Control method and device of alternating current and direct current power conversion device based on clustering training

Abstract

The invention provides a control method and device of an alternating current and direct current power conversion device based on clustering training, and relates to the technical field of power systems. The cluster training-based AC / DC power conversion device control method comprises the steps of obtaining state data of an AC / DC power conversion device for a preset operation and maintenance event; processing the state data by using the operation and maintenance management and control model, and generating a control parameter vector for a preset operation and maintenance event; and indicating a control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector so as to realize an expected operation and maintenance result of the preset operation and maintenance event. According to the invention, the AC / DC power conversion device can be controlled efficiently and accurately, so that the expected regulation and control effect can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a control method and device for an AC / DC power conversion device based on cluster training. Background Art

[0002] With the vigorous development of the energy internet and the large-scale integration of renewable energy, AC / DC conversion systems, as the core link in energy transmission and distribution, are facing challenges. Technological advancements and improvements in equipment performance are key to promoting efficient energy utilization. Voltage source converter (VSC)-based AC / DC power conversion systems (such as VSC-HVDC), representing a new generation of flexible transmission technologies, have become a crucial support for smart grids and distributed energy systems due to their flexibility, controllability, and fast response speed. However, the control algorithms for VSC devices are highly complex, involving issues such as multivariable nonlinear coupling, PWM modulation strategy optimization, and dynamic characteristic modeling of power electronic devices. Traditional control methods (such as PID control and vector control) have limitations when dealing with time-varying system parameters, load disturbances, and multi-objective optimization requirements, including insufficient robustness, low energy conversion efficiency, and significant discrepancies between theoretical and actual operating performance.

[0003] To overcome these bottlenecks, AI-based UPH (unified power flow) control systems have emerged. AI technology, through data-driven modeling methods (such as deep learning and reinforcement learning), can automatically extract system dynamic characteristics, optimize control parameters, and achieve multi-objective collaborative optimization (such as efficiency, stability, and response speed). For example, using reinforcement learning algorithms to adjust PWM modulation strategies online can significantly improve system robustness; using deep neural networks to approximate complex nonlinear mapping relationships can compensate for the shortcomings of traditional modeling; and a hybrid decision-making mechanism that combines expert knowledge bases with AI algorithms can balance control accuracy and real-time requirements. However, the application of AI in UPH control still faces challenges such as data quality, algorithm interpretability, and engineering implementation.

[0004] Therefore, how to efficiently and accurately control the AC / DC power conversion device to achieve the expected regulation effect is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The present invention provides a control method and device for an AC / DC power conversion device based on cluster training, which are used to solve the above-mentioned defects in the prior art and efficiently and accurately control the AC / DC power conversion device to achieve the expected regulation effect.

[0006] The present invention provides a control method for an AC / DC power conversion device based on cluster training, comprising the following steps.

[0007] Acquire status data of an AC / DC power conversion device for a preset operation and maintenance event; process the status data using an operation and maintenance management and control model to generate a control parameter vector for the preset operation and maintenance event; instruct a control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve an expected operation and maintenance result of the preset operation and maintenance event; wherein the operation and maintenance management and control model is a machine learning model trained based on a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data are historical status data of the AC / DC power conversion device for the preset operation and maintenance event; the label of the sample data is generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

[0008] According to a control method for an AC / DC power conversion device based on cluster training provided by the present invention, the operation and maintenance management and control model is a batch-input batch-output neural network model, the status data includes status data for a plurality of preset operation and maintenance events, and the control parameter vector includes control parameter vectors for the plurality of preset operation and maintenance events; the use of the operation and maintenance management and control model to process the status data to generate the control parameter vectors for the preset operation and maintenance events includes: inputting the status data for the plurality of preset operation and maintenance events into the batch-input batch-output neural network model, and obtaining the control parameter vectors for the plurality of preset operation and maintenance events output by the batch-input batch-output neural network model.

[0009] According to a control method for an AC / DC power conversion device based on cluster training provided by the present invention, the training sample data set is obtained in the following manner: multiple groups of historical status data of the AC / DC power conversion device for the preset operation and maintenance events are obtained; using H2 and H∞ control algorithms, a control parameter vector corresponding to each group of the historical status data is obtained; wherein the control parameter vector is used to enable the AC / DC power conversion device to achieve the expected operation and maintenance result of the preset operation and maintenance event under the control of its control system; the training data set is obtained based on the multiple groups of historical status data and the control parameter vector corresponding to each group of the historical status data.

[0010] According to a control method for an AC / DC power conversion device based on clustering training provided by the present invention, the training data set is obtained based on the multiple groups of historical state data and the control parameter vectors corresponding to each group of the historical state data, including: using a linear weighted algorithm and a genetic algorithm to obtain multiple groups of sample data based on the multiple groups of historical state data and the control parameter vectors corresponding to each group of the historical state data; wherein the multiple groups of sample data include a first proportion of positive samples and a second proportion of negative samples; and the training sample data set is obtained based on the multiple groups of sample data and the control parameter vectors corresponding to each group of the sample data.

[0011] According to a control method for an AC / DC power conversion device based on clustering training provided by the present invention, the training sample data set is obtained based on the multiple groups of sample data and the control parameter vector corresponding to each group of the sample data, including: using a clustering algorithm to obtain a label for each group of the sample data based on the control parameter vectors corresponding to all the sample data; and obtaining the training sample data set based on each group of the sample data and its corresponding label.

[0012] According to a control method for an AC / DC power conversion device based on cluster training provided by the present invention, the method further includes: utilizing a reinforcement learning algorithm to optimize a control parameter vector based on state data accumulated during operation of the AC / DC power conversion device to achieve the expected operation and maintenance results under the preset operation and maintenance events; utilizing the accumulated state data and the corresponding control parameter vectors to establish an expert knowledge base; and utilizing the data in the expert knowledge base to retrain the operation and maintenance management model to obtain an optimized operation and maintenance management model.

[0013] The present invention also provides a control device for an AC / DC power conversion device based on cluster training, comprising the following modules: An acquisition module is used to obtain status data of an AC / DC power conversion device for preset operation and maintenance events; a generation module is used to process the status data using an operation and maintenance management and control model to generate a control parameter vector for the preset operation and maintenance event; a control module is used to instruct the control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve the expected operation and maintenance result of the preset operation and maintenance event; wherein the operation and maintenance management and control model is a machine learning model trained based on a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data are historical status data of the AC / DC power conversion device for the preset operation and maintenance event; the label of the sample data is generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

[0014] The present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the control method for an AC / DC power conversion device based on clustering training as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the control method of the AC / DC power conversion device based on cluster training as described above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned control methods for an AC / DC power conversion device based on cluster training.

[0017] The present invention provides a control method and device for an AC / DC power conversion device based on clustering training. The method and device obtain status data of the AC / DC power conversion device for a preset operation and maintenance event; process the status data using an operation and maintenance control model to generate a control parameter vector for the preset operation and maintenance event; and instruct the control system of the AC / DC power conversion device to control the AC / DC power conversion device based on the control parameter vector to achieve the expected operation and maintenance results of the preset operation and maintenance event. Because the operation and maintenance control model is obtained by learning a training sample data set based on a machine learning algorithm; the sample data included in the training sample data set is the historical status data of the AC / DC power conversion device for the preset operation and maintenance event, and the labels of the sample data are generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data, the sample data in the training sample data set has stronger interpretability and is not affected by data instability factors. The operation and maintenance control model obtained based on the training sample data set can efficiently and accurately generate the optimal control parameter vector for the preset operation and maintenance event, and then, based on the control parameter vector, the AC / DC power conversion device can be accurately and reliably controlled. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 It is a flow chart of a control method for an AC / DC power conversion device based on cluster training provided by the present invention.

[0020] Figure 2 It is a flowchart of the method for obtaining a sample data set provided by the present invention.

[0021] Figure 3 This is a schematic diagram of sample interpolation / extrapolation generation provided by the present invention.

[0022] Figure 4 It is a cluster analysis schematic diagram provided by the present invention.

[0023] Figure 5 It is a schematic diagram of retraining using a knowledge base provided by the present invention.

[0024] Figure 6It is a structural schematic diagram of a control device for an AC / DC power conversion device based on cluster training provided by the present invention.

[0025] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0026] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0027] The following combination Figure 1-Figure 5 The present invention describes a control method for an AC / DC power conversion device based on clustering training.

[0028] Figure 1 is a flow chart of a control method for an AC / DC power conversion device based on cluster training provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 101: Acquire status data of an AC / DC power conversion device for a preset operation and maintenance event.

[0029] AC / DC power converters are responsible for converting AC to DC, or vice versa, to enable the transmission and distribution of electrical energy between different voltage levels or systems. AC / DC power converters are key components of energy internet and renewable energy systems. They use power electronics technology to achieve flexible conversion and control of electrical energy to meet the needs of different application scenarios. In some embodiments, the AC / DC power converter is a voltage source converter, such as a VSC-HVDC (Voltage Source Converter based High Voltage Direct Current).

[0030] Preset O&M events are events that occur during the operation of an AC / DC power converter and require specific control strategies or response measures. These events may include, but are not limited to, failures in the power system to which the AC / DC power converter is connected, sudden load changes, voltage fluctuations, and so on. For example, a preset O&M event might be a short circuit failure in the power system to which the AC / DC power converter is connected.

[0031] Status data is a series of parameters or variables describing the current operating state of an AC / DC power converter. These parameters include key parameters such as voltage, current, power, and temperature, reflecting the converter's operating status under different operating conditions. This data may include, but is not limited to, input bus voltage and current, output bus voltage and current, bus active and reactive power, and system volatility and stability parameters. For example, if a drop in input bus voltage is detected, this voltage value and its rate of change are considered part of the status data.

[0032] During implementation, status data of the AC / DC power conversion device can be obtained through a variety of common methods, not limited to the descriptions in this specification. For example, sensor monitoring can be used to install voltage sensors, current sensors, temperature sensors, vibration sensors, and other sensors at key locations of the AC / DC power conversion device (such as the input / output busbars, power modules, and capacitor banks) to collect real-time data on physical quantities such as the device's voltage, current, temperature, and vibration. These sensors transmit this data to a monitoring system via wired or wireless means for subsequent analysis.

[0033] For example, remote monitoring and data collection can be performed on AC / DC power conversion devices using a SCADA system (Supervisory Control and Data Acquisition). SCADA systems can integrate multiple communication protocols (such as Modbus and IEC 61850) to read real-time status data from the device's PLC (Programmable Logic Controller) or embedded controller, including digital and analog values, fault alarms, and more.

[0034] Step 102: Process the status data using the operation and maintenance control model to generate a control parameter vector for a preset operation and maintenance event.

[0035] A control parameter vector is a set of control parameters generated by processing status data through the operation and maintenance management model in response to pre-defined operation and maintenance events (such as system failures and sudden load changes). The control parameter vector is used to control the operation of the AC / DC power converter to achieve the desired operation and maintenance results. The control parameter vector typically includes information on multiple dimensions, such as voltage adjustment, current limit, and power factor setting, which together determine the AC / DC power converter's response to specific operation and maintenance events.

[0036] For example, in an AC / DC power conversion system, the operation and maintenance control model generates the following control parameter vector for a short circuit: Voltage adjustment: +5% (indicates that when the load increases, the output voltage is increased by 5% to maintain stability); current limit value: 120% rated current (indicates that when the load increases, the maximum current allowed is 120% of the rated current); power factor setting: 0.95 (indicates that when the load increases, the power factor is adjusted to 0.95 to improve efficiency).

[0037] The operation and maintenance control model is a machine learning model trained based on a training sample dataset. In the specific implementation process, the operation and maintenance control model can be built based on a variety of machine learning models and is not limited by the description in this specification. For example, the control model can be built based on the transformer model. Considering the device performance and sample quantity limitations, the original transformer model structure can be simplified based on the distillation principle.

[0038] The input of the operation and maintenance control model is state data, and the output is the control parameter vector.

[0039] In some embodiments, the operation and maintenance control model is a batch input and batch output neural network model (for example, a batch input and batch output neural network model built based on the Transformer model architecture), the status data includes status data for a variety of preset operation and maintenance events, and the control parameter vector includes control parameter vectors for a variety of preset operation and maintenance events.

[0040] In the above embodiment, the status data for multiple preset operation and maintenance events can be batch-inputted into the batch-output neural network model to obtain control parameter vectors for the multiple preset operation and maintenance events output by the batch-input-batch-output neural network model.

[0041] The training sample data set includes multiple sets of sample data, which are historical status data of the AC / DC power conversion device for preset operation and maintenance events; the labels of the sample data are generated by clustering the control parameter vectors corresponding to the multiple sets of sample data. For a detailed description of the training sample data set, see Figure 2 The relevant content in will not be repeated here.

[0042] During the specific implementation process, the loss function can be used to measure the difference between the predicted value of the sample data based on the initial machine learning model and the corresponding label of the sample data; the parameters of the initial machine learning model are adjusted using a preset optimization algorithm (for example, a gradient descent optimization algorithm) to minimize the value of the loss function until the training termination condition is reached (the loss function converges, or the preset number of training times is reached, etc.) to obtain an operation and maintenance control model.

[0043] Step 103 : Instruct the control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve the expected operation and maintenance result of the preset operation and maintenance event.

[0044] The expected operation and maintenance result is the expected system operation state or operation and maintenance effect after the control system (for example, the VSC UPH PWM control system) of the AC-DC power conversion device controls the device according to the generated control parameter vector. The expected operation and maintenance result is set for a preset operation and maintenance event (such as a fault, a load change, etc.), and is intended to ensure stable, efficient and safe operation of the system.

[0045] For example, for a short-circuit fault event, the expected operation and maintenance result includes: quickly isolating the fault, the control system should cut off the fault current within a very short time (such as milliseconds) after detecting the short-circuit fault, to prevent the fault from expanding; protecting the equipment, to ensure that the current during the fault does not exceed a certain percentage (such as 130%) of the rated current of the equipment, to prevent the equipment from overheating or being damaged; restoring system stability, after the fault is cleared, the system should be able to quickly restore stable operation, and the voltage and current should gradually return to normal values.

[0046] In some embodiments, an enhanced learning algorithm can also be used to optimize the control parameter vector based on the accumulated state data in the operation of the AC-DC power conversion device, to achieve the expected operation and maintenance result under the preset operation and maintenance event; the accumulated state data and the control parameter vector corresponding thereto are used to establish an expert knowledge base including high-quality samples; the data in the expert knowledge base is used to retrain the operation and maintenance control model, to obtain an optimized operation and maintenance control model.

[0047] For example, based on the enhanced learning algorithm A3C (Asynchronous Advantage Actor-Critic, asynchronous advantage actor-critic algorithm), the control and management strategies embodied by the use control parameter vector of the AC-DC power conversion device for various operation and maintenance events can be summarized based on the accumulated state data in the operation of the AC-DC power conversion device, and the UPH operation expert knowledge base can be established based on the accumulated state data and the control parameter vector corresponding thereto, and the corresponding operation and maintenance control model can be optimized based on the knowledge in the expert knowledge base, so that the operation and maintenance control model can continue to find the optimal UPH topology and parameters.

[0048] As shown in FIG. 1, Figure 5 As shown in FIG. 1,

[0049] Figure 2is a flow chart of the method for obtaining a sample data set provided by the present invention, such as Figure 2 As shown, the method includes the following: Step 201: Acquire multiple groups of historical status data of an AC / DC power conversion device for preset operation and maintenance events.

[0050] For a detailed description of the AC / DC power conversion device, preset operation and maintenance events, and status data, see Figure 1 The relevant content in will not be repeated here.

[0051] Step 202: Utilize the H2 and H∞ control algorithms to obtain a control parameter vector corresponding to each set of historical state data; wherein the control parameter vector is used to enable the AC / DC power conversion device to achieve the expected operation and maintenance results of the preset operation and maintenance events under the control of its control system.

[0052] During implementation, the H2 control algorithm can be used to process historical state data. This algorithm focuses on optimizing system performance by minimizing quadratic performance metrics (such as the variance or energy of the system output) to seek the optimal control strategy for pre-set operational events. This results in a set of optimal control parameter vectors that ensure the AC / DC power converter system achieves the expected operational outcomes for the pre-set operational events. Simultaneously, the H∞ control algorithm is used to process historical state data. This algorithm emphasizes system robustness—the ability of a system to maintain stable operation in the face of uncertainty or disturbances. By optimizing the H∞ norm, a controller is designed, resulting in another set of control parameter vectors that enhance system robustness. By combining the results of the H2 and H∞ control algorithms, a corresponding control parameter vector is generated for each set of historical state data.

[0053] The hybrid H2 / H∞ algorithm described above, based on the H2 and H∞ control of AC / DC power conversion devices (e.g., VSC-HVDC), needs to consider the following constraints: power balance, including active and reactive power balance in the input and output areas, as well as transmission and conversion losses; voltage and current balance, based on the Wikihof voltage and current balance formula; and virtual flux balance, where virtual is a virtual parameter formed based on the integration of voltage or current.

[0054] At the same time, the H2 and H∞ control algorithms need to achieve the following optimization objectives: minimization of system losses, maximization of power factor, minimization of voltage and current fluctuations, minimization of voltage and current ghosting, input and output voltage shape requirements (sinusoidal), and network stability time.

[0055] For Voltage Source Converterbased High Voltage Direct Current (VSC-HVDC) transmission equipment, the related actions and control strategies implemented using control parameter vectors mainly include: VSC-HVDC rectifier part, VSC-HVDC inverter part, VSC-HVDC buck circuit, and VSC-HVDC boost circuit.

[0056] Specifically, the relevant control algorithm is: a PWM control algorithm based on H2 / H∞, which can obtain the relevant optimal parameters for system operation through matrix solution, ensure the robust and efficient operation of the AC / DC power conversion device control system, greatly reduce the risk of system disturbances, and thus ensure the accuracy and reliability of the relevant model.

[0057] Afterwards, the effectiveness of the control parameter vectors generated above can be evaluated through theoretical calculations and power simulation systems, and the state data with confidence greater than a threshold or meeting specific constraints and their corresponding control parameter vectors are retained as original samples.

[0058] The original sample not only considers the performance optimization of the system, but also takes into account the robustness requirements of the system, providing an important control basis for the subsequent operation and maintenance management model to achieve efficient and stable operation of AC / DC power conversion devices under complex working conditions.

[0059] Step 203: Using a linear weighted algorithm and a genetic algorithm, obtain multiple sets of sample data according to multiple sets of historical state data and the control parameter vector corresponding to each set of historical state data; wherein the multiple sets of sample data include positive samples in a first proportion and negative samples in a second proportion.

[0060] In the specific implementation process, a linear weighted algorithm (such as Figure 3 The interpolation / extrapolation sample generation algorithm shown in FIG2 is used to generate multiple sets of new sample data based on multiple sets of historical state data and the control parameter vector corresponding to each set of historical state data. This can be expressed as follows.

[0061] (1) in, , , is the state data sequence vector or tensor, + =1, (in the case of extrapolation, >1). This formula can be extended to multidimensional vector data, where the sum of all values ​​is 1.

[0062] For the control parameter vector corresponding to the historical state data, perform the above linear weighted processing to obtain the label corresponding to the new sample data.

[0063] Then, a certain amount of random noise disturbance (including 0-mean uniform distribution or 0-mean Gaussian distribution) is superimposed on each new sample data, and its validity is tested, and the sample data that passes the constraint verification is retained.

[0064] Afterwards, the improved genetic algorithm (NSGA-II) can be used to iteratively process the sample data that has passed the constraint verification to generate a first proportion of positive samples and a second proportion of negative samples. In order to increase the percentage of positive samples, a special sample (weight) modification algorithm can be designed. In addition, if necessary, probabilistic sampling technology can be used to ensure sample diversity and generate more positive samples.

[0065] Positive examples correspond to data from AC / DC power converters operating normally in the power system (e.g., stable power transmission and voltage fluctuations within acceptable limits). Negative examples correspond to data from abnormal or faulty power system conditions (e.g., voltage exceeding limits, power imbalance, etc.). By properly setting the ratio of positive to negative examples, the sample data comprehensively covers typical AC / DC power converter operating scenarios, providing a representative data foundation for subsequent training of operation and maintenance control models, thereby improving the model's ability to identify the operating status of AC / DC power converters and control accuracy.

[0066] In some embodiments, when attributes of a few or specific sample data do not satisfy the constraint conditions, targeted sample modification steps or algorithms may be designed to ensure that the sample ultimately satisfies the constraint conditions.

[0067] Step 204: Using a clustering algorithm, obtain a label for each group of sample data based on the control parameter vectors corresponding to all sample data.

[0068] A clustering algorithm is an unsupervised learning method whose core goal is to partition samples in a dataset into clusters (or groups) such that samples within the same cluster are highly similar, while samples from different clusters are less similar. In the embodiments provided herein, a clustering algorithm is used to automatically discover underlying patterns or categories in the data based on a control parameter vector of the sample data, thereby assigning labels to the sample data.

[0069] During the specific implementation process, the control parameter vectors corresponding to all sample data can be collected to form the input data set of the clustering algorithm. According to the characteristics and requirements of the state data, a suitable clustering algorithm is selected, such as K-Means, hierarchical clustering, DBSCAN, etc. In the control system of AC / DC power conversion devices, it is necessary to consider the real-time performance, scalability and robustness to noise data of the algorithm at the same time. The control parameter vector is input into the clustering algorithm, and the clustering process is executed to obtain K clusters. A unique label is assigned to each cluster to represent the common characteristics or categories of the sample data in the cluster. For example, cluster 1 can be labeled as "high power stable operation", cluster 2 can be labeled as "low power fluctuating operation", and so on.

[0070] The labels obtained from clustering are used as supervisory information for the sample data to construct a labeled training dataset. The training sample dataset is used to train the operation and maintenance control model, enabling it to learn control behaviors under different control strategies or operating effects represented by control parameter vectors.

[0071] In some embodiments, the number of sample clusters Can be based on the number of faults in the power system And the normal state number k of the power system is set, that is, , . (2) For example, the number of faults (fault_num): 5 (DC side overvoltage, AC side harmonic interference, converter module failure, load mutation, communication interruption); the number of normal states (k): 5 (steady-state operation, light-load operation, heavy-load operation, dynamic response, energy-saving mode); the number of clusters set: 5 + 5 = 10.

[0072] The specific clustering process is as follows: a) Use the nearest distance to divide the clusters and calculate the cluster centers (k-means algorithm), with 10-30 iterations.

[0073] b) Use mean-shift algorithm to change the center of the cluster, with 30-50 iterations and a mean-shift window of 5*5.

[0074] c) Re-divide the clusters and calculate the cluster centers.

[0075] d) When the change in cluster center is lower than the threshold or reaches the number of iterations (100 times), the clustering performance is evaluated based on density average or cluster shape (considering the maximum sample value, minimum sample value, average sample value and sample median), and the variance of density / number of nodes (density histogram) within all clusters can be minimized, and the geometric shape is optimal considering the symmetry of the cluster shape.

[0076] e) Gradually increase the number of clusters and repeat the above iterations. During the iteration process, discrete samples that are far from the cluster center can be clustered together into separate clusters (boundary clusters), and internal clusters can be clustered based on the nearest distance criterion to ensure the cluster density aggregation of internal clusters and the appropriate cluster sample distribution of internal clusters. However, in the training of the operation and maintenance control model, boundary clusters are not considered, and only the modeling of internal clusters is considered. The calculation formula for the formation of relevant boundary clusters is: (3) f) is the existing cluster center, is the maximum amplitude value of the sample, is the number of clusters.

[0077] In the embodiments provided by the present invention, Figure 4 As shown in the figure, by adaptively clustering the control parameter vectors corresponding to the sample data and using the clustering results as labels for the sample data, the relevant sample data is made more interpretable and is not affected by the instability of the sample data output space. The training samples obtained based on the above method can effectively improve the inference speed of the operation and maintenance control model, reduce the complexity of distributed operation equipment and improve its reliability.

[0078] Step 205: Obtain a training sample data set based on each set of sample data and its corresponding label.

[0079] In a specific implementation process, each set of sample data and its corresponding label can be used as a set of labeled sample data in a training sample data set.

[0080] In the embodiments provided by the present invention, by combining mathematical calculation control with system operation theory derivation, historical state data is processed based on the H2 / H∞ control method to generate high-quality original samples, ensuring the adequacy of positive data samples and providing a reliable foundation for subsequent modeling. A linear weighted algorithm and a genetic algorithm are used to generate new samples based on the original samples, effectively expanding the sample space and enhancing the generalization ability of the model. By introducing an appropriate amount of noise, the uncertainty of AC / DC power conversion devices in actual operation is simulated, improving the robustness of the operation and maintenance control model to noisy data and avoiding the occurrence of overfitting. This allows for the efficient and low-cost acquisition of abundant high-quality sample data.

[0081] In the embodiment provided by the present invention, the control parameter vectors corresponding to all sample data are clustered, and the labels of similar samples are classified into one category, which can effectively reduce data redundancy, reduce the computational complexity of the operation and maintenance control model during training and inference, and effectively improve the processing efficiency of the operation and maintenance control model for status data.

[0082] The control device of the AC / DC power conversion device based on clustering training provided by the present invention is described below. The control device of the AC / DC power conversion device based on clustering training described below and the control method of the AC / DC power conversion device based on clustering training described above can be referenced to each other.

[0083] Figure 6 Schematic diagram of the structure of the control device of the AC / DC power conversion device based on cluster training provided by the present invention. Figure 6 As shown, the device 600 includes the following modules.

[0084] The acquisition module 610 is used to obtain status data of the AC / DC power conversion device for a preset operation and maintenance event.

[0085] The generating module 620 is configured to process the state data using the operation and maintenance control model to generate a control parameter vector for the preset operation and maintenance event.

[0086] The control module 630 is configured to instruct the control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve the expected operation and maintenance result of the preset operation and maintenance event.

[0087] In which, the operation and maintenance management model is a machine learning model obtained by training based on a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data are historical status data of the AC / DC power conversion device for the preset operation and maintenance events; the labels of the sample data are generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

[0088] Figure 7 An example of a physical structure diagram of an electronic device is shown below. Figure 7As shown, the electronic device may include: a processor (processor) 710 , a communication interface (Communications Interface) 720 , a memory (memory) 730 and a communication bus 740 , wherein the processor 710 , the communication interface 720 and the memory 730 communicate with each other via the communication bus 740 . The processor 710 can call the logic instructions in the memory 730 to execute a control method for an AC / DC power conversion device based on clustering training, the method including: obtaining status data of the AC / DC power conversion device for a preset operation and maintenance event; processing the status data using an operation and maintenance management model to generate a control parameter vector for the preset operation and maintenance event; instructing the control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve the expected operation and maintenance result of the preset operation and maintenance event; wherein the operation and maintenance management model is a machine learning model obtained by training based on a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data is the historical status data of the AC / DC power conversion device for the preset operation and maintenance event; the label of the sample data is generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

[0089] Furthermore, the logic instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0090] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the control method of the AC / DC power conversion device based on clustering training provided by the above methods. The method includes: obtaining status data of the AC / DC power conversion device for a preset operation and maintenance event; processing the status data using an operation and maintenance management model to generate a control parameter vector for the preset operation and maintenance event; instructing the control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve the expected operation and maintenance result of the preset operation and maintenance event; wherein the operation and maintenance management model is a machine learning model obtained by training a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data is the historical status data of the AC / DC power conversion device for the preset operation and maintenance event; the label of the sample data is generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

[0091] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the control method of the AC / DC power conversion device based on clustering training provided by the above-mentioned methods, the method comprising: obtaining status data of the AC / DC power conversion device for a preset operation and maintenance event; processing the status data using an operation and maintenance management model to generate a control parameter vector for the preset operation and maintenance event; instructing the control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve the expected operation and maintenance result of the preset operation and maintenance event; wherein the operation and maintenance management model is a machine learning model obtained by training a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data is the historical status data of the AC / DC power conversion device for the preset operation and maintenance event; the label of the sample data is generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

[0092] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0093] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 various embodiments of the present invention.

Claims

1. A control method for an AC / DC power conversion device based on cluster training, characterized in that: include: Obtaining status data of the AC / DC power conversion device for preset operation and maintenance events; Processing the state data using an operation and maintenance control model to generate a control parameter vector for the preset operation and maintenance event; instructing a control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve an expected operation and maintenance result of the preset operation and maintenance event; In which, the operation and maintenance management model is a machine learning model obtained by training based on a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data are historical status data of the AC / DC power conversion device for the preset operation and maintenance events; the labels of the sample data are generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

2. The control method of an AC / DC power conversion device based on cluster training according to claim 1, characterized in that: The operation and maintenance management model is a batch input and batch output neural network model, the state data includes state data for multiple preset operation and maintenance events, and the control parameter vector includes control parameter vectors for the multiple preset operation and maintenance events; The step of processing the state data using the operation and maintenance control model to generate a control parameter vector for the preset operation and maintenance event includes: The status data for a plurality of preset operation and maintenance events are input into the batch input and batch output neural network model to obtain control parameter vectors for the plurality of preset operation and maintenance events output by the batch input and batch output neural network model.

3. The control method of an AC / DC power conversion device based on cluster training according to claim 1 or 2, characterized in that: The training sample dataset is obtained by: Acquiring multiple sets of historical status data of the AC / DC power conversion device for the preset operation and maintenance events; Using H2 and H∞ control algorithms, a control parameter vector corresponding to each set of the historical state data is obtained; wherein the control parameter vector is used to enable the AC / DC power conversion device to achieve the expected operation and maintenance result of the preset operation and maintenance event under the control of its control system; The training data set is obtained according to the multiple groups of historical state data and the control parameter vector corresponding to each group of historical state data.

4. The control method of an AC / DC power conversion device based on cluster training according to claim 3, characterized in that: The obtaining of the training data set according to the multiple groups of historical state data and the control parameter vector corresponding to each group of the historical state data includes: Using a linear weighted algorithm and a genetic algorithm, based on the multiple sets of historical state data and the control parameter vector corresponding to each set of the historical state data, multiple sets of sample data are obtained; wherein the multiple sets of sample data include positive samples in a first ratio and negative samples in a second ratio; The training sample data set is obtained according to the multiple groups of sample data and the control parameter vector corresponding to each group of sample data.

5. The control method of an AC / DC power conversion device based on cluster training according to claim 4, characterized in that: The step of obtaining the training sample data set according to the multiple groups of sample data and the control parameter vector corresponding to each group of sample data includes: Using a clustering algorithm, according to the control parameter vectors corresponding to all the sample data, a label of each group of the sample data is obtained; The training sample data set is obtained according to each group of the sample data and its corresponding label.

6. The control method of an AC / DC power conversion device based on cluster training according to claim 4, characterized in that: The method further comprises: Utilizing a reinforcement learning algorithm, based on state data accumulated during the operation of the AC / DC power conversion device, to optimize a control parameter vector to achieve an expected operation and maintenance result under the preset operation and maintenance event; Using the accumulated state data and the corresponding control parameter vectors, an expert knowledge base is established; The operation and maintenance control model is retrained using the data in the expert knowledge base to obtain an optimized operation and maintenance control model.

7. A control device for an AC / DC power conversion device based on cluster training, characterized in that: include: An acquisition module, configured to acquire status data of the AC / DC power conversion device for preset operation and maintenance events; A generation module, configured to process the state data using an operation and maintenance control model to generate a control parameter vector for the preset operation and maintenance event; a control module, configured to instruct a control system of the AC / DC power conversion device to control the AC / DC power conversion device according to the control parameter vector to achieve an expected operation and maintenance result of the preset operation and maintenance event; In which, the operation and maintenance management model is a machine learning model obtained by training based on a training sample data set; the training sample data set includes multiple groups of sample data, and the sample data are historical status data of the AC / DC power conversion device for the preset operation and maintenance events; the labels of the sample data are generated by clustering the control parameter vectors corresponding to each of the multiple groups of sample data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the control method of the AC / DC power conversion device based on cluster training as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the control method of the AC / DC power conversion device based on cluster training as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the control method of the AC / DC power conversion device based on cluster training as claimed in any one of claims 1 to 6 is implemented.

Images