Low load rate efficiency optimization method and system of photovoltaic optimizer

Abstract

The invention relates to the technical field of load optimization, and discloses a low-load-rate efficiency optimization method and system for a photovoltaic optimizer. The method comprises the following steps: monitoring the working state of the photovoltaic optimizer in real time to obtain working state data; based on the working state data, the working mode of the photovoltaic optimizer is analyzed and recognized, and low-load-rate working mode characteristics are obtained; performing dynamic optimization calculation on the operation parameters of the photovoltaic optimizer to obtain an optimization parameter strategy; based on the optimization parameter strategy, carrying out adaptive adjustment on the soft switching topology of the photovoltaic optimizer to obtain soft switching control parameters; according to the soft switching control parameters, performing dynamic reconstruction on a multi-phase parallel structure of the photovoltaic optimizer to obtain target parallel configuration; and performing continuous iteration on the efficiency optimization process of the photovoltaic optimizer based on the target parallel configuration to obtain a low-load-rate efficiency optimization control instruction. According to the invention, the efficiency of low-load rate efficiency optimization of the photovoltaic optimizer is improved.

Description

Technical Field

[0001] The present application relates to the technical field of load optimization, and in particular to a method and system for optimizing the efficiency of a photovoltaic optimizer at a low load rate. Background Art

[0002] Photovoltaic optimizers are an important part of solar power generation systems, used to optimize the output power of each solar panel and improve the power generation efficiency of the entire system. Traditional photovoltaic optimizers mainly focus on efficiency optimization under standard test conditions, using fixed control strategies and topologies. These optimizers usually work better at medium and high load rates and can effectively improve the overall power generation of the system. With the continuous development of photovoltaic technology, many photovoltaic systems have begun to adopt intelligent tracking and dynamic adjustment technologies to adapt to different lighting conditions and load requirements.

[0003] However, existing photovoltaic optimizers often have the problem of reduced efficiency under low load rate working conditions. This is because at low load rates, the inherent losses of the optimizer, such as switching loss and conduction loss, account for a relatively high proportion, resulting in a decrease in overall efficiency. In addition, traditional optimizers lack accurate identification and targeted optimization of low load rate working modes, and cannot dynamically adjust their operating parameters and topology according to real-time working conditions. These deficiencies lead to poor power generation efficiency of photovoltaic systems under low light or low power demand, affecting the overall performance and economic benefits of the system. Summary of the invention

[0004] The present application provides a low-load rate efficiency optimization method and system for a photovoltaic optimizer, which are used to improve the efficiency of low-load rate efficiency optimization of a photovoltaic optimizer.

[0005] In a first aspect, the present application provides a low-load efficiency optimization method for a photovoltaic optimizer, the low-load efficiency optimization method for a photovoltaic optimizer comprising: real-time monitoring of the working state of the photovoltaic optimizer to obtain working state data; based on the working state data, analyzing and identifying the working mode of the photovoltaic optimizer to obtain low-load working mode characteristics; according to the low-load working mode characteristics, dynamically optimizing and calculating the operating parameters of the photovoltaic optimizer to obtain an optimization parameter strategy; based on the optimization parameter strategy, adaptively adjusting the soft switching topology of the photovoltaic optimizer to obtain soft switching control parameters; according to the soft switching control parameters, dynamically reconstructing the multi-phase parallel structure of the photovoltaic optimizer to obtain a target parallel configuration; based on the target parallel configuration, continuously iterating the efficiency optimization process of the photovoltaic optimizer to obtain a low-load efficiency optimization control instruction.

[0006] In a second aspect, the present application provides a low load rate efficiency optimization system for a photovoltaic optimizer, the low load rate efficiency optimization system for a photovoltaic optimizer comprising:

[0007] A monitoring module is used to monitor the working status of the photovoltaic optimizer in real time and obtain working status data;

[0008] An identification module, used to analyze and identify the working mode of the photovoltaic optimizer based on the working status data, and obtain the characteristics of the low load rate working mode;

[0009] A calculation module, used to dynamically optimize and calculate the operating parameters of the photovoltaic optimizer according to the characteristics of the low load rate working mode to obtain an optimization parameter strategy;

[0010] A regulating module, used for adaptively regulating the soft switch topology of the photovoltaic optimizer based on the optimization parameter strategy to obtain soft switch control parameters;

[0011] A reconstruction module, used for dynamically reconstructing the multi-phase parallel structure of the photovoltaic optimizer according to the soft switch control parameters to obtain a target parallel configuration;

[0012] The iteration module is used to continuously iterate the efficiency optimization process of the photovoltaic optimizer based on the target parallel configuration to obtain a low load efficiency optimization control instruction.

[0013] In the technical solution provided by the present application, the real-time monitoring of the working state of the photovoltaic optimizer provides an accurate and timely data basis for the entire optimization process, ensuring the accuracy of subsequent optimization decisions. The working mode analysis and identification based on these working state data enable the system to quickly and accurately determine whether it is currently in a low load rate state, and extract the corresponding working mode characteristics, laying the foundation for targeted optimization. According to the identified low load rate working mode characteristics, the system can perform dynamic optimization calculations to obtain the optimization parameter strategy that best suits the current working state. This real-time adjustment capability greatly improves the adaptability of the system in a changing environment. The adaptive adjustment of the soft switching topology further reduces the switching loss and improves the overall efficiency of the system, especially under low load rate conditions, this optimization effect is more significant. The dynamic reconstruction of the multi-phase parallel structure provides the system with a more flexible power management solution, which can dynamically adjust the number of parallel phases according to the load conditions, which not only ensures efficiency, but also optimizes the current distribution. Finally, through the continuous iteration of the efficiency optimization process, the system can continuously learn and improve, adapt to changes in various working conditions, and maintain an efficient operating state for a long time. This comprehensive, dynamic and intelligent optimization method enables the photovoltaic optimizer to maintain a high working efficiency under low load conditions, effectively solving the problem of reduced efficiency of traditional optimizers at low load rates, and improving the efficiency of low load rate efficiency optimization of the photovoltaic optimizer. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0015] Figure 1 A schematic diagram of an embodiment of a method for optimizing the efficiency of a photovoltaic optimizer at a low load rate in an embodiment of the present application;

[0016] Figure 2 Schematic diagram of an embodiment of a low load efficiency optimization system for a photovoltaic optimizer in an embodiment of the present application. DETAILED DESCRIPTION

[0017] Embodiments of the present application provide a method and system for optimizing the efficiency of a photovoltaic optimizer at a low load rate. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data passed in this way can be interchanged where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0018] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the low load rate efficiency optimization method of the photovoltaic optimizer in the embodiment of the present application includes:

[0019] Step S101, real-time monitoring of the working state of the photovoltaic optimizer to obtain working state data;

[0020] Step S102: Analyze and identify the working mode of the photovoltaic optimizer based on the working status data to obtain the low load rate working mode characteristics;

[0021] Step S103: dynamically optimizing and calculating the operating parameters of the photovoltaic optimizer according to the characteristics of the low load rate working mode to obtain an optimization parameter strategy;

[0022] Step S104: Based on the optimization parameter strategy, the soft switch topology of the photovoltaic optimizer is adaptively adjusted to obtain soft switch control parameters;

[0023] Step S105, dynamically reconstructing the multi-phase parallel structure of the photovoltaic optimizer according to the soft switch control parameters to obtain a target parallel configuration;

[0024] Step S106: Based on the target parallel configuration, the efficiency optimization process of the photovoltaic optimizer is continuously iterated to obtain a low load efficiency optimization control instruction.

[0025] It is understandable that the execution subject of the present application may be a low load efficiency optimization system of a photovoltaic optimizer, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0026] Specifically, the working state of the photovoltaic optimizer is monitored in real time to obtain working state data. In this process, multiple parameters such as input and output voltage and current, switching frequency, duty cycle and temperature are collected. For voltage and current data, filtering is used to obtain smooth values ​​to reduce the impact of instantaneous fluctuations. For example, using a moving average filter to process the sampled data can effectively reduce the interference of high-frequency noise. Based on the processed data, the input and output power and real-time efficiency values ​​are calculated, and the efficiency state flag is obtained by threshold comparison. Subsequently, based on the working state data, the working mode of the photovoltaic optimizer is analyzed and identified to obtain the low load rate working mode characteristics. In this step, the output power value is first normalized to obtain the load rate value, and the low load rate state is determined by threshold comparison. Cluster analysis is performed on the working state data in the low load rate time period to extract the mode feature vector. The feature vector is classified using the support vector machine algorithm to obtain the working mode category. This method can effectively identify different low load rate working modes and provide a basis for subsequent optimization.

[0027] According to the identified low load rate working mode characteristics, the operating parameters of the photovoltaic optimizer are dynamically optimized and calculated to obtain the optimization parameter strategy. This process uses the particle swarm optimization algorithm to iteratively update the particle swarm and continuously optimize the objective function to finally obtain the optimal solution. The optimization objective function takes into account multiple factors such as efficiency, stability and response speed to ensure the best performance under low load rate conditions. Based on the optimization parameter strategy, the soft switching topology of the photovoltaic optimizer is adaptively adjusted to obtain the soft switching control parameters. This step includes the analysis of the soft switching circuit topology, the construction of an equivalent model, and the sensitivity analysis of key parameters. The soft switching performance is improved by optimizing the switching timing, adjusting the dead time and the resonance parameters. For example, by accurately controlling the turn-on and turn-off moments of the switch tube, the switching loss can be significantly reduced and the system efficiency can be improved.

[0028] According to the soft switch control parameters, the multi-phase parallel structure of the photovoltaic optimizer is dynamically reconstructed to obtain the target parallel configuration. This process first analyzes the current distribution characteristics of each phase, calculates the current imbalance, and then dynamically evaluates the number of parallel phases and designs the phase interleaving control strategy. The optimal configuration of the multi-phase parallel structure is achieved by optimizing the power distribution ratio and designing a dynamic balancing algorithm. This dynamic reconstruction method can flexibly adjust the number of parallel phases and phase difference according to the load conditions, effectively improving the system efficiency under low load rates. Finally, based on the target parallel configuration, the efficiency optimization process of the photovoltaic optimizer is continuously iterated to obtain the low load efficiency optimization control instructions. This step uses the long short-term memory network algorithm to learn the historical optimization results and establish an efficiency prediction model. Through the reinforcement learning method, the parameter adjustment strategy is continuously optimized to achieve continuous improvement in efficiency.

[0029] For example, when a photovoltaic optimizer is working at a low load rate, through real-time monitoring, the input voltage is 30V, the input current is 1A, the output voltage is 220V, and the output current is 0.13A. After data processing and analysis, it is determined that the current load rate is 30%, which belongs to the low load rate working state. Based on this working mode feature, the optimization algorithm calculates that the optimal switching frequency should be adjusted from the original 50kHz to 35kHz, and the dead time is optimized from 200ns to 150ns. At the same time, the multi-phase parallel structure is adjusted from the original 4 phases to 2 phases, and the phase difference is set to 90°. Through the dynamic optimization and structural reconstruction of these parameters, the efficiency of the system at a load rate of 30% is increased from the original 92% to 95%, achieving the efficiency optimization goal under low load rate conditions.

[0030] In the embodiment of the present application, the real-time monitoring of the working state of the photovoltaic optimizer provides an accurate and timely data basis for the entire optimization process, ensuring the accuracy of subsequent optimization decisions. The working mode analysis and identification based on these working state data enable the system to quickly and accurately determine whether it is currently in a low load rate state, and extract the corresponding working mode characteristics, laying the foundation for targeted optimization. According to the identified low load rate working mode characteristics, the system can perform dynamic optimization calculations to obtain the most suitable optimization parameter strategy for the current working state. This real-time adjustment capability greatly improves the adaptability of the system in a changing environment. The adaptive adjustment of the soft switch topology further reduces the switching loss and improves the overall efficiency of the system, especially under low load rate conditions, this optimization effect is more significant. The dynamic reconstruction of the multi-phase parallel structure provides the system with a more flexible power management solution, which can dynamically adjust the number of parallel phases according to the load conditions, which not only ensures efficiency, but also optimizes the current distribution. Finally, through the continuous iteration of the efficiency optimization process, the system can continuously learn and improve, adapt to changes in various working conditions, and maintain an efficient operating state for a long time. This comprehensive, dynamic and intelligent optimization method enables the photovoltaic optimizer to maintain a high working efficiency under low load conditions, effectively solving the problem of reduced efficiency of traditional optimizers at low load rates, and improving the efficiency of low load rate efficiency optimization of the photovoltaic optimizer.

[0031] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0032] (1) Sampling the input voltage of the photovoltaic optimizer to obtain input voltage data, and filtering the input voltage data to obtain a smoothed input voltage value;

[0033] (2) sampling the input current of the photovoltaic optimizer to obtain input current data, and filtering the input current data to obtain a smoothed input current value;

[0034] (3) sampling the output voltage of the photovoltaic optimizer to obtain output voltage data, and filtering the output voltage data to obtain a smoothed output voltage value;

[0035] (4) sampling the output current of the photovoltaic optimizer to obtain output current data, and filtering the output current data to obtain a smoothed output current value;

[0036] (5) calculating the input power of the photovoltaic optimizer according to the smoothed input voltage value and the smoothed input current value to obtain an input power value, and calculating the output power of the photovoltaic optimizer according to the smoothed output voltage value and the smoothed output current value to obtain an output power value;

[0037] (6) Calculating the efficiency of the photovoltaic optimizer according to the input power value and the output power value to obtain a real-time efficiency value, and performing a threshold comparison on the real-time efficiency value to obtain an efficiency status mark;

[0038] (7) measuring the switching frequency of the photovoltaic optimizer to obtain switching frequency data, and performing statistical analysis on the switching frequency data to obtain switching frequency characteristics;

[0039] (8) measuring the duty cycle of the photovoltaic optimizer to obtain duty cycle data, and performing statistical analysis on the duty cycle data to obtain duty cycle characteristics;

[0040] (9) Collecting the temperature of the photovoltaic optimizer to obtain temperature data, and performing trend analysis on the temperature data to obtain temperature change characteristics;

[0041] (10) The real-time efficiency value, efficiency status mark, switching frequency characteristics, duty cycle characteristics and temperature change characteristics are combined to obtain working status data.

[0042] Specifically, the working state of the system is monitored comprehensively and accurately, and the voltage and current at the input and output ends of the photovoltaic optimizer are sampled. The sampling process uses a high-precision analog-to-digital converter to obtain raw data at a high sampling frequency. In order to eliminate high-frequency noise and interference that may be introduced during the sampling process, the collected raw data is filtered. Common filtering methods include moving average filtering, median filtering or Kalman filtering. Taking moving average filtering as an example, by taking the average value of n consecutive sampling points, data fluctuations can be effectively smoothed to obtain more stable voltage and current values. The filtered smoothed input voltage value and smoothed input current value are used to calculate the input power of the photovoltaic optimizer. Similarly, the smoothed output voltage value and smoothed output current value are used to calculate the output power. The power calculation uses a simple multiplication operation, but because the filtered data is used, the power value obtained can better reflect the actual working state of the system and avoid the influence of instantaneous fluctuations on the calculation results.

[0043] Based on the calculated input power and output power, the real-time efficiency of the photovoltaic optimizer can be further calculated. Efficiency calculation is a key indicator for evaluating system performance, especially under low load conditions, where even a small change in efficiency may have a significant impact on overall performance. The calculated real-time efficiency value is compared with the preset efficiency threshold to obtain an efficiency status flag. This flag is used to indicate whether the current system is in an efficient operating state, providing a decision basis for subsequent optimization control. In addition to voltage, current and power efficiency, the switching frequency and duty cycle of the photovoltaic optimizer are also important monitoring parameters. The switching frequency is usually measured by the frequency counting method, which counts the rising or falling edges of the switching signal within a certain time window and then converts it to a frequency value. For the measurement of the duty cycle, the time measurement method can be used to calculate the proportion of the time that the switching signal is in the high level state to the entire cycle. After statistical analysis of these data, the switching frequency characteristics and duty cycle characteristics that reflect the switching characteristics of the system can be obtained.

[0044] Temperature is another key factor affecting the performance of photovoltaic systems. The temperature of the photovoltaic optimizer is collected in real time, usually using temperature sensors such as thermocouples or thermistors. After trend analysis of temperature data, the temperature change characteristics can be obtained, which is of great significance for predicting system performance changes and implementing temperature compensation control. Finally, multi-dimensional data such as real-time efficiency value, efficiency status mark, switching frequency characteristics, duty cycle characteristics and temperature change characteristics are combined to form complete working status data. These comprehensive data provide a comprehensive information basis for subsequent working mode analysis and optimization strategy formulation.

[0045] For example, the data collection and processing process of a photovoltaic optimizer under low load rate working state is as follows: First, the input voltage is sampled by a high-precision ADC at a sampling frequency of 10kHz to obtain a set of raw data. Assume that the average value of 100 consecutive data points obtained by sampling is 30V and the standard deviation is 0.5V. After 10-point moving average filtering, the smoothed input voltage value is 30.1V, and the standard deviation is reduced to 0.2V. The same method is applied to the sampling and processing of input current, output voltage and output current. Assume that the smoothed input current value obtained after processing is 1A, the output voltage is 220V, and the output current is 0.13A. Based on this calculation, the input power is 30.1W and the output power is 28.6W. The real-time efficiency calculation result is 95%. Comparing this efficiency value with the preset low load rate efficiency threshold of 90%, the efficiency status mark is obtained as "high efficiency".

[0046] At the same time, the switching frequency is measured, and 35 rising edges are counted in a 1ms time window, which is converted to a switching frequency of 35kHz. The duty cycle measurement result is 40%. The temperature data collected by the temperature sensor is processed and shows that the temperature has slowly risen from 35°C to 37°C in the past 10 minutes. These data are combined to form a multi-dimensional working status data set, which contains information on voltage, current, power, efficiency, switching characteristics and temperature.

[0047] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0048] (1) Normalizing the output power value in the working status data to obtain a load factor value, and performing a threshold comparison on the load factor value to obtain a low load factor mark;

[0049] (2) According to the low load rate mark, the working status data is divided into time windows to obtain the low load rate time period, and the working status data in the low load rate time period is clustered and analyzed to obtain the working mode cluster;

[0050] (3) Extracting features from the working mode cluster to obtain a mode feature vector, and classifying the mode feature vector using a support vector machine algorithm to obtain a working mode category;

[0051] (4) According to the working mode category, historical working status data is retrieved to obtain similar working condition data, and similar working condition data is statistically analyzed to obtain a target working parameter set;

[0052] (5) Perform association rule mining on the target working parameter set to obtain parameter optimization rules, and evaluate the current working status based on the parameter optimization rules to obtain parameter adjustment suggestions;

[0053] (6) Based on the working mode category, target working parameter set and parameter adjustment suggestions, a low load rate working mode description is constructed to obtain the low load rate working mode characteristics.

[0054] Specifically, the output power value in the working state data is normalized to obtain the load rate value. Normalization is to divide the output power value by the rated power of the photovoltaic optimizer so that optimizers of different capacities can be compared under the same standard. The obtained load rate value is then compared with the preset low load rate threshold to determine whether it is currently in a low load rate state. This low load rate mark is an important basis for subsequent analysis. According to the low load rate mark, the working state data is divided into time windows to obtain a low load rate time period. The purpose of time window division is to aggregate continuous low load rate working states together to facilitate subsequent pattern recognition. For the working state data in the divided low load rate time period, a cluster analysis method is used for processing. Cluster analysis is an unsupervised learning method that can classify similar data points into one category. Commonly used clustering algorithms include K-means, DBSCAN, etc. Through cluster analysis, different working mode clusters can be obtained, each cluster representing a specific low load rate working mode. Feature extraction is performed on the obtained working mode cluster to obtain a mode feature vector. The purpose of feature extraction is to convert high-dimensional raw data into low-dimensional feature representation, which not only retains the key information of the original data but also reduces the complexity of subsequent processing. Common feature extraction methods include principal component analysis (PCA), linear discriminant analysis (LDA), etc. The extracted feature vectors are then classified using the support vector machine (SVM) algorithm. SVM is a powerful classification algorithm that can construct a hyperplane in a high-dimensional space to effectively separate data of different categories. Through SVM classification, the working mode category can be obtained, that is, the current low-load working state can be classified into a predefined type.

[0055] According to the determined working mode category, the historical working status data is retrieved to find the historical data under similar working conditions. This step utilizes database technology and similarity calculation methods, such as Euclidean distance or cosine similarity. Statistical analysis is performed on the retrieved similar working condition data to obtain the target working parameter set. This parameter set reflects the optimal parameter combination for system operation under similar working conditions. Association rule mining is performed on the target working parameter set to obtain parameter optimization rules. Association rule mining is a data mining technology used to discover the association relationship between data items. Commonly used algorithms include the Apriori algorithm and the FP-growth algorithm. Through association rule mining, the mutual influence relationship between different parameters can be discovered, providing guidance for parameter optimization. According to the parameter optimization rules obtained by mining, the current working status is evaluated to obtain parameter adjustment suggestions.

[0056] Finally, the working mode category, target working parameter set and parameter adjustment suggestions are integrated to construct a low-load working mode description and obtain the low-load working mode feature. This feature is a multi-dimensional description that includes the category of the current working state, the optimal parameter combination and the parameter direction that needs to be adjusted.

[0057] For example, the rated power of a photovoltaic optimizer is 1000W, and the current output power is 200W. After normalization, the load rate is 20%. Assuming that the preset low load rate threshold is 30%, the current state is marked as low load rate. In the past 2 hours, the system has been in a low load rate state. K-means clustering analysis is performed on the working state data within these 2 hours, and the number of clusters is set to 3, and 3 working mode clusters are obtained. PCA feature extraction is performed on each cluster, and the first 5 principal components are selected as feature vectors. These feature vectors are input into the pre-trained SVM model, and the classification results show that the current working mode belongs to the "weak light and low temperature" category. Based on this category, 100 groups of data with similar working conditions are retrieved from the historical database. Statistical analysis is performed on these 100 groups of data to obtain the target working parameter set, including the optimal switching frequency of 35kHz, the optimal duty cycle of 45%, etc. The association rule mining is performed through the FP-growth algorithm, and it is found that in this working mode, the switching frequency has a strong correlation with the efficiency. According to the current switching frequency of 50kHz, the adjustment suggestion of reducing the switching frequency is obtained. Finally, the "weak light and low temperature" operating mode category, the target operating parameter set containing the optimal parameters, and the adjustment suggestion of "reducing the switching frequency" are combined to form a complete low load rate operating mode feature description.

[0058] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0059] S1. According to the characteristics of the low-load working mode, a parameter optimization objective function is constructed to obtain an optimization target model, and constraint conditions are set for the optimization target model to obtain an optimization problem description;

[0060] S2. Based on the description of the optimization problem, initialize the parameters of the particle swarm optimization algorithm to obtain the initial particle swarm, and evaluate the fitness of the initial particle swarm to obtain the initial target solution;

[0061] S3, iteratively update the initial particle swarm to obtain a set of candidate solutions, and sort the candidate solution set by fitness to obtain an iterative target solution;

[0062] S4, comparing the iterative target solution with the initial target solution to obtain a global target solution, and updating the particle swarm according to the global target solution to obtain an updated particle swarm;

[0063] S5. Repeat steps S1 to S4 according to a preset number of iterations until the termination condition is met and the target optimization solution is obtained;

[0064] S6. Generate operating parameter adjustment instructions based on the target optimization solution to obtain an optimization parameter strategy.

[0065] Specifically, according to the characteristics of the low-load working mode, a parameter optimization objective function is constructed to obtain an optimization target model. This objective function usually includes the weighted sum of multiple indicators such as efficiency, stability and response speed. For example, efficiency can be used as the main optimization target, while considering output voltage fluctuation and system response time as constraints. Constraints are set for the optimization target model, including the value range of each parameter, system safety operation restrictions, etc., so as to obtain a complete description of the optimization problem. Based on the description of the optimization problem, the parameters of the particle swarm optimization algorithm are initialized. The particle swarm optimization algorithm is a group intelligence optimization method that simulates the foraging behavior of bird flocks. The initialization process includes setting the number of particles, the maximum number of iterations, the learning factor and the inertia weight. Each particle represents a set of possible optimization parameter combinations, such as switching frequency, duty cycle, etc. The initial particle swarm is usually generated in a random manner to cover the entire parameter search space. The fitness of the initial particle swarm is evaluated, and the objective function value corresponding to each particle is calculated to obtain the initial target solution.

[0066] Subsequently, the initial particle swarm is iteratively updated. During the update process, each particle adjusts its speed and position according to its own optimal position and the global optimal position. This process simulates the behavior of mutual learning and information sharing among individuals in the group. After the update, a set of candidate solutions is obtained, which are sorted by fitness to obtain the optimal solution of the current iteration, i.e., the iterative target solution. The iterative target solution is compared with the initial target solution, and the better solution is selected as the global target solution. The entire particle swarm is updated according to the global target solution, guiding the search to focus on more promising areas. This process is repeated until the preset number of iterations is reached or the termination condition is met, such as the improvement amplitude is less than the threshold for multiple consecutive iterations. The optimal solution finally obtained is the target optimization solution.

[0067] According to the target optimization solution, specific operating parameter adjustment instructions are generated to form an optimization parameter strategy. These instructions will guide the photovoltaic optimizer on how to adjust its operating parameters to achieve optimal efficiency under low load conditions. For example, a photovoltaic optimizer needs to optimize two key parameters, switching frequency and duty cycle, in low load mode. The objective function is set to maximize efficiency while considering the constraint that the output voltage fluctuation does not exceed ±5%. Initialize the particle swarm, set the number of particles to 50, and the maximum number of iterations to 100. Each particle contains two dimensions: switching frequency (range 20-60kHz) and duty cycle (range 30%-70%).

[0068] After initialization, 50 particles are randomly generated, for example, the initial position of particle 1 is (35kHz, 45%), particle 2 is (42kHz, 52%), etc. The fitness of these 50 particles is evaluated, assuming that the efficiency of particle 1 is 93.5%, the efficiency of particle 2 is 94.2%, and both meet the voltage fluctuation constraint. The initial global optimal solution is 94.2%. Entering the iteration process, each particle updates the speed and position according to its own best position and the global best position. For example, particle 1 may be updated to (38kHz, 48%), and the efficiency is recalculated to 94.0%. All updated particles are reordered, assuming that the new optimal solution is 94.5%, and the corresponding parameters are (40kHz, 50%). This new solution replaces the original global optimal solution and becomes the reference point for the next round of iterations. After 100 iterations, or when the efficiency improvement is less than 0.1% for 10 consecutive iterations, the algorithm terminates. Assume that the optimal solution finally obtained is an efficiency of 95.2%, the corresponding switching frequency is 39kHz, and the duty cycle is 51%. Based on this result, a parameter adjustment instruction is generated: the switching frequency is adjusted to 39kHz and the duty cycle is set to 51%.

[0069] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0070] (1) According to the optimization parameter strategy, the topological structure of the soft switching circuit is analyzed to obtain a topological feature description, and a soft switching equivalent model is constructed based on the topological feature description to obtain equivalent circuit parameters;

[0071] (2) Perform sensitivity analysis on the equivalent circuit parameters to obtain a key parameter set, and design a soft switching adjustment strategy based on the key parameter set to obtain a parameter adjustment scheme;

[0072] (3) Based on the parameter adjustment scheme, the timing of zero voltage switching and zero current switching is optimized to obtain the switching timing arrangement, and the dead time is calculated according to the switching timing arrangement to obtain the dead time control parameter;

[0073] (4) dynamically adjusting and calculating the parameters of the resonant circuit to obtain the resonant parameter value, and designing the resonant circuit control strategy according to the resonant parameter value to obtain the resonant control parameter;

[0074] (5) Based on the dead zone control parameters and the resonant control parameters, a predictive control algorithm is designed to obtain a soft switching prediction model, and a control sequence is generated through the soft switching prediction model to obtain predictive control parameters;

[0075] (6) The dead zone control parameters, resonant control parameters and predictive control parameters are integrated to form a soft switching control strategy and obtain soft switching control parameters.

[0076] Specifically, according to the optimization parameter strategy, the topological structure of the soft switching circuit is analyzed to obtain the topological feature description. This process involves a detailed study of the layout and connection methods of the switching devices, resonant elements and auxiliary circuits in the circuit. The topological feature description includes information such as the basic structure of the circuit, the type and number of switching devices, and the composition of the resonant network. Based on these topological features, a soft switching equivalent model is constructed, which simplifies the complex physical structure and uses equivalent circuit parameters to represent the electrical characteristics of the system. Equivalent circuit parameters usually include equivalent inductance, equivalent capacitance, equivalent resistance, etc. Subsequently, a sensitivity analysis is performed on the equivalent circuit parameters to determine which parameters have the greatest impact on the system performance, thereby obtaining a set of key parameters. Sensitivity analysis methods include local sensitivity analysis and global sensitivity analysis, which quantify the importance of each parameter by calculating the degree of impact of parameter changes on the output. Based on the obtained key parameter set, a soft switching adjustment strategy is designed to form a parameter adjustment plan. This plan clarifies the parameters that need to be adjusted and their adjustment direction and amplitude.

[0077] Based on the parameter adjustment scheme, the timing of zero voltage switching (ZVS) and zero current switching (ZCS) is optimized. ZVS and ZCS are two main implementation methods of soft switching technology. By switching when the voltage or current of the switching device is zero, the switching loss is significantly reduced. The goal of timing optimization is to ensure that the switching device switches state at the best time. After optimization, the switching timing arrangement is obtained, and the dead time is calculated based on it. The dead time refers to the time interval between the main switch turning off and the auxiliary switch turning on, which is crucial for the realization of soft switching. The calculated dead time control parameters will be used for the subsequent control strategy formulation. Dynamic adjustment and calculation of the parameters of the resonant circuit is another important part of soft switching optimization. The resonant circuit is usually composed of inductors and capacitors, and its parameters directly affect the implementation effect of soft switching. The dynamic adjustment process takes into account factors such as load changes and input voltage fluctuations, and the optimal resonant parameter values ​​are obtained through real-time calculation. According to these parameter values, the resonant circuit control strategy is designed to obtain the resonant control parameters, which will guide the real-time adjustment of the resonant circuit.

[0078] Based on the dead zone control parameters and resonant control parameters obtained above, a predictive control algorithm is designed. Predictive control is an advanced control method that can predict future behavior based on the current state of the system and historical data, and make control decisions accordingly. By establishing a soft switching prediction model, the voltage and current change trends of the switching device can be predicted in advance, so as to control the switching timing more accurately. The control sequence generated by the prediction model forms the predictive control parameters, which will be used to adjust the switching operation in real time. Finally, the dead zone control parameters, resonant control parameters and predictive control parameters are integrated to form a complete soft switching control strategy and obtain the final soft switching control parameters. These parameters comprehensively consider the switching timing, resonant characteristics and system dynamic response, and can achieve the optimal control of soft switching.

[0079] For example, a photovoltaic optimizer adopts LLC resonant topology. Through analysis, its equivalent circuit contains an equivalent inductor Le = 100μH, an equivalent capacitor Ce = 50nF, and an equivalent resistor Re = 10Ω. The sensitivity analysis results show that the equivalent inductor Le has the greatest impact on system performance, followed by the equivalent capacitor Ce. The adjustment scheme designed based on this is: give priority to adjusting the resonant inductor value, and fine-tune the resonant capacitor at the same time. The timing optimization calculation shows that at a switching frequency of 35kHz, the ZVS optimal turn-on time of the main switch is when the voltage drops to 10V, and the ZCS optimal turn-on time of the auxiliary switch is when the current drops to 0.5A. Based on this calculation, the optimal dead time is 300ns. The dynamic adjustment results of the resonant parameters show that at the current 20% load rate, the optimal resonant inductor value is 90μH and the resonant capacitor value is 48nF.

[0080] Based on these parameters, the predictive control algorithm builds a model that can predict the voltage and current changes in the next five switching cycles. The model predicts that the switch voltage may rise rapidly within 200ns when the load changes suddenly, and generates a control sequence that triggers the switch 50ns in advance. The soft switching control parameter set includes: dead time 300ns, resonant inductor 90μH, resonant capacitor 48nF, switching frequency 35kHz, and predictive control instructions triggered 50ns in advance when the load changes. The application of this set of parameters has increased the efficiency of the photovoltaic optimizer from the original 92% to 95.5% under 20% load conditions, significantly improving the performance at low load rates.

[0081] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0082] (1) According to the soft switch control parameters, the current of each phase of the multi-phase parallel structure is analyzed to obtain the current distribution characteristics, and the current imbalance is calculated based on the current distribution characteristics to obtain the phase imbalance index;

[0083] (2) According to the phase imbalance index, the number of parallel phases is dynamically evaluated to obtain the target number of phases, and a phase staggered control strategy is designed based on the target number of phases to obtain the phase difference set value;

[0084] (3) Optimizing the power distribution of each parallel phase to obtain the power distribution ratio, and designing a dynamic balancing algorithm based on the power distribution ratio to obtain the balancing control parameters;

[0085] (4) Based on the target number of phases and the phase difference setting value, a phase switching strategy is designed to obtain a phase switching sequence, and a soft start and soft shutdown scheme is designed according to the phase switching sequence to obtain a switching control parameter;

[0086] (5) By balancing control parameters and switching control parameters, a multi-phase parallel dynamic control model is constructed to obtain a parallel control strategy, and phase control instructions are generated according to the parallel control strategy to obtain the target parallel configuration.

[0087] Specifically, according to the soft switch control parameters, the current of each phase of the multi-phase parallel structure is analyzed to obtain the current distribution characteristics. This process involves current sampling and data processing of each parallel phase, and the distribution of current between each phase is described by statistical analysis methods, such as calculating the mean and standard deviation. Based on the current distribution characteristics, the current imbalance is calculated to obtain the phase imbalance index. The current imbalance is usually expressed as the deviation of each phase current from the average current, reflecting the uniformity of current distribution in the parallel structure. According to the calculated phase imbalance index, the number of parallel phases is dynamically evaluated. The evaluation process considers factors such as the current load level, the degree of imbalance, and the system efficiency target, and determines the number of parallel phases that best suits the current working state, that is, the target number of phases, through an optimization algorithm. Based on the target number of phases, a phase interleaving control strategy is designed. Phase interleaving technology is a commonly used control method in multi-phase parallel systems. By staggering the conduction time of each phase switch, the input and output ripples can be reduced and the system efficiency can be improved. The core of the phase interleaving control strategy is to determine the phase difference between each phase, that is, the phase difference set value.

[0088] The power distribution of each parallel phase is optimized and calculated to obtain the power distribution ratio. This step aims to achieve load balancing between phases and improve the overall efficiency of the system. The optimization calculation takes into account factors such as the current carrying capacity, heat distribution, and efficiency curve of each phase, and uses a nonlinear optimization algorithm to solve the optimal distribution scheme. According to the calculated power distribution ratio, a dynamic balancing algorithm is designed to obtain the balancing control parameters. The dynamic balancing algorithm achieves dynamic power distribution by adjusting the duty cycle or reference current of each phase in real time. Based on the target number of phases and the phase difference setting value determined previously, a phase switching strategy is designed. The phase switching strategy determines how to smoothly transition between different numbers of parallel phases, including the increase and decrease of phases. Through the optimization algorithm, a phase switching sequence is obtained, which specifies the start-up or shutdown order of each phase during the load change process. According to the phase switching sequence, a soft start and soft shutdown scheme is designed to obtain the switching control parameters. Soft start and soft shutdown technology can reduce the current impact during the phase switching process, protect system components, and improve reliability.

[0089] Finally, a multi-phase parallel dynamic control model is constructed by integrating the balancing control parameters and switching control parameters. This model describes the dynamic behavior of the multi-phase system under different working conditions, including phase number changes, power distribution adjustments, etc. Based on this model, a parallel control strategy is formulated, and specific phase control instructions are generated, and finally the target parallel configuration is obtained. For example, a photovoltaic optimizer adopts a 4-phase parallel structure and is dynamically reconstructed under low load rate working conditions. First, the current of the four parallel phases is sampled by a high-precision current sensor with a sampling frequency of 100kHz and data collected within 1ms. After data processing, the average current of each phase is 1.2A, 1.1A, 0.9A, and 0.8A respectively. The current imbalance is calculated, and the phase imbalance index is 15%.

[0090] Based on this imbalance indicator, combined with the current load rate of 20%, the target number of phases is determined to be 2 through a dynamic evaluation algorithm. Subsequently, a phase interleaving control strategy is designed, and the optimal phase difference is calculated to be 180 degrees. The power distribution of the two parallel phases is optimized, and the optimal power distribution ratio is calculated to be 55:45 considering the current distribution characteristics. According to this distribution ratio, a dynamic balancing algorithm is designed to obtain the balancing control parameters: the duty cycle of phase 1 is adjusted to 52%, and the duty cycle of phase 2 is adjusted to 48%. The phase switching strategy determines the shutdown of phases 3 and 4, and the switching order is to shut down phase 4 first and then phase 3. The soft shutdown scheme is designed to linearly reduce the duty cycle of phases 3 and 4 to 0 within 50μs, while gradually increasing the duty cycle of phases 1 and 2. The final dynamic control model comprehensively considers phase adjustment, power distribution and switching process, and the generated control instructions include: activating phase 1 and phase 2, setting the phase difference to 180 degrees, phase 1 duty cycle to 52%, phase 2 duty cycle to 48%, and completing soft shutdown of phase 3 and phase 4 within 1ms. The implementation of this parallel configuration has increased the efficiency of the photovoltaic optimizer from the original 91% to 94.5% at a load rate of 20%, while reducing the output current ripple by 40%, significantly improving the performance and stability at low load rates.

[0091] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0092] (1) According to the target parallel configuration, an efficiency optimization evaluation model is constructed to obtain efficiency evaluation indicators, and historical optimization results are analyzed based on the efficiency evaluation indicators to obtain the optimization effect trend;

[0093] (2) The optimization effect trend is learned through the long short-term memory network algorithm to obtain the efficiency prediction model, and the efficiency prediction model is used to predict the efficiency of different parameter combinations to obtain the parameter-efficiency mapping relationship;

[0094] (3) Based on the parameter-efficiency mapping relationship, a reinforcement learning environment is designed to obtain a learning environment model, and the reinforcement learning agent is trained through the learning environment model to obtain an optimized strategy function;

[0095] (4) Generate parameter adjustment actions through the optimization strategy function to obtain a parameter adjustment sequence, and execute the parameter adjustment sequence to perform actual optimization attempts to obtain optimization result data;

[0096] (5) Evaluate the optimization result data, update the reinforcement learning agent, obtain an updated optimization strategy function, and generate new parameter adjustment actions based on the updated optimization strategy function to obtain low-load efficiency optimization control instructions.

[0097] Specifically, an efficiency optimization evaluation model is constructed according to the target parallel configuration. The model comprehensively considers factors such as input power, output power, and switching loss to calculate the system efficiency. Efficiency evaluation indicators usually include conversion efficiency, power factor, etc. Based on these indicators, a time series analysis of historical optimization results is performed to obtain the optimization effect trend. This trend reflects the changing law of system efficiency over time and different optimization strategies. Subsequently, the long short-term memory network (LSTM) algorithm is used to learn the optimization effect trend. LSTM is a special recurrent neural network that can effectively process long sequence data and capture long-term dependencies. Through LSTM learning, an efficiency prediction model is established. The input of this model includes the current working state, environmental parameters, etc., and the output is the predicted system efficiency. The model is used to predict the efficiency of different parameter combinations and obtain the parameter-efficiency mapping relationship. This mapping relationship describes the impact of various parameter settings on system efficiency.

[0098] Based on the parameter-efficiency mapping relationship, a reinforcement learning environment is designed. The reinforcement learning environment simulates the working process of the PV optimizer, including the state space (such as load factor, temperature, etc.), action space (such as switching frequency, dead time adjustment, etc.) and reward function (based on the degree of efficiency improvement). Through this learning environment model, the reinforcement learning agent is trained. The reinforcement learning agent uses algorithms such as Q-learning or deep Q network (DQN) to learn the optimal parameter adjustment strategy by continuously interacting with the environment. During the training process, the agent explores different parameter adjustment actions and updates its strategy based on the rewards obtained. Finally, the optimization policy function is obtained, which can select the optimal parameter adjustment action based on the current state. Using the trained optimization policy function, a sequence of parameter adjustment actions is generated. This sequence specifies how to gradually adjust the system parameters to improve efficiency. These parameter adjustments are executed, actual optimization attempts are made, and the optimization result data is recorded. The optimization result data includes information such as efficiency changes and power fluctuations before and after the adjustment.

[0099] Evaluate the optimization result data, calculate the actual efficiency improvement, and compare it with the expected effect. According to the evaluation results, update the value function or policy network of the reinforcement learning agent to obtain the updated optimization policy function. This update process uses experience replay technology to add new experience data to the training set and retrain the agent. Based on the updated optimization policy function, generate new parameter adjustment actions to form low-load efficiency optimization control instructions. For example, a photovoltaic optimizer performs efficiency optimization at a load rate of 20%. First, build an efficiency evaluation model, considering factors such as input power, output power, and switching loss. Analyzing the optimization data of the past week, it is found that the efficiency has gradually increased from the initial 91% to 94.5%, showing a fluctuating upward trend.

[0100] The LSTM network is used to learn this trend. The input includes daily load rate, ambient temperature, optimization parameters, etc., and the output is the next day's predicted efficiency. The trained LSTM model achieves an average prediction error of 0.2% on the test set. The model is used to predict different parameter combinations. For example, at 20% load rate and 25℃ ambient temperature, the combined prediction efficiency of switching frequency 35kHz and dead time 300ns is 94.7%. A reinforcement learning environment is constructed. The state space includes load rate (10%-30%), temperature (15℃-35℃), etc., and the action space includes switching frequency adjustment (±5kHz), dead time adjustment (±50ns), etc. The reward function is designed as the percentage of efficiency improvement. The DQN algorithm is used to train the agent. After 10,000 interactions, the optimized policy function is obtained. Apply the optimized policy function to generate a parameter adjustment sequence: reduce the switching frequency from 40kHz to 35kHz, and increase the dead time from 250ns to 300ns. After performing these adjustments, the actual measured efficiency increased from 94.5% to 94.8%. The evaluation results show that the actual improvement is basically consistent with the prediction. Update the reinforcement learning agent, add the new data to the experience pool, and retrain for 100 rounds. The updated policy function generates new control instructions: keep the switching frequency unchanged at 35kHz and further increase the dead time to 320ns. This process continues to optimize the efficiency performance of the system at low load rates. With this method, the efficiency of the photovoltaic optimizer at 20% load rate finally stabilizes at around 95%, significantly improving the energy conversion efficiency under low load rate working conditions.

[0101] The above describes the low load rate efficiency optimization method of the photovoltaic optimizer in the embodiment of the present application. The following describes the low load rate efficiency optimization system of the photovoltaic optimizer in the embodiment of the present application. Figure 2 In one embodiment of the present application, a low load efficiency optimization system of a photovoltaic optimizer includes:

[0102] The monitoring module 201 is used to monitor the working state of the photovoltaic optimizer in real time and obtain working state data;

[0103] An identification module 202 is used to analyze and identify the working mode of the photovoltaic optimizer based on the working status data to obtain the low load rate working mode characteristics;

[0104] A calculation module 203 is used to dynamically optimize and calculate the operating parameters of the photovoltaic optimizer according to the characteristics of the low load rate working mode to obtain an optimization parameter strategy;

[0105] An adjustment module 204, configured to adaptively adjust the soft switch topology of the photovoltaic optimizer based on the optimization parameter strategy to obtain soft switch control parameters;

[0106] A reconstruction module 205, configured to dynamically reconstruct the multi-phase parallel structure of the photovoltaic optimizer according to the soft switch control parameters to obtain a target parallel configuration;

[0107] The iteration module 206 is used to continuously iterate the efficiency optimization process of the photovoltaic optimizer based on the target parallel configuration to obtain a low load efficiency optimization control instruction.

[0108] Through the cooperation of the above components, the real-time monitoring of the working status of the photovoltaic optimizer provides an accurate and timely data basis for the entire optimization process, ensuring the accuracy of subsequent optimization decisions. The working mode analysis and identification based on these working status data enable the system to quickly and accurately determine whether it is currently in a low load rate state, and extract the corresponding working mode characteristics, laying the foundation for targeted optimization. According to the identified low load rate working mode characteristics, the system can perform dynamic optimization calculations to obtain the most suitable optimization parameter strategy for the current working state. This real-time adjustment capability greatly improves the adaptability of the system in a changing environment. The adaptive adjustment of the soft switching topology further reduces the switching loss and improves the overall efficiency of the system, especially under low load rate conditions, this optimization effect is more significant. The dynamic reconstruction of the multi-phase parallel structure provides the system with a more flexible power management solution, which can dynamically adjust the number of parallel phases according to the load conditions, which not only ensures efficiency but also optimizes the current distribution. Finally, through the continuous iteration of the efficiency optimization process, the system can continuously learn and improve, adapt to changes in various working conditions, and maintain a high-efficiency operating state for a long time. This comprehensive, dynamic and intelligent optimization method enables the photovoltaic optimizer to maintain a high working efficiency under low load conditions, effectively solving the problem of reduced efficiency of traditional optimizers at low load rates, and improving the efficiency of low load rate efficiency optimization of the photovoltaic optimizer.

[0109] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing the low load efficiency of a photovoltaic optimizer, characterized in that: The low load rate efficiency optimization method of the photovoltaic optimizer comprises: Monitor the working status of the photovoltaic optimizer in real time to obtain working status data; Based on the working status data, the working mode of the photovoltaic optimizer is analyzed and identified to obtain the characteristics of the low load rate working mode; According to the characteristics of the low load rate working mode, the operating parameters of the photovoltaic optimizer are dynamically optimized and calculated to obtain an optimization parameter strategy; Based on the optimization parameter strategy, the soft switch topology of the photovoltaic optimizer is adaptively adjusted to obtain the soft switch control parameters; According to the soft switch control parameters, dynamically reconstructing the multi-phase parallel structure of the photovoltaic optimizer to obtain a target parallel configuration; Based on the target parallel configuration, the efficiency optimization process of the photovoltaic optimizer is continuously iterated to obtain a low load efficiency optimization control instruction.

2. The low load efficiency optimization method of a photovoltaic optimizer according to claim 1, characterized in that: The real-time monitoring of the working state of the photovoltaic optimizer to obtain working state data includes: Sampling the input voltage of the photovoltaic optimizer to obtain input voltage data, and filtering the input voltage data to obtain a smoothed input voltage value; Sampling the input current of the photovoltaic optimizer to obtain input current data, and filtering the input current data to obtain a smoothed input current value; Sampling the output voltage of the photovoltaic optimizer to obtain output voltage data, and filtering the output voltage data to obtain a smoothed output voltage value; Sampling the output current of the photovoltaic optimizer to obtain output current data, and filtering the output current data to obtain a smoothed output current value; The input power of the photovoltaic optimizer is calculated according to the smoothed input voltage value and the smoothed input current value to obtain an input power value, and the output power of the photovoltaic optimizer is calculated according to the smoothed output voltage value and the smoothed output current value to obtain an output power value; Calculating the efficiency of the photovoltaic optimizer according to the input power value and the output power value to obtain a real-time efficiency value, and performing a threshold comparison on the real-time efficiency value to obtain an efficiency status flag; Measuring the switching frequency of the photovoltaic optimizer to obtain switching frequency data, and performing statistical analysis on the switching frequency data to obtain switching frequency characteristics; Measuring the duty cycle of the photovoltaic optimizer to obtain duty cycle data, and performing statistical analysis on the duty cycle data to obtain duty cycle characteristics; Collecting the temperature of the photovoltaic optimizer to obtain temperature data, and performing trend analysis on the temperature data to obtain temperature change characteristics; The real-time efficiency value, the efficiency status flag, the switching frequency characteristic, the duty cycle characteristic and the temperature change characteristic are combined to obtain the working status data.

3. The low load efficiency optimization method of a photovoltaic optimizer according to claim 1, characterized in that: The working mode of the photovoltaic optimizer is analyzed and identified based on the working status data to obtain the low load rate working mode characteristics, including: Normalizing the output power value in the working status data to obtain a load factor value, and performing a threshold comparison on the load factor value to obtain a low load factor mark; According to the low load rate mark, the working state data is divided into time windows to obtain a low load rate time period, and the working state data in the low load rate time period is clustered and analyzed to obtain a working mode cluster; Extracting features from the working mode cluster to obtain a mode feature vector, and classifying the mode feature vector using a support vector machine algorithm to obtain a working mode category; According to the working mode category, historical working state data is retrieved to obtain similar working condition data, and statistical analysis is performed on the similar working condition data to obtain a target working parameter set; Performing association rule mining on the target working parameter set to obtain parameter optimization rules, and evaluating the current working state according to the parameter optimization rules to obtain parameter adjustment suggestions; According to the working mode category, the target working parameter set and the parameter adjustment suggestion, a low load rate working mode description is constructed to obtain the low load rate working mode characteristics.

4. The low load efficiency optimization method of a photovoltaic optimizer according to claim 1, characterized in that: According to the low load rate working mode characteristics, the operating parameters of the photovoltaic optimizer are dynamically optimized and calculated to obtain an optimization parameter strategy, including: S1. According to the characteristics of the low-load rate working mode, a parameter optimization objective function is constructed to obtain an optimization target model, and constraint conditions are set for the optimization target model to obtain an optimization problem description; S2. Based on the optimization problem description, the parameters of the particle swarm optimization algorithm are initialized to obtain an initial particle swarm, and the fitness of the initial particle swarm is evaluated to obtain an initial target solution; S3, iteratively updating the initial particle swarm to obtain a candidate solution set, and sorting the candidate solution set by fitness to obtain an iterative target solution; S4, comparing the iterative target solution with the initial target solution to obtain a global target solution, and updating the particle swarm according to the global target solution to obtain an updated particle swarm; S5. Repeat steps S1 to S4 according to a preset number of iterations until the termination condition is met and the target optimization solution is obtained; S6. Generate an operating parameter adjustment instruction according to the target optimization solution to obtain the optimization parameter strategy.

5. The low load efficiency optimization method of a photovoltaic optimizer according to claim 1, characterized in that: Based on the optimization parameter strategy, the soft switch topology of the photovoltaic optimizer is adaptively adjusted to obtain the soft switch control parameters, including: According to the optimization parameter strategy, the topological structure of the soft switch circuit is analyzed to obtain a topological feature description, and a soft switch equivalent model is constructed based on the topological feature description to obtain equivalent circuit parameters; Performing sensitivity analysis on the equivalent circuit parameters to obtain a key parameter set, and designing a soft switch adjustment strategy based on the key parameter set to obtain a parameter adjustment scheme; Based on the parameter adjustment scheme, the timing of the zero voltage switch and the zero current switch is optimized to obtain the switch timing arrangement, and the dead time is calculated according to the switch timing arrangement to obtain the dead time control parameter; Dynamically adjust and calculate the parameters of the resonant circuit to obtain a resonant parameter value, and design a resonant circuit control strategy based on the resonant parameter value to obtain a resonant control parameter; Based on the dead zone control parameter and the resonance control parameter, a predictive control algorithm is designed to obtain a soft switching prediction model, and a control sequence is generated through the soft switching prediction model to obtain a predictive control parameter; The dead zone control parameter, the resonance control parameter and the prediction control parameter are integrated to form a soft switching control strategy and obtain the soft switching control parameter.

6. The low load efficiency optimization method of a photovoltaic optimizer according to claim 1, characterized in that: The method of dynamically reconstructing the multi-phase parallel structure of the photovoltaic optimizer according to the soft switch control parameters to obtain a target parallel configuration includes: According to the soft switch control parameters, the current of each phase of the multi-phase parallel structure is analyzed to obtain the current distribution characteristics, and the current imbalance is calculated based on the current distribution characteristics to obtain the phase imbalance index; According to the inter-phase imbalance index, dynamically evaluate the number of parallel phases to obtain a target number of phases, and design a phase staggered control strategy based on the target number of phases to obtain a phase difference setting value; Optimizing and calculating the power distribution of each parallel phase to obtain a power distribution ratio, and designing a dynamic balancing algorithm according to the power distribution ratio to obtain a balancing control parameter; Based on the target number of phases and the phase difference setting value, a phase switching strategy is designed to obtain a phase switching sequence, and a soft start and soft shutdown scheme is designed according to the phase switching sequence to obtain a switching control parameter; A multi-phase parallel dynamic control model is constructed by using the balancing control parameters and the switching control parameters to obtain a parallel control strategy, and a phase control instruction is generated according to the parallel control strategy to obtain the target parallel configuration.

7. The low load efficiency optimization method of a photovoltaic optimizer according to claim 6, characterized in that: The efficiency optimization process of the photovoltaic optimizer is continuously iterated based on the target parallel configuration to obtain a low load efficiency optimization control instruction, including: According to the target parallel configuration, an efficiency optimization evaluation model is constructed to obtain an efficiency evaluation index, and based on the efficiency evaluation index, historical optimization results are analyzed to obtain an optimization effect trend; The optimization effect trend is learned by a long short-term memory network algorithm to obtain an efficiency prediction model, and the efficiency prediction is performed on different parameter combinations by the efficiency prediction model to obtain a parameter-efficiency mapping relationship; Based on the parameter-efficiency mapping relationship, a reinforcement learning environment is designed to obtain a learning environment model, and a reinforcement learning agent is trained through the learning environment model to obtain an optimization strategy function; Generate parameter adjustment actions through the optimization strategy function to obtain a parameter adjustment sequence, and execute the parameter adjustment sequence to perform actual optimization attempts to obtain optimization result data; The optimization result data is evaluated, the reinforcement learning agent is updated, an updated optimization strategy function is obtained, and a new parameter adjustment action is generated based on the updated optimization strategy function to obtain the low-load efficiency optimization control instruction.

8. A low load efficiency optimization system for a photovoltaic optimizer, characterized in that: A method for optimizing the efficiency of a photovoltaic optimizer at a low load rate according to any one of claims 1 to 7, wherein the system for optimizing the efficiency of a photovoltaic optimizer at a low load rate comprises: A monitoring module is used to monitor the working status of the photovoltaic optimizer in real time and obtain working status data; An identification module, used to analyze and identify the working mode of the photovoltaic optimizer based on the working status data, and obtain the characteristics of the low load rate working mode; A calculation module, used to dynamically optimize and calculate the operating parameters of the photovoltaic optimizer according to the characteristics of the low load rate working mode to obtain an optimization parameter strategy; A regulating module, used for adaptively regulating the soft switch topology of the photovoltaic optimizer based on the optimization parameter strategy to obtain soft switch control parameters; A reconstruction module, used for dynamically reconstructing the multi-phase parallel structure of the photovoltaic optimizer according to the soft switch control parameters to obtain a target parallel configuration; The iteration module is used to continuously iterate the efficiency optimization process of the photovoltaic optimizer based on the target parallel configuration to obtain a low load efficiency optimization control instruction.

Images