation control is within the ideal range throughout the module switching process.
[0047] like Figure 8 The diagram shown is a connection block diagram of the module dynamic reordering of a power system balancing scheduling method based on fuzzy health determination according to an embodiment of the present invention; Figure 8 In order to improve the compatibility of balanced scheduling and ensure the synchronization of scheduling states, in one embodiment of the present invention, the balanced scheduling method includes the following steps: In step S80, it is determined whether the current power system has experienced one of the following three situations: power-on initialization, the health score of any power module is lower than the preset warning score threshold, or the number of power modules changes. In step S81, if the current power system experiences one of the following three situations: power-on initialization, the health score of any of the power modules is lower than the preset warning score threshold, or the number of power modules changes, the step of collecting the electrical operating parameters of each power module is re-executed.
[0048] The system determines whether the current power system is experiencing one of three conditions: power-on initialization, any power module's health score falling below a preset warning score threshold, or a change in the number of power modules. When any of these conditions are met, the system re-executes the steps of collecting electrical operating parameters of each power module, initiating a new round of equivalent aging time calculation, health assessment, and TOPSIS decision ranking. This ensures that the scheduling system can dynamically perceive changes in system status—whether it's initial power-on, module performance degradation below the warning threshold, or an increase or decrease in the number of parallel modules—and can immediately restart the entire process reordering, ensuring that the power-on sequence and power-off sequence always match the real-time system status, thereby achieving true adaptive dynamic balanced scheduling.
[0049] To demonstrate the intuitive effect of the power system balancing scheduling method based on fuzzy health determination of the present invention, a measured control experiment was set up in one embodiment of the present invention. The power-on output power was 37.6kW. The control group used traditional fixed power-on / off sequence scheduling, while the experimental group used the balancing scheduling method of the present invention. The experimental conditions included a brand-new four-channel parallel 18.8kW power module, running continuously for 12 hours per day for 14 consecutive days. The equivalent aging time of the control group (traditional fixed timing) after 14 days was obtained, as shown in Table 5 below. The equivalent aging time of the experimental group (balancing scheduling of the present invention) after 14 days was obtained, as shown in Table 6 below. Table 5
[0050] Table 6
[0051] In Tables 5 and 6, the control group uses traditional fixed power-on / off sequence scheduling, where the equivalent aging time deviation between the two power modules is as high as 165 hours, indicating polarized module aging. The experimental group, however, uses the balanced scheduling of this invention, where the equivalent aging time deviation between the two power modules is only 8 hours, and the aging degree of each module is basically synchronized. In the traditional scheme, the performance of the prioritized working modules degrades rapidly, while idle modules maintain excellent long-term operating conditions. Under this invention, the distribution range of health scores for all modules is narrowed, and performance degradation is more uniform. Based on device lifespan model extrapolation, the balanced scheduling method of this invention increases the average mean time between failures (MTBF) of the power system by more than 42%. It also enables synchronized aging and replacement of all power modules, reducing maintenance costs.
[0052] The second aspect of this invention provides a balanced scheduling system for a power system based on fuzzy health decision-making. The power system includes multiple parallel power modules. The scheduling system includes a data acquisition unit, an equivalent aging calculation unit, a fuzzy evaluation matrix construction unit, a health score calculation unit, a TOPSIS decision-making unit, and a PID power closed-loop unit. Specifically, the data acquisition unit collects the electrical operating parameters of each power module; the equivalent aging calculation unit uses an adaptive duration estimation algorithm to obtain the equivalent aging duration of each power module based on the electrical operating parameters, and calculates the equivalent duration deviation of each power module accordingly; the fuzzy evaluation matrix construction unit constructs an evaluation index set based on the electrical operating parameters, and generates a fuzzy evaluation matrix based on the evaluation index set; the health score calculation unit uses the AHP (Analytic Hierarchy Process) method to obtain the health score of each power module based on the fuzzy evaluation matrix; and the TOPSIS decision-making unit uses the TOPSIS algorithm to calculate the equivalent duration deviation and... A decision matrix is constructed based on health scores. The normalized decision matrix obtains positive and negative ideal solutions. The comprehensive proximity of each power module is calculated based on the positive and negative ideal solutions. The modules are then sorted in descending order according to their comprehensive proximity to generate the current power-on sequence, and the current power-off sequence is generated in reverse order. The comprehensive proximity is the ratio of the Euclidean distance from the power module to the negative ideal solution to the sum of the Euclidean distances to the positive and negative ideal solutions. A PID power closed-loop unit is used to perform power closed-loop processing for each power module based on the current power-on and power-off sequences using a PID algorithm to achieve intelligent balanced scheduling of the power system.
[0053] The above technical solution involves first collecting the physical runtime of each module during start-up and shutdown, as well as raw parameters such as output voltage, IGBT junction temperature, and ambient temperature under different load conditions under rated input. An adaptive exponential smoothing algorithm is used to calculate the equivalent aging time of each module and determine the duration deviation, quantifying the aging deviation between modules. Based on the raw parameters, a four-dimensional evaluation index set is constructed, including output linear margin, IGBT temperature rise margin, no-load standby loss, and protection threshold offset. Each index is mapped to an interval using benefit-type and cost-type membership functions, generating a fuzzy evaluation matrix. A pairwise comparison judgment matrix is constructed using the Saaty 1-9 scaling method. After consistency verification, the index weight vector is obtained through eigenvector normalization. The fuzzy evaluation matrix and the weight vector are then fused to calculate the health score of each module. A two-dimensional decision matrix is constructed, with the health score as a benefit attribute and the duration deviation as a cost attribute. After vector normalization, positive and negative ideal solutions are determined. The Euclidean distance from each module to the positive and negative ideal solutions is calculated, and the comprehensive proximity of each module is obtained through a comprehensive proximity formula. Dynamic power-on sequences are generated in descending order, and power-off sequences are generated in reverse order. During module switching, a PID algorithm is used for power closed-loop processing. Through the synergy of gradual increase / decrease mode and fine segmented variable step size closed-loop regulation, a smooth transition of output power is achieved. Three types of trigger conditions are monitored: power-on initialization, module health score below the warning threshold, and changes in the number of modules. If any one of these conditions is met, the entire process is re-executed for reordering. This method achieves synchronous balancing of the working losses and aging levels of all parallel power modules, significantly extends the mean time between failures (MTBF) of the entire system, reduces output power fluctuations during module switching, reduces the probability of sudden module failures, simplifies operation and maintenance, and saves costs.
[0054] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solutions of the present invention, and these simple modifications all fall within the protection scope of the present invention. Furthermore, it should be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.
Claims
1. A power system balanced scheduling method based on fuzzy health determination, wherein the power system comprises multiple power modules connected in parallel, characterized in that, The balanced scheduling method includes: Collect the electrical operating parameters of each power module; An adaptive duration estimation algorithm is used to obtain the equivalent aging duration of each power module based on electrical operating parameters, and the equivalent duration deviation of each power module is calculated accordingly. The health score of each power module was obtained by using the fuzzy evaluation matrix and the AHP (Analytic Hierarchy Process). The TOPSIS algorithm is used to construct a decision matrix based on the equivalent duration deviation and health score. The normalized decision matrix is used to obtain the positive ideal solution and the negative ideal solution. The comprehensive proximity of each power module is obtained based on the positive ideal solution and the negative ideal solution. The comprehensive proximity is the ratio of the Euclidean distance from the power module to the negative ideal solution to the sum of the Euclidean distances to the positive ideal solution and the negative ideal solution. The current power-on sequence is generated by sorting each power module in descending order based on their overall proximity, and the current power-off sequence is generated in reverse order. Based on the current power-on and power-off sequences, a PID algorithm is used to perform closed-loop power processing for each power module.
2. The balanced scheduling method according to claim 1, characterized in that, The electrical operating parameters include at least one of the following: the physical duration of operation during each start-up and shutdown, the voltage output under different load conditions at rated input, the junction temperature of the IGBT device, the voltage and current of the input DC bus, and the actual OVP trigger voltage.
3. The balanced scheduling method according to claim 2, characterized in that, An adaptive exponential smoothing cumulative duration estimation algorithm is used to obtain the equivalent aging duration of each power module based on electrical operating parameters, and the equivalent duration deviation of each power module is calculated accordingly, including: Obtain the physical duration of operation for each power module during each start-up and shutdown; Based on the physical duration of operation during each start-up and shutdown, an adaptive duration estimation algorithm is used to obtain the equivalent aging duration of the power module using formula (1). ,(1) in, For the power module index, Index for the number of runs, For the first The power module is in the first Equivalent aging time for each run For the first The power module is in the first Equivalent aging time for each run For exponential smoothing forgetting factor, This is the correction factor for power-on surge loss. For the first The power module is the first Physical duration of each run; The equivalent duration deviation of each power module is obtained using formulas (2)-(3). ,(2) ,(3) in, This serves as the baseline for the overall average equivalent aging of the power supply system. This represents the total number of power modules. For the first Equivalent aging time of each power module For the first Equivalent duration deviation of each power module.
4. The balanced scheduling method according to claim 2, characterized in that, The balanced scheduling method further includes: An evaluation index set is constructed based on electrical operating parameters. The evaluation index set includes the output linear margin of benefit-type indicators and the temperature rise margin of IGBT devices, and the no-load standby loss and protection threshold offset of cost-type indicators. The load regulation rate is obtained using formula (4) based on the voltage output and nominal rated output voltage of the power module under different load conditions at rated input. The output linear margin of the power module is obtained using formula (5). ,(4) ,(5) in, For the load regulation rate of the power module, This refers to the maximum voltage output of the power module under rated input conditions. This is the minimum voltage output of the power module under rated input. The nominal rated output voltage, This is the factory-permitted maximum load regulation rate; To output linear margin; The temperature rise margin of the IGBT device is obtained based on the junction temperature of the IGBT device in the power module using formula (6). ,(6) in, This refers to the junction temperature limit of IGBT devices. This refers to the junction temperature of the IGBT device. The no-load standby loss is obtained using formula (7) based on the voltage and current of the input DC bus of the power supply module. ,(7) in, This refers to the power loss during idle standby. The input voltage is the DC bus input voltage. For the input DC bus current; The protection threshold offset is obtained using formula (8) based on the actual OVP trigger voltage of the power module. ,(8) in, To protect the threshold offset, This is the actual OVP trigger voltage of the power module. This refers to the nominal OVP threshold of the power module. The benefit-type indicators and cost-type indicators of each power module are mapped according to formulas (9)-(10) to obtain the fuzzy membership degree of the indicators. ,(9) ,(10) in, For fuzzy membership functions of benefit-type indicators, These are the original values for the current benefit-type indicators. This represents the historical minimum value of the current benefit-related indicators. This represents the historical maximum value of the current benefit-related indicators. For cost-type indicators, fuzzy membership function. These are the original values for the current cost-related indicators. This is the historical minimum value of the current cost-related indicator. This represents the historical maximum value of the current cost-related indicator. The fuzzy evaluation matrix of the power system is constructed based on the fuzzy membership degrees of all indicators of all power modules.
5. The balanced scheduling method according to claim 4, characterized in that, Based on the fuzzy evaluation matrix, the health score of each power module is obtained using the AHP (Analytic Hierarchy Process) method, including: The Saaty1–9 scaling method was used to construct pairwise comparison judgment matrices between the four indicators of each power module; Based on pairwise comparisons, the maximum eigenvalue is obtained using formula (11) from the judgment matrix. ,(11) in, The largest eigenvalue, For the number of indicators, For indexing indicators, This is a pairwise comparison judgment matrix. The initial weight vector, The first step is to compare the product of the pairwise comparison judgment matrix and the initial weight vector. One portion, For the first The current weight of each indicator; The consistency ratio is obtained using formula (12) based on the largest eigenvalue; ,(12) in, To determine the consistency index of the matrix by comparing each pair of matrices, To determine the consistency ratio of the matrices by comparing them pairwise, This is the average random consistency index value based on the number of indicators; Determine if the consistency ratio is greater than the preset verification threshold. If the consistency ratio is not greater than the preset verification threshold, reconstruct the pairwise comparison judgment matrix; If the consistency ratio is greater than the preset verification threshold, the pairwise comparison judgment matrix is normalized using formula (13) to obtain the corresponding index weight vector. ,(13) in, For the indicator weight vector, To output the index weight vector components corresponding to the linear residual, These are the weighted vector components of the index corresponding to the temperature rise margin of IGBT devices. These are the weighted vector components of the index corresponding to idle standby power loss. To protect the index weight vector components corresponding to the threshold offset, For the first The indicator weight vector of each indicator; The health score of each power module is obtained using formula (14) based on the fuzzy evaluation matrix and the index weight vector. ,(14) in, For the first Health score of each power module For fuzzy membership functions, For the first The power module's first Fuzzy membership degree of each indicator.
6. The balanced scheduling method according to claim 1, characterized in that, The TOPSIS algorithm is used to construct a decision matrix based on equivalent duration deviation and health score. The normalized decision matrix yields positive and negative ideal solutions. Based on these solutions, the overall proximity score of each power module is calculated, including: Based on the equivalent duration deviation and health score, a decision matrix is constructed using formula (15). ,(15) in, For the decision matrix, For the first The health score of the power module is the first one. The first attribute of a power module For the first The equivalent duration deviation of the power module is the first... The second attribute of a power module, This represents the total number of power modules; The decision matrix is vector normalized using formula (16). ,(16) in, The elements of the normalized decision matrix are the _th _th_ element. The power module's first One attribute, The original element of the decision matrix is the i-th element. The power module's first Attributes The positive and negative ideal solutions in the power supply system are determined using formula (17). ,(17) in, This is the ideal solution in the power supply system. For the normalized first Health score of each power module For the normalized first Equivalent duration deviation of each power module To find the maximum value function, The function is for finding the minimum value; The Euclidean distances from each power module to the positive and negative ideal solutions are obtained using formula (18). ,(18) in, For the first The Euclidean distance from each power module to the ideal solution. For the first The Euclidean distance from each power module to the negative ideal solution. The first component of the positive ideal solution is the maximum health score among all normalized power modules. The second component of the positive ideal solution is the minimum equivalent duration deviation among all power modules after normalization. The first component of the negative ideal solution is the minimum health score among all power modules after normalization. The second component of the negative ideal solution is the maximum value of the equivalent duration deviation among all power modules after normalization. The overall proximity of each power module is obtained using formula (19) based on the Euclidean distance of each power module. ,(19) in, For the first Overall proximity of the power modules.
7. The balanced scheduling method according to claim 1, characterized in that, Based on the current power-on and power-off sequences, a PID algorithm is used to perform power closed-loop processing for each power module, including: Obtain the target power and determine whether the current power system is in power output ramp-up mode or power output ramp-down mode; Once the power system is confirmed to be in power output ramp-up mode, the corresponding power modules are turned on sequentially according to the current power-on sequence, and the power output is increased synchronously with a preset fixed ramp-up step size, and the current sampling power is obtained in real time. Determine whether the current sampling power is not less than the difference between the target power and the preset first threshold; When the current sampling power is not less than the difference between the target power and the preset first threshold, exit the power output gradual increase mode and switch to the fine segmented closed-loop adjustment stage to adjust the power until the output power reaches the target power. The power system is determined to be in power output easing mode. The power output is reduced by a preset fixed easing step size, and the corresponding power modules are shut down synchronously in the current shutdown sequence. The current sampled power is obtained in real time. Determine whether the current sampling power is not greater than the sum of the target power and the preset first threshold; When the current sampling power is not greater than the sum of the target power and the preset first threshold, the power output slow-down mode is exited and the fine segmented closed-loop adjustment stage is entered to adjust the power until the output power reaches the target power.
8. The balanced scheduling method according to claim 7, characterized in that, The power is adjusted by entering a fine-tuned, segmented closed-loop control phase until the output power reaches the target power, including: The sampling power is obtained according to the preset frequency; The power deviation between the sampled power and the target power is obtained according to formula (20). ,(20) in, For the first Power deviation per sampling period For the first The absolute value of the power deviation over each sampling period. For the target power, For the first Sampling power per sampling period; The power adjustment step size is determined using formula (21) based on the absolute value of the power deviation. ,(21) in, For the first The adjustment step size for each sampling period, For symbolic functions, This is an adjustment function based on the absolute value of the power deviation; Adjust the output power according to formula (22) until the output power reaches the target power. ,(22) in, For the first Output power per sampling period For the first Output power per sampling period.
9. The balanced scheduling method according to claim 7, characterized in that, The balanced scheduling method includes: Determine whether the current power system is experiencing one of three conditions: power-on initialization, the health score of any of the power modules is lower than a preset warning score threshold, or a change in the number of power modules. If any of the following three conditions occur in the current power system: power-on initialization, health score of any power module falling below the preset warning score threshold, or change in the number of power modules, the step of collecting electrical operating parameters of each power module shall be re-executed.
10. A power system balancing scheduling system based on fuzzy health determination, the power system comprising multiple parallel power modules, characterized in that, The scheduling system includes: The acquisition unit is used to collect the electrical operating parameters of each power module; The equivalent aging calculation unit is used to obtain the equivalent aging time of each power module based on the electrical operating parameters using an adaptive duration estimation algorithm, and to calculate the equivalent duration deviation of each power module accordingly. The fuzzy evaluation matrix construction unit is used to construct an evaluation index set based on electrical operating parameters and generate a fuzzy evaluation matrix based on the evaluation index set. The health score calculation unit is used to obtain the health score of each power module based on the fuzzy evaluation matrix using the AHP (Analytic Hierarchy Process). The TOPSIS decision unit is used to construct a decision matrix based on the equivalent duration deviation and health score using the TOPSIS algorithm, normalize the decision matrix to obtain positive and negative ideal solutions, calculate the comprehensive proximity of each power module based on the positive and negative ideal solutions, sort the power modules in descending order based on the comprehensive proximity to generate the current power-on sequence, and generate the current power-off sequence in reverse order. The comprehensive proximity is the ratio of the Euclidean distance from the power module to the negative ideal solution to the sum of the Euclidean distances to the positive and negative ideal solutions. The PID power closed-loop unit is used to perform power closed-loop processing for each power module based on the current power-on and power-off sequences using a PID algorithm, so as to achieve intelligent balanced scheduling of the power system.
