Fault early warning regulation method and device based on energy storage converter and electronic equipment

Abstract

Embodiments of the present disclosure disclose a fault early warning regulation method and device based on energy storage converter and electronic equipment. A specific implementation of the method includes: performing spectrum graph data conversion processing on a set of converter-related physical information to obtain a set of converter signal spectrum graphs; using a spectrum graph feature extraction model to obtain a set of converter signal feature vectors; generating a set of converter fault deviation information according to the set of converter signal feature vectors and a set of operating state information; obtaining a set of early warning level information according to the set of operating state information; performing fault positioning identification on the set of converter fault deviation information to obtain a set of converter fault position information; and regulating the energy storage converter at different levels according to the set of converter fault position information and the set of early warning level information. The implementation can reduce the number of unplanned shutdowns of the power equipment in which the energy storage converter is located, reduce the damage degree of the energy storage converter, and improve the stability of the new energy station and the utilization rate of new energy.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and more specifically to a fault early warning and control method, apparatus, and electronic equipment based on energy storage converters. Background Technology

[0002] Energy storage converters are core components in renewable energy power plants, enabling bidirectional energy conversion between DC energy storage batteries and the AC power grid. The power devices inside these converters, such as IGBTs (Insulated Gate Bipolar Transistors) and capacitors, operate under high voltage and high current conditions for extended periods, making them prone to failure. For fault early warning and control based on energy storage converters, the common approach is as follows: First, sensors monitor the converter's operating data, including voltage, current, and temperature. Then, a supervised learning model trained on historical fault data and preset fixed thresholds are used to determine faults. Finally, an early warning message is issued based on the fault determination results.

[0003] However, in practice, it has been found that when using the above methods for fault early warning of energy storage converters, the following technical problems often exist: First, because the "normal" state of the energy storage converter changes dynamically with operating conditions such as load rate, SOC, and ambient temperature, the preset fixed threshold cannot be adaptively adjusted. False alarms are easily generated under harsh operating conditions such as heavy load and high temperature, while false alarms may be missed under stable operating conditions such as light load and low temperature. Second, the supervised learning model heavily relies on a large number of labeled fault samples for training, while fault samples are scarce and limited in type in actual industrial scenarios, resulting in insufficient generalization ability of the model and reduced recognition accuracy. This leads to low fault early warning accuracy of energy storage converters, making it difficult to effectively reduce the number of unplanned outages of power equipment where energy storage converters are located, and thus failing to improve the stability of new energy power plants and the utilization rate of new energy.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the present disclosure concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure propose a fault early warning and control method, apparatus, and electronic device based on energy storage converters to solve one or more of the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide a fault early warning and control method based on an energy storage converter, comprising: acquiring a set of converter-related physical information and a set of operating status information of the energy storage converter; performing spectrum diagram data conversion processing on the aforementioned set of converter-related physical information to obtain a set of converter signal spectrum diagrams; using a spectrum diagram feature extraction model to perform feature extraction processing on the aforementioned set of converter signal spectrum diagrams to obtain a set of converter signal feature vectors; generating a set of converter fault deviation information based on the aforementioned set of converter signal feature vectors and the aforementioned set of operating status information; determining the early warning level information of each converter fault deviation information in the aforementioned set of converter fault deviation information based on the aforementioned set of operating status information to obtain a set of early warning level information; performing fault location identification on the aforementioned set of converter fault deviation information to obtain a set of converter fault location information; and performing different levels of control on the aforementioned energy storage converter based on the aforementioned set of converter fault location information and the aforementioned set of early warning level information.

[0008] Secondly, some embodiments of this disclosure provide a fault early warning and control device based on an energy storage converter, comprising: an acquisition unit configured to acquire a set of converter-related physical information and an operating status information set of the energy storage converter; a conversion unit configured to perform spectrum data conversion processing on the aforementioned set of converter-related physical information to obtain a converter signal spectrum set; a feature extraction unit configured to perform feature extraction processing on the aforementioned set of converter signal spectrum using a spectrum feature extraction model to obtain a set of converter signal feature vectors; and a generation unit configured to generate a fault early warning and control device based on the aforementioned set of converter signal spectrum data. The converter signal feature vector set and the aforementioned operating status information set are used to generate a converter fault deviation information set; the determination unit is configured to determine the warning level information of each converter fault deviation information in the aforementioned converter fault deviation information set according to the aforementioned operating status information set, thereby obtaining a warning level information set; the identification unit is configured to perform fault location identification on the aforementioned converter fault deviation information set, thereby obtaining a converter fault location information set; and the control unit is configured to perform different levels of control on the aforementioned energy storage converter according to the aforementioned converter fault location information set and the aforementioned warning level information set.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, such that when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any implementation of the first aspect.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method as described in any implementation of the first aspect.

[0011] The above embodiments of this disclosure have the following beneficial effects: The fault early warning and control method based on energy storage converters in some embodiments of this disclosure can reduce the number of unplanned outages of power equipment where energy storage converters are located, and improve the stability and utilization rate of new energy power plants. Specifically, the reason why it is difficult to effectively reduce the frequency of unplanned outages of power equipment where energy storage converters are located, and thus cannot improve the stability and utilization rate of new energy power plants, is that: since the "normal" state of energy storage converters changes dynamically with operating conditions such as load rate, SOC, and ambient temperature, the preset fixed threshold cannot be adaptively adjusted, and false alarms are easily generated under harsh operating conditions such as heavy load and high temperature, while false alarms may be missed under stable operating conditions such as light load and low temperature; at the same time, the supervised learning model relies heavily on a large number of labeled fault samples for training, while fault samples are scarce and limited in type in actual industrial scenarios, resulting in insufficient generalization ability of the model and reduced recognition accuracy, resulting in low fault early warning accuracy of energy storage converters, making it difficult to effectively reduce the number of unplanned outages of power equipment where energy storage converters are located, and thus unable to improve the stability and utilization rate of new energy power plants. Based on this, some embodiments of the present disclosure of fault early warning and control based on energy storage converters can first acquire the converter-related physical information set and operating status information set of the energy storage converter. Here, the acquired converter-related physical information set provides synchronous multi-dimensional data for subsequent spectrum diagram data conversion; the acquired operating status information set serves as the data foundation for subsequently generating accurate converter deviation information and early warning level information. Secondly, the aforementioned converter-related physical information set is processed by spectrum diagram data conversion to obtain a converter signal spectrum diagram set. Here, the time-domain signal is converted into a frequency-domain representation. The aforementioned converter signal spectrum diagram can significantly amplify the weak frequency characteristic changes caused by potential faults, while suppressing random noise interference in the time-domain signal, providing higher quality and more sensitive data for subsequent feature extraction. Thirdly, the aforementioned converter signal spectrum diagram set is processed by a spectrum diagram feature extraction model to obtain a converter signal feature vector set. Here, a self-supervised learning spectrum diagram feature extraction model is used to extract redundant feature representations from the converter signal spectrum diagram, overcoming the dependence on prior fault knowledge and improving the generalization ability for unknown fault types. Next, based on the aforementioned converter signal feature vector set and operating status information set, a converter fault deviation information set is generated. Here, converter fault deviation information corresponding to the operating status information is generated for the converter signal feature vectors, reducing the possibility of inaccurate generated converter fault deviation information due to different operating states of the energy storage converter. Subsequently, based on the aforementioned operating status information set, the warning level information for each converter fault deviation information in the aforementioned converter fault deviation information set is determined, resulting in a warning level information set. Here, adaptive warning thresholds are set for converter fault deviation information under different operating states, improving the accuracy of warnings.Then, fault location identification is performed on the aforementioned converter fault deviation information set to obtain the converter fault location information set. Here, the obtained converter fault location information can pinpoint the source of the anomaly to a specific component or circuit module, shortening fault investigation time and improving the accuracy and precision of fault warnings for energy storage converters. Finally, based on the aforementioned converter fault location information set and the aforementioned warning level information set, different levels of regulation are applied to the aforementioned energy storage converter. Here, by comprehensively considering the converter fault location information set and the warning level information set, different levels of regulation are triggered on the aforementioned energy storage converter, reducing the number of shutdowns caused by sudden failures of the energy storage converter, thereby reducing the number of shutdowns of the power equipment where the energy storage converter is located. Therefore, this fault warning and regulation method based on energy storage converters can reduce the number of unplanned outages of the power equipment where the energy storage converter is located, improving the stability of new energy power plants and the utilization rate of new energy. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of the fault early warning and control method based on energy storage converter according to the present disclosure; Figure 2 This is a schematic diagram of the structure of some embodiments of the fault early warning and control device based on the energy storage converter according to the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0016] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0017] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0018] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0019] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Figure 1 A flowchart 100 is shown, illustrating some embodiments of the fault early warning and control method based on an energy storage converter according to this disclosure. The fault early warning and control method based on an energy storage converter includes the following steps: Step 101: Obtain the converter-related physical information set and operating status information set of the energy storage converter.

[0021] In some embodiments, the executing entity (e.g., an electronic device) of the above-mentioned new energy power station energy storage converter fault early warning method can acquire the converter-related physical information set and operating status information set of the energy storage converter through wired or wireless connection. The converter-related physical information set can be physical quantity data collected by sensors deployed on the energy storage converter components. The converter-related physical information set may include, but is not limited to, at least one of the following: electrical parameters of the AC side of the energy storage converter, electrical parameters of the DC side of the energy storage converter, and IGBT module temperature signals. The operating status information set can be data and parameters of the operating status of the energy storage converter during operation. The operating status information set may include, but is not limited to, at least one of the following: SOC (State of Charge) and load factor. In practice, the executing entity can first, through deployed high-frequency voltage sensors, current sensors, and temperature sensors, simultaneously collect the electrical parameters of the AC and DC sides of the energy storage converter and the IGBT module temperature signals within a preset time window. Then, the aforementioned electrical parameters and temperature signals are determined as the relevant physical information of the converter. The aforementioned preset time window can be a pre-set time period for collecting the aforementioned physical information of the converter. At the same time, the corresponding operating status information within the aforementioned preset time window is read through the communication interface of the controller of the aforementioned energy storage converter or the site monitoring system.

[0022] Step 102: Perform spectrum data conversion processing on the converter-related physical information set to obtain the converter signal spectrum set.

[0023] In some embodiments, the execution entity may perform spectrum data conversion processing on the aforementioned converter-related physical information set to obtain a converter signal spectrum set. The converter signal spectrum in the aforementioned converter signal spectrum set may be a two-dimensional data representation of the energy distribution characteristics of the aforementioned converter-related physical information in the frequency domain.

[0024] In some optional implementations of certain embodiments, the above-mentioned spectrum data conversion processing of the converter-related physical information set to obtain the converter signal spectrum set may include the following steps: The first step is to preprocess the aforementioned converter-related physical information set to obtain a preprocessed converter-related physical information set. In practice, the aforementioned execution entity can use an FIR (Finite Impulse Response) filter to preprocess the aforementioned converter-related physical information set to obtain a preprocessed converter-related physical information set.

[0025] The second step involves performing frequency domain transformation on the preprocessed converter-related physical information set to obtain the converter signal spectrum atlas. In practice, the execution entity can use the FFT (Fast Fourier Transform) algorithm to perform frequency domain transformation on the preprocessed converter-related physical information set to obtain the converter signal spectrum atlas.

[0026] Step 103: Using the spectrum feature extraction model, perform feature extraction processing on the converter signal spectrum set to obtain the converter signal feature vector set.

[0027] In some embodiments, the aforementioned execution entity can utilize a spectrogram feature extraction model to perform feature extraction processing on the aforementioned converter signal spectrogram set, obtaining a converter signal feature vector set. The converter signal feature vectors in the aforementioned converter signal feature vector set can characterize the redundancy-free numerical feature representation of the aforementioned converter signal spectrogram. The aforementioned spectrogram feature extraction model can be a model that takes the aforementioned converter signal spectrogram as input, performs redundancy-free feature extraction on the aforementioned converter signal spectrogram, and outputs converter signal feature vectors. For example, the aforementioned spectrogram feature extraction model can be a ResNet-18 (Residual Network-18) with the classification layer removed.

[0028] In some optional implementations of certain embodiments, the above-described spectrogram feature extraction model is obtained through the following steps: The first step is to obtain a set of historical physical information related to the energy storage converter. This set of historical physical information related to the energy storage converter is a dataset that does not include fault information. The historical physical information related to the energy storage converter in this set can be converter-related physical information obtained before the current time. As an example, the implementation of the above step can be referred to in step 101, and will not be repeated here.

[0029] The second step involves performing frequency domain transformation on the historical physical information set of the energy storage converter to obtain a historical converter signal spectrum map. The historical converter signal spectrum map in this set can be a two-dimensional data representation of the energy distribution characteristics of the historical physical information of the energy storage converter in the frequency domain. As an example, the implementation of the above step can be referenced from the implementation of step 102, and will not be repeated here.

[0030] The third step involves data augmentation of the aforementioned historical converter signal spectrum dataset to obtain a first sample spectrum dataset and a second sample spectrum dataset. The data augmentation methods may include, but are not limited to, at least one of the following: random cropping, noise injection, and frequency masking. The first sample spectrum dataset and the second sample spectrum dataset may be obtained by applying different data augmentations to the same historical converter signal spectrum dataset. For example, the first sample spectrum dataset may be obtained by injecting noise into the historical converter signal spectrum dataset, and the corresponding second sample spectrum dataset may be obtained by applying frequency masking to the same historical converter signal spectrum dataset.

[0031] Fourth, based on the first and second sample spectrogram sets, perform the following training steps: Sub-step 1 involves inputting the first sample spectrogram set and the second sample spectrogram set into the initial spectrogram feature extraction model to obtain a first spectrogram feature vector set and a second spectrogram feature vector set. The first spectrogram feature vector in the first spectrogram feature vector set represents the numerical feature representation of the first sample spectrogram. The second spectrogram feature vector in the second spectrogram feature vector set represents the numerical feature representation of the second sample spectrogram. The initial spectrogram feature extraction model can be a model composed of a spectrogram encoder layer and a spectrogram projection layer connected in series. The spectrogram encoder layer can be a ResNet-18 that removes the classification layer. The spectrogram projection layer can be a three-layer MLP (Multi-Layer Perceptron).

[0032] Sub-step 2 involves determining the cross-correlation matrix of the first and second spectral feature vector sets. This cross-correlation matrix can be a matrix describing the relationship between the first and second spectral feature vector sets. For example, it can be the cross-correlation matrix of the first and second spectral feature vector sets.

[0033] Sub-step 3 involves inputting the feature cross-correlation matrix into the spectrogram feature extraction loss function to obtain the feature loss function value. This spectrogram feature extraction loss function can be the loss function of the Barlow Twins algorithm. The feature loss function value characterizes the quality of the feature representation learned by the initial spectrogram feature extraction model.

[0034] Sub-step 4: In response to determining that the feature loss function value meets the preset loss value condition, the initial spectrogram feature extraction model is identified as the trained spectrogram feature extraction model. The preset loss value condition can be that the relative rate of change of the largest loss function value within a consecutive preset training period is less than or equal to a preset rate of change threshold. The preset training period can be 15 training periods. The relative rate of change of the loss function value can be the ratio of the difference between the feature loss function value of the previous training period and the feature loss function value of the current training period to the feature loss function value of the previous training period. The preset rate of change threshold can be a pre-set threshold used to determine whether the above training step should be executed again. For example, the preset rate of change threshold can be 0.001. In practice, the executing entity can identify the spectrogram encoder layer of the initial spectrogram feature extraction model as the trained spectrogram feature extraction model.

[0035] Fifth, in response to the determination that the feature loss function value does not meet the aforementioned preset loss value condition, the aforementioned historical converter signal spectrum dataset is re-processed with data augmentation to obtain a first target sample spectrum dataset and a second target sample spectrum dataset. These are used as the first and second sample spectrum datasets. Additionally, the relevant parameters in the initial spectrum feature extraction model are adjusted to obtain an adjusted spectrum feature extraction model, which is then used as the initial spectrum feature extraction model to repeat the aforementioned training steps. In practice, the aforementioned execution entity can employ the gradient backpropagation algorithm to adjust the parameters of the initial spectrum feature extraction model based on the aforementioned feature loss function value.

[0036] Step 104: Generate a converter fault deviation information set based on the converter signal feature vector set and the operating status information set.

[0037] In some embodiments, the execution entity may generate a converter fault deviation information set based on the converter signal feature vector set and the operating status information set. The converter fault deviation information in the converter fault deviation information set can characterize the degree of abnormality of the converter signal feature vectors.

[0038] In some optional implementations of certain embodiments, generating the converter fault deviation information set based on the converter signal feature vector set and the operating status information set may include the following steps: The first step involves inputting the aforementioned converter signal feature vector set into the converter fault coding layer of the trained converter fault identification model to obtain a fault coding feature vector set. This converter fault identification model further includes a converter fault bottleneck layer and a converter fault conditional decoding layer. The converter fault identification model can take the aforementioned converter signal feature vector set and the aforementioned operating state information set as inputs, reconstruct the features of the converter signal feature vector set based on the operating state information set, and output a converter fault deviation information set. For example, the converter fault identification model can be a conditional autoencoder model that stacks three ELM-AE (Extreme Learning Machine Auto-Encoder). The converter fault coding layer can take the aforementioned converter signal feature vector set as input, perform random projection and nonlinear transformation on the converter signal feature vector set, and output a fault coding feature vector set. The fault coding feature vectors in the fault coding feature vector set can be converter signal feature vectors after removing unimportant details and noise information.

[0039] The second step involves inputting the aforementioned fault-coded feature vector set into the converter fault bottleneck layer to obtain a fault-compressed feature vector set. The converter fault bottleneck layer can be a model that takes the aforementioned fault-coded feature vector set as input, compresses the information of the fault-coded feature vectors, and outputs the aforementioned fault-compressed feature vector set. The fault-compressed feature vectors in the aforementioned fault-compressed feature vector set can be lower-dimensional and more compact feature representations compared to the aforementioned fault-coded feature vectors.

[0040] The third step involves encoding the aforementioned set of operating status information to obtain a set of operating status vectors. These operating status vectors can be numerical representations of the operating status information, standardized to the same numerical scale. For example, the operating status vectors could be [0.7, 0.9]. Here, 0.7 in the operating status vector could represent a battery remaining charge of 70%, and 0.9 could represent the converter's current load rate of 90%. In practice, the executing entity can use normalization processing to transform the aforementioned operating status information into operating status vectors.

[0041] The fourth step is to concatenate the above-mentioned operating status vector set and the above-mentioned fault compressed feature vector set to obtain the concatenated feature vector set.

[0042] The fifth step involves inputting the aforementioned spliced ​​feature vector set into the converter fault condition decoding layer to obtain a fault reconstruction feature vector set. The converter fault condition decoding layer can be a model that takes the spliced ​​feature vector set as input, reconstructs the fault compression feature vectors included in the spliced ​​feature vectors based on the operating state vectors, and outputs the fault reconstruction feature vector set. The fault reconstruction feature vectors in this set can be feature representations of the converter-related physical information of the energy storage converter in the absence of potential fault features.

[0043] The sixth step involves generating a converter fault deviation information set based on the aforementioned fault reconstruction feature vector set and the aforementioned converter signal feature vector set. In practice, the executing entity can determine the mean square error of each fault reconstruction feature vector in the aforementioned fault reconstruction feature vector set and the mean square error of the converter signal feature vector corresponding to the fault reconstruction feature vector in the aforementioned converter signal feature vector set, using these as converter fault deviation information to obtain the converter fault deviation information set.

[0044] Step 105: Based on the operating status information set, determine the warning level information for each converter fault deviation information in the converter fault deviation information set to obtain the warning level information set.

[0045] In some embodiments, the aforementioned executing entity can determine the warning level information of each converter fault deviation information in the aforementioned converter fault deviation information set based on the aforementioned operating status information set, thereby obtaining a warning level information set. The warning level information in the aforementioned warning level information set can characterize the potential fault risk level of the aforementioned energy storage converter within a corresponding time window. The aforementioned warning level information may include, but is not limited to, at least one of the following: converter fault deviation information and warning level identifiers. The aforementioned warning level identifiers can be identifiers indicating whether the aforementioned energy storage converter has a potential fault within the corresponding time window, and the different risk levels of any existing potential faults. For example, the aforementioned warning level identifiers may include: a normal identifier, a level 3 warning identifier, a level 2 warning identifier, and a level 1 warning identifier. The normal identifier can characterize that the aforementioned energy storage converter is operating well and has no potential faults within the corresponding time window. The level 3 warning identifier can characterize that the aforementioned energy storage converter has a slight anomaly in its operating status within the corresponding time window. The level 2 warning identifier can characterize that the aforementioned energy storage converter has a clear and significant anomaly in its operating status within the corresponding time window. The level 1 warning identifier can characterize that the aforementioned energy storage converter has a serious anomaly in its operating status within the corresponding time window.

[0046] In some optional implementations of certain embodiments, determining the warning level information of each converter fault deviation information in the converter fault deviation information set based on the above-mentioned operating status information set to obtain the warning level information set may include the following steps: The first step, for each converter fault deviation information in the above converter fault deviation information set, is to perform the following determination steps: Sub-step 1: Determine the operating status information corresponding to the above converter fault deviation information as the target operating status information.

[0047] Sub-step 2 involves inputting the aforementioned target operating status information into the trained converter fault dynamic threshold model to obtain a first warning threshold and a second warning threshold. The first warning threshold can be a threshold value used to preliminarily determine whether the warning level indicator corresponding to the converter fault deviation information reaches the level three warning indicator under the aforementioned target operating status information. The second warning threshold can be a threshold value used to preliminarily determine whether the warning level indicator corresponding to the converter fault deviation information reaches the level two warning indicator under the aforementioned target operating status information. The converter fault dynamic threshold model can take operating status information as input, perform nonlinear mapping and quantile regression prediction on the operating status information, and output the first and second warning thresholds. For example, the aforementioned converter fault dynamic threshold model can be a LightGBM (Light Gradient Boosting Machine) model.

[0048] Sub-step 3 involves inputting the aforementioned target operating state information into the trained converter prediction distribution model to obtain the target mean and target standard deviation. The target mean characterizes the converter fault deviation information corresponding to a normal energy storage converter under the target operating state information. The target standard deviation characterizes the confidence level of the target mean. The converter prediction distribution model can be a model that takes operating state information as input, performs probabilistic prediction and uncertainty quantification on the operating state information, and outputs the target mean and target standard deviation. For example, the converter prediction distribution model could be a GPR (Gaussian Process Regression) model.

[0049] Sub-step 4: Generate a target conservative threshold based on the aforementioned target mean and target standard deviation. This target conservative threshold can be determined by whether the warning level indicator corresponding to the converter fault deviation information reaches the maximum critical value of the first warning indicator, even when the converter fault dynamic threshold model has errors. In practice, the executing entity can add three times the aforementioned target standard deviation to the aforementioned target mean to determine the target conservative threshold.

[0050] Sub-step 5: The first warning threshold, the second warning threshold, and the target conservative threshold are determined as the fault threshold group for the converter fault deviation information.

[0051] Sub-step 6: Based on the converter signal feature vector corresponding to the aforementioned converter fault deviation information and the preset mean value of the converter signal feature vector, a converter signal feature vector deviation value is generated. The preset mean value of the converter signal feature vector can be a pre-defined mean vector representing a set of historically normal converter signal feature vectors. This set of historically normal converter signal feature vectors can be at least one converter signal feature vector located before the current time. The converter signal feature vector deviation value characterizes the overall deviation between the aforementioned converter signal feature vector and the preset mean value. In practice, the executing entity can determine the converter signal feature vector deviation value as the Mahalanobis distance between the aforementioned converter signal feature vector and the preset mean value of the converter signal feature vector.

[0052] Sub-step 7: Based on the aforementioned converter signal characteristic vector deviation value and the preset converter signal characteristic vector fault threshold, generate a converter signal fault identifier. The converter signal fault identifier can be a binary indicator indicating whether the operating mode of the energy storage converter has changed significantly. The converter signal fault identifier can be a normal converter signal identifier or an abnormal converter signal identifier. The preset converter signal characteristic vector fault threshold can be a pre-set critical value used to determine whether the converter signal characteristic vector corresponding to the aforementioned converter signal characteristic vector deviation value has deviated significantly. In practice, the executing entity can generate a converter signal abnormal identifier as a converter signal fault identifier in response to determining that the aforementioned converter signal characteristic vector deviation value is greater than or equal to the preset converter signal characteristic vector fault threshold, and generate a converter signal normal identifier as a converter signal fault identifier in response to determining that the aforementioned converter signal characteristic vector deviation value is less than the preset converter signal characteristic vector fault threshold.

[0053] Sub-step 8: Based on the aforementioned converter fault deviation information, the aforementioned fault threshold group, and the aforementioned converter signal fault identifier, determine the warning level information of the aforementioned converter fault deviation information. In practice, the executing entity may first, in response to determining that the aforementioned converter fault deviation information is less than the first warning threshold included in the aforementioned fault threshold group and that the aforementioned converter signal fault identifier is a converter signal normal identifier, determine the warning level identifier of the aforementioned converter deviation information as a normal identifier, as the target warning level information. Then, in response to determining that the aforementioned converter fault deviation information is greater than or equal to the first warning threshold included in the aforementioned fault threshold group and less than the second warning threshold included in the aforementioned fault threshold group, or that the aforementioned converter signal fault identifier is a converter signal abnormal identifier, determine the warning level identifier of the aforementioned converter deviation information as a level three warning identifier, as the target warning level information. Next, in response to determining that the aforementioned converter fault deviation information is greater than or equal to the second warning threshold included in the aforementioned fault threshold group and less than or equal to the target conservative threshold included in the aforementioned fault threshold group, and simultaneously that the aforementioned converter signal fault identifier is a converter signal abnormality identifier, the warning level identifier of the aforementioned converter deviation information is determined to be a level two warning identifier, as the target warning level information. Then, in response to determining that the aforementioned converter fault deviation information is greater than or equal to the aforementioned target conservative threshold, and that the aforementioned converter signal fault identifier is a converter signal abnormality identifier, the warning level identifier of the aforementioned converter deviation information is determined to be a level one warning identifier, as the target warning level information. Finally, the converter fault deviation information corresponding to the aforementioned target warning level identifier and the aforementioned target warning level information is determined to be the warning level information.

[0054] Step 106: Perform fault location identification on the converter fault deviation information set to obtain the converter fault location information set.

[0055] In some embodiments, the aforementioned execution entity can perform fault location identification on the aforementioned converter fault deviation information set to obtain a converter fault location information set. The converter fault location information in the aforementioned converter fault location information set may be the location information of components in the aforementioned energy storage converter that have potential faults.

[0056] In addressing the technical problems mentioned above, the application scenario—energy storage converters in new energy power plants located in remote areas—often presents the following technical problem: Due to limited maintenance resources in remote areas (e.g., few maintenance personnel, difficulty in equipment allocation, and long travel times to and from the power plant), existing methods for fault location in energy storage converters rely solely on a single source of sensor data. This leads to inaccurate location information, resulting in prolonged troubleshooting and confirmation time for maintenance personnel after an early warning is issued. Consequently, unplanned outages at new energy power plants increase, leading to increased power generation losses. Furthermore, inaccurate location information prevents maintenance personnel from repairing the actual fault point, allowing potential faults in the energy storage converter to worsen during subsequent operation, ultimately causing damage and increasing the converter failure rate. To address the following characteristics required for this application scenario: accurate location results, the ability to integrate multi-source information for comprehensive decision-making, and a certain degree of generalization capability for unknown faults, we have decided to adopt the following solution.

[0057] In some optional implementations of certain embodiments, the above-mentioned fault location identification of the converter fault deviation information set to obtain the converter fault location information set may include the following steps: The first step is to filter the aforementioned warning level information set to obtain a filtered warning level information set, which serves as the warning information set to be located. In practice, the implementing entity can filter at least one warning level information whose corresponding warning level identifier is not a normal identifier from the aforementioned warning level information set, thus obtaining a filtered warning level information set, which serves as the warning information set to be located.

[0058] The second step is to perform the following generation steps for each of the aforementioned early warning messages in the set of early warning messages to be located: Sub-step 1: The converter signal feature vector and converter fault deviation information corresponding to the above-mentioned early warning information to be located are determined as the target converter signal feature vector and target converter fault deviation information.

[0059] Sub-step 2: Based on the fault reconstruction feature vector corresponding to the fault deviation information of the target converter and the signal feature vector of the target converter, a sensor contribution set is generated. The sensor contribution in this set characterizes the degree of correlation between the physical quantity data collected by the sensors and the anomaly of the energy storage converter. In practice, the executing entity can first determine the difference between the signal feature vector of the target converter and the fault reconstruction feature vector as the reconstruction error vector. Then, according to a preset sensor feature dimension mapping table, the reconstruction error vectors are grouped to obtain a grouped reconstruction error vector set. Next, the sum of the squares of each element in each grouped reconstruction error vector in the grouped set is determined as the sensor dimension error value set. The sensor dimension error values ​​in this set characterize the degree of anomaly of the physical information detected by the sensors. Then, the sum of each sensor dimension error value in the set is determined as the total sensor dimension error value. Finally, the ratio of each sensor dimension error value in the set to the total sensor dimension error value is determined as the sensor contribution set.

[0060] Sub-step 3 involves performing sensor location processing on the sensors corresponding to the aforementioned sensor contribution set to obtain preliminary location information. This preliminary location information represents a preliminary judgment result indicating the presence of a potential abnormal location in the energy storage converter. In practice, the executing entity can first filter at least one sensor contribution value from the aforementioned sensor contribution set that is greater than or equal to a preset sensor contribution value threshold, as the target sensor contribution set. This preset sensor contribution value threshold can be a pre-set critical value used to determine whether a sensor contribution value is a target sensor contribution value. For example, the preset sensor contribution value threshold could be 0.1. Then, the sensor location information corresponding to each target sensor contribution value in the target sensor contribution set is obtained, as the sensor location information set. This sensor location information can be the installation location information of the sensor in the energy storage converter. Finally, the sensor location information set and the target sensor contribution set are used to determine the preliminary location information.

[0061] Sub-step 4 involves performing similarity matching processing on the target converter signal feature vector based on a preset fault feature database to obtain database location information. The preset fault feature database can be a pre-defined database storing processed historical fault types and their corresponding converter signal feature vectors. The database location information can be fault type information corresponding to the target converter signal feature vector. For example, the fault type information could be IGBT aging. In practice, the executing entity can first determine the cosine similarity between the target converter signal feature vector and the converter signal feature vector corresponding to each historical fault type, using this as a converter signal similarity value set. Then, it can select the converter signal similarity value with the largest value from the set as the target converter signal similarity value. Finally, in response to determining that the target converter signal similarity value is greater than or equal to a preset database location threshold, the target converter signal similarity value and the corresponding historical fault type are determined as database location information. The aforementioned preset database location threshold can be a pre-set threshold used to determine whether to adopt the historical fault type corresponding to the target converter signal similarity value as the database location information.

[0062] Sub-step 5 involves adding the sensor operating status information set corresponding to the aforementioned early warning information to the preset physical topology map, along with the aforementioned preliminary positioning information and the aforementioned database positioning information, to obtain the target physical topology map. The sensor operating status information in the aforementioned sensor operating status information set can be the real-time operating parameters of the aforementioned sensors. For example, the aforementioned sensor operating status information can include, but is not limited to, at least one of the following: sensor runtime and sensor sampling frequency. The aforementioned preset physical topology map can be a pre-defined weighted graph representing the connection relationships of each component of the aforementioned energy storage converter using nodes and edges. The node types of the aforementioned preset physical topology map can include, but are not limited to, at least one of the following: sensor nodes, component nodes, and fault type nodes. The aforementioned sensor nodes can be nodes composed of sensors deployed on the aforementioned energy storage converter components. The aforementioned component nodes can be nodes composed of components included in the aforementioned energy storage converter. The aforementioned fault type nodes can be nodes composed of historical fault types. The edge weights included in the aforementioned preset physical topology map can be scalar values ​​determined based on the physical connection strength, signal propagation path, or historical fault correlation of each part of the aforementioned energy storage converter. The scalar values ​​of the aforementioned edge weights can be manually preset values. For example, the aforementioned fault type node could be an IGBT aging node, the aforementioned component node could be an IGBT module, and the aforementioned sensor node could be the IGBT module temperature sensor. Therefore, there would be a weighted edge between the IGBT aging fault type node and the IGBT module component node. There would also be a weighted edge between the IGBT aging fault type node and the IGBT module temperature sensor sensor node. The aforementioned target physical topology map could be a preset physical topology map with added node information but unchanged topology structure. In practice, the aforementioned execution entity can first numerically encode the aforementioned sensor operating status information set to obtain a set of sensor additional feature vectors. The sensor additional feature vectors in the aforementioned sensor additional feature vector set can characterize the node status information of the aforementioned sensor nodes. Then, the sensor additional feature vector set and the target sensor contribution set included in the preliminary positioning information are added to the features of the sensor nodes corresponding to the aforementioned preset physical topology map, and the target converter signal similarity value included in the database positioning information is added to the features of the fault type node corresponding to the aforementioned preset physical topology map, resulting in a preset physical topology map after the addition, which serves as the target physical topology map.

[0063] Sub-step 6 involves inputting the target physical topology map into the node information aggregation layer of the trained localization information aggregation model to obtain a set of node aggregated feature vectors and a set of edge feature vectors. The localization information aggregation model further includes an attention weighting layer, a localization information classification layer, and an output layer. The node aggregated feature vectors in the node aggregated feature vector set can be node representations obtained by aggregating the feature information of each node itself and its neighbors. The edge feature vectors in the edge feature vector set can characterize the strength and type of connections between nodes. The localization information aggregation model can be a model that takes the target physical topology map as input, performs hierarchical feature learning and information fusion on the target physical topology map, and outputs a set of node failure probabilities. The node information aggregation layer can be a model that takes the target physical topology map as input, performs adjacency message aggregation and feature update processing on each node in the target physical topology map through graph convolution operations, and obtains a set of node aggregated feature vectors and a set of edge feature vectors. For example, the node information aggregation layer can be a GCN (Graph Convolutional Network) model.

[0064] Sub-step 7 involves inputting the aforementioned node aggregated feature vector set and edge feature vector set into the aforementioned attention weighting layer to obtain a node weighted feature vector set. The node weighted feature vectors in the node weighted feature vector set can be the node feature representations after the importance weights of the aforementioned node aggregated feature vectors have been adjusted through an attention mechanism. The aforementioned attention weighting layer can be a model that takes the node aggregated feature vector set and edge feature vector set as input, calculates the attention scores between nodes, and performs a weighted summation of the feature vectors of each node's neighboring nodes to obtain the node weighted feature vector set. For example, the aforementioned attention weighting layer can be a GAT (Graph Attention Network) model.

[0065] Sub-step 8 involves inputting the aforementioned node weighted feature vector set into the location information classification layer to obtain the node failure probability set. The node failure probabilities in this set can be predicted probabilities of component failures within the energy storage converter. The location information classification layer can be a model that takes the node weighted feature vector set as input, performs linear transformation and nonlinear activation on the set, then normalizes it, and outputs the node failure probability set. This location information classification layer can include a classifier with a Softmax function. For example, it can be a model composed of an MLP (Multi-Layer Perceptron) and a Softmax activation function.

[0066] Sub-step 9 involves filtering out converter component information corresponding to node fault probabilities that meet preset filtering conditions from the aforementioned node fault probability set, using this information as converter fault location information. The preset filtering conditions can be conditions where the node fault probability is greater than a preset component node probability threshold. This preset component node probability threshold can be a pre-set critical value used to determine whether the converter component information corresponding to the aforementioned node fault probabilities is converter fault location information. For example, the preset component node probability threshold could be 0.7.

[0067] The second step involves adjusting the energy storage converter at different levels based on the obtained converter fault location information set and the aforementioned early warning level information set. As an example, the implementation of this step can refer to the implementation method of step 107, and will not be repeated here.

[0068] The above-mentioned technical solution and related content, as an inventive point of this disclosure, solves the second technical problem, "leading to increased power generation loss and increased converter failure rate." Factors leading to increased power generation loss and increased converter failure rate are often as follows: Due to limited maintenance resources in remote areas, existing methods for fault location of energy storage converters rely solely on a single information source of sensor data for fault analysis and location, resulting in inaccurate location information. Consequently, after issuing an early warning, inaccurate location information prolongs the time for maintenance personnel to troubleshoot and confirm faults, leading to increased unplanned outages at new energy power plants and increased power generation loss. Simultaneously, inaccurate location information prevents maintenance personnel from repairing the actual fault point, allowing potential faults in the energy storage converter to worsen during subsequent operation, causing damage to the energy storage converter and increasing the converter failure rate. Solving these factors can reduce unplanned outages at new energy power plants and lower power generation loss and converter failure rate. To achieve this effect, this disclosure first filters the aforementioned early warning level information set to obtain a filtered early warning level information set as the early warning information set to be located. Here, warning level information representing normal conditions is removed from the warning level information set to avoid wasting computational resources on invalid data. Then, for each warning information to be located in the above warning information set, the following generation steps are performed: First, the converter signal feature vector and converter fault deviation information corresponding to the above warning information to be located are determined as the target converter signal feature vector and target converter fault deviation information. Second, a sensor contribution set is generated based on the fault reconstruction feature vector and target converter signal feature vector corresponding to the above target converter fault deviation information. Here, the sensor contribution transforms scattered sensor data into quantitative information on the degree of correlation with the anomaly, providing basic data for subsequent sensor location processing. Third, sensor location processing is performed on the sensors corresponding to the above sensor contribution set to obtain preliminary location information. Here, the sensor contribution is mapped to the specific physical location of the sensor, providing preliminary fault location reference and input data for the subsequent construction of the target physical topology map. Fourth, similarity matching processing is performed on the above target converter signal feature vector according to the preset fault feature database to obtain database location information. Here, the database location information is obtained based on historical experience and fault modes. This allows for rapid identification of known fault types, preventing inaccurate positioning due to the limited source of initial location information. It also provides input data for the subsequent construction of the target physical topology map. The fifth step involves adding the sensor operating status information set corresponding to the aforementioned early warning information, the aforementioned preliminary location information, and the aforementioned database location information to the preset physical topology map to obtain the target physical topology map.Here, the target physical topology map associates numerical information such as sensor operating status information, preliminary positioning information, and database positioning information with specific physical connections, providing a structured input containing data features and physical constraints for subsequent positioning steps. The sixth step inputs the target physical topology map to the node information aggregation layer of the trained positioning information aggregation model, obtaining a set of node aggregated feature vectors and an edge feature vector set. This positioning information aggregation model also includes an attention weighting layer, a positioning information classification layer, and an output layer. Here, the node information aggregation layer uses graph convolution operations to aggregate the feature information of each node and its neighbors, capturing the deep correlation between "sensor signal - component - fault type," avoiding positioning errors caused by fragmented analysis of preliminary positioning information and database positioning information. The seventh step inputs the node aggregated feature vector set and the edge feature vector set to the attention weighting layer, obtaining a set of node weighted feature vectors. Here, the attention weighting layer performs weighted summation on the node aggregated feature vectors, allowing the model to automatically focus on key information and reduce interference from non-critical information in the positioning results. Step 8: Input the aforementioned node weighted feature vector set into the location information classification layer to obtain the node fault probability set. Here, the node weighted feature vector set is mapped to the node fault probability set. The node fault probability quantifies the confidence level of the output conclusion of the aforementioned location information aggregation model, refining the location results to specific components of the energy storage converter, improving the accuracy of the location information, and providing basic data for subsequent screening. Step 9: Filter the converter component information corresponding to the node fault probabilities that meet the preset screening conditions from the aforementioned node fault probability set, as the converter fault location information. Here, the output converter fault location information is the location of the faulty component with the highest confidence. Accurate fault location information can improve maintenance efficiency, reduce downtime, and thus reduce power generation loss. It can also reduce the occurrence of incorrect or missed repairs during maintenance, lowering the converter damage rate.

[0069] Step 107: Based on the converter fault location information set and the early warning level information set, perform different levels of regulation on the energy storage converter.

[0070] In some embodiments, the aforementioned executing entity can perform different levels of regulation on the energy storage converter based on the aforementioned converter fault location information set and the aforementioned warning level information set. These different levels of regulation may include, but are not limited to, at least one of the following: monitoring regulation operation, maintenance regulation operation, and emergency regulation operation. The monitoring regulation operation may include, but is not limited to, at least one of the following: reducing the active power output or input setpoint of the energy storage converter, or reducing the switching frequency of the IGBTs and diodes. The maintenance regulation operation may include, but is not limited to, at least one of the following: switching the operating mode of the energy storage converter to a restricted state (e.g., prohibiting charging and allowing only discharging), disconnecting the faulty auxiliary component circuit and activating the backup circuit, or notifying maintenance personnel to perform maintenance on the energy storage converter. The emergency regulation operation may include, but is not limited to, at least one of the following: performing a soft shutdown on the energy storage converter, blocking remote and local operation commands for the energy storage converter, or notifying maintenance personnel to perform maintenance on the energy storage converter.

[0071] In addressing the technical challenges mentioned above, the application scenario—energy storage converters at the grid connection point of renewable energy power plants—often presents the following technical problem: At renewable energy power plants, due to the random fluctuations of renewable energy sources such as wind and solar power, the power grid frequently experiences short-term frequency shifts and voltage transients. These changes, transmitted to the grid connection point of the energy storage converter, cause temporary anomalies in its voltage, current, and other measurement signals. This leads to misjudging these discrete random disturbances caused by the external power grid environment as internal faults of the converter itself, resulting in an increased false alarm rate. Furthermore, because potential faults are often concealed and progressive, the accuracy of single-time-window early warning systems for identifying potential faults is low, reducing the accuracy of fault warnings for energy storage converters. This leads to an increase in unplanned outages at renewable energy power plants caused by energy storage converter failures, increasing wind / solar curtailment rates, and ultimately reducing the stability and utilization rate of renewable energy power plants. Considering the following characteristics required for this application scenario—high anti-interference capability and sensitivity to potential faults—we have decided to adopt the following solution.

[0072] In some optional implementations of certain embodiments, the above-mentioned regulation of the energy storage converter at different levels based on the converter fault location information set and the warning level information set includes: The first step involves responding to the presence of an emergency warning level in the aforementioned warning level information set, and then performing emergency control operations on the energy storage converter according to a preset control rule engine. This preset control rule engine can be a pre-defined rule execution system that automatically matches and triggers different levels of control strategies based on the warning level identifier corresponding to the energy storage converter and the fault location information. The warning level information representing the emergency level can include a Level 1 warning identifier. In practice, the executing entity can first match the Level 1 warning identifier and the corresponding converter fault location information with the control rules included in the preset control rule engine to obtain an emergency control instruction set. The emergency control instructions in this set can be executable computer instructions for performing emergency control operations on the energy storage converter. Then, the emergency control instruction set is executed to perform emergency control operations on the energy storage converter.

[0073] The second step, in response to the absence of warning level information representing an emergency level in the aforementioned warning level information set, is to execute the following steps based on the aforementioned converter fault location information set and the aforementioned warning level information set: Sub-step 1: Based on the aforementioned warning level information set, generate a set of warning level weight coefficients. The warning level weight coefficients in this set can be numerical weights corresponding to the aforementioned warning level information. In practice, the executing entity can retrieve the warning level weight coefficients for each warning level information in the aforementioned warning level information set from a preset warning weight mapping database to obtain the set of warning level weight coefficients. This preset warning weight mapping database can be a pre-defined database that defines the weight coefficients corresponding to different warning levels. For example, the warning level weight coefficient corresponding to a Level 2 warning identifier could be 1.5, and the warning level weight coefficient corresponding to a Level 3 warning identifier could be 1.

[0074] Sub-step 2 involves performing dynamic threshold standardization on the aforementioned converter fault deviation information set to obtain a standardized converter fault deviation information set. In practice, the executing entity can first perform the following second determination step for each converter fault deviation information in the aforementioned converter fault deviation information set: First, in response to determining that the warning level identifier corresponding to the aforementioned converter fault deviation information is a level 2 warning identifier, the second warning threshold corresponding to the aforementioned converter fault deviation information is used as the dynamic threshold of the aforementioned converter fault deviation information. Second, in response to determining that the warning level identifier corresponding to the aforementioned converter fault deviation information is not a level 2 warning identifier, the first warning threshold corresponding to the aforementioned converter fault deviation information is used as the dynamic threshold of the aforementioned converter fault deviation information. Third, the difference between the aforementioned converter fault deviation information and the aforementioned dynamic threshold, and the ratio of the difference to the aforementioned dynamic threshold, are determined as the normalized converter fault deviation information.

[0075] Sub-step 3 involves determining the aforementioned early warning level weight coefficient set, the aforementioned standardized downstream converter fault deviation value set, and the aforementioned converter fault location information set as an early warning trend information set. The early warning trend information in this set can be the comprehensive characteristic information of the aforementioned energy storage converter within the corresponding time window.

[0076] Sub-step 4: Based on the aforementioned early warning trend information set, perform the following steps: Operation step 1 involves time-series sampling of the early warning trend information set to obtain the target early warning trend information and the sampled early warning trend information set. The target early warning trend information can be the early warning trend information corresponding to the time window furthest from the current time in the aforementioned early warning trend information set. The sampled early warning trend information set can be the early warning trend information set after removing the target early warning trend information.

[0077] Operation step 2: Update the converter trend warning sequence according to the target warning trend information to obtain the updated converter trend warning sequence. The aforementioned converter trend warning sequence can be at least one warning trend information prior to the time corresponding to the target warning trend information, arranged in the order in which the converter trend warning information was obtained. For example, the aforementioned converter trend warning sequence can be a queue containing warning trend information for 150 time windows. In practice, the executing entity can update the converter trend warning sequence according to the target warning trend information based on the first-in, first-out principle to obtain the updated converter trend warning sequence.

[0078] Step 3 involves performing time-series analysis on the updated converter trend warning sequence to obtain the converter fault severity value. This fault severity value characterizes the severity of potential faults in the energy storage converter. In practice, the execution entity can use the CUSUM (Cumulative Sum) algorithm to perform time-series analysis on the updated converter trend warning sequence to obtain the converter fault severity value.

[0079] Step 4 involves performing time-series aggregation processing on the converter fault location information set corresponding to the updated converter trend warning sequence to obtain the target converter fault location information. This target converter fault location information can be the location information of the component most likely to exhibit potential fault characteristics among the components included in the energy storage converter. In practice, the executing entity can first determine the weighted fault probability set by multiplying the node fault probability set corresponding to the converter fault location information set and the warning level weight coefficient set. Then, each weighted fault probability with the same converter fault location information in the weighted fault probability set is grouped into a set of weighted fault probability groups. Next, the weighted fault probabilities in each of the weighted fault probability groups are accumulated to obtain a comprehensive fault probability score set. The comprehensive fault probability score in the comprehensive fault probability score set can be a numerical representation of the frequency of a component's occurrence in different time windows and the severity of its corresponding warning level. Finally, the converter fault location information corresponding to the largest comprehensive fault probability score in the comprehensive fault probability score set is determined as the target converter fault location information.

[0080] Operation step 5: In response to determining that the converter fault severity value is greater than or equal to a first preset warning threshold, the energy storage converter is subject to focused control based on the target converter fault location information and the preset warning rule engine. The first preset warning threshold can be a pre-set threshold used to determine whether the converter fault severity value reaches the critical value for focused control operation. In practice, the executing entity can first, in response to determining that the converter fault severity value is greater than or equal to the first preset warning threshold, determine the warning identifier corresponding to the energy storage converter as a level three warning identifier. Then, the level three warning identifier and the target converter fault location information are matched with the control rules included in the preset control rule engine to obtain a focused control instruction set. The focused control instructions in the focused control instruction set can be executable computer instructions for focusing control operations on the energy storage converter. Finally, the focused control instruction set is executed to perform emergency control operations on the energy storage converter.

[0081] Operation step 6: In response to determining that the converter fault severity value is greater than or equal to the second preset warning threshold, maintenance and control operations are performed on the energy storage converter based on the target converter fault location information and the preset warning rule engine. The first preset warning threshold is less than the second preset warning threshold. The second preset warning threshold can be a pre-set threshold used to determine whether the converter fault severity value has reached the critical value for maintenance and control operations. In practice, the executing entity can first, in response to determining that the converter fault severity value is greater than or equal to the second preset warning threshold, identify the warning identifier corresponding to the energy storage converter as a secondary warning identifier. Then, the secondary warning identifier and the target converter fault location information are matched with the control rules included in the preset control rule engine to obtain a maintenance and control instruction set. The maintenance and control instructions in the instruction set can be executable computer instructions for performing maintenance and control operations on the energy storage converter. Finally, the maintenance and control instruction set is executed to perform maintenance and control operations on the energy storage converter. Operation step 7: In response to determining that the number of sampled early warning trend information items included in the sampled early warning trend information set is less than or equal to a first preset value, the above operation steps are terminated. The first preset value can be a pre-set threshold value used to determine whether the above operation steps should continue. The first preset value can be 0.

[0082] Sub-step 5: In response to the determination that the number of sampled early warning trend information included in the sampled early warning trend information set is greater than the first preset value, the updated converter trend early warning sequence is determined as the converter trend early warning sequence, and the sampled early warning trend information set is determined as the early warning trend information set, so as to execute the above operation steps again.

[0083] The above-mentioned technical solution and its related contents, as an inventive point of the embodiments of this disclosure, solve the third technical problem: "the number of unplanned shutdowns of new energy power stations caused by energy storage converter failures increases, thereby reducing the stability of new energy power stations and the utilization rate of new energy." Factors contributing to the increased frequency of unplanned outages at renewable energy power plants due to energy storage converter failures, thereby reducing their stability and renewable energy utilization, are often as follows: Due to the random fluctuations in renewable energy sources such as wind and solar power, the power grid frequently experiences short-term frequency deviations and voltage transients. These changes, transmitted to the grid connection point of the energy storage converter, cause temporary anomalies in its voltage, current, and other measurement signals. This leads to early warning judgments based on a single time window often misinterpreting these discrete random disturbances caused by the external grid environment as internal faults within the converter itself, resulting in an increased false alarm rate. Simultaneously, the concealed and gradual nature of potential faults reduces the accuracy of single-time-window early warning judgments in identifying potential faults, further decreasing the accuracy of fault warnings for energy storage converters. This leads to an increase in the frequency of unplanned outages at renewable energy power plants caused by energy storage converter failures, thus reducing their stability and renewable energy utilization. Addressing these factors can reduce the frequency of unplanned outages at renewable energy power plants caused by energy storage converter failures, thereby improving their stability and reducing the reduction in renewable energy utilization. To achieve this effect, this disclosure firstly, in response to the existence of warning level information representing an emergency level in the aforementioned warning level information set, performs emergency control operations on the aforementioned energy storage converter according to a preset warning rule engine. Here, emergency control operations are directly performed on warning level information representing an emergency level, ensuring a rapid response to severe faults. Then, in response to the absence of warning level information representing an emergency level in the aforementioned warning level information set, based on the aforementioned converter fault location information set and the aforementioned warning level information set, the following steps are performed: First, a warning level weight coefficient set is generated based on the aforementioned warning level information set. Second, the aforementioned converter fault deviation information set is subjected to dynamic threshold standardization processing to obtain a standardized converter fault deviation information set. Third, the aforementioned warning level weight coefficient set, the aforementioned standardized converter fault deviation value set, and the aforementioned converter fault location information set are determined as a warning trend information set. Here, different warning levels are assigned weights, and different dynamic thresholds are selected according to the warning level during the standardization process, providing accurate input data for subsequent time series analysis and time series aggregation processing. Fourth step, based on the above early warning trend information set, perform the following operations: Sub-step 1, perform time-series sampling on the early warning trend information set to obtain the target early warning trend information and the sampled early warning trend information set.Sub-step 2: Update the converter trend warning sequence according to the target warning trend information to obtain the updated converter trend warning sequence. The updated sequence can be at least one warning trend information from a time point prior to the target warning trend information. Here, the converter trend warning sequence is updated according to the first-in, first-out principle, ensuring the system always analyzes based on the latest, continuous data within a fixed time span, providing a data foundation for subsequent time-series analysis. Simultaneously, discrete interference signals can be diluted within the sequence. Sub-step 3: Perform time-series analysis on the updated converter trend warning sequence to obtain the converter fault severity value. Here, the updated converter trend warning sequence is accumulated and calculated, accumulating small, persistent positive deviations while ignoring brief random fluctuations. This reduces the system's sensitivity to instantaneous interference from the external power grid (reducing false alarms) and increases its sensitivity to potential faults (reducing missed alarms and providing early warnings). Sub-step 4: Perform time-series aggregation processing on the converter fault location information set corresponding to the updated converter trend warning sequence to obtain the target converter fault location information. Here, the warning level weights are combined with the fault probability information of each time window and weighted cumulatively. This integrates the frequency and severity information of the fault location, allowing the most persistent and significant fault source to be identified from multiple possible abnormal locations. Sub-step 5, in response to determining that the converter fault severity value is greater than or equal to the first preset warning threshold, performs a monitoring and control operation on the energy storage converter based on the target converter fault location information and the preset warning rule engine. Sub-step 6, in response to determining that the converter fault severity value is greater than or equal to the second preset warning threshold, performs a maintenance and control operation on the energy storage converter based on the aforementioned target converter fault location information and the preset warning rule engine, wherein the first preset warning threshold is less than the second preset warning threshold. Here, comparing the converter fault severity value with two incremental preset thresholds triggers different levels of control operations, further filtering interference and achieving hierarchical management of potential fault characteristics of different degrees. Sub-step 7: In response to determining that the number of sampled early warning trend information items included in the sampled early warning trend information set is less than or equal to a first preset value, the updated converter trend early warning sequence is determined as the converter trend early warning sequence. Fourth step: In response to determining that the number of sampled early warning trend information items included in the sampled early warning trend information set is greater than a first preset value, the updated converter trend early warning sequence is determined as the converter trend early warning sequence, and the sampled early warning trend information set is determined as the early warning trend information set, to repeat the above operation steps. Here, a loop mechanism controls the flow to process all time window data to be analyzed one by one, ensuring the completeness of the analysis conclusions and improving the fault early warning accuracy of the energy storage converter.Therefore, the above technical solution can reduce the number of unplanned outages of new energy power plants caused by energy storage converter failures, and improve the stability of new energy power plants and reduce the reduction in new energy utilization.

[0084] The above embodiments of this disclosure have the following beneficial effects: The fault early warning and control method based on energy storage converters in some embodiments of this disclosure can reduce the number of unplanned outages of power equipment where energy storage converters are located, and improve the stability and utilization rate of new energy power plants. Specifically, the reason why it is difficult to effectively reduce the frequency of unplanned outages of power equipment where energy storage converters are located, and thus cannot improve the stability and utilization rate of new energy power plants, is that: since the "normal" state of energy storage converters changes dynamically with operating conditions such as load rate, SOC, and ambient temperature, the preset fixed threshold cannot be adaptively adjusted, and false alarms are easily generated under harsh operating conditions such as heavy load and high temperature, while false alarms may be missed under stable operating conditions such as light load and low temperature; at the same time, the supervised learning model relies heavily on a large number of labeled fault samples for training, while fault samples are scarce and limited in type in actual industrial scenarios, resulting in insufficient generalization ability of the model and reduced recognition accuracy, resulting in low fault early warning accuracy of energy storage converters, making it difficult to effectively reduce the number of unplanned outages of power equipment where energy storage converters are located, and thus unable to improve the stability and utilization rate of new energy power plants. Based on this, some embodiments of the present disclosure of fault early warning and control based on energy storage converters can first acquire the converter-related physical information set and operating status information set of the energy storage converter. Here, the acquired converter-related physical information set provides synchronous multi-dimensional data for subsequent spectrum diagram data conversion; the acquired operating status information set serves as the data foundation for subsequently generating accurate converter deviation information and early warning level information. Secondly, the aforementioned converter-related physical information set is processed through spectrum diagram data conversion to obtain a converter signal spectrum diagram set. Here, the time-domain signal is converted into a frequency-domain representation. The aforementioned converter signal spectrum diagram can significantly amplify the weak frequency characteristic changes caused by potential faults, while suppressing random noise interference in the time-domain signal, providing higher quality and more sensitive data for subsequent feature extraction. Thirdly, a spectrum diagram feature extraction model is used to perform feature extraction processing on the aforementioned converter signal spectrum diagram set to obtain a converter signal feature vector set. Here, a self-supervised learning spectrum diagram feature extraction model is used to extract redundant feature representations from the converter signal spectrum diagram, overcoming the dependence on prior fault knowledge and improving the generalization ability for unknown fault types. Next, based on the aforementioned converter signal feature vector set and operating status information set, a converter fault deviation information set is generated. Here, converter fault deviation information corresponding to the operating status information is generated for the converter signal feature vectors, reducing the possibility of inaccurate generated converter fault deviation information due to different operating states of the energy storage converter. Subsequently, based on the aforementioned operating status information set, the warning level information for each converter fault deviation information in the aforementioned converter fault deviation information set is determined, resulting in a warning level information set. Here, adaptive warning thresholds are set for converter fault deviation information under different operating states, improving the accuracy of warnings.Then, fault location identification is performed on the aforementioned converter fault deviation information set to obtain the converter fault location information set. Here, the obtained converter fault location information can pinpoint the source of the anomaly to a specific component or circuit module, shortening fault investigation time and improving the accuracy and precision of fault early warning for energy storage converters. Finally, based on the aforementioned converter fault location information set and the aforementioned early warning level information set, different levels of regulation are applied to the aforementioned energy storage converter. Here, by comprehensively combining the converter fault location information set and the early warning level information set to trigger different levels of early warning, overreaction to minor faults can be reduced, thereby reducing unnecessary shutdowns. Therefore, this fault early warning and regulation method based on energy storage converters can reduce the number of unplanned outages of power equipment where energy storage converters are located, improving the stability of new energy power plants and the utilization rate of new energy sources.

[0085] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a fault early warning and control device based on an energy storage converter. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, this fault early warning and control device based on energy storage converter can be specifically applied to various electronic devices.

[0086] like Figure 2 As shown, a fault early warning and control device 200 based on an energy storage converter includes: an acquisition unit 201, a conversion unit 202, a feature extraction unit 203, a generation unit 204, a determination unit 205, a positioning unit 206, and a control unit 207. The acquisition unit 201 is configured to acquire a set of converter-related physical information and an operating status information set of the energy storage converter. The conversion unit 202 is configured to perform spectrum diagram data conversion processing on the aforementioned converter-related physical information set to obtain a converter signal spectrum diagram set. The feature extraction unit 203 is configured to perform feature extraction processing on the aforementioned converter signal spectrum diagram set using a spectrum diagram feature extraction model to obtain a converter signal feature vector set. The generation unit 204 is configured to generate a converter fault deviation information set based on the aforementioned converter signal feature vector set and the aforementioned operating status information set. The determination unit 205 is configured to determine the early warning level information for each converter fault deviation information in the aforementioned converter fault deviation information set based on the aforementioned operating status information set, to obtain an early warning level information set. The positioning unit 206 is configured to: perform fault location identification on the aforementioned converter fault deviation information set to obtain a converter fault location information set. The control unit 207 is configured to: perform different levels of control on the aforementioned energy storage converter based on the aforementioned converter fault location information set and the aforementioned early warning level information set.

[0087] It is understandable that the various units and references described in the fault early warning and control device 200 based on the energy storage converter are... Figure 1 The steps described in the method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the fault early warning and control device 200 based on energy storage converter and the units contained therein, and will not be repeated here.

[0088] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device (e.g., an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0089] like Figure 3 As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0090] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0091] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0092] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0093] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0094] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: acquire a set of converter-related physical information and an operating status information set for the energy storage converter; perform spectrum diagram data conversion processing on the aforementioned converter-related physical information set to obtain a converter signal spectrum diagram set; use a spectrum diagram feature extraction model to perform feature extraction processing on the aforementioned converter signal spectrum diagram set to obtain a converter signal feature vector set; generate a converter fault deviation information set based on the aforementioned converter signal feature vector set and the aforementioned operating status information set; determine the warning level information for each converter fault deviation information in the aforementioned converter fault deviation information set based on the aforementioned operating status information set to obtain a warning level information set; perform fault location identification on the aforementioned converter fault deviation information set to obtain a converter fault location information set; and perform different levels of regulation on the aforementioned energy storage converter based on the aforementioned converter fault location information set and the aforementioned warning level information set.

[0095] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0096] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0097] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, a conversion unit, a feature extraction unit, a generation unit, a determination unit, a positioning unit, and a control unit. The names of these units do not necessarily limit the specific unit; for example, an acquisition unit may also be described as "a unit that acquires a set of converter-related physical information and an operating status information set of an energy storage converter."

[0098] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0099] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A fault early warning and control method based on an energy storage converter, comprising: Acquire relevant physical information sets and operating status information sets of the energy storage converter, as well as historical relevant physical information sets of the energy storage converter, wherein the historical relevant physical information sets of the energy storage converter are datasets that do not include fault information; The relevant physical information set and the historical relevant physical information set of the energy storage converter are subjected to spectrum graph data conversion processing to obtain a converter signal spectrum graph set and a historical converter signal spectrum graph set. The spectrum graph data conversion processing includes: The relevant physical information set is preprocessed to obtain the preprocessed relevant physical information set; Frequency domain transformation is performed on the preprocessed relevant physical information set and the historical relevant physical information set of the energy storage converter to obtain the converter signal spectrum map set and the historical converter signal spectrum map set. The converter signal spectrum atlas is processed using a spectrum feature extraction model to obtain a converter signal feature vector set. The spectrum feature extraction model is obtained through the following steps: Data augmentation is performed on the historical converter signal spectrum set to obtain a first sample spectrum set and a second sample spectrum set. Based on the first sample spectrogram set and the second sample spectrogram set, the following training steps are performed: The first sample spectrum set and the second sample spectrum set are input into the initial spectrum feature extraction model to obtain the first spectrum feature vector set and the second spectrum feature vector set. Determine the cross-correlation matrix of the first and second spectral eigenvector sets; The feature cross-correlation matrix is ​​input into the spectrogram feature extraction loss function to obtain the feature loss function value; In response to the determination that the feature loss function value meets the preset loss value condition, the initial spectrogram feature extraction model is determined as the trained spectrogram feature extraction model; In response to the determination that the feature loss function value does not meet the preset loss value condition, the historical converter signal spectrum map set is re-processed with data augmentation to obtain a first target sample spectrum map set and a second target sample spectrum map set, which are used as the first sample spectrum map set and the second sample spectrum map set. The relevant parameters in the initial spectrum map feature extraction model are adjusted to obtain an adjusted spectrum map feature extraction model, which is used as the initial spectrum map feature extraction model, so as to execute the training step again. Based on the converter signal feature vector set and the operating status information set, a converter fault deviation information set is generated; Based on the operating status information set, determine the warning level information of each converter fault deviation information in the converter fault deviation information set to obtain the warning level information set; The converter fault deviation information set is used to perform fault location identification to obtain the converter fault location information set. Based on the converter fault location information set and the early warning level information set, the energy storage converter is regulated at different levels.

2. The method according to claim 1, wherein, The step of generating a converter fault deviation information set based on the converter signal feature vector set and the operating status information set includes: The converter signal feature vector set is input into the converter fault coding layer of the trained converter fault identification model to obtain the fault coding feature vector set. The converter fault identification model further includes a converter fault bottleneck layer and a converter fault condition decoding layer. The fault coding feature vector set is input into the converter fault bottleneck layer to obtain the fault compressed feature vector set; The set of operating status information is encoded to obtain a set of operating status vectors; The running state vector set and the fault compressed feature vector set are concatenated to obtain a concatenated feature vector set. The spliced ​​feature vector set is input into the converter fault condition decoding layer to obtain the fault reconstruction feature vector set; A converter fault deviation information set is generated based on the fault reconstruction feature vector set and the converter signal feature vector set.

3. The method according to claim 1, wherein, The step involves determining the warning level information for each converter fault deviation information in the converter fault deviation information set based on the operating status information set, thereby obtaining a warning level information set, including: For each converter fault deviation information in the converter fault deviation information set, the following determination steps are performed: The operating status information corresponding to the converter fault deviation information is determined as the target operating status information; The target operating status information is input into the trained converter fault dynamic threshold model to obtain the first warning threshold and the second warning threshold. The target operating status information is input into the trained converter prediction distribution model to obtain the target mean and target standard deviation; A target conservative threshold is generated based on the target mean and the target standard deviation; The first warning threshold, the second warning threshold, and the target conservative threshold are determined as the fault threshold group for the converter fault deviation information; Based on the converter signal feature vector corresponding to the converter fault deviation information and the preset mean value of the converter signal feature vector, a converter signal feature vector deviation value is generated. A converter signal fault identifier is generated based on the converter signal feature vector deviation value and the preset converter signal feature vector fault threshold. Based on the converter fault deviation information, the fault threshold group, and the converter signal fault identifier, the warning level information of the converter fault deviation information is determined.

4. A fault early warning and control device based on an energy storage converter, comprising: The acquisition unit is configured to acquire a set of relevant physical information and a set of operating status information of the energy storage converter, as well as a set of historical relevant physical information of the energy storage converter, wherein the set of historical relevant physical information of the energy storage converter is a dataset that does not include fault information. The conversion unit is configured to perform spectrum data conversion processing on the relevant physical information set and the historical relevant physical information set of the energy storage converter to obtain a converter signal spectrum set and a historical converter signal spectrum set, wherein the spectrum data conversion processing includes: The relevant physical information set is preprocessed to obtain the preprocessed relevant physical information set; Frequency domain transformation is performed on the preprocessed relevant physical information set and the historical relevant physical information set of the energy storage converter to obtain the converter signal spectrum map set and the historical converter signal spectrum map set. The feature extraction unit is configured to perform feature extraction processing on the converter signal spectrum set using a spectrum feature extraction model to obtain a converter signal feature vector set. The spectrum feature extraction model is obtained through the following steps: Data augmentation is performed on the historical converter signal spectrum set to obtain a first sample spectrum set and a second sample spectrum set. Based on the first sample spectrogram set and the second sample spectrogram set, the following training steps are performed: The first sample spectrum set and the second sample spectrum set are input into the initial spectrum feature extraction model to obtain the first spectrum feature vector set and the second spectrum feature vector set. Determine the cross-correlation matrix of the first and second spectral eigenvector sets; The feature cross-correlation matrix is ​​input into the spectrogram feature extraction loss function to obtain the feature loss function value; In response to the determination that the feature loss function value meets the preset loss value condition, the initial spectrogram feature extraction model is determined as the trained spectrogram feature extraction model; In response to the determination that the feature loss function value does not meet the preset loss value condition, the historical converter signal spectrum map set is re-processed with data augmentation to obtain a first target sample spectrum map set and a second target sample spectrum map set, which are used as the first sample spectrum map set and the second sample spectrum map set. The relevant parameters in the initial spectrum map feature extraction model are adjusted to obtain an adjusted spectrum map feature extraction model, which is used as the initial spectrum map feature extraction model, so as to execute the training step again. The generation unit is configured to generate a converter fault deviation information set based on the converter signal feature vector set and the operating status information set; The determining unit is configured to determine the warning level information of each converter fault deviation information in the converter fault deviation information set based on the operating status information set, thereby obtaining a warning level information set; The positioning unit is configured to perform fault location identification on the converter fault deviation information set to obtain the converter fault location information set. The control unit is configured to control the energy storage converter at different levels based on the converter fault location information set and the early warning level information set.

5. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-3.

6. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-3.

Images