A fish-light complementary power station intelligent operation and maintenance system

Abstract

The present application relates to the field of fish-light complementary power station operation and maintenance technology, and more particularly to a fish-light complementary power station intelligent operation and maintenance system, which comprises: a multi-source data acquisition layer for acquiring multi-source data of the fish-light complementary power station and transmitting the data to an edge computing layer; the edge computing layer is used for preprocessing the multi-source data and transmitting the preprocessed multi-source data to a cloud service platform; the cloud service platform is used for operation and maintenance analysis based on the preprocessed multi-source data, and then issuing early warning and generating operation instructions according to the maintenance scheme, and transmitting the operation instructions to the operation and maintenance execution layer; the operation and maintenance execution layer is used for executing the operation instructions to realize the real maintenance of the fish-light complementary power station. The present application forms an integrated operation and maintenance perspective of fish and light based on the multi-source data of the fish-light complementary power station, can consider the optimized operation and maintenance scheme based on the multi-source data of photovoltaic power generation and breeding activities, and improves the economy and intelligent degree of the fish-light complementary power station operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance technology for solar-fishery complementary power plants, and in particular to a smart operation and maintenance system for solar-fishery complementary power plants. Background Technology

[0002] Solar-aquaculture hybrid power stations represent a comprehensive "photovoltaic+" utilization model. By installing photovoltaic arrays above aquaculture ponds, they achieve three-dimensional resource development, enabling power generation on the upper level and aquaculture on the lower level. This model offers multiple benefits, including improving land utilization efficiency, optimizing energy structure, and increasing industrial added value. However, this model also faces unique complexities during operation: the photovoltaic power generation system and the aquaculture system are closely coupled in physical space, and their operational states influence each other. Therefore, the operation and maintenance objectives must simultaneously consider both power generation and aquaculture activities.

[0003] Currently, some operation and maintenance (O&M) solutions for photovoltaic (PV) power plants exist, but there are shortcomings in the O&M management of solar-aquaculture hybrid power plants. Firstly, traditional PV power plant O&M solutions focus on monitoring power generation equipment, lacking consideration for aquaculture activities, resulting in data silos and failing to establish an integrated O&M perspective for both "fishery" and "solar" activities. Secondly, the level of intelligence in O&M decision-making is limited. Existing PV power plant O&M solutions mostly rely on threshold alarms and periodic inspections, lacking intelligent decision-making methods based on multi-dimensional data, leading to uneconomical O&M timing or delayed responses. Thirdly, the formulated maintenance plans are usually based on experience, unable to predict the potential chain reactions between the O&M process and power generation and aquaculture, resulting in certain implementation risks and a degree of uncertainty. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a smart operation and maintenance system for a solar-fishery hybrid power station.

[0005] To achieve the above objectives, this invention provides a smart operation and maintenance system for a solar-fishery hybrid power station. The system includes: a multi-source data acquisition layer, an edge computing layer, a cloud service platform, and an operation and maintenance execution layer. The multi-source data acquisition layer collects multi-source data from the solar-fishery hybrid power station and transmits this data to the edge computing layer via the Internet of Things (IoT). The multi-source data includes monitoring data, image data, and weather forecasts. The edge computing layer preprocesses the multi-source data and transmits the preprocessed data to the cloud service platform via the IoT. The cloud service platform performs operation and maintenance analysis based on the preprocessed multi-source data, issues early warnings, generates work instructions based on maintenance plans, and sends these instructions to the operation and maintenance execution layer via the IoT. The operation and maintenance analysis includes determining the appearance defects of equipment in the power station, the current and future operating status of the equipment, and evaluating the feasibility of maintenance plans. The operation and maintenance execution layer executes the work instructions to achieve actual maintenance of the solar-fishery hybrid power station. This invention forms an integrated operation and maintenance perspective of "fishing" and "solar" based on multi-source data of fishery-solar complementary power stations. It can obtain the optimal operation and maintenance solution by comprehensively considering photovoltaic power generation and aquaculture activities based on multi-source data, thereby improving the economy and intelligence of operation and maintenance of fishery-solar complementary power stations.

[0006] Optionally, the monitoring data includes equipment operating parameters, environmental monitoring data, photovoltaic module power generation, grid electricity price, and aquaculture water indicators, and the image data is images of the power station equipment.

[0007] Optionally, the equipment operating parameters include at least three categories: photovoltaic module operating parameters, inverter operating parameters, and combiner box operating parameters. The photovoltaic module operating parameters include at least the photovoltaic module tilt angle, photovoltaic module output voltage, and photovoltaic module output current. The inverter operating parameters include at least the inverter input current, inverter output current, inverter input voltage, and inverter output voltage. The combiner box operating parameters include at least the combiner box input current, combiner box input voltage, combiner box output current, and combiner box output voltage. The environmental monitoring data includes at least five environmental indicators: ambient temperature, ambient humidity, irradiance, wind speed, and rainfall. The aquaculture water indicators include at least three indicators: dissolved oxygen concentration, water pH value, and algae concentration.

[0008] Optionally, the images of the power station equipment include at least images of photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes.

[0009] Optionally, the cloud service platform is specifically used for:

[0010] When the aquaculture water indicators are abnormal, a water abnormality warning is issued, a water maintenance plan is set, and then a water maintenance instruction is generated and sent to the operation and maintenance execution layer after manual verification.

[0011] Image recognition models are used to determine whether photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes have appearance defects.

[0012] Based on the photovoltaic module operating parameters and the environmental monitoring data, the first classification model is used to determine the photovoltaic module operating status, which includes normal function, operational failure, and combined pollution impact.

[0013] The inverter operating status and the combiner box operating status are determined based on the inverter operating parameters, the combiner box operating parameters, and the environmental monitoring data. Both the inverter operating status and the combiner box operating status include normal function and operational failure.

[0014] When the equipment is functioning normally, the predicted values ​​of environmental indicators are obtained by using weather forecasts, and then the predicted values ​​of equipment operating parameters are obtained by using a long short-term memory network to determine the future operating status of photovoltaic modules, inverters and combiner boxes;

[0015] When the aforementioned combined pollution occurs, the timing of maintenance is determined using the photovoltaic module power generation, the grid electricity price, the weather forecast, and the aquaculture water indicators.

[0016] If the equipment has the aforementioned appearance defects, operational malfunctions, and combined pollution effects, an equipment maintenance warning will be issued and an equipment maintenance plan will be set.

[0017] The equipment maintenance plan is run on the digital twin model of the solar-fishery hybrid power station, and its feasibility is then evaluated.

[0018] The feasible equipment maintenance plan is manually verified, and then a work instruction is generated and sent to the operation and maintenance execution layer.

[0019] Optionally, the step of determining whether photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes have appearance defects based on image recognition models includes the following steps:

[0020] The type of appearance defect is determined and the image data is labeled, thereby constructing an image dataset;

[0021] An image recognition model is constructed using the image dataset and a convolutional neural network, and the image recognition model is used to identify the appearance defects in real time.

[0022] Optionally, determining the inverter operating status and the combiner box operating status based on the inverter operating parameters, the combiner box operating parameters, and the environmental monitoring data includes the following steps:

[0023] The inverter conversion efficiency is calculated using the inverter operating parameters, and then the second classification model is used to determine the inverter operating status in combination with the environmental monitoring data.

[0024] The junction box loss is calculated using the junction box operating parameters, and then the operating status of the junction box is determined using a third classification model in conjunction with the environmental monitoring data.

[0025] Optionally, when the equipment is functioning normally, environmental indicator predictions are obtained using weather forecasts, and then long short-term memory networks are used to obtain predicted equipment operating parameters to determine the future operating status of the photovoltaic modules, inverters, and combiner boxes, including the following steps:

[0026] When the equipment is functioning normally, environmental indicator predictions are obtained using weather forecasts.

[0027] Based on the predicted values ​​of the environmental indicators, the equipment operating parameters, and the environmental monitoring data, a long short-term memory network is used to obtain the predicted values ​​of the equipment operating parameters.

[0028] The predicted values ​​of the environmental indicators and the predicted values ​​of the equipment operating parameters are used to determine the future operating status of the photovoltaic modules, inverters, and combiner boxes.

[0029] Optionally, when the combined pollution occurs, determining the timing of maintenance using the photovoltaic module's power generation, the grid electricity price, the weather forecast, and the aquaculture water indicators includes the following steps:

[0030] The peak power of the photovoltaic module under standard test conditions is corrected based on the environmental monitoring data, and then the standard power generation is calculated. At the same time, the power generation of the photovoltaic module within one day after cleaning is used as the reference power generation, and the ratio of the reference power generation to the standard power generation is used as the performance evaluation index.

[0031] A daily reference value for the power generation of the solar-fishery hybrid power station is calculated based on the power generation of the photovoltaic modules and the performance evaluation indicators.

[0032] In response to the combined pollution impact, after each operation and maintenance, the daily power generation loss is calculated in real time using the reference value of the power generation of the fishery-solar hybrid power station, the power generation of the photovoltaic modules, and the grid electricity price, and then the cumulative power generation loss is calculated.

[0033] When the aforementioned combined pollution occurs and the dissolved oxygen concentration and algae concentration are abnormal, the water body maintenance command will be executed first to restore the water body to normal.

[0034] If the combined pollution still exists on the second day after the water body maintenance instruction is executed, and if the cumulative power generation loss is greater than the pollution removal cost and the weather forecast shows that the rainfall in the next two days is lower than the rainfall threshold, maintenance must be carried out immediately. Otherwise, the long short-term memory network is used to obtain the power generation forecast and electricity price forecast within the future preset time window.

[0035] After obtaining the predicted power generation value and the predicted electricity price value, calculate the predicted cumulative power generation loss value for the Kth day in the future;

[0036] If the predicted cumulative power generation loss is greater than the pollution removal cost and the weather forecast shows that the rainfall from the Kth day to the K+2th day is lower than the rainfall threshold, then it is determined that maintenance will be carried out on the Kth day; otherwise, monitoring will continue.

[0037] Optionally, the process of running the equipment maintenance scheme on the digital twin model of the solar-fishery hybrid power station and then evaluating its feasibility includes the following steps:

[0038] The equipment maintenance scheme is run on the digital twin model to obtain simulation results;

[0039] Operation and maintenance evaluation indicators are obtained based on simulation results. These indicators include at least operation and maintenance time indicators, operation and maintenance cost indicators, performance recovery indicators, and aquaculture impact indicators calculated based on the aquaculture water body indicators.

[0040] The feasibility score of the equipment maintenance plan is calculated by weighted summation based on the operation and maintenance evaluation indicators, thereby determining the feasibility of the equipment maintenance plan.

[0041] The present invention has at least the following beneficial effects:

[0042] 1. This invention performs operation and maintenance analysis based on multi-source data such as equipment operation parameters, environmental monitoring data, photovoltaic module power generation, grid electricity price, aquaculture water indicators, and power station equipment images of the fishery-solar complementary power station, forming an integrated operation and maintenance perspective of "fishery" and "solar", which improves the scenario adaptability of the solution.

[0043] 2. This invention determines whether there are any external defects in the equipment in the solar-fishery hybrid power station, as well as the current and future operating status of the equipment, and then formulates a maintenance plan. This can alleviate the problems of uneconomical maintenance timing or delayed response caused by threshold alarms and regular inspections, reduce the blindness of maintenance, and improve the economy and intelligence of maintenance.

[0044] 3. This invention uses a digital twin model to virtually simulate maintenance plans, and then calculates the feasibility score of the maintenance plan and performs manual verification to determine the final operation and maintenance plan, which further alleviates the blindness of operation and maintenance and improves the operation and maintenance effect. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the framework of a smart operation and maintenance system for a solar-fishery hybrid power station according to an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram of the operation process of the cloud service platform according to an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram illustrating the process of determining the timing of maintenance using photovoltaic module power generation, grid electricity price, weather forecast, and aquaculture water indicators, according to an embodiment of the present invention. Detailed Implementation

[0049] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0050] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0051] It should be noted in advance that, in one alternative embodiment, except for independent descriptions, the same symbols or letters appearing in all formulas have the same meaning.

[0052] In one optional embodiment, please refer to Figure 1The present invention provides a smart operation and maintenance system for a solar-fishery hybrid power station. The system includes: a multi-source data acquisition layer 1, an edge computing layer 2, a cloud service platform 3, and an operation and maintenance execution layer 4. The multi-source data acquisition layer 1 is used to collect multi-source data from the solar-aquaculture hybrid power station and transmit the multi-source data to the edge computing layer 2 via the Internet of Things (IoT). The multi-source data includes monitoring data, image data, and weather forecasts. The monitoring data includes equipment operating parameters, environmental monitoring data, photovoltaic module power generation, grid electricity price, and aquaculture water indicators. The image data consists of images of the power station equipment. The edge computing layer 2 is used to preprocess the multi-source data and transmit the preprocessed multi-source data to the cloud service platform 3 via the IoT. The cloud service platform 3 is used to perform operation and maintenance analysis based on the preprocessed multi-source data, thereby issuing early warnings and generating work instructions based on the maintenance plan. The work instructions are then sent to the operation and maintenance execution layer 4 via the IoT. The operation and maintenance analysis includes judging the appearance defects of the equipment in the solar-aquaculture hybrid power station, the current and future operating status of the equipment, and the feasibility evaluation of the maintenance plan. The operation and maintenance execution layer 4 includes devices such as aerators, unmanned boats, drones, and mobile communication devices (such as mobile phones of operation and maintenance personnel) to execute work instructions and realize the actual maintenance of the solar-aquaculture hybrid power station.

[0053] The equipment operating parameters must include at least three categories: photovoltaic module operating parameters, inverter operating parameters, and combiner box operating parameters. Specifically, photovoltaic module operating parameters must include at least three parameters: photovoltaic module tilt angle, photovoltaic module output voltage, and photovoltaic module output current. Inverter operating parameters must include at least four parameters: inverter input current, inverter output current, inverter input voltage, and inverter output voltage. Combiner box operating parameters must include at least four parameters: combiner box input current, combiner box input voltage, combiner box output current, and combiner box output voltage. Environmental monitoring data must include at least five environmental indicators: ambient temperature, ambient humidity, irradiance, wind speed, and rainfall. Aquaculture water indicators must include at least three indicators: dissolved oxygen concentration, water pH value, and algae concentration. Power station equipment images must include at least images of photovoltaic modules, photovoltaic support structures, inverters, and combiner boxes.

[0054] The multi-source data acquisition layer 1 forms a sensor network through tilt sensors, voltage sensors, current sensors, temperature sensors, humidity sensors, wind speed sensors, thermopile sensors, rain gauges, dissolved oxygen sensors, water quality pH sensors, water quality blue-green algae sensors, and high-definition cameras deployed in the fishery-solar hybrid power station. This network is used to collect real-time equipment operating parameters, environmental monitoring data, aquaculture water indicators, and images of the power station equipment. Daily photovoltaic module power generation is obtained through smart meters, and a data interface is established with the local power grid company to obtain real-time grid electricity prices. Weather forecasts are also obtained through the fishery-solar hybrid power station's weather forecasting system. The sensor deployment method and density can be adjusted according to actual needs.

[0055] In the operation and maintenance management of solar-aquaculture hybrid power plants, a systematic numbering system can be used to achieve efficient data collection and fault location. Specifically, each aquaculture area is assigned a unique identifier number, and the photovoltaic array is divided into several sub-regions based on the area of ​​the aquaculture area, each assigned the same number as the area it is located in. This "area-photovoltaic area" numbering binding allows data collection to be carried out by geographical unit partitioning. Furthermore, a three-level coding rule of "area-equipment type-serial number" is adopted for key equipment such as inverters and combiner boxes in the power plant to form a complete equipment positioning coordinate system. In the data collection stage, drones equipped with high-definition cameras are automatically dispatched daily to conduct inspections along a preset route. The drones use onboard GPS and a visual positioning system to match the current shooting location in real time. While acquiring equipment images, the drones automatically embed the aquaculture area number, photovoltaic area number, and equipment number into the image metadata in the form of digital watermarks, and simultaneously record auxiliary information such as shooting time and environmental parameters. This combination of numbering system and automated collection method is conducive to quickly locating abnormal areas and improving the efficiency of monitoring and operation and maintenance.

[0056] After receiving multi-source data, edge computing layer 2 immediately preprocesses it. For time-series data such as equipment operating parameters, environmental monitoring data, and aquaculture water indicators, edge computing layer 2 performs the following operations in sequence: smoothing short-term fluctuations through Kalman filtering while preserving long-term trends; based on... The principles include outlier identification and removal; resampling to the same time interval, such as 5 minutes; missing value filling using linear interpolation; and standardization using Z-score normalization. For photovoltaic module power generation and grid electricity prices, edge computing layer 2 requires two preprocessing operations: outlier identification and removal, and missing value filling. For power station equipment images, edge computing layer 2 needs to achieve image noise reduction through spatial domain filtering (mean filtering, Gaussian filtering, median filtering) and frequency domain filtering (Fourier transform), image enhancement through histogram equalization and sharpening filtering, and image size uniformity to 224×224 pixels through cropping.

[0057] In an optional embodiment, please refer to Figure 2 The cloud service platform 3 will execute the following steps:

[0058] S1. When the aquaculture water indicators are abnormal, a water abnormality warning is issued, a water maintenance plan is set, and then a water maintenance instruction is generated and sent to the operation and maintenance execution layer after manual verification.

[0059] Specifically, in this embodiment, the normal functional ranges of three aquaculture water indicators—dissolved oxygen concentration, water pH value, and algae concentration—are determined based on the types of aquatic products in the solar-aquaculture hybrid power station. For any given water area, if any one or more of the measured dissolved oxygen concentration, water pH value, and algae concentration are outside their respective normal functional ranges, it indicates an abnormality in the aquaculture water indicators. In this case, the cloud service platform 3 issues a water abnormality warning using its built-in early warning module, specifically a voice warning, based on the abnormal aquaculture water indicators. For example, when the water pH value in water area numbered 01 is outside its normal functional range and exceeds the upper limit of the normal functional range, the early warning module issues a voice warning stating, "The water pH value in water area numbered 01 is too high."

[0060] Furthermore, while issuing an alert for water body anomalies, the cloud service platform 3 automatically sets up a water body maintenance plan. The water body maintenance plan is set up as follows: When the dissolved oxygen concentration is abnormal, the start and stop of the aerators in the corresponding water area are controlled to restore the dissolved oxygen concentration to the normal functional range. For example, when the dissolved oxygen concentration in water area numbered 01 is lower than 5 mg / L, the water body maintenance plan is to start the aerator in water area numbered 01 until its dissolved oxygen concentration is restored to the normal functional range; when the water pH value is abnormal, acidic solution (such as dilute hydrochloric acid) or alkaline solution (such as quicklime water) is added to the corresponding water area for neutralization, or the water is changed to stabilize the water pH value within a suitable range; when the algae concentration is abnormal, algaecides are added to the corresponding water area or algae are manually removed to stabilize the algae concentration within the normal functional range.

[0061] The generated maintenance plan is sent to the administrator's mobile phone via push notification. On one hand, push notifications allow administrators to confirm the water body maintenance plan, and then the cloud service platform 3 uses the confirmed plan to generate work instructions and distribute them to the operation and maintenance execution layer. On the other hand, for some maintenance operations, such as manual algae removal, which require human intervention, push notifications can remind relevant personnel to promptly participate in the maintenance.

[0062] S2. Based on the image recognition model, determine whether there are appearance defects in photovoltaic modules, photovoltaic brackets, inverters and combiner boxes.

[0063] Step S2 specifically includes the following steps:

[0064] S21. Determine the type of the appearance defect and label the image data to construct an image dataset.

[0065] Specifically, in this embodiment, visual defects in photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes refer to functional failures caused by physical structural damage, material aging, or environmental stress. Common visual defects include, but are not limited to, broken photovoltaic module glass, deformed photovoltaic module frames, cracked photovoltaic module backsheets, deformed photovoltaic bracket structures, corroded photovoltaic brackets, cracked inverter casings, and cracked combiner box casings. Many of these visual defects can be identified using image recognition models. Therefore, this embodiment acquires images of power plant equipment and uses LabelImg software to label the damage types, then uses the labeled images to construct an image dataset.

[0066] S22. Construct an image recognition model using the image dataset and a convolutional neural network, and use the image recognition model to identify the appearance defects in real time.

[0067] Specifically, in this embodiment, the YOLOv8 algorithm is used to construct an image recognition model, and the image dataset is divided into a training set and a validation set in a 7:3 ratio to complete the training and validation of the image recognition model. The trained and validated image recognition model can then be used to identify appearance defects of various devices in real time. In other alternative embodiments, image recognition models can be constructed separately for different devices.

[0068] S3. Based on the photovoltaic module operating parameters and the environmental monitoring data, use the first classification model to determine the photovoltaic module operating status, which includes normal function, operational failure, and combined pollution impact.

[0069] Specifically, in this embodiment, unlike typical photovoltaic power plants, the photovoltaic modules in a solar-fishery hybrid power plant may be affected by various pollutants, primarily suppressing their power generation capacity through "shading." These pollutants include bird droppings, insect carcasses, dust, water vapor condensate, dissolved oxygen, and algae. Multiple pollutants may coexist, hence the term "compound pollution." Compound pollution leads to a reduction in photogenerated carriers, causing a proportional decrease in the photovoltaic module's output current (inversely proportional to the degree of shading), while its impact on output voltage is generally smaller. In contrast, typical operational faults such as increased series resistance, hot spots, and localized short circuits often cause a non-uniform decrease in the photovoltaic module's output current, which is more severe in high-current regions, and the output voltage also drops significantly, especially near the maximum power point. Furthermore, the performance of photovoltaic modules is also affected by environmental factors such as temperature, irradiance, and humidity. Therefore, a classification model can be constructed to identify the operating status of the photovoltaic modules.

[0070] This embodiment uses a one-dimensional convolutional neural network (1D-CNN) to construct the first classification model. It boasts high computational efficiency, is suitable for processing high-frequency sampled data, and meets real-time requirements. The input to the first classification model is multivariate time-series data, including environmental monitoring data and photovoltaic module operating parameters, with a time window length of 60 minutes. The convolutional modules of the first classification model sequentially include: a first convolutional layer with 64 kernels, a kernel size of 3, a stride of 1, and padding set to "same," using the ReLU activation function to extract local temporal features; a first pooling layer employing max pooling, with a pooling window size of 2×1 and a stride of 2; a second convolutional layer with 128 kernels, a kernel size of 3, a stride of 1, padding set to "same," and using the ReLU activation function to further extract higher-level temporal patterns; and a second pooling layer employing max pooling, with a pooling window size of 2×1 and a stride of 2. The fully connected modules of the first classification model consist of: a flattening layer, used to flatten the output of the second pooling layer into a 1D vector; and a fully connected layer with 256 neurons, using the ReLU activation function to fuse temporal and inter-variable features. The output layer of the first classification model has three independent sigmoid activation units, corresponding to the probabilities of normal function, operational failure, and combined contamination, respectively. The loss function uses binary cross-entropy loss, supporting multi-label learning. An output of (0,0,0) indicates normal function; (1,0,0) indicates operational failure; (0,1,0) indicates combined contamination; and (1,1,0) indicates both operational failure and combined contamination.

[0071] A dataset was constructed using environmental monitoring data and photovoltaic module operating parameters under different operating conditions of the photovoltaic modules. This dataset was then divided into a training set and a validation set in a 7:3 ratio to train and validate the primary classification model. The trained and validated primary classification model can then be used to determine the operating status of the photovoltaic modules.

[0072] S4. Determine the inverter operating status and combiner box operating status based on the inverter operating parameters, combiner box operating parameters, and environmental monitoring data. The inverter operating status and combiner box operating status both include normal function and operating fault.

[0073] Step S4 specifically includes the following steps:

[0074] S41. Calculate the inverter conversion efficiency using the inverter operating parameters, and then use the second classification model to determine the inverter operating status in conjunction with the environmental monitoring data.

[0075] Specifically, in this embodiment, the inverter's performance is affected by environmental factors such as temperature and humidity, and its conversion efficiency typically decreases significantly or fluctuates drastically when an inverter experiences an operational failure. Therefore, this embodiment uses inverter operating parameters to calculate the inverter's conversion efficiency, and then uses 1D-CNN to construct a second classification model. The input to this model is multivariate time-series data, including ambient temperature, ambient humidity, and inverter conversion efficiency, with a time window length of 60 minutes. The convolutional module of the second classification model sequentially includes: a first convolutional layer with 32 kernels, a kernel size of 5, a stride of 1, and using the ReLU activation function; a first pooling layer using max pooling with a pooling window size of 2×1; a second convolutional layer with 64 kernels, a kernel size of 3, a stride of 1, and using the ReLU activation function; and a second pooling layer with a pooling size of 2. The fully connected modules of the second classification model consist of: a flattening layer, used to flatten the output of the second pooling layer into a 1D vector; and a fully connected layer with 128 neurons, using the ReLU activation function and a dropout rate of 0.5. The output layer of the second classification model has one independent Sigmoid activation unit, used to output the probability of the inverter functioning normally and malfunctioning, and the loss function is binary cross-entropy loss.

[0076] The inverter conversion efficiency satisfies the following relationship:

[0077]

[0078] in, For inverter conversion efficiency, This is the inverter input voltage. Input current to the inverter, This is the inverter output voltage. This is the inverter output current.

[0079] Furthermore, a dataset was constructed using ambient temperature, ambient humidity, and inverter conversion efficiency under different inverter operating conditions. This dataset was then divided into a training set and a validation set in a 7:3 ratio to train and validate the secondary classification model. The trained and validated secondary classification model can then be used to determine the inverter's operating status.

[0080] S42. Calculate the junction box loss using the junction box operating parameters, and then use the third classification model to determine the junction box operating status in conjunction with the environmental monitoring data.

[0081] Specifically, in this embodiment, the performance of the combiner box is affected by environmental factors such as temperature and humidity, and when the combiner box malfunctions, its losses typically increase abnormally. Therefore, this embodiment calculates the combiner box losses and constructs a third classification model based on the construction method of the second classification model to determine the operating status of the combiner box. The combiner box losses satisfy the following relationship:

[0082]

[0083] in, For junction box losses, Input voltage to the combiner box. Input current to the combiner box, This is the output voltage of the combiner box. This is the output current of the combiner box.

[0084] S5. When the equipment is functioning normally, the predicted values ​​of environmental indicators are obtained by using weather forecasts, and then the predicted values ​​of equipment operating parameters are obtained by using a long short-term memory network to determine the future operating status of photovoltaic modules, inverters and combiner boxes.

[0085] Step S5 specifically includes the following steps:

[0086] S51. When the equipment is functioning normally, use weather forecasts to obtain predicted values ​​of environmental indicators.

[0087] Specifically, in this embodiment, environmental indicator forecasts for the next three days of the solar-aquaculture hybrid power station can be obtained using weather forecasts. The reason this embodiment only obtains environmental indicator forecasts for the next three days is to improve the accuracy of equipment operating status predictions through short-term forecasting.

[0088] S52. Based on the predicted values ​​of the environmental indicators, the equipment operating parameters, and the environmental monitoring data, use a long short-term memory network to obtain the predicted values ​​of the equipment operating parameters.

[0089] Specifically, in this embodiment, environmental monitoring data and photovoltaic (PV) module operating parameters are used to construct an environment-PV module time series. Each environment-PV module time series is 60 bytes long with a time step of 5 minutes. Each time step includes eight features: ambient temperature, ambient humidity, irradiance, wind speed, rainfall, PV module tilt angle, PV module output voltage, and PV module output current. Multiple environment-PV module time series are used to construct a PV module indicator prediction dataset, which is then divided into a training set and a validation set in a 7:3 ratio to train and validate a Long Short-Term Memory (LSTM) network, resulting in a PV module indicator prediction model. Once the PV module indicator prediction model is obtained, the predicted environmental indicators can be used to predict the PV module tilt angle, PV module output voltage, and PV module output current over a future period.

[0090] It should be noted that although aquaculture water indicators also affect the function of photovoltaic modules, and their introduction may improve the accuracy of predictions, the introduction of aquaculture water indicators also means that aquaculture water indicators must be predicted, which will introduce more uncertainty. In general, aquaculture water indicators are not the main factors affecting the performance of photovoltaic modules. Therefore, aquaculture water indicators were not considered when constructing the photovoltaic module indicator prediction model.

[0091] More specifically, the photovoltaic module performance prediction model employs a two-layer LSTM structure and two fully connected layers, using mean squared error as the loss function, the Adam optimizer, and a learning rate of 0.001. The first LSTM layer contains 128 neurons, with "return_sequences=True", dropout of 0.2, and tanh activation. The second LSTM layer contains 64 neurons, with "return_sequences=False", dropout of 0.1, and also tanh activation. The output of the second LSTM layer enters the first fully connected layer. The first fully connected layer has 32 neurons and uses ReLU activation. The second fully connected layer is the output layer, with 3 neurons and a linear activation function, used to output the photovoltaic module tilt angle, output voltage, and output current.

[0092] Furthermore, referring to the construction method of the photovoltaic module index prediction dataset, a combiner box index prediction dataset can be constructed using ambient temperature, ambient humidity, combiner box input current, combiner box input voltage, combiner box output current, and combiner box output voltage. Similarly, an inverter index prediction dataset can be constructed using ambient temperature, ambient humidity, inverter input current, inverter output current, inverter input voltage, and inverter output voltage. Even further, the combiner box index prediction model and inverter index prediction model can be obtained by referring to the acquisition method of the photovoltaic module index prediction model.

[0093] However, it should be noted that in a solar-fishery hybrid power station, the output of the photovoltaic modules is usually the input of the combiner box, while the output of the combiner box is generally the input of the inverter. Therefore, based on the prediction of the output voltage and current of the photovoltaic modules, only the output voltage and current of the combiner box need to be predicted; similarly, based on the prediction of the output voltage and current of the combiner box, only the output voltage and current of the inverter need to be predicted. In fact, both the combiner box indicator prediction model and the inverter indicator prediction model only need to predict these two indicators.

[0094] S53. Use the predicted values ​​of the environmental indicators and the predicted values ​​of the equipment operating parameters to determine the future operating status of the photovoltaic modules, inverters and combiner boxes.

[0095] Specifically, in this embodiment, after obtaining the predicted values ​​of environmental indicators and the predicted values ​​of the equipment operating parameters, the first classification model, the second classification model and the third classification model can be used to determine the future operating status of the photovoltaic modules, inverters and combiner boxes.

[0096] S6. When the combined pollution occurs, the timing of maintenance is determined by using the photovoltaic module power generation, the grid electricity price, the weather forecast, and the aquaculture water indicators.

[0097] Please see Figure 3 Step S6 specifically includes the following steps:

[0098] S61. Based on the environmental monitoring data, correct the peak power of the photovoltaic module under standard test conditions, and then calculate the standard power generation. At the same time, take the power generation of the photovoltaic module within one day after cleaning as the reference power generation, and take the ratio of the reference power generation to the standard power generation as the performance evaluation index.

[0099] Specifically, in this embodiment, after each cleaning of the photovoltaic module, the power generation of the photovoltaic module within one day after cleaning is obtained through a smart meter and used as a reference power generation. Simultaneously, the peak power of the photovoltaic module under standard test conditions is corrected using the irradiance and temperature data within one day after cleaning, and recorded as the standard output power. The standard output power specifically satisfies the following relationship:

[0100]

[0101] Where P is the standard output power. G represents the peak power of the photovoltaic module under standard test conditions, and G is the irradiance measured by the thermopile sensor. Irradiance under standard test conditions. The power temperature coefficient is approximately [value missing]. For monocrystalline silicon modules, the power temperature coefficient is approximately [value missing]. T is the ambient temperature measured by the temperature sensor. The ambient temperature is the temperature under standard test conditions.

[0102] By calculating the standard output power at different times within one day after cleaning the photovoltaic modules, the corresponding time-power curve can be obtained. Integrating this time-power curve yields the standard power generation within one day after cleaning. The ratio of the reference power generation to the standard power generation is then used as a performance evaluation index for the photovoltaic modules. This performance evaluation index measures the power generation capacity of clean photovoltaic modules under performance degradation.

[0103] S62. Calculate the reference value of power generation of the solar-fishery hybrid power station daily based on the power generation of the photovoltaic modules and the performance evaluation indicators.

[0104] Specifically, in this embodiment, for any photovoltaic module, its power generation is divided by the latest performance evaluation index to obtain the reference power generation of that photovoltaic module on that day. Furthermore, the sum of the reference power generation of all photovoltaic modules on that day is used as the reference value for the power generation of the solar-fishery hybrid power station.

[0105] S63. In response to the combined pollution impact, after each operation and maintenance, the daily power generation loss is calculated in real time using the reference value of the power generation of the fishery-solar complementary power station, the power generation of the photovoltaic module, and the grid electricity price, and then the cumulative power generation loss is calculated.

[0106] Specifically, in this embodiment, the day after each maintenance operation of the solar-aquaculture hybrid power station in response to the combined pollution impact is designated as the first cumulative day. The daily power generation loss of the solar-aquaculture hybrid power station is calculated from the first cumulative day, and then the cumulative power generation loss from the most recent maintenance operation in response to the combined pollution impact to the present is calculated. The cumulative power generation loss specifically satisfies the following relationship:

[0107]

[0108] in, This represents the cumulative power generation loss since the most recent operation and maintenance of the solar-aquaculture hybrid power station in response to the combined effects of pollution. Let be the average grid electricity price on the i-th cumulative day. Here is the reference value for the power generation of the solar-fishery hybrid power station on the i-th cumulative day, and N is the number of photovoltaic modules. For the cumulative power generation of the photovoltaic module on the i-th day, This refers to the daily power generation of the solar-fishery hybrid power station. It should be noted that since the grid electricity price may fluctuate on the same day, resulting in multiple grid electricity prices, this formula does not directly use the grid electricity price, but rather the average grid electricity price for each day. If the grid electricity price on a particular day does not fluctuate, then the average grid electricity price for that day equals the grid electricity price.

[0109] Of course, the cumulative power generation loss can be calculated by region, that is, the cumulative power generation loss of each photovoltaic power generation area can be calculated, thereby realizing regional operation and maintenance, improving the accuracy of operation and maintenance and reducing operation and maintenance costs.

[0110] S64. When the combined pollution occurs and the dissolved oxygen concentration and algae concentration are abnormal, the water body maintenance command shall be executed first to restore the water body to normal.

[0111] Specifically, in this embodiment, dissolved oxygen concentration affects the heat transfer efficiency of the water. High DO water has a higher convective heat transfer coefficient, potentially accelerating heat dissipation from the back of the module; however, if DO is too low, heat dissipation efficiency decreases, potentially leading to an increase in photovoltaic module temperature. Excessive algae concentration can reduce water transmittance, negatively impacting underwater photovoltaic modules and reducing the reflectivity of sunlight across the water surface. This reduces the amount of scattered light that would otherwise be reflected to the back of the photovoltaic module, indirectly affecting the module's overall light utilization efficiency. Although the initial abnormalities in dissolved oxygen and algae concentrations may not have a significant impact on the photovoltaic module, the suppression effect is not significant when the combined pollution effect is first detected. Furthermore, water maintenance commands must be executed to restore the water quality when dissolved oxygen and algae concentrations are abnormal. Therefore, the inhibitory effect of dissolved oxygen and algae concentrations on the photovoltaic module's power generation efficiency can be prioritized, and the surface of the photovoltaic module can be temporarily left untreated.

[0112] S65. If the combined pollution still exists on the second day after the water body maintenance instruction is executed, and if the cumulative power generation loss is greater than the pollution removal cost and the weather forecast shows that the rainfall in the next two days is lower than the rainfall threshold, maintenance must be carried out immediately. Otherwise, the long short-term memory network is used to obtain the predicted power generation and electricity price within the future preset time window.

[0113] Specifically, in this embodiment, if there is no compound pollution impact on the second day after the execution of the water body maintenance order, then this execution of the water body maintenance order is considered the most recent operation and maintenance (O&M) of the solar-aquaculture hybrid power station in response to the compound pollution impact. If compound pollution impact still exists on the second day after the execution of the water body maintenance order, and if the cumulative power generation loss exceeds the cost of pollution removal and the weather forecast indicates that the rainfall in the next two days is below the rainfall threshold, then it is determined that immediate O&M of the photovoltaic modules is required. In drier areas, the rainfall threshold can be set to 10mm, while in humid areas, the rainfall threshold can be appropriately increased to avoid over-reliance on natural cleaning.

[0114] If the combined pollution impact still exists on the second day after the water body maintenance order is executed, the cumulative power generation loss is greater than the pollution removal cost, and the rainfall in the next two days is not less than the rainfall threshold, then this rainfall will be regarded as the most recent operation and maintenance of the fishery-solar hybrid power station in response to the combined pollution impact, and the cumulative power generation loss needs to be recalculated after this rainfall.

[0115] If the combined pollution still exists on the second day after the water body maintenance order is executed, and the cumulative power generation loss is less than the pollution removal cost, then a long short-term memory network is used to construct a power generation prediction model and an electricity price prediction model to obtain the power generation prediction value and electricity price prediction value within a preset time window in the future. The preset time window is the next three days.

[0116] Based on the collected or predicted time series data of ambient temperature, humidity, irradiance, wind speed, and rainfall, daily average temperature, humidity, irradiance, wind speed, and rainfall are calculated respectively. Then, using the daily average temperature, humidity, irradiance, wind speed, rainfall, and power generation data from a solar-fishery hybrid power station, a power generation prediction dataset is constructed. In this dataset, each time series is 20 bytes long with a time step of one day, and each time step includes six features: daily average temperature, humidity, irradiance, wind speed, rainfall, and power generation. Next, the power generation prediction dataset is divided into a training set and a validation set in a 7:3 ratio to train and validate the LSTM model, resulting in the power generation prediction model.

[0117] The power generation prediction model employs a two-layer LSTM structure and a fully connected layer, using mean squared error as the loss function. The Adam optimizer is selected, and the learning rate is set to 0.001. The first LSTM layer extracts local features, while the second LSTM layer captures global temporal relationships. Setting `return_sequences=True` ensures that the output of the first LSTM layer can be passed to the second layer. Each LSTM layer contains 100 neurons to fully capture long-term dependencies in the time series and uses the tanh activation function. Finally, the fully connected layer maps the output of the LSTM layers to the predicted value, i.e., the predicted power generation value. After obtaining the predicted power generation value at the previous time point, this predicted value can be used as a known condition, combined with environmental indicator predictions, to predict the predicted power generation value at the next time point.

[0118] A power price prediction dataset was constructed based on the collected grid electricity prices. Each time series in the dataset is 60 bytes long, with a time step of 3 hours, and each time step contains only one feature: the grid electricity price. The dataset was divided into a training set and a validation set in a 7:3 ratio to train and validate the LSTM model, resulting in the grid electricity price prediction model. The model employs a single-layer LSTM structure with one fully connected layer. The LSTM layer contains 50 neurons, and other settings are the same as the power generation prediction model. Similarly, after obtaining the predicted electricity price value for the previous time point, this value can be used as a known condition to predict the predicted electricity price value for the next time point.

[0119] This solution calculates cumulative power generation losses and combines them with pollution removal costs and rainfall conditions to determine the maintenance time. This can alleviate the problems of uneconomical maintenance timing or delayed response caused by threshold alarms and regular inspections, reduce the blindness of maintenance, and improve maintenance efficiency and intelligence.

[0120] S66. After obtaining the predicted power generation value and the predicted electricity price value, calculate the predicted cumulative power generation loss value for the Kth day in the future.

[0121] Specifically, in this embodiment, after obtaining the predicted power generation and electricity price, the cumulative power generation loss from the time of the most recent operation and maintenance of the solar-fishery hybrid power station in response to the combined pollution impact to the Kth day in the future can be calculated with reference to step S63; that is, the predicted cumulative power generation loss. Furthermore, according to step S65, K=3.

[0122] S67. If the cumulative power generation loss prediction value is greater than the pollution removal cost and the weather forecast shows that the rainfall in the next K days to the next K+2 days is lower than the rainfall threshold, then it is determined that maintenance will be carried out on the next K days; otherwise, monitoring will continue.

[0123] Specifically, in this embodiment, if the predicted cumulative power generation loss is greater than the pollution removal cost and the weather forecast indicates that the rainfall from day K to day K+2 is below the rainfall threshold, then it is determined that maintenance will be performed on day K; otherwise, monitoring and forecasting will continue without taking any action. This is beneficial for improving maintenance efficiency and intelligence.

[0124] S7. If the equipment has the aforementioned appearance defects, operational malfunctions, and combined pollution effects, an equipment maintenance warning will be issued and an equipment maintenance plan will be set.

[0125] Specifically, in this embodiment, if any of the following abnormal equipment states exist—appearance defects, operational malfunctions, or combined pollution effects—the cloud service platform 3, based on the abnormal equipment and its number, or the number of its location, uses its built-in early warning module to issue a real-time equipment maintenance warning via voice alert. For example, when the image recognition model identifies that the casing of the first inverter in the water area numbered 01 is damaged, a voice prompt is issued: "The casing of the first inverter in the water area numbered 01 is damaged." Or, if combined pollution effects still exist on the second day after the water body maintenance instruction is executed in the photovoltaic power generation area numbered 01, and if its cumulative power generation loss exceeds the cost of pollution removal and the weather forecast shows that the rainfall in the next two days is below the rainfall threshold, a voice prompt is issued: such as "Combined pollution effects still exist in the photovoltaic power generation area numbered 01!" At this time, the management personnel can first conduct an on-site inspection to determine the specific cause of the malfunction based on the location or area of ​​the equipment with appearance defects, operational malfunctions, or combined pollution effects, and then set at least one equipment maintenance plan on the cloud service platform 3, such as personnel scheduling, maintenance routes, and repair plans.

[0126] If equipment experiences operational malfunctions or combined pollution effects within the next three days, the cloud service platform 3 will issue a preventative maintenance warning via voice alert using its built-in warning module, based on the malfunctioning equipment and its serial number, or the serial number of its location. For example, if the photovoltaic modules still experience combined pollution effects on the second day after the water body maintenance instruction is executed, and the predicted cumulative power generation loss two days later will exceed the cost of pollution removal, the voice warning will be: "Combined pollution effects still exist in photovoltaic power generation area numbered 01! Combined pollution effects maintenance is expected to be carried out in two days!" Similarly, management personnel can conduct on-site inspections and set up preventative equipment maintenance plans on the cloud service platform 3 to prepare for maintenance in advance.

[0127] This solution forms an integrated operation and maintenance perspective of "fishing" and "solar" based on multi-source data. It can simultaneously identify and predict the appearance defects and operational failures of equipment in fishery-solar complementary power stations, as well as whether there is mixed pollution suppression in photovoltaic modules. In this way, it can issue early warnings in a timely manner, remind relevant personnel to maintain the equipment, and alleviate the problems of uneconomical operation and maintenance timing or delayed response caused by threshold alarms and regular inspections.

[0128] S8. Run the equipment maintenance plan on the digital twin model of the solar-fishery hybrid power station, and then evaluate its feasibility.

[0129] Step S8 specifically includes the following steps:

[0130] S81. Run the equipment maintenance scheme on the digital twin model of the solar-fishery hybrid power station to obtain simulation results.

[0131] The construction method of the digital twin model can refer to existing technologies, so only a brief description is given below. Specifically, in this embodiment, a digital twin model of a solar-aquaculture hybrid power station is constructed on cloud service platform 3 based on COMSOL Multiphysics software, and the three-dimensional deviation between the digital twin model and the actual power station is compared by laser scanning. In the digital twin model, a single diode model is used to describe the photovoltaic power generation behavior, a dissolved oxygen balance model (Streetter-Phelps model) is used to describe the aquaculture water ecology, the Monod equation is used to describe the relationship between algal growth rate and light and nutrients (such as nitrogen and phosphorus), CFD (Computational Fluid Dynamics) is used to simulate water flow, and the influence of component transmittance on the water light field is simulated based on the radiative transfer equation (RTE). An LSTM is trained in Python to obtain a pH prediction model (using a water quality pH time series of length 50 and a time step of 5 minutes as input to predict future water quality pH values). Using the "multiphysics coupling" function of COMSOL Multiphysics software, other models other than the pH prediction model can be directly coupled. In the simulation, the model function of the pH prediction model is called through LiveLink.

[0132] Then, the equipment maintenance scheme was run on the digital twin model of the solar-aquaculture power station to obtain simulation results.

[0133] S82. Obtain operation and maintenance evaluation indicators based on simulation results. The operation and maintenance evaluation indicators include at least operation and maintenance time indicators, operation and maintenance cost indicators, performance recovery indicators, and aquaculture impact indicators calculated based on the aquaculture water body indicators.

[0134] Specifically, in this embodiment, the operation and maintenance time indicator is the reciprocal of the operation and maintenance time, and the operation and maintenance cost indicator is the reciprocal of the operation and maintenance cost. The performance recovery indicator and the aquaculture impact indicator satisfy the following relationships:

[0135]

[0136]

[0137] in, These are performance recovery indicators for equipment, primarily targeting operational failures; The output power of the photovoltaic module before maintenance. The output power of the photovoltaic modules after maintenance. The conversion efficiency before inverter maintenance. The conversion efficiency after inverter maintenance. This refers to the losses before the junction box maintenance. This refers to the losses after maintenance of the combiner box; Indicators affecting aquaculture; To maintain the value of the kth aquaculture water body indicator; K is the number of aquaculture water body indicators, and in this embodiment K=3; This represents the optimal value for the k-th aquaculture water indicator. It needs to be determined based on the type of aquatic product.

[0138] It should be noted that, taking photovoltaic modules as an example, if they are not maintained or are replaced directly, then... We directly take 1. The same applies to the inverter and combiner box.

[0139] Furthermore, after obtaining each operation and maintenance evaluation index, the maximum and minimum value normalization method is used to normalize each operation and maintenance evaluation index.

[0140] S83. Calculate the feasibility score of the equipment maintenance plan using a weighted summation method based on the operation and maintenance evaluation indicators, and then determine the feasibility of the equipment maintenance plan.

[0141] Specifically, in this embodiment, the feasibility score of the equipment maintenance plan satisfies the following relationship:

[0142]

[0143] Where F is the feasibility score, and t is the normalized operation and maintenance time indicator. As time weight, Here, c represents the cost weight, and c is the normalized operation and maintenance cost indicator. To restore the weight of power generation, For normalized performance recovery metrics, The weighting of aquaculture influences This is a normalized indicator of the impact of aquaculture. , , and The settings can be customized based on the relevant personnel's emphasis on time, cost, power generation, and aquaculture activities, or they can be obtained through expert evaluation.

[0144] Furthermore, if the feasibility score of the equipment maintenance plan exceeds 80 points, it is deemed feasible.

[0145] S9. Manually verify the feasible equipment maintenance plan, generate a work instruction, and send the work instruction to the operation and maintenance execution layer.

[0146] Specifically, in this embodiment, firstly, for feasible equipment maintenance plans, managers can perform secondary verification to confirm their feasibility. Secondly, the equipment maintenance plan that passes the secondary verification is determined as the final operation and maintenance plan; if multiple equipment maintenance plans pass the secondary verification, the one with the highest feasibility score is selected as the final operation and maintenance plan. If no equipment maintenance plan passes the secondary verification, the equipment maintenance plan is reset and virtual simulation is performed on the digital twin model.

[0147] Then, for the finalized operation and maintenance plan, the cloud service platform 3 will generate operation instructions and operation and maintenance information for the maintenance equipment, which together constitute the work instructions. The operation instructions for the maintenance equipment include information such as the start time, route, and stop time for drones, unmanned surface vessels, etc.; the operation instructions for the maintenance equipment also include information such as the start time, route, and stop time for drones, unmanned surface vessels, etc., based on GPS and BeiDou dual-mode positioning. Finally, the operation instructions are directly distributed to each maintenance device via the Internet of Things, while the operation and maintenance information is sent to mobile communication devices (the maintenance personnel's mobile phones) via push notifications.

[0148] It should be noted that in some cases, the actions described in the specification can be performed in different orders and still achieve the desired results. In this embodiment, the order of steps is given only to make the embodiment clearer and easier to explain, and is not intended to limit it.

[0149] In summary, this solution offers at least the following advantages: First, the system performs operation and maintenance analysis based on multi-source data, including equipment operating parameters, environmental monitoring data, photovoltaic module power generation, grid electricity prices, aquaculture water indicators, and images of power station equipment. This creates an integrated operation and maintenance perspective for both "fishery" and "solar" aspects, improving the solution's adaptability to various scenarios. Second, by determining whether equipment in the fishery-solar hybrid power station exhibits current external defects and assessing its current and future operating status, the system formulates maintenance plans. This alleviates the problems of uneconomical maintenance timing or delayed response caused by threshold alarms and periodic inspections, reducing the blindness of operation and maintenance and improving its economy and intelligence. Third, by virtually simulating maintenance plans on a digital twin model and then calculating the feasibility score and manually verifying the plans, the system determines the final operation and maintenance plan, further reducing the blindness of operation and maintenance and improving its effectiveness.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A smart operation and maintenance system for a solar-fishery hybrid power station, characterized in that, include: Multi-source data acquisition layer, edge computing layer, cloud service platform, and operation and maintenance execution layer; The multi-source data acquisition layer is used to collect multi-source data from the solar-fishery hybrid power station and transmit the multi-source data to the edge computing layer via the Internet of Things. The multi-source data includes monitoring data, image data, and weather forecasts. The monitoring data includes equipment operating parameters, environmental monitoring data, photovoltaic module power generation, grid electricity price, and aquaculture water indicators; the image data is images of the power station equipment. The equipment operating parameters include at least three categories: photovoltaic module operating parameters, inverter operating parameters, and combiner box operating parameters. The photovoltaic module operating parameters include at least the photovoltaic module tilt angle, photovoltaic module output voltage, and photovoltaic module output current. The inverter operating parameters include at least the inverter input current, inverter output current, inverter input voltage, and inverter output voltage. The combiner box operating parameters include at least the combiner box input current, combiner box input voltage, combiner box output current, and combiner box output voltage. The environmental monitoring data includes at least five environmental indicators: ambient temperature, ambient humidity, irradiance, wind speed, and rainfall. The aquaculture water indicators include at least three indicators: dissolved oxygen concentration, water pH value, and algae concentration. The images of the power station equipment include at least images of photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes; The edge computing layer is used to preprocess the multi-source data and transmit the preprocessed multi-source data to the cloud service platform via the Internet of Things. The cloud service platform is used to perform operation and maintenance analysis based on the preprocessed multi-source data, thereby issuing early warnings and generating work instructions based on the maintenance plan, and sending the work instructions to the operation and maintenance execution layer through the Internet of Things. The operation and maintenance analysis includes judging the appearance defects of the equipment in the power plant, the current and future operating status of the equipment, and the feasibility evaluation of the maintenance plan. The cloud service platform is specifically used for: When the aquaculture water indicators are abnormal, a water abnormality warning is issued, a water maintenance plan is set, and then a water maintenance instruction is generated and sent to the operation and maintenance execution layer after manual verification. Image recognition models are used to determine whether photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes have appearance defects. Based on the photovoltaic module operating parameters and the environmental monitoring data, the first classification model is used to determine the photovoltaic module operating status, which includes normal function, operational failure, and combined pollution impact. The inverter operating status and the combiner box operating status are determined based on the inverter operating parameters, the combiner box operating parameters, and the environmental monitoring data. Both the inverter operating status and the combiner box operating status include normal function and operational failure. When the equipment is functioning normally, the predicted values ​​of environmental indicators are obtained by using weather forecasts, and then the predicted values ​​of equipment operating parameters are obtained by using a long short-term memory network to determine the future operating status of photovoltaic modules, inverters and combiner boxes; When the aforementioned combined pollution occurs, the timing of maintenance is determined using the photovoltaic module power generation, the grid electricity price, the weather forecast, and the aquaculture water indicators. If the equipment has the aforementioned appearance defects, operational malfunctions, and combined pollution effects, an equipment maintenance warning will be issued and an equipment maintenance plan will be set. The equipment maintenance plan is run on the digital twin model of the solar-fishery hybrid power station, and its feasibility is then evaluated. The feasible equipment maintenance plan is manually verified, and then a work instruction is generated and sent to the operation and maintenance execution layer. The operation and maintenance execution layer is used to execute the operation instructions to realize the actual maintenance of the fishery-solar hybrid power station.

2. The intelligent operation and maintenance system for a solar-fishery hybrid power station according to claim 1, characterized in that, The method for determining whether photovoltaic modules, photovoltaic brackets, inverters, and combiner boxes have appearance defects based on image recognition models includes the following steps: The type of appearance defect is determined and the image data is labeled, thereby constructing an image dataset; An image recognition model is constructed using the image dataset and a convolutional neural network, and the image recognition model is used to identify the appearance defects in real time.

3. The intelligent operation and maintenance system for a solar-fishery hybrid power station according to claim 2, characterized in that, Determining the inverter's operating status and the combiner box's operating status based on the inverter's operating parameters, the combiner box's operating parameters, and the environmental monitoring data includes the following steps: The inverter conversion efficiency is calculated using the inverter operating parameters, and then the second classification model is used to determine the inverter operating status in combination with the environmental monitoring data. The junction box loss is calculated using the junction box operating parameters, and then the operating status of the junction box is determined using a third classification model in conjunction with the environmental monitoring data.

4. The intelligent operation and maintenance system for a solar-fishery hybrid power station according to claim 3, characterized in that, When the equipment is functioning normally, it uses weather forecasts to obtain predicted values ​​of environmental indicators, and then uses a long short-term memory network to obtain predicted values ​​of equipment operating parameters to determine the future operating status of photovoltaic modules, inverters, and combiner boxes. This includes the following steps: When the equipment is functioning normally, environmental indicator predictions are obtained using weather forecasts. Based on the predicted values ​​of the environmental indicators, the equipment operating parameters, and the environmental monitoring data, a long short-term memory network is used to obtain the predicted values ​​of the equipment operating parameters. The predicted values ​​of the environmental indicators and the predicted values ​​of the equipment operating parameters are used to determine the future operating status of the photovoltaic modules, inverters, and combiner boxes.

5. The intelligent operation and maintenance system for a solar-fishery hybrid power station according to claim 4, characterized in that, When the combined pollution occurs, the timing of maintenance is determined using the photovoltaic module's power generation, the grid electricity price, weather forecasts, and aquaculture water indicators, including the following steps: The peak power of the photovoltaic module under standard test conditions is corrected based on the environmental monitoring data, and then the standard power generation is calculated. At the same time, the power generation of the photovoltaic module within one day after cleaning is used as the reference power generation, and the ratio of the reference power generation to the standard power generation is used as the performance evaluation index. A daily reference value for the power generation of the solar-fishery hybrid power station is calculated based on the power generation of the photovoltaic modules and the performance evaluation indicators. In response to the combined pollution impact, after each operation and maintenance, the daily power generation loss is calculated in real time using the reference value of the power generation of the fishery-solar hybrid power station, the power generation of the photovoltaic modules, and the grid electricity price, and then the cumulative power generation loss is calculated. When the aforementioned combined pollution occurs and the dissolved oxygen concentration and algae concentration are abnormal, the water body maintenance command will be executed first to restore the water body to normal. If the combined pollution still exists on the second day after the water body maintenance instruction is executed, and if the cumulative power generation loss is greater than the pollution removal cost and the weather forecast shows that the rainfall in the next two days is lower than the rainfall threshold, maintenance must be carried out immediately. Otherwise, the long short-term memory network is used to obtain the power generation forecast and electricity price forecast within the future preset time window. After obtaining the predicted power generation value and the predicted electricity price value, calculate the predicted cumulative power generation loss value for the Kth day in the future; If the predicted cumulative power generation loss is greater than the pollution removal cost and the weather forecast shows that the rainfall from the Kth day to the K+2th day is lower than the rainfall threshold, then it is determined that maintenance will be carried out on the Kth day; otherwise, monitoring will continue.

6. The intelligent operation and maintenance system for a solar-fishery hybrid power station according to claim 5, characterized in that, The equipment maintenance plan is then run on the digital twin model of the solar-fishery hybrid power station, and its feasibility is evaluated, including the following steps: The equipment maintenance scheme was run on the digital twin model of the solar-fishery hybrid power station to obtain simulation results; Operation and maintenance evaluation indicators are obtained based on simulation results. These indicators include at least operation and maintenance time indicators, operation and maintenance cost indicators, performance recovery indicators, and aquaculture impact indicators calculated based on the aquaculture water body indicators. The feasibility score of the equipment maintenance plan is calculated by weighted summation based on the operation and maintenance evaluation indicators, thereby determining the feasibility of the equipment maintenance plan.

Images