Photovoltaic power station launching method and apparatus, and storage medium and device

Through human-computer interaction and preprocessing and execution strategy generation of large language models, the problem of cumbersome and professionalism of photovoltaic power station opening steps is solved, and convenient and efficient intelligent station opening is achieved, improving user experience.

WO2024227450A1PCT designated stage expired Publication Date: 2024-11-07SHANGHAI SIGEYUAN INTELLIGENT TECH CO LTD

Patent Information

Application Number
PCT/CN2024/101760
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2024-04-29
Filing Date
2024-06-26
Publication Date
2024-11-07

AI Technical Summary

Technical Problem

The existing photovoltaic power station opening methods are cumbersome and the professionalism requirements for operators are high, resulting in the inability to open the station smoothly, seriously affecting the user experience.

Method used

Using human-computer interaction, preprocessing input information through a large language model, generating execution strategies and controlling the execution command of the photovoltaic power station. If an abnormality occurs, the processing will be cycled until the station is opened. Combined with voice input, QR code scanning or text input, use recursion The task model automatically diagnoses and resolves exceptions.

Benefits of technology

It has achieved convenient, efficient and intelligent website opening, simplified the website opening process, and improved user experience and operation efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024101760_07112024_PF_FP_ABST
    Figure CN2024101760_07112024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of photovoltaics. Disclosed are a photovoltaic power station launching method and apparatus, and a storage medium and a device. The photovoltaic power station launching method comprises: S1, inputting a photovoltaic power station launching instruction; S2, preprocessing the launching instruction or a launching abnormality, and outputting preprocessed information to a large language model; S3, the large language model generating an execution policy according to the preprocessed information, and outputting an execution command; and S4, controlling a photovoltaic power station to execute the execution command, and if the photovoltaic power station successfully completes the execution command, completing launching, or controlling the photovoltaic power station to execute the execution command, and if a launching abnormality occurs during the process of the photovoltaic power station executing the execution command, returning to S2; and executing the cycle until the launching of the photovoltaic power station is completed. The present invention solves the problem of existing photovoltaic power station launching methods involving cumbersome steps and having high professional requirements for operators, thus resulting in it not being possible to smoothly launch a station and consequently seriously affecting the user experience.
Need to check novelty before this filing date? Find Prior Art

Description

Photovoltaic power station operation method, device, storage medium and equipment Technical Field

[0001] The present invention relates to the field of photovoltaic technology, and in particular to a photovoltaic power station operation method, device, storage medium and equipment. Background Art

[0002] A photovoltaic power station refers to a power generation system that utilizes solar energy and adopts special materials such as crystalline silicon panels, inverters and other electronic components. It is connected to the power grid and transmits electricity to the grid.

[0003] As shown in Figure 1, existing remote commissioning methods for photovoltaic power plants are cumbersome and require a high level of operator expertise. Traditional commissioning requires numerous parameters, many of which are highly specialized and difficult for installers to understand, making them prone to errors. Numerous inherent constraints exist between these parameters, such as the range of settings and the impact of setting value A on value B. These constraints can lead to issues with smooth commissioning, severely impacting the user experience. With the increasing adoption and development of intelligent technology, it is necessary to develop simpler and more efficient solutions to address these issues.

[0004] Summary of the Invention

[0005] Purpose of the invention: To provide a photovoltaic power station startup method, device, storage medium and equipment to solve the above-mentioned problems existing in the prior art.

[0006] The present invention provides a photovoltaic power station startup method, comprising: S1: inputting a photovoltaic power station startup instruction; S2: preprocessing the startup instruction or startup exception, and outputting preprocessing information to a large language model; S3: the large language model generating an execution strategy according to the preprocessing information, and outputting an execution command; S4: controlling the photovoltaic power station to execute the execution command, and if the photovoltaic power station successfully completes the execution command, the startup is completed; or, controlling the photovoltaic power station to execute the execution command, and if the photovoltaic power station encounters a startup exception during the execution of the command, returning to S2, and looping until the startup of the photovoltaic power station is completed.

[0007] Preferably, the method of inputting the photovoltaic power station startup instruction includes: voice input, QR code scanning input or text input.

[0008] Preferably, the preprocessing of the station start-up instruction or station start-up exception and outputting the preprocessing information to the large language model also includes: obtaining a standard information set of the photovoltaic power station; matching the information of the standard information set of the photovoltaic power station and the station start-up instruction to construct preprocessing information that can be recognized by the large language model.

[0009] Preferably, the pre-processing information is a preliminary execution strategy for handling site opening exceptions. When a site opening exception occurs, the user manually inputs a temporary instruction to modify the site opening parameters to modify the parameters of the preliminary execution strategy. The temporary instruction is generated by the large language model to generate an execution strategy for resolving the site opening exception.

[0010] Preferably, the method of inputting the site opening instruction includes: voice input, QR code scanning input or text input, and the method of inputting the temporary instruction includes: voice input, QR code scanning input or text input.

[0011] Preferably, the preprocessing of the photovoltaic power station start-up instruction or photovoltaic power station start-up exception and outputting the preprocessing information to the large language model also includes: obtaining a photovoltaic power station standard information set; matching the information of the photovoltaic power station standard information set and the start-up exception information to construct preprocessing information that can be recognized by the large language model.

[0012] Preferably, the large language model generates an execution strategy based on the preprocessing information, and outputs an execution command, further comprising: generating a diagnostic command based on the station opening instruction or station opening anomaly.

[0013] Preferably, it further includes performing diagnostic operations on the photovoltaic power station. If the diagnosis is abnormal, the large language model generates a preliminary execution strategy for solving the corresponding abnormality based on the feedback abnormality, and performs loop processing in sequence by adopting a recursive task model. During the processing, the preliminary execution strategy is modified and adjusted through the input of the temporary instructions until all abnormalities are resolved.

[0014] Preferably, the recursive task model is used to perform cyclic processing in sequence until all abnormal situations are resolved, including: for an execution action with multiple subtasks, the execution action is disassembled, and the system recursively performs automatic execution of each subtask.

[0015] Preferably, the photovoltaic power station commissioning method further comprises: feeding back commissioning completion or commissioning abnormality to the client.

[0016] Preferably, an execution strategy for resolving site launch anomalies may be manually generated; or, some parameters may be manually set and combined with a large language model to jointly generate an execution strategy for resolving site launch anomalies.

[0017] Preferably, a virtual power station mapped to the actual power station is created before step S1, and the virtual power station includes: device name, device connection relationship, display interface, dynamic effect and device panel.

[0018] Preferably, the establishment of the large language model includes the following training process:

[0019] Acquire photovoltaic-related basic data and instruction data to construct a domain data set, and annotate the domain data set to obtain an annotated domain data set;

[0020] Training the pre-trained language model using the domain dataset to update the pre-trained language model, and training the updated pre-trained language model using the annotated domain dataset to obtain a supervised fine-tuning model;

[0021] The supervised fine-tuning model adds a classification mechanism, and then combines manual feedback to score and rank the output of the supervised fine-tuning model after adding classification, forming a reward model with a scoring mechanism;

[0022] The domain dataset is used as input to train the reinforcement learning model, the obtained output result is sent to the reward model for scoring, and the obtained answer score is fed back to the reinforcement learning model for adjustment and reinforcement of the reinforcement learning model;

[0023] The obtained adjusted and reinforced reinforcement learning model is the final large language model.

[0024] Preferably, when the domain dataset trains the pre-trained language model, part of the parameter layers of the pre-trained language model are frozen, and during training, the domain dataset only updates parameters of another part of the parameter layers in the pre-trained language model.

[0025] Preferably, when the annotated domain dataset is used to train the updated pre-trained language model, the model data is updated through a back-propagation algorithm to obtain the supervised fine-tuning model.

[0026] Preferably, when the reinforcement learning model is adjusted and reinforced, a KL divergence constraint with the supervised fine-tuning model is added to the loss function of the reinforcement learning model to prevent the reinforcement learning model from overfitting.

[0027] Preferably, the photovoltaic-related basic data includes one or more of the user manual of the photovoltaic power station, frequently asked questions and answers, technical documents, online materials, and user feedback information, and the instruction data is obtained by parsing the system instruction set and interface parameters of the photovoltaic power station.

[0028] The present invention provides a photovoltaic power station commissioning device, comprising: an input module for client inputting information; a preprocessing module for preprocessing commissioning instructions or commissioning anomalies and outputting preprocessing information to a large language model; an artificial intelligence module embedded with the large language model, the large language model generating an execution strategy based on the preprocessing information and outputting an execution command; and a control execution module for controlling the photovoltaic power station to execute the execution command issued by the artificial intelligence module and feeding back the execution result.

[0029] Preferably, the input information is a site opening instruction or a temporary instruction, wherein the execution result is completion of site opening or site opening exception. When the site opening is abnormal, the user manually inputs a temporary instruction to modify the site opening parameters, and the temporary instruction generates an execution strategy for resolving the site opening exception through the large language model.

[0030] Preferably, the preprocessing module includes: a storage unit for storing a standard information set of a photovoltaic power station; an information matching unit for matching the pre-stored standard information set of a photovoltaic power station with the information received by the preprocessing module to construct preprocessing information that can be recognized by a large language model.

[0031] Preferably, the control execution module includes a diagnostic unit for performing diagnostic operations on the photovoltaic power station and feeding back diagnostic results.

[0032] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program is configured to execute any one of the above-mentioned photovoltaic power station startup methods when running.

[0033] The present invention provides an electronic device, comprising: a memory and a processor, wherein the memory stores a computer program, wherein the processor is configured to run the computer program to execute any one of the above-mentioned photovoltaic power station startup methods.

[0034] Beneficial effect: In the embodiment of the present application, a human-computer interaction method is adopted to collect input information from the client; the input information is preprocessed by an input preprocessing module, and the preprocessed information is sent to an artificial intelligence module; the startup execution strategy is automatically analyzed and generated by the large language model, and various problems encountered during the startup of the photovoltaic power station are intelligently solved, thereby achieving the technical effect of convenient, efficient and intelligent startup, and further solving the technical problem that the existing photovoltaic power station startup method has cumbersome steps and high professional requirements for operators, resulting in an inability to start the station smoothly, thereby seriously affecting the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] FIG1 is a schematic diagram of an existing remote photovoltaic power station control;

[0036] FIG2 is a schematic diagram of a photovoltaic power station start-up device according to the present application;

[0037] FIG3 is a flow chart of a method for commissioning a photovoltaic power station according to the present application;

[0038] FIG4 is a schematic diagram of the process of establishing a large language model according to the present application;

[0039] Figure 5 is a schematic diagram comparing the graphics card memory usage of the traditional training method and the graphics card memory usage after using ZeRO graphics memory optimization technology;

[0040] FIG6 is a schematic diagram of a workflow of a photovoltaic power station startup method according to one embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0042] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0043] Furthermore, the terms "installed," "disposed," "provided with," "connected," "connected," and "socketed" should be interpreted broadly. For example, they can refer to fixed connections, removable connections, or integral structures; mechanical connections or electrical connections; direct connections or indirect connections through an intermediary; or internal communication between two devices, elements, or components. Those skilled in the art will understand the specific meanings of these terms in the present invention based on specific circumstances.

[0044] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0045] In order to achieve the above-mentioned purpose, the present application provides a photovoltaic power station startup device. As shown in Figure 2, the photovoltaic power station startup device includes: an input module, a preprocessing module, an artificial intelligence module and a control execution module. Among them, the input module is used for the client to input information, and the input module can be various terminal devices such as the user's mobile phone and computer. The input module is connected to the preprocessing module. After receiving the information input by the input module, the preprocessing module preprocesses it and outputs the preprocessed information to the artificial intelligence module; the artificial intelligence module is embedded with a large language model, and the large language model generates an execution strategy based on the preprocessing information and outputs an execution command; the control execution module is used to control the photovoltaic power station to execute the execution command issued by the artificial intelligence module and feedback the execution result.

[0046] Among them, the execution result is the completion of the site opening or the site opening exception. When the site opening is abnormal, the user manually inputs a temporary instruction to modify the site opening parameters, and the temporary instruction generates an execution strategy for resolving the site opening exception through the large language model.

[0047] This application uses relevant data from the photovoltaic field to train the large language model to improve the model's understanding and application capabilities of professional knowledge in the photovoltaic field. The optimized large language model is accurately and effectively embedded in the photovoltaic station commissioning process to achieve intelligent interaction between users and the system, provide intelligent problem identification and solving functions, and enable users to ask questions or instructions in natural language. The system can accurately understand and give corresponding answers or operational suggestions.

[0048] The power station commissioning method described in this application can be applied to all aspects of the photovoltaic power station commissioning process, including equipment fault diagnosis, user consultation and answering, data monitoring and analysis, and other aspects. By integrating the large language model with the photovoltaic power station commissioning process, it can achieve the rapid application and promotion of intelligent technologies in the photovoltaic field, providing strong support for the intelligent development of the photovoltaic field.

[0049] In one embodiment, as shown in FIG4 , the establishment of the large language model includes the following training process:

[0050] Acquire photovoltaic-related basic data and instruction data to construct a domain data set, and annotate the domain data set to obtain an annotated domain data set;

[0051] Training the pre-trained language model using the domain dataset to update the pre-trained language model, and training the updated pre-trained language model using the annotated domain dataset to obtain a supervised fine-tuning model;

[0052] The supervised fine-tuning model adds a classification mechanism, and then combines manual feedback to score and rank the output of the supervised fine-tuning model after adding classification, forming a reward model with a scoring mechanism;

[0053] The domain dataset is used as input to train the reinforcement learning model, the obtained output result is sent to the reward model for scoring, and the obtained answer score is fed back to the reinforcement learning model for adjustment and reinforcement of the reinforcement learning model;

[0054] The obtained adjusted and reinforced reinforcement learning model is the final large language model.

[0055] In this embodiment, the establishment of the domain dataset is data preparation, which includes data collection, data cleaning, data labeling, data processing and dataset division. Specifically:

[0056] Data Collection: First, determine the source of data collection. PV-related basic data may include, but is not limited to, PV system user manuals, FAQs, technical documentation, online resources (such as forum posts, professional websites, forums, and social media related to PV systems, which capture general domain knowledge and provide more real-time, specific Q&A), and user feedback. Command data is obtained by parsing the system command set and interface parameters of a specific PV power plant.

[0057] Take the common questions and answers as an example to construct the basic data. For example, the questions corresponding to the natural language input are used as the content, and the corresponding answers are used as the target. The following are some examples of content and corresponding targets:

[0058] Taking the system instruction set and interface parameters of a photovoltaic power station as an example, the corresponding system instructions are input in natural language as the content, and the corresponding interface parameters are used as the target. The following are some examples of content and corresponding targets:

[0059] Data cleaning: The collected basic data may contain a large amount of noise and invalid information, so data cleaning and preprocessing are required. Data cleaning includes steps such as removing duplicate data, correcting erroneous information, and filtering out irrelevant information to ensure data quality and accuracy.

[0060] Data labeling: After data cleaning, it needs to be labeled to accurately understand the meaning and relationships of the data during model training. Data labeling may include pairing questions and answers, classifying and labeling instruction data, etc., to ensure that the labeled domain dataset can provide effective supervision signals for the model.

[0061] Data processing: Depending on the specific training requirements of each model, the data may need to be further processed and converted. Such further processing and conversion may include operations such as word segmentation, vectorization, and serialization of text data to facilitate model input and processing.

[0062] Dataset Partitioning: Finally, the domain dataset needs to be divided into training, validation, and test sets. Similarly, the labeled domain dataset also needs to be divided into training, validation, and test sets for model training, evaluation, and testing. Typically, the training set is used for model training, the validation set is used to adjust the model's hyperparameters and structure, and the test set is used to ultimately evaluate the model's performance and generalization capabilities.

[0063] Through the above steps, data related to photovoltaic system questions and answers or instructions can be effectively prepared, providing a reliable foundation for subsequent model training.

[0064] Preferably, when the domain dataset is used to train the pre-trained language model, part of the parameter layers of the pre-trained language model are frozen. During training, the domain dataset only updates the parameters of another part of the parameter layers in the pre-trained language model. Then, when the updated pre-trained language model is trained using the annotated domain dataset, the model data is updated using the back propagation algorithm to obtain the supervised fine-tuning model.

[0065] When the reinforcement learning model is adjusted and reinforced, a KL divergence constraint with the supervised fine-tuning model is added to the loss function of the reinforcement learning model to prevent the reinforcement learning model from overfitting.

[0066] During the training of large language models, each graphics card is primarily occupied by two types of data: model state data and data other than model state. Model state data consists of model parameters (fp16), model gradients (fp16), and Adam state (fp32 momentum and fp32 variance). The remaining data consists of activation values, temporary buffers, and some graphics memory fragments. Among the model state parameters, optimizer state parameters account for the largest proportion. To optimize model state storage, ZeRO (Zero Redundancy Optimizer) technology is used to split the model parameters into multiple shards, allowing each worker node to store only a portion of the parameters. Specifically, the optimizer state is split 1 / N and distributed across N graphics cards. During forward and backward propagation, only the gradients are needed, not the optimizer. The remaining gradients are concatenated to form the complete gradient for backward propagation. After forward and backward propagation are complete, the optimizer is required to update the parameters. At this point, the optimizer parameters and complete gradients from multiple graphics cards are aggregated to perform the parameter update.

[0067] Figure 5 compares GPU memory usage with traditional training methods and after using ZeRO memory optimization technology. Generally speaking, in traditional data parallel processing, each process uses the same parameters for training and maintains a complete copy of the optimizer states, which also consumes a large amount of memory (1 Float16 = 2 bytes, 1 Float32 = 4 bytes). In a mixed-precision scenario, taking a model with Ψ as an example and the Adam optimizer, Adam needs to store: 1) a copy of the parameters and gradients in Float16 precision, each of which takes up 2Ψ bytes of memory; and 2) a copy of the parameters, momentum, and variance in Float32 precision, corresponding to three copies of 4Ψ, for a total memory usage of 2Ψ + 2Ψ + 3*4Ψ = 16Ψ bytes. Using ZeRO-1 technology, this can be converted to (2Ψ + 2Ψ + 3*4Ψ / n) bytes of memory. For example, when n = 3, only 8Ψ bytes of memory are required. In addition, in order to improve training efficiency and solve the problem of floating-point underflow, mixed precision is used for training. During training, the gradients, parameters, and data used in forward propagation and backpropagation are all float16 type, and the optimizer saves a copy of float32 type parameters. When updating the parameters, the update results are saved as float32 type parameters, and then the float32 type parameters are converted to float16 type for forward propagation and backpropagation.

[0068] Users can activate the app and enter the system settings through the client's input module. This activation method includes, but is not limited to, voice input, QR code scanning, or text input. Voice input involves users entering natural language commands, such as "open site, open site, or open site." The app then collects keywords from these specific natural language commands in real time, enabling accurate and efficient activation.

[0069] Specifically, the user wakes up the built-in AI assistant function in the app through voice wake-up, and the AI ​​assistant makes voice announcements, such as: Welcome users to use the AI ​​assistant; the user installer interacts with the AI ​​assistant through voice, telling it that he wants to open a site or open a site, and then the system enters the initialization state.

[0070] The preprocessing module includes: a storage unit for storing a standard information set of a photovoltaic power station, which includes user information, photovoltaic power station documents, product descriptions, photovoltaic industry technical information, etc.; an information matching unit for matching the pre-stored standard information set of a photovoltaic power station with the information received by the preprocessing module to construct preprocessed information that can be recognized by a large language model.

[0071] As shown in Figure 6, the AI ​​assistant collects voice input information and uses the app's built-in context (context semantics) model to match the input natural language with the content of the photovoltaic power station standard information set in the preset photovoltaic database, verify or correct the information, thereby obtaining more accurate information and improving the accuracy of subsequent model processing. At the same time, an input preprocessing module is used to combine user input with the system context and perform a series of processing to obtain more accurate and concise user input information. This is used to match the pre-stored photovoltaic power station standard information set with the information received by the preprocessing module, construct preprocessed information that can be recognized by the large language model, and send this processed information to the large language model for data processing. The large language model refers to a large language model, including but not limited to CHATGPT, and can be other models such as Alibaba Cloud Tongyi Qianwen Large Language Model, Huawei Pangu AI Large Model, etc. The large language model can automatically generate an execution strategy based on the preprocessed information and output an execution command.

[0072] The control execution module connects to the application programming interface (API) of a terminal, cloud, or on-site management device, controls the photovoltaic power station to execute the execution commands issued by the artificial intelligence module, and provides feedback on the execution results. The control execution module includes a series of components, including a photovoltaic inverter. The control execution module controls the photovoltaic power station to execute the execution commands. If the photovoltaic power station successfully completes the execution commands, the station is started. Preferably, the control execution module also includes a diagnostic unit for performing diagnostic operations on the photovoltaic power station and providing feedback on the diagnostic results.

[0073] The photovoltaic power station commissioning device of the present invention implements a system in which, during the commissioning process, the user feeds back information to the AI ​​assistant through natural language, and the AI ​​assistant feeds back the input information to the data processing model; a large language model generates a commissioning strategy and outputs a commissioning execution command, and adopts human-computer interaction with a natural language input device. There is no need to manually input various parameters following the complicated commissioning process, and the App intelligently and automatically completes each commissioning process, which not only simplifies the remote commissioning process but also greatly improves the efficiency of commissioning and the user experience.

[0074] According to another embodiment of the present invention, a photovoltaic power station commissioning method is provided, comprising:

[0075] S1: Enter the PV power station startup instructions. Users can choose voice input, QR code scanning, or text input. Specifically, the user uses voice to wake up the app's built-in AI assistant, which will announce something like "Welcome to the AI ​​Assistant." The user or installer uses voice interaction to tell the AI ​​assistant "I want to start a station" or "Start a station," which will initiate the system initialization state.

[0076] S2: Preprocess the station opening instructions or station opening exceptions and output the preprocessing information to the large language model;

[0077] The user-entered station command information is preprocessed through the preprocessing module, and the voice input information is collected through the AI ​​assistant. The app's built-in context (context semantics) model is used to match the input natural language with the content in the preset photovoltaic database, verify or correct the information, thereby obtaining more accurate information and improving the accuracy of subsequent model processing. At the same time, the input preprocessing module (input preproccesser) is used to combine user input with the system context and perform a series of processing to obtain more accurate and concise user input information. The processed information is then sent to the large language model. The large language model refers to a large language model, including but not limited to CHATGPT, and can be other models such as Alibaba Cloud Tongyi Qianwen Large Language Model, Huawei Pangu AI Large Model, etc.

[0078] According to an embodiment of the present invention, preferably, the input information is preprocessed by an input preprocessing module, and the preprocessed information is sent to a large language model, including: obtaining a standard information set of a photovoltaic power station; matching the information of the standard information set of the photovoltaic power station and the station start-up instruction to construct preprocessed information that can be recognized by the large language model.

[0079] S3: The large language model generates an execution strategy based on the preprocessed information and outputs execution commands;

[0080] The large language model CHATGPT generates a startup execution strategy based on pre-processed information such as maximum power point and other parameters, and converts the execution strategy into executable tasks such as photovoltaic power station startup steps, parameter debugging, and value setting, and outputs execution commands.

[0081] S4: Control the photovoltaic power station to execute the execution command. If the photovoltaic power station successfully completes the execution command, the station is started; or, control the photovoltaic power station to execute the execution command. If the photovoltaic power station encounters a station start-up exception during the execution of the command, return to S2 and execute the loop until the photovoltaic power station is started.

[0082] Among them, when the site opening exception occurs, the user manually inputs a temporary instruction to modify the site opening parameters. The temporary instruction is generated by the large language model to generate an execution strategy for solving the site opening exception. The way to input the temporary instruction includes: voice input, QR code scanning input or text input.

[0083] After receiving the execution command output by the artificial intelligence module, the control execution module controls the photovoltaic power station to execute the execution command. If the photovoltaic power station successfully completes the execution command, the station is commissioned. The control execution module can feedback the commissioning completion information to the client.

[0084] After receiving the execution command output by the artificial intelligence module, the control execution module controls the photovoltaic power station to execute the command. If the photovoltaic power station encounters a startup anomaly during the execution of the command, such as a network disconnection, insufficient voltage, or system upgrade, the control execution module reports the technical issue to the preprocessing module. The preprocessing module retrieves the photovoltaic power station standard information set and matches the startup anomaly information with the information in the photovoltaic power station standard information set to construct preprocessing information recognizable by the large language model to form a preliminary execution strategy. The large language model CHATGPT generates a strategy to resolve the network disconnection based on the preprocessing information, such as the network disconnection. It converts the request for network connection into executable tasks such as the photovoltaic power station's action process and value settings, and outputs an execution command. The control execution module initiates the network connection settings based on the network connection execution command to resolve the network disconnection. After the network disconnection is resolved, the system continues to automatically execute the startup action until the startup is complete. During the process of resolving the startup anomaly, the preliminary execution strategy may require further modification and adjustment. This can be achieved by manually inputting temporary instructions. The temporary instructions can be modified by extracting keywords from the large language model. The user manually inputs the temporary instruction to modify the parameters of the preliminary execution strategy to generate a final execution strategy that resolves the site opening anomaly.

[0085] The volt-ampere characteristic of a photovoltaic power station is between that of a voltage source and a current source. It is a very unstable new energy source, and its output power is very sensitive to the environment, climate, and load. At present, photovoltaic power stations are generally started by repeatedly trying to turn the photovoltaic inverter on and off. The photovoltaic inverter can only operate normally when the output power of the photovoltaic cell array can meet the minimum power required for the operation of the photovoltaic inverter. The large language model can generate diagnostic commands based on the station start-up instructions or station start-up anomalies, and perform diagnostic operations on the photovoltaic power station. If the diagnosis is abnormal, the large language model generates an execution strategy to resolve the corresponding anomaly based on the feedback of the anomaly, and uses a recursive task model to perform loop processing in sequence until all anomalies are resolved. In this embodiment, the recursive task model performs loop processing in sequence until all anomalies are resolved, including: for an execution action with multiple subtasks, the execution action is disassembled, and the system recursively performs automatic execution of each subtask.

[0086] Another embodiment of the photovoltaic power station commissioning method of the present invention can also report commissioning anomalies to the client. Users can choose to manually generate an execution strategy to resolve commissioning anomalies; alternatively, they can manually set certain parameters and combine them with a large language model to jointly generate an execution strategy to resolve commissioning anomalies.

[0087] The photovoltaic power station startup method of the present invention can also create a virtual power station mapped to the actual power station before the user inputs the photovoltaic power station startup instruction. The virtual power station includes: device name, device connection relationship, display interface, dynamic effect and device panel, thereby completing the complete interface display of the photovoltaic power station startup, providing a basis for subsequent convenient operations, and facilitating users to view the photovoltaic power station startup progress and handle encountered situations in real time.

[0088] From the above description, it can be seen that this application achieves the following technical effects:

[0089] In an embodiment of the present application, a human-computer interaction method is adopted to collect input information from the client; the input information is preprocessed by an input preprocessing module, and the preprocessed information is sent to an artificial intelligence module; the startup execution strategy is automatically analyzed and generated by the large language model, and various problems encountered during the startup of the photovoltaic power station are intelligently solved, thereby achieving the technical effect of convenient, efficient and intelligent startup, and further solving the technical problem that the existing photovoltaic power station startup method has cumbersome steps and high professional requirements for operators, resulting in an inability to start the station smoothly, thereby seriously affecting the user experience.

[0090] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program is configured to execute a photovoltaic power station startup method when running.

[0091] Computer readable storage medium:

[0092] If the module integrated into the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned embodiment method by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the above-mentioned steps of displaying the above-mentioned memory power consumption optimization and processing the memory power consumption optimization.

[0093] The term "computer program" includes computer program code, which may be in source code, object code, executable file, or some intermediate form. Computer-readable media may include any entity or device capable of carrying computer program code, recording media, USB flash drives, removable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunications signals, and software distribution media. It should be noted that the content of computer-readable media may be appropriately expanded or reduced based on the requirements of legislation and patent practice within a jurisdiction. For example, in some jurisdictions, based on legislation and patent practice, computer-readable media does not include electric carrier signals or telecommunications signals.

[0094] One embodiment of the present application further provides an electronic device for storing the photovoltaic power station commissioning method described above. The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the computer program to perform the photovoltaic power station commissioning method.

[0095] Obviously, those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computing device. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0096] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the scope of protection of the present invention.

Claims

1. A photovoltaic power station startup method, characterized in that: include: S1: Input the photovoltaic power station start-up command; S2: pre-process the station opening instruction or station opening exception, and output the pre-processing information to the large language model; S3: The large language model generates an execution strategy based on the preprocessing information and outputs the execution command; S4: Control the photovoltaic power station to execute the execution command. If the photovoltaic power station successfully completes the execution command, the station startup is completed; or, control the photovoltaic power station to execute the execution command. If the photovoltaic power station encounters a station startup exception during the execution of the command, return to S2 and execute in a loop until the photovoltaic power station startup is completed.

2. The photovoltaic power station startup method according to claim 1, characterized in that: The preprocessing of the station opening instruction or station opening exception and outputting the preprocessing information to the large language model also includes: Obtain the standard information set of photovoltaic power plants; The information of the photovoltaic power station standard information set and the station start-up instruction are matched to construct pre-processed information that can be recognized by the large language model.

3. The photovoltaic power station startup method according to claim 1, characterized in that: The preprocessing of the photovoltaic power station start-up instruction or photovoltaic power station start-up exception and outputting the preprocessing information to the large language model further includes: Obtain the standard information set of photovoltaic power plants; The information of the photovoltaic power station standard information set and the station start-up abnormal information are matched and processed to construct pre-processed information that can be recognized by the large language model.

4. The photovoltaic power station startup method according to claim 3, characterized in that: The preprocessing information is a preliminary execution strategy for handling site opening exceptions. When a site opening exception occurs, a user manually inputs a temporary instruction to modify site opening parameters to modify the parameters of the preliminary execution strategy. The temporary instruction is generated by the large language model to generate an execution strategy for resolving the site opening exception.

5. The photovoltaic power station startup method according to claim 1, characterized in that: The method of inputting the site opening instruction includes: voice input, QR code scanning input or text input, and the method of inputting the temporary instruction includes: voice input, QR code scanning input or text input.

6. The photovoltaic power station startup method according to claim 1, characterized in that: The large language model generates an execution strategy according to the preprocessing information, and outputs an execution command, which further includes: generating a diagnosis command according to the station opening instruction or station opening exception.

7. The photovoltaic power station startup method according to claim 6, characterized in that: It further includes performing diagnostic operations on the photovoltaic power station. If the diagnosis is abnormal, the large language model generates a preliminary execution strategy for solving the corresponding abnormality based on the feedback abnormal situation, and performs loop processing in sequence by adopting a recursive task model. During the processing, the preliminary execution strategy is modified and adjusted through the input of the temporary instruction until all abnormal situations are resolved.

8. The photovoltaic power station startup method according to claim 7, characterized in that: The recursive task model is used to perform cyclic processing in sequence until all abnormal situations are resolved, including: For an execution action with multiple subtasks, the execution action is disassembled, and the system recursively executes each subtask automatically.

9. The photovoltaic power station startup method according to claim 1, characterized in that: Before step S1, a virtual power station mapped with the actual power station is created, and the virtual power station includes: device name, device connection relationship, display interface, dynamic effect and device panel.

10. The photovoltaic power station startup method according to claim 1, characterized in that: The establishment of the large language model includes the following training process: Acquire photovoltaic-related basic data and instruction data to construct a domain data set, and annotate the domain data set to obtain an annotated domain data set; The pre-trained language model is trained by the domain dataset to update the pre-trained language model, and the updated pre-trained language model is trained by the labeled domain dataset to obtain a supervised fine-tuning model; The supervised fine-tuning model adds a classification mechanism, and then combines manual feedback to the output of the supervised fine-tuning model after the classification is added for scoring and ranking, thereby forming a reward model with a scoring mechanism; The domain dataset is used as input to train the reinforcement learning model, the obtained output result is sent to the reward model for scoring, and the obtained answer score is fed back to the reinforcement learning model for adjusting and strengthening the reinforcement learning model; The obtained adjusted and reinforced reinforcement learning model is the final large language model.

11. The photovoltaic power station startup method according to claim 10, characterized in that: When the domain dataset is used to train the pre-trained language model, part of the parameter layers of the pre-trained language model are frozen, and when training is performed, the domain dataset only updates parameters of another part of the parameter layers in the pre-trained language model.

12. The photovoltaic power station startup method according to claim 10 or 11, characterized in that: When the annotated domain dataset is used to train the updated pre-trained language model, the model data is updated through a back-propagation algorithm to obtain the supervised fine-tuning model.

13. The photovoltaic power station startup method according to claim 10, characterized in that: When the reinforcement learning model is adjusted and reinforced, a KL divergence constraint with the supervised fine-tuning model is added to the loss function of the reinforcement learning model to prevent the reinforcement learning model from overfitting.

14. The photovoltaic power station startup method according to claim 10, characterized in that: The photovoltaic-related basic data includes one or more of the user manual of the photovoltaic power station, frequently asked questions and answers, technical documents, network materials, and user feedback information. The instruction data is obtained by parsing the system instruction set and interface parameters of the photovoltaic power station.

15. Photovoltaic power station start-up device, characterized in that: include: Input module, used for client to input information; The preprocessing module preprocesses the station opening instructions or station opening exceptions and outputs the preprocessing information to the large language model; An artificial intelligence module is embedded with the large language model, and the large language model generates an execution strategy according to the preprocessing information and outputs an execution command; The control execution module controls the photovoltaic power station to execute the execution command issued by the artificial intelligence module and feeds back the execution result.

16. The photovoltaic power station startup device according to claim 15, characterized in that: The input information is a site opening instruction or a temporary instruction, wherein the execution result is completion of site opening or site opening exception. When the site opening is abnormal, the user manually inputs a temporary instruction to modify the site opening parameters, and the temporary instruction generates an execution strategy for solving the site opening exception through the large language model.

17. The photovoltaic power station startup device according to claim 15, characterized in that: The preprocessing module comprises: A storage unit, used for storing a photovoltaic power station standard information set; The information matching unit is used to match the information received by the pre-processing module with the pre-stored photovoltaic power station standard information set to construct pre-processed information that can be recognized by the large language model.

18. The photovoltaic power station startup device according to claim 15, characterized in that: The control execution module includes a diagnosis unit, which is used to perform diagnosis operations on the photovoltaic power station and feed back diagnosis results.

19. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the photovoltaic power station startup method according to any one of claims 1 to 15 when running.

20. Electronic equipment, including: A memory and a processor, characterized in that a computer program is stored in the memory, wherein the processor is configured to run the computer program to execute the photovoltaic power station startup method according to any one of claims 1 to 15.

Citation Information

Patent Citations

  • GNSS receiver control method and device based on dialect recognition, equipment and medium

    CN112233671A

  • Multi-assistant natural language input processing

    US20210090575A1

Cited By

  • Numerical reasoning large model fine tuning method, system and equipment and storage medium

    CN120409708A

  • Process parameter optimization diagnosis method fusing large language model and structured model

    CN120805035A

  • A diagnostic method for process parameter optimization that integrates large language models and structured models

    CN120805035B