Remote photovoltaic power generation operation and maintenance management and control system based on cloud technology

Abstract

The invention relates to the technical field of photovoltaic power generation operation and maintenance, and discloses a remote photovoltaic power generation operation and maintenance management and control system based on a cloud technology, which constructs a digital twin virtual model of a photovoltaic power station, realizes multi-person cooperative remote operation, establishes an intention reasoning conflict management system, and deploys a multi-level conflict solving system. And an augmented reality remote cooperation technology is integrated. Through the system, the fault processing efficiency is improved, the multi-person cooperation ability is enhanced, cooperation conflicts are reduced, immersive operation and maintenance experience is achieved, the system reliability is improved, many technical problems existing in traditional photovoltaic operation and maintenance are solved, intelligent management and optimization of the whole life cycle of a photovoltaic power station can be achieved, and the system reliability is improved. The energy utilization efficiency and the economic benefit are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation operation and maintenance, and more specifically, it relates to a remote photovoltaic power generation operation and maintenance control system based on cloud technology. Background Art

[0002] With the continuous expansion of the scale of the photovoltaic industry and the continuous expansion of the distribution range, the efficient operation and maintenance of photovoltaic power stations face many challenges. The traditional operation and maintenance of photovoltaic power stations mainly adopts a combination of on-site operation and maintenance and remote monitoring, but there are many technical problems in such an operation and maintenance mode:

[0003] The traditional photovoltaic operation and maintenance interface is abstract and complex, and often uses forms such as two-dimensional tables and curve charts to display the operation status and fault information of equipment. It is difficult for operation and maintenance personnel to intuitively understand the equipment status and fault location, resulting in a long time for fault understanding and decision-making, and affecting the processing efficiency. Especially in large-scale or distributed photovoltaic power stations, the number of equipment is large and widely distributed, and it is difficult to form a global understanding of the complex system only by planar data representation.

[0004] The efficiency of remote guidance for on-site repair is low. When complex faults occur on-site and expert guidance is required, the traditional operation and maintenance system usually can only provide limited remote support through voice or video calls, and cannot achieve accurate visual marking and operation guidance, resulting in low communication efficiency and easy misunderstanding and deviation in the repair process. It is difficult for the expert team to conduct real-time collaborative diagnosis and decision-making, prolonging the fault response time.

[0005] When multiple people cooperate to operate the system, it is difficult to effectively prevent and solve intention conflicts and operation interferences. The traditional operation and maintenance system lacks the ability to understand the operation intentions of users. When multiple experts operate on a certain problem at the same time, intention conflicts and operation interferences are likely to occur, affecting the cooperation efficiency and decision-making quality.

[0006] The traditional operation and maintenance system lacks a mechanism for multiple people to cooperate simultaneously, and it is difficult for remote expert teams to integrate the professional knowledge of multiple parties to solve complex problems. Especially in the operation and maintenance of large-scale photovoltaic power station groups across regions and organizations, this problem is more prominent.

[0007] Existing technologies have tried to use means such as video conferencing and remote monitoring to improve the effect of remote operation and maintenance, but these technologies are usually independent of each other and lack an effective integration and coordination mechanism. At the same time, with the development of emerging technologies such as digital twin, cloud computing, and augmented reality, new possibilities are provided to solve the above problems, but there is currently no systematic solution for the field of photovoltaic power generation operation and maintenance to organically combine these technologies and solve the conflict management problem in the process of multi-person cooperation.

[0008] Therefore, there is an urgent need for a remote photovoltaic power generation operation and maintenance control system that can integrate technologies such as digital twins, intention reasoning, conflict management, and augmented reality to improve operation and maintenance efficiency and solve the above technical problems. Summary of the Invention

[0009] The present invention provides a remote photovoltaic power generation operation and maintenance control system based on cloud technology, which solves the technical problems in related technologies such as abstract and complex photovoltaic operation and maintenance interfaces, low efficiency of remote guidance for on-site maintenance, intention conflicts in multi-person collaborative operations, and lack of multi-person simultaneous collaboration mechanisms.

[0010] The present invention provides a remote photovoltaic power generation operation and maintenance control system based on cloud technology, including:

[0011] A digital twin virtual model module that constructs virtual models of photovoltaic modules, busbar boxes, inverters, and transformers using multi-physics field simulation technology, and collects real-time data through Internet of Things devices to drive the update of the virtual model status;

[0012] A multi-person collaborative remote operation module that uses an operation transformation algorithm to ensure distributed consistency, controls collaborative behaviors through permission management, and supports voice, video, and annotation sharing;

[0013] An intention reasoning conflict management module that infers potential intentions from user behavior sequences and contexts based on a Bayesian intention reasoning model, evaluates multi-user intention Figure 1 consistency, and analyzes potential conflict points;

[0014] A multi-level conflict resolution module that adopts automatic reconciliation, provides suggestions, or guides negotiation resolution strategies according to the severity and type of conflicts;

[0015] An augmented reality remote collaboration module that combines augmented reality with the digital twin model to support remote experts in providing precise guidance to on-site maintenance personnel.

[0016] Further, in the construction of the virtual model, each device is represented by a unique identifier, the device status set contains multiple status parameters, and the preprocessed real-time data is mapped to the corresponding status parameters of the virtual model through a data mapping function.

[0017] Further, the system also includes a visualization engine. For continuous parameters such as temperature and power, a heat map rendering algorithm is applied to generate a heat map covering the surface of the virtual model, and the status parameter values are converted into corresponding RGBA color values through a color mapping function.

[0018] Further, in the operation transformation algorithm, operations are defined. When multiple users issue operations simultaneously, the equivalent operation of one operation after another operation is calculated through an operation transformation function, and an operation transformation path is constructed through an operation history record and a status vector.

[0019] Furthermore, the permission management adopts a fine-grained permission management model, where permissions are defined as a function of users, operation objects, and operation behaviors, and the function of users, operation objects, and operation behaviors is calculated based on user roles, time constraints, and context conditions.

[0020] Furthermore, the Bayesian intention inference model is calculated through the relationship between posterior probability, likelihood probability, and prior probability, and infers potential intentions from the user behavior sequence and context.

[0021] Furthermore, the intention model library in the field of photovoltaic operation and maintenance contains a predefined set of intentions, and optimizes the log-likelihood function based on historical operation data through the maximum likelihood estimation method.

[0022] Furthermore, the multi-level conflict resolution system includes a conflict resolution strategy selection function, which selects corresponding resolution strategies according to the severity and type of conflicts.

[0023] Furthermore, the augmented reality remote collaboration technology uses a hybrid marker strategy to achieve spatial registration, identifies preset markers from the on-site image stream through a marker recognition algorithm, establishes a transformation matrix between coordinate systems, and accurately superimposes virtual content onto the view of the actual device.

[0024] A storage medium stores non-transitory computer-readable instructions that, when executed by a computer, can execute the modules in the above-mentioned remote photovoltaic power generation operation and maintenance control system based on cloud technology.

[0025] The beneficial effects of the present invention are as follows:

[0026] By constructing a remote photovoltaic power generation operation and maintenance control system based on cloud technology, the present invention solves technical problems such as complex interface abstraction, low efficiency of remote guidance, multi-person collaborative conflict management, and lack of multi-person collaboration mechanism in traditional photovoltaic operation and maintenance, and achieves multiple technical effects.

[0027] Through the application of the digital twin virtual model and visualization technology, the system provides an intuitive three-dimensional display of device status, enabling operation and maintenance personnel to quickly understand the device status and fault location, and reducing the fault diagnosis time.

[0028] The multi-person collaborative remote operation system based on the operation conversion algorithm and the permission management system solves the inconsistency problem caused by multi-person simultaneous operation and enhances the team collaboration ability.

[0029] Through the Bayesian intention inference model and the multi-level conflict resolution system, the system can actively identify and manage potential conflicts in the multi-user collaboration process, reduce the incidence rate of collaboration conflicts, and improve the team collaboration efficiency.

[0030] The integration of augmented reality remote collaboration technology enables precise visual guidance from remote experts to on-site maintenance personnel, improves the remote processing rate, and reduces the need for experts to travel on-site. Description of the Drawings

[0031] Figure 1 It is a module diagram of a remote photovoltaic power generation operation and maintenance management and control system based on cloud technology of the present invention. Detailed Implementation Manner

[0032] Now, the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0033] In at least one embodiment of the present invention, a remote photovoltaic power generation operation and maintenance management and control system based on cloud technology is disclosed, as Figure 1 shown, including:

[0034] A digital twin virtual model module that constructs virtual models of photovoltaic modules, busbar boxes, inverters, and transformers using multi-physics field simulation technology, and collects real-time data through Internet of Things devices to drive the update of the virtual model status;

[0035] Specifically, it includes the following sub-steps:

[0036] Sub-step 1.1: Construct an accurate virtual model of a photovoltaic power station based on multi-physics field simulation technology;

[0037] Using 3D modeling technology combined with multi-physics field simulation technologies such as thermodynamics, optics, and electrical engineering, construct an accurate virtual model including devices such as photovoltaic modules, busbar boxes, inverters, and transformers, and form a digital twin basic model that can reflect actual physical characteristics. Each device in the virtual model is represented by a unique identifier indicating, where represents the device serial number, and the device status set :

[0038] ;

[0039] where represents the th state parameter, such as temperature, voltage, current, etc., represents the total number of state parameters;

[0040] Sub-step 1.2: Drive the virtual model state update with real-time data collected by IoT devices;

[0041] Deploy a data collection gateway to obtain real-time operation data of photovoltaic devices from on-site IoT devices through MQTT, Modbus or other industrial communication protocols:

[0042] ;

[0043] where represents the collected data at time point , and the collection frequency is , represents the number of time points;

[0044] Perform denoising, interpolation, and normalization on the collected data through the data preprocessing algorithm to obtain the normalized data set .

[0045] Establish a data mapping function to map the preprocessed real-time data to the corresponding state parameters of the virtual model, realizing the dynamic update of the virtual model state. The mapping function can be expressed as:

[0046] ;

[0047] where represents the th data point after preprocessing, represents the th state parameter in the virtual model, and the mapping relationship is established based on the device identifier and parameter type;

[0048] Sub-step 1.3: Intuitively display the device operation state by combining visualization technologies such as heat maps;

[0049] Based on the updated virtual model state parameters , build a visualization engine to achieve multi-dimensional data visualization. For continuous parameters such as temperature and power, apply the heat map rendering algorithm:

[0050] ;

[0051] where represents the device state set, represents the visualization result, generating a heat map covering the surface of the virtual model;

[0052] The color mapping function for heat map generation is:

[0053] ;

[0054] where is the status parameter value, is the corresponding RGBA color value, where R represents the red channel value, G represents the green channel value, and B represents the blue channel value, represents the transparency value. By setting the threshold parameter adjust the color gradient, where represents the minimum threshold, represents the warning threshold, represents the critical threshold, represents the maximum threshold, making the abnormal state more prominent visually.

[0055] For discrete parameters such as device status and alarms, use the symbol marking method:

[0056] ;

[0057] where is the status parameter, and the icon type, marking color, and marking size are visually displayed in the form of icons on the 3D model.

[0058] The multi-person collaborative remote operation module uses an operation transformation algorithm to ensure distributed consistency, controls collaborative behavior through permission management, and supports voice, video, and annotation sharing;

[0059] Specifically, it includes the following sub-steps:

[0060] Sub-step 2.1: Ensure distributed consistency based on the operation transformation algorithm;

[0061] Use the operation transformation algorithm to achieve the consistency of distributed users operating the digital twin model simultaneously. The operation is defined as:

[0062] ;

[0063] where type represents the operation type, location represents the operation location, and user represents the operating user.

[0064] When multiple users issue operations simultaneously the operation transformation function transforms the operation into the equivalent operation after the operation to ensure eventual consistency;

[0065] where represents the original operation of the th user, represents the original operation of the th user, represents the total number of users' original operations, is the transformed operation.

[0066] Through the operation history and the state vector where represents the operation sequence number of the th user, represents the total number of operation sequence numbers of the customer), construct an operation conversion path to ensure that all clients finally converge to the same state.

[0067] In this embodiment, the operation conversion algorithm is applied to the multi-person collaborative operation scenario of the digital twin model of the photovoltaic power station. For example, when operation and maintenance expert A marks a component fault point on the digital twin model :

[0068] ;

[0069] where the mark represents the mark operation type, represents operation and maintenance expert A, "temperature anomaly" represents the mark content, and at the same time operation and maintenance expert B also marks another problem at a similar position :

[0070] ;

[0071] where represents operation and maintenance expert B, "voltage fluctuation" represents the mark content, and the operation conversion algorithm will automatically handle the mutual influence of these two operations and display these two marks consistently on the interfaces of all collaborators, without any one mark being overwritten or lost due to concurrent operations. In actual implementation, different conversion strategies are designed for different types of operations:

[0072] For mark operations, a position merging strategy is adopted. When the distance between two marks is less than the threshold (representing the mark distance threshold), they are combined and displayed in a visually distinguishable manner;

[0073] For parameter adjustment operations, a weighted average strategy is adopted. According to the user weight (representing the weight value of the th user), calculate the final value of the parameter:

[0074] ;

[0075] where is the parameter value set by the th user, represents the number of users participating in the operation;

[0076] For control instruction operations, a priority strategy is adopted. Determine the operation priority according to the user role, and high-priority operations overwrite low-priority operations.

[0077] Sub-step 2.2: Control collaboration behavior through the permission management system;

[0078] Establish a fine-grained permission management model, and the permissions are defined as:

[0079] , where represents the user, represents the operation object, represents the operation behavior, represents the user role, represents the time constraint, represents the context condition.

[0080] The permission check algorithm executes the following logic: When the user attempts to perform an operation on the object and at this time, the system calculates the permission function based on the user role , the current time and the context . The result is a boolean value indicating whether the operation is allowed.

[0081] Through the role hierarchy (where represents the permission inclusion relationship, represents the total number of roles) and the operation set (where represents the th operation type, represents the total number of operation types), construct the permission matrix (indicating the -role pair -operation permission configuration matrix), to achieve flexible permission configuration.

[0082] Sub-step 2.3: Combine real-time communication technology to support voice, video, and annotation sharing;

[0083] Deploy a real-time communication system based on WebRTC to support peer-to-peer voice (indicating the voice data stream) and video (indicating the video data stream) transmission, and control the communication delay within (indicating the maximum allowable delay time). The video stream encoding adopts an adaptive bitrate algorithm:

[0084] ;

[0085] where is the original video stream, is the maximum bandwidth limit, is the minimum quality requirement, is the encoded video stream.

[0086] Build a shared annotation system, where the annotation is defined as:

[0087] ;

[0088] where the timestamp represents the creation time. The annotation real-time synchronization algorithm is based on the publish-subscribe mode to ensure that all users can instantly see the annotations created by other users.

[0089] The intention inference conflict management module infers potential intentions from the user behavior sequence and context based on the Bayesian intention inference model, evaluates the multi-user intention Figure 1 consistency, and analyzes potential conflict points;

[0090] Specifically, it includes the following sub-steps:

[0091] Sub-step 3.1: Infer potential intentions from the user behavior sequence and context through the Bayesian intention inference model;

[0092] Build a Bayesian intention inference model, based on the user behavior sequence (where represents the -th user behavior, represents the total number of user behaviors) and context information (representing the context information of the current operating environment) to infer the user's potential intention (representing the user's operation intention). The intention inference probability model is expressed as:

[0093] ;

[0094] where, is the posterior probability of the intention given the behavior sequence and context, is the likelihood probability of observing the behavior sequence given the intention and context, is the prior probability of the given context schematic diagram, represents proportionality.

[0095] By learning the parameters (where represents the -th model parameter, represents the total number of model parameters) from the historical operation data, optimize the likelihood function to improve the intention inference accuracy.

[0096] In this embodiment, the Bayesian intention inference model establishes an intention model specific to the field of photovoltaic operation and maintenance by analyzing the historical operation behaviors of photovoltaic operation and maintenance personnel. In the specific implementation, a set of common intention sets for photovoltaic operation and maintenance is first defined:

[0097] ;

[0098] Among them represents the fault diagnosis intention, represents the performance analysis intention, represents the maintenance plan intention, represents the emergency response intention, and the behavior patterns that the user may execute under each intention are obtained by statistically analyzing historical data.

[0099] For example, when the system observes that the operation and maintenance personnel have continuously executed operation sequences such as viewing component temperature data, viewing historical power generation, and adjusting the heat map threshold, the Bayesian intention inference model will calculate the posterior probabilities of various possible intentions:

[0100] ;

[0101] Among them represents the posterior probability that the user's intention is fault diagnosis under the condition of observing the user behavior sequence and the context ; represents the likelihood probability of observing the user execute the behavior sequence under the condition that the user's intention is fault diagnosis and the context is ; represents the prior probability that the user's intention is fault diagnosis under the context ; represents the normalization factor, which sums over all possible intentions ; represents the index in the intention set, traversing all possible intention types.

[0102] ;

[0103] Among them represents the posterior probability that the user's intention is performance analysis under the condition of observing the user behavior sequence and the context ; represents the likelihood probability of observing the user execute the behavior sequence under the condition that the user's intention is performance analysis and the context is ; represents the prior probability that the user's intention is performance analysis under the context ; denotes the normalization factor, which is the same as the above formula.

[0104] Based on the calculation results, the system determines that the most likely current intention of the user is performance analysis ( ) rather than fault diagnosis ( ), and predicts the possible subsequent operation requirements of the user based on this, and prepares relevant data or tools in advance.

[0105] The parameter learning of the model adopts the maximum likelihood estimation method, based on the labeled historical operation dataset :

[0106] ;

[0107] where denotes the th historical data sample, which includes the behavior sequence, the corresponding intention, and the context information, denotes the total number of historical data samples, and optimizes the log-likelihood function through the gradient descent algorithm:

[0108] ;

[0109] where denotes the log-likelihood function, which is used to evaluate the quality of the model parameters ; denotes the set of model parameters, which need to be learned through the optimization algorithm; denotes the total number of historical data samples; denotes the th user behavior sequence in the th historical data sample; denotes the user intention in the th historical data sample; denotes the context information in the th historical data sample; denotes the probability of observing the behavior sequence under the conditions of the given intention , context .

[0110] In practical applications, in order to handle the dynamic changes in the photovoltaic operation and maintenance scenario, the model adopts an online learning method, and continuously adjusts the parameters according to the feedback of the operation and maintenance personnel, so that the accuracy of intention inference increases with the increase of the system usage time.

[0111] Sub-step 3.2: Evaluate the consistency of multi-user intentions;

[0112] For multiple users in the system (where denotes the a user, the set of intents (representing the total number of users) where represents the th user's intent) for consistency evaluation. Construct an intent similarity matrix , where the element represents the intent similarity between user and user .

[0113] The intent similarity function is defined as:

[0114] ;

[0115] where and are the vector representations of the i-th and j-th user intents respectively, represents the dot product of the two intent vectors, represents the norm of the intent vector , represents the norm of the intent vector .

[0116] Based on the intent similarity matrix, group the users through a clustering algorithm ;

[0117] where represents the intent similarity matrix, is the similarity threshold, is the th group of users with similar intents, represents the number of groups.

[0118] Sub-step 3.3: Analyze potential conflict points;

[0119] For operations between different intent groups and , construct a conflict detection function ;

[0120] where and are operations from different intent groups respectively.

[0121] The conflict severity is classified as , and the conflict type is classified as ;

[0122] Through the prediction function ;

[0123] Based on the current intent and the operation history Predict possible future operation sequences, where represents possible future operations, represents the th possible operation in the prediction, represents the number of predicted operations, and potential conflict points are detected in advance.

[0124] Multi-level conflict resolution module, which adopts automatic reconciliation, provides suggestions or guides negotiation resolution strategies according to the severity and type of conflicts;

[0125] Specifically, it includes the following sub-steps:

[0126] Sub-step 4.1: Adopt an automatic reconciliation strategy according to the severity and type of conflicts;

[0127] Construct a conflict resolution strategy selection function ;

[0128] Select a suitable resolution strategy according to the severity and type of conflicts.

[0129] For conflicts with low severity, adopt an automatic reconciliation algorithm , merge the conflicting operations and into a consistent operation , where represents the merged operation.

[0130] The automatic reconciliation algorithm is based on operation priorities (where represents an operation, represents the priority value) and an operation dependency graph (where is a set of operations, is a set of dependency relationships), ensuring that the merged operation sequence satisfies causal consistency and intent retention.

[0131] In the photovoltaic operation and maintenance scenario, the automatic reconciliation algorithm is mainly applied to the situation where multiple operation and maintenance personnel simultaneously fine-tune the same or related equipment parameters. For example, when two operation and maintenance personnel simultaneously adjust the working mode parameters of the same inverter, the system will automatically adopt corresponding reconciliation strategies according to the conflict detection results:

[0132] For parameter setting conflicts, such as when two users set the inverter power factor to 0.95 and 0.98 respectively, the system uses a weighted average algorithm to automatically calculate the merged value, or selects the operation of the high-priority user according to the user weights;

[0133] For resource occupancy conflicts, such as two users simultaneously attempting to control the same PTZ camera, the system adopts a round-robin algorithm to allocate control time periods for different users, ensuring that each user can complete the necessary observation tasks;

[0134] For display layout conflicts, such as two users having different preferences for the layout of the same view, the system provides personalized views for each user through view cloning technology while maintaining the consistency of the underlying data.

[0135] In the specific implementation, the conflict resolution strategy selection function adopts a decision tree algorithm to construct a mapping relationship between conflict features and the best resolution strategies through a training data set. During actual operation, this function can quickly select the most appropriate solution according to the specific features of the current conflict.

[0136] Sub-step 4.2: Provide suggestions for medium-severity conflicts;

[0137] For medium-severity conflicts, the system generates a set of suggestions , where each suggestion (representing the th suggestion) includes a conflict description (representing a detailed description of the conflict), possible solutions (representing the proposed solution methods), and expected results (representing the expected effects of adopting this solution), representing the total number of suggestions.

[0138] The suggestion generation algorithm is based on a historical conflict resolution case library , where represents the th historical conflict case. Through case-based reasoning, the most similar historical case to the current conflict is found, and suggestions suitable for the current situation are generated based on its solution, representing the total number of historical conflict cases.

[0139] The suggestions are presented to the user through non-intrusive interface elements. The user can choose to accept the suggestions , and the system automatically applies the solution; or reject the suggestions, triggering a higher-level conflict resolution mechanism.

[0140] Sub-step 4.3: Guide negotiation for high-severity conflicts;

[0141] For high-severity conflicts, the system starts a guided negotiation process ;

[0142] where represents the set of conflicting users, represents the set of conflict intentions, represents the current context, and a solution is reached through negotiation.

[0143] Construct a structured negotiation interface, including the following elements:

[0144] Conflict visualization view (representing the visual display of conflicts), showing the causes and impacts of conflicts; Intent expression area (representing the interface area where users express their intentions), allowing users to clearly express their operation intentions; Solution proposal area (representing the interface area where solutions are proposed), where users can propose solutions to resolve conflicts; Voting mechanism ;

[0145] Among them represents the set of users participating in the vote, represents the set of proposals to be voted on, and the decision represents the voting result, making a collective decision on the proposed solutions.

[0146] The negotiation process is controlled by a negotiation guidance algorithm ;

[0147] Among them represents the current negotiation state, and the prompt represents the generated guidance prompt, generating appropriate guidance prompts according to the current negotiation state to help users reach a consensus.

[0148] An augmented reality remote collaboration module, which combines augmented reality with a digital twin model to support remote experts in providing precise guidance to on-site maintenance personnel;

[0149] Specifically, it includes the following sub-steps:

[0150] Sub-step 5.1: Combine augmented reality technology with the digital twin model;

[0151] Construct an augmented reality and digital twin fusion engine to achieve spatial registration of the virtual model and the actual device. Through a marker recognition algorithm ;

[0152] Among them represents the on-site image stream, represents the set of recognized markers, recognizing a preset marker set from the on-site image stream to establish a transformation matrix between the world coordinate system and the camera coordinate system ;

[0153] Based on the transformation matrix and the three-dimensional geometric information of the digital twin model (representing the geometric data of the digital twin model), calculate the augmented reality overlay layer, where represents the augmented reality overlay layer, Represents an augmented reality rendering function that precisely superimposes virtual content onto the view of the actual device.

[0154] Spatial registration accuracy evaluation function Calculates the error between the overlay and the actual device;

[0155] Where the error represents the registration error value, and when the error exceeds the threshold (representing the maximum allowable error), it triggers the re-registration process.

[0156] In the actual application of a photovoltaic power station, the augmented reality and digital twin fusion engine adopts a hybrid marking strategy, combining natural feature points and artificial marking points for registration. The specific implementation includes:

[0157] Pre-install specific AR marking patterns on key devices such as photovoltaic modules, busbar boxes, and inverters. These marks have high contrast and unique geometric features, facilitating recognition by computer vision algorithms;

[0158] For devices where marks cannot be installed, the system uses the geometric features of the device itself (such as corners, interfaces, nameplates, etc.) as natural marking points, and establishes a corresponding relationship with the corresponding reference points in the digital twin model through feature point matching algorithms;

[0159] In scenarios with poor lighting conditions or limited viewing angles, the system combines the IMU (Inertial Measurement Unit) data worn by on-site workers to achieve sensor fusion-based spatial positioning.

[0160] Mark recognition algorithm Based on the improved SIFT / ORB feature detection and RANSAC matching methods, it has been specifically optimized for the interference such as reflection and shadow unique to the photovoltaic scene, and can maintain a high recognition rate under various lighting conditions. When enough matching points are recognized, the system solves the camera pose through the PnP algorithm (Perspective-n-Point) and establishes an accurate spatial transformation relationship.

[0161] Sub-step 5.2: Support remote experts to provide precise guidance to on-site maintenance workers;

[0162] Construct a remote guidance channel, and the expert-side view (representing the operation interface of the remote expert) includes a digital twin model, a live real-time image stream, and augmented reality interaction tools ;

[0163] Where the pointer represents a virtual pointer tool.

[0164] The guidance content generation module converts the expert input (representing the operation input of the expert) into augmented reality guidance elements (indicating the generated AR guidance elements), including virtual pointers x, y, z (where , , represent three-dimensional space coordinates), the position and content of the annotation toolkit, and the type, position, and status of the three-dimensional model tool.

[0165] Guidance element synchronization algorithm Considering network latency (indicating the network transmission latency time), optimize the guidance elements (indicating the original guidance elements created by experts) for transmission and display timing, and generate optimized guidance elements (indicating the guidance elements after timing optimization), ensuring that the guidance received by on-site personnel is time-consistent and spatially accurate.

[0166] In the remote collaboration application of a photovoltaic power station, the specific implementation of the remote guidance system includes:

[0167] The expert-side interface adopts a partitioned layout. The real-time video from the first-person perspective of on-site maintenance personnel is displayed on the left, and the digital twin model of the corresponding location is displayed on the right. The two maintain a spatial correspondence relationship, and the expert can operate on either view;

[0168] The expert can draw on the interface using a stylus, gestures, or a mouse, and the system automatically converts the input on the two-dimensional screen into guidance elements in three-dimensional space. For example, when the expert draws an arrow on the screen pointing to the radiator of the inverter, the system calculates the intersection point of the ray and the three-dimensional model and accurately places a three-dimensional arrow in the actual space;

[0169] For complex operations, the expert can select standard operation animations (such as "tighten screws", "replace fuse", etc.) from the system's preset model library and position them on the on-site equipment through simple drag-and-drop. The on-site personnel can then see the animated guidance floating on the actual equipment.

[0170] To address the impact of network latency on the timing of remote guidance, the guidance element synchronization algorithm adopts timestamp compensation and predictive rendering techniques. The system monitors the network round-trip latency in real-time, adds timestamps to the guidance elements, and performs appropriate time alignment at the receiving end according to the current latency situation to ensure that the display of the guidance elements is synchronized with the on-site situation.

[0171] Sub-step 5.3: Improve communication efficiency through visual annotation, step guidance, etc.;

[0172] Build a visual annotation system that supports multiple annotation types , and the annotation is defined as:

[0173] ,

[0174] Among them, the marked position represents the position coordinates in three-dimensional space, and the importance represents the importance level of the mark.

[0175] The step guidance module is based on a predefined maintenance process template (where represents the total number of steps), and generates a customized maintenance step sequence for the current fault type (indicating the fault type that needs to be processed currently). (Among them, represents the customized maintenance step sequence). Each step includes an operation guide (indicating the detailed operation instructions for this step), required tools (indicating the list of tools required to complete this step), precautions (indicating the safety warnings and precautions for this step), and completion criteria (indicating the judgment criteria for completing this step).

[0176] The interactive feedback mechanism allows on-site maintenance personnel to provide real-time feedback on the current status through gestures (indicating the gesture input of on-site personnel) or voice (indicating the voice command input of on-site personnel). The feedback recognition function (where represents the feedback recognition function, and the feedback represents the structured feedback information after recognition) converts these inputs into structured feedback information, based on which experts can adjust the guidance content.

[0177] In an embodiment of the present invention, an application example of the foregoing remote photovoltaic power generation operation and maintenance management and control system based on cloud technology is provided:

[0178] This example focuses on the actual application of a large photovoltaic power station group. The total installed capacity of this photovoltaic power station group reaches 500 MW, including 5 photovoltaic power stations distributed in different geographical locations, with a total floor area of about 1,000 hectares and more than 100,000 pieces of equipment, including photovoltaic modules, busbar boxes, inverters, transformers, box transformers, cables, etc. The operation and maintenance team consists of a headquarters expert team (10 people) and on-site operation and maintenance personnel at each power station (3 - 5 people per station). In the traditional operation and maintenance mode, when complex faults occur in the power station, the expert team needs to travel long distances to the site for diagnosis and treatment, and the average fault response time exceeds 48 hours, seriously affecting the power generation efficiency.

[0179] After applying this technical solution, a cloud-based remote photovoltaic power generation operation and maintenance management and control system is established, realizing the efficient collaboration between the headquarters experts and on-site personnel at each power station. As shown in Table 1, the basic information and operation and maintenance requirements of the power station group are given:

[0180] Table 1 Basic Information and Operation and Maintenance Requirements of Photovoltaic Power Station Group

[0181]

[0182] Implementation process:

[0183] Construction of the digital twin virtual model system for the photovoltaic power station;

[0184] For the above-mentioned photovoltaic power station group, first, accurate three-dimensional geographical information and equipment layout data of the power station were obtained through drone aerial photography and laser scanning technology. Combining CAD drawings and BIM models, an accurate three-dimensional model containing all key equipment was constructed. As shown in Table 2, the hierarchical structure and accuracy indicators of the digital twin model are presented:

[0185] Table 2 Construction of the digital twin model

[0186]

[0187] In actual implementation, a data acquisition gateway was deployed inside the PV-03 power station, connecting various on-site devices through industrial Ethernet and wireless sensor networks, and a real-time acquisition system containing 23,890 data points was established. Using the edge computing architecture, after preliminary data processing at the site, the key information was transmitted to the cloud through 4G / 5G networks. As shown in Table 3, the data acquisition and preprocessing are presented:

[0188] Table 3 Data acquisition and preprocessing

[0189]

[0190] Implementation of the multi-person collaborative remote operation system;

[0191] A multi-person collaborative operation platform was deployed at the headquarters operation and maintenance center, equipped with a large-scale splicing display wall and multiple workstations, supporting up to 10 experts to remotely collaborate simultaneously. As shown in Table 4, the user role settings and permission configurations of the system are presented:

[0192] Table 4 User roles and permission configurations

[0193]

[0194] In the process of handling an inverter failure at the PV-04 power station on a certain occasion, the collaborative operation data recorded by the system is shown in Table 5:

[0195] Table 5 Example of collaborative operation records (inverter failure handling)

[0196]

[0197] The system processed the conflicts generated during the above collaboration process in real time. Based on the Bayesian intention reasoning model, it was recognized that the intention of equipment expert A was to "solve the problem of overheating hardware", while the intention of equipment expert B was to "adjust the power grid adaptability". Through the conflict resolution algorithm, the system automatically suggested performing a restart operation first and then adjusting the protection parameters, ultimately achieving the coordination and unity of the intentions of the two experts.

[0198] Application of augmented reality remote collaboration technology;

[0199] During the repair of a primary circuit fault at PV-02 power station, equipment experts at the headquarters guided on-site technicians through an augmented reality remote collaboration system. The on-site technicians were equipped with AR glasses and handheld tablet devices, and the experts provided guidance through a remote operation interface. As shown in Table 6, the key processes of this remote collaboration are presented:

[0200] Table 6 Record of augmented reality remote collaboration process

[0201]

[0202] Verification of technical effects:

[0203] The application of this embodiment in a photovoltaic power station group has produced certain technical effects, which are mainly reflected in two aspects: the improvement of fault handling efficiency and the enhancement of the collaboration ability of operation and maintenance personnel.

[0204] Improvement of fault handling efficiency;

[0205] By tracking and recording the fault handling data for six months before and after the implementation of the system, the efficiency comparison shown in Table 7 was obtained:

[0206] Table 7 Comparison of fault handling efficiency before and after system implementation

[0207]

[0208] Enhancement of multi-person collaboration ability;

[0209] After the implementation of the system, through questionnaire surveys and system log analysis, the changes in the collaboration ability of the operation and maintenance team were evaluated, and the results are shown in Table 8:

[0210] Table 8 Comparison of multi-person collaboration ability before and after system implementation

[0211]

[0212] The embodiments of the present invention have been described above. However, these embodiments are not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the scope of protection of this embodiment.

Claims

1. A remote photovoltaic power generation operation and maintenance control system based on cloud technology, characterized in that, Including: A digital twin virtual model module that constructs virtual models of photovoltaic modules, busbar boxes, inverters, and transformers using multi-physics field simulation technology, and collects real-time data through Internet of Things devices to drive the update of the virtual model status; A multi-person collaborative remote operation module that uses an operation transformation algorithm to ensure distributed consistency, controls collaborative behaviors through permission management, and supports voice, video, and annotation sharing; An intention inference conflict management module that infers potential intentions from user behavior sequences and contexts based on a Bayesian intention inference model, evaluates the consistency of multi-user intentions, and analyzes potential conflict points; A multi-level conflict resolution module that adopts automatic reconciliation, provides suggestions, or guides negotiation resolution strategies according to the severity and type of conflicts; An augmented reality remote collaboration module that combines augmented reality with the digital twin model to support remote experts in providing precise guidance to on-site maintenance personnel.

2. The remote photovoltaic power generation operation and maintenance control system based on cloud technology according to claim 1, wherein In the construction of the virtual model, each device is represented by a unique identifier, and the device status set contains multiple status parameters. The preprocessed real-time data is mapped to the corresponding status parameters of the virtual model through a data mapping function.

3. A remote photovoltaic power generation operation and maintenance management and control system based on cloud technology according to claim 1, characterized in that, The system also includes a visualization engine. For continuous parameters such as temperature and power, a heat map rendering algorithm is applied to generate a heat map covering the surface of the virtual model, and the status parameter values are converted into corresponding RGBA color values through a color mapping function.

4. A remote photovoltaic power generation operation and maintenance control system based on cloud technology according to claim 1, characterized in that, In the operation transformation algorithm, operations are defined. When multiple users issue operations simultaneously, the equivalent operation of one operation after another operation is calculated through an operation transformation function, and an operation transformation path is constructed through an operation history record and a state vector.

5. A remote photovoltaic power generation operation and maintenance management and control system based on cloud technology according to claim 1, characterized in that, The permission management adopts a fine-grained permission management model, and the permission is defined as a function of the user, the operation object, and the operation behavior. The function of the user, the operation object, and the operation behavior is calculated based on the user role, time constraints, and context conditions.

6. The remote photovoltaic power generation operation and maintenance control system based on cloud technology according to claim 1, characterized in that The Bayesian intention inference model is calculated through the relationship between posterior probability, likelihood probability, and prior probability, and infers potential intentions from user behavior sequences and contexts.

7. The remote photovoltaic power generation operation and maintenance control system based on cloud technology according to claim 1, wherein, The intention model library in the field of photovoltaic operation and maintenance contains a predefined set of intentions, and the log-likelihood function is optimized based on historical operation data through the maximum likelihood estimation method.

8. A remote photovoltaic power generation operation and maintenance management and control system based on cloud technology according to claim 1, characterized in that, The multi-level conflict resolution system includes a conflict resolution strategy selection function that selects corresponding resolution strategies according to the severity and type of conflicts.

9. The remote photovoltaic power generation operation and maintenance control system based on cloud technology according to claim 1, characterized in that The augmented reality remote collaboration technology adopts a hybrid marker strategy to achieve spatial registration, identifies preset markers from the on-site image stream through a marker recognition algorithm, establishes a transformation matrix between coordinate systems, and precisely superimposes virtual content on the view of the actual device.

10. A storage medium stores non - transitory computer - readable instructions, characterized in that, When the non-transitory computer-readable instructions are executed by a computer, they can execute the modules in a remote photovoltaic power generation operation and maintenance management and control system based on cloud technology as described in any one of claims 1-9.

Images