Photovoltaic inverter fault early warning method and system based on long short-term memory network

Abstract

The invention discloses a photovoltaic inverter fault early warning method and system based on a long short-term memory network, and the method comprises the steps: collecting the operation time sequence data of a photovoltaic inverter, carrying out the data preprocessing, and constructing a training data set; constructing a deep neural network model based on a long short-term memory network, training the deep neural network model by using the training data set, inputting a historical operation parameter sequence, and outputting an operation trend prediction result in a period of time in the future; calculating whether the deviation between the prediction result and the actual operation value exceeds a preset threshold value or not, and judging whether to trigger early warning or not in combination with a preset time duration condition; and when early warning is triggered, an alarm signal is output to a monitoring system or operation and maintenance personnel, and early fault early warning is realized. According to the method, the deep learning model is constructed by using the inverter operation time sequence data, and advanced prediction of multiple types of faults is realized, so that the stability and the maintenance efficiency of the system are improved.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic inverter fault prediction technology, specifically to a photovoltaic inverter fault early warning method and system based on long short-term memory networks. Background Technology

[0002] With the large-scale development of photovoltaic power plants, inverters, as core power electronic devices, directly affect power generation efficiency and system safety through their operational stability. Inverter faults typically manifest as insulation failures, overheating, aging filter capacitors, and grid disturbances. Current common methods rely on fixed thresholds, expert rules, or traditional statistical models for state assessment, which suffer from low accuracy and poor generalization ability, especially in terms of early warning capabilities under long-term operating conditions. In recent years, deep learning technology has been gradually promoted in power equipment condition assessment, among which Long Short-Term Memory (LSTM) networks have become a research hotspot for equipment fault prediction due to their excellent performance in time series modeling. Summary of the Invention

[0003] To address this, this invention provides a photovoltaic inverter fault early warning method and system based on long short-term memory networks, in order to solve the problems of low accuracy, poor generalization ability, and insufficient early warning capability for long-term operation in existing photovoltaic inverter fault state judgment technologies.

[0004] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions: According to a first aspect of the present invention, an embodiment of the present invention provides a method for early warning of photovoltaic inverter faults based on long short-term memory networks, the method comprising: Collect and preprocess the operating sequence data of the photovoltaic inverter to construct a training dataset; A deep neural network model based on a long short-term memory network is constructed, and the deep learning model is trained using a training dataset. The input is a sequence of historical operating parameters, and the output is a prediction result of the operating trend for a future period of time. Calculate whether the deviation between the predicted result and the actual operating value exceeds a preset threshold, and determine whether to trigger an early warning based on the preset time duration condition; When an alert is triggered, an alarm signal is output to the monitoring system or maintenance personnel to provide early warning of faults.

[0005] Furthermore, the operation sequence data of the photovoltaic inverter is collected, specifically including: Key operating parameters of the photovoltaic inverter are collected, including DC input voltage and current, AC output voltage, current, temperature, frequency, power factor, and ambient temperature, forming multi-dimensional time series data.

[0006] Further, data preprocessing specifically includes: The collected photovoltaic inverter runtime sequence data is standardized, missing value imputation is performed, outlier removal is performed, and sliding window slicing is applied to form a training dataset suitable for model training.

[0007] Furthermore, a deep neural network model based on a long short-term memory network is constructed, and the deep learning model is trained using a training dataset, specifically including: Build a deep neural network model containing one or more LSTM units, each layer containing multiple hidden units, supporting multi-step prediction output; during the training phase, use the mean squared error loss function and the Adam optimizer for training.

[0008] Furthermore, a deep neural network model based on long short-term memory networks is constructed, specifically including: The deep neural network model adopts a two-layer stacked LSTM structure, with 64 hidden units in each layer, the activation function being tanh, and the output layer being a fully connected regression layer.

[0009] Furthermore, the model is trained using the training dataset, specifically including: By updating and incrementally training the deep learning model online, it can adapt to changes in operating conditions under long-term inverter operation.

[0010] Furthermore, the deviation between the predicted result and the actual operating value is calculated to see if it exceeds a preset threshold, and a warning is triggered based on a preset time duration condition. Specifically, this includes: If the deviation between the predicted value and the actual operating value exceeds a preset threshold and continues for a preset duration, an early warning will be triggered.

[0011] Furthermore, the method also includes: Deploy deep learning models on edge computing devices or main control systems to achieve real-time online prediction and alarm linkage.

[0012] According to a second aspect of the present invention, an embodiment of the present invention provides a photovoltaic inverter fault early warning system based on a long short-term memory network, the system comprising: The dataset construction module is used to collect the operating sequence data of the photovoltaic inverter and perform data preprocessing to build the training dataset. The model building module is used to build a deep neural network model based on a long short-term memory network, and to train the deep learning model using a training dataset. It takes a sequence of historical operating parameters as input and outputs a prediction result of the operating trend for a future period of time. The fault threshold determination module is used to calculate whether the deviation between the predicted result and the actual operating value exceeds a preset threshold, and to determine whether to trigger an early warning based on a preset time duration condition. The early warning module is used to output alarm signals to the monitoring system or maintenance personnel when an early warning is triggered, so as to realize early warning of faults.

[0013] According to three aspects, one embodiment provides an electronic device, the device comprising: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to perform the steps of a photovoltaic inverter fault early warning method based on a long short-term memory network as described in any of the preceding claims.

[0014] According to a fourth aspect, one embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a photovoltaic inverter fault early warning method based on a long short-term memory network as described in any of the preceding claims.

[0015] Compared with existing technologies, the photovoltaic inverter fault early warning method and system based on long short-term memory networks provided by this invention have the following beneficial effects: 1. This invention utilizes Long Short-Term Memory (LSTM) networks to perform deep modeling of photovoltaic inverter operating sequence data, which can effectively identify potential abnormal trends and has higher prediction accuracy and robustness compared to methods based on fixed thresholds.

[0016] 2. This invention proposes a mechanism that combines fault precursor trend prediction with dynamic threshold early warning, which can issue early warning signals within a time window of minutes to hours before a fault occurs, providing maintenance personnel with sufficient response time and reducing the risk of sudden downtime.

[0017] 3. The method of the present invention can adapt to the special operating conditions of inverters in complex environments such as high altitude and cold weather, and solves the problem of insufficient generalization ability of traditional models, thus having stronger adaptability and engineering applicability.

[0018] 4. The model of this invention can be deployed on edge computing devices or power plant monitoring systems to achieve real-time online operation and adaptive updates, thereby reducing system operation and maintenance costs and improving the overall power generation efficiency and safety of photovoltaic power plants. Attached Figure Description

[0019] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic diagram illustrating the overall workflow of a photovoltaic inverter fault early warning method based on a long short-term memory network, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a photovoltaic inverter fault early warning system based on a long short-term memory network, provided as an embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0021] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0022] The first embodiment of this invention provides a photovoltaic inverter fault early warning method based on a long short-term memory network. It utilizes inverter runtime timing data to construct a deep learning model, enabling early prediction of multiple types of faults, thereby improving system stability and maintenance efficiency. The following is a combination of... Figure 1 Please provide a detailed explanation.

[0023] like Figure 1 As shown, in step S100, the operating sequence data of the photovoltaic inverter is collected and preprocessed to construct a training dataset.

[0024] The above steps specifically include: S110, Data Acquisition: Acquire key operating parameters of the photovoltaic inverter, including DC input voltage and current, AC output voltage, current, temperature, frequency, power factor, ambient temperature, etc., to form multi-dimensional time series data.

[0025] In this embodiment, the data acquisition module is deployed at the inverter controller or in the communication concentrator, and the acquisition frequency is 1 minute or higher. Training samples are constructed using a sliding window mechanism (e.g., predicting the trend of the next 10 minutes based on data from the past 60 minutes).

[0026] S120, Data Preprocessing: Standardize the raw data, impute missing values, remove outliers, and perform sliding window slicing to form a sample set suitable for model training.

[0027] like Figure 1 As shown, in step S200, a deep neural network model based on a long short-term memory network is constructed, and the deep learning model is trained using a training dataset. The historical running parameter sequence is input, and the running trend prediction result for a future period of time is output.

[0028] The above steps specifically include: S210, Model Building: Construct a deep neural network structure containing one or more LSTM units, each layer containing multiple hidden units, to enhance the ability to model long-term series. Input the sequence of running parameters and output the running trend over a future period of time, supporting multi-step prediction output.

[0029] In this embodiment, the model employs a two-layer stacked LSTM structure, with each layer containing 64 hidden units and the activation function being tanh. The output layer is a fully connected regression layer. After model training, the weights are exported and integrated into the online system.

[0030] S220, Model Training: During the training phase, the mean squared error loss function and the Adam optimizer are used for training.

[0031] In practical applications, supervised training can be performed using fault label data, or an unsupervised prediction model can be combined with a reconstruction error strategy to improve the detection capability for unknown fault types. The model supports online updates and incremental training to adapt to changes in operating conditions under long-term inverter operation.

[0032] like Figure 1 As shown, in step S300, it is calculated whether the deviation between the predicted result and the actual running value exceeds a preset threshold, and whether an early warning is triggered is determined in combination with the preset time duration condition.

[0033] The above steps specifically include: In this embodiment, an early warning signal is triggered when the deviation between the model's predicted value and the actual value continuously exceeds a set threshold (e.g., the deviation exceeds 10% for 10 consecutive minutes). The warning mechanism uses a combination of dynamic threshold and time window to avoid misjudgment due to single-point fluctuations and improve the stability of the warning.

[0034] like Figure 1 As shown, in step S400, when an early warning is triggered, an alarm signal is output to the monitoring system or maintenance personnel to achieve early fault warning.

[0035] In this embodiment, the method also includes system integration and deployment: deploying the model on edge computing devices or main control systems to achieve real-time online prediction and alarm linkage.

[0036] The photovoltaic inverter fault early warning method based on long short-term memory network proposed in this invention has the following advantages: (1) It has the ability to identify the trend of fault precursors and can provide early warnings at the minute to hour level; (2) It has stronger robustness and adaptability compared to threshold-based static strategies; (3) It supports continuous self-learning and upgrading of the model and has online training capabilities; (4) It can be deployed in the actual system of centralized or distributed photovoltaic power stations.

[0037] Corresponding to the photovoltaic inverter fault early warning method based on long short-term memory networks disclosed in the above embodiments, this invention also discloses a photovoltaic inverter fault early warning system based on long short-term memory networks, such as... Figure 2 As shown, it specifically includes: The dataset construction module is used to collect the operating sequence data of the photovoltaic inverter and perform data preprocessing to build the training dataset. The model building module is used to build a deep neural network model based on a long short-term memory network, and to train the deep learning model using a training dataset. It takes a sequence of historical operating parameters as input and outputs a prediction result of the operating trend for a future period of time. The fault threshold determination module is used to calculate whether the deviation between the predicted result and the actual operating value exceeds a preset threshold, and to determine whether to trigger an early warning based on a preset time duration condition. The early warning module is used to output alarm signals to the monitoring system or maintenance personnel when an early warning is triggered, so as to realize early warning of faults.

[0038] It should be noted that for a detailed description of the photovoltaic inverter fault early warning system based on long short-term memory network provided in the embodiments of the present invention, please refer to the relevant description of the photovoltaic inverter fault early warning method based on long short-term memory network provided in the embodiments of the present invention, which will not be repeated here.

[0039] In addition, embodiments of the present invention also provide an electronic device, the device comprising: a processor and a memory; the memory being used to store one or more program instructions; the processor being used to execute one or more program instructions to perform the steps of a photovoltaic inverter fault early warning method based on a long short-term memory network as described in any of the preceding embodiments.

[0040] It should be noted that for a detailed description of an electronic device provided in the embodiments of the present invention, please refer to the relevant description of a photovoltaic inverter fault early warning method based on a long short-term memory network provided in the embodiments of this application, which will not be repeated here.

[0041] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a photovoltaic inverter fault early warning method based on a long short-term memory network as described in any of the preceding claims.

[0042] It should be noted that for a detailed description of the computer-readable storage medium provided in the embodiments of the present invention, please refer to the relevant description of the photovoltaic inverter fault early warning method based on long short-term memory network provided in the embodiments of this application, which will not be repeated here.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for early warning of photovoltaic inverter faults based on long short-term memory networks, characterized in that, The method includes: Collect and preprocess the operating sequence data of the photovoltaic inverter to construct a training dataset; A deep neural network model based on a long short-term memory network is constructed, and the deep neural network model is trained using a training dataset. The input is a sequence of historical operating parameters, and the output is a prediction result of the operating trend for a future period of time. Calculate whether the deviation between the predicted result and the actual operating value exceeds a preset threshold, and determine whether to trigger an early warning based on the preset time duration condition; When an alert is triggered, an alarm signal is output to the monitoring system or maintenance personnel to provide early warning of faults.

2. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 1, characterized in that, Collecting the operating sequence data of the photovoltaic inverter, specifically including: Key operating parameters of the photovoltaic inverter are collected, including DC input voltage and current, AC output voltage, current, temperature, frequency, power factor, and ambient temperature, forming multi-dimensional time series data.

3. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 1, characterized in that, Data preprocessing specifically includes: The collected photovoltaic inverter runtime sequence data is standardized, missing value imputation is performed, outlier removal is performed, and sliding window slicing is applied to form a training dataset suitable for model training.

4. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 1, characterized in that, Constructing a deep neural network model based on a long short-term memory network, and training the deep learning model using a training dataset, specifically including: Build a deep neural network model containing one or more LSTM units, each layer containing multiple hidden units, supporting multi-step prediction output; during the training phase, use the mean squared error loss function and the Adam optimizer for training.

5. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 4, characterized in that, Constructing a deep neural network model based on Long Short-Term Memory (LSTM) networks specifically includes: The deep neural network model adopts a two-layer stacked LSTM structure, with 64 hidden units in each layer, the activation function being tanh, and the output layer being a fully connected regression layer.

6. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 1, characterized in that, The model is trained using the training dataset, specifically including: By updating and incrementally training the deep learning model online, it can adapt to changes in operating conditions under long-term inverter operation.

7. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 1, characterized in that, Calculate whether the deviation between the predicted result and the actual operating value exceeds a preset threshold, and determine whether to trigger an early warning based on a preset time duration condition, specifically including: If the deviation between the predicted value and the actual operating value exceeds a preset threshold and continues for a preset duration, an early warning will be triggered.

8. The photovoltaic inverter fault early warning method based on long short-term memory network according to claim 1, characterized in that, The method further includes: Deploy deep learning models on edge computing devices or main control systems to achieve real-time online prediction and alarm linkage.

9. A photovoltaic inverter fault early warning system based on a long short-term memory network, characterized in that, The system includes: The dataset construction module is used to collect the operating sequence data of the photovoltaic inverter and perform data preprocessing to build the training dataset. The model building module is used to build a deep neural network model based on a long short-term memory network, and to train the deep learning model using a training dataset. It takes a sequence of historical operating parameters as input and outputs a prediction result of the operating trend for a future period of time. The fault threshold determination module is used to calculate whether the deviation between the predicted result and the actual operating value exceeds a preset threshold, and to determine whether to trigger an early warning based on a preset time duration condition. The early warning module is used to output alarm signals to the monitoring system or maintenance personnel when an early warning is triggered, so as to realize early warning of faults.

10. An electronic device, characterized in that, The device includes: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to perform the steps of a photovoltaic inverter fault early warning method based on a long short-term memory network as described in any one of claims 1 to 8.

Images