Photovoltaic tracking method and system based on deep learning optimization scheduling

Abstract

The invention discloses a photovoltaic tracking method and system based on deep learning optimization scheduling. The method comprises the following steps: a photovoltaic tracking system and an environment data acquisition device acquire initial information; after the initial information is obtained, a tracking camera of the environmental data acquisition device obtains the relative position of the sun by using a sun-oriented movement trajectory tracking algorithm model, then calculates a tracking angle according to the position information of the sun and drives the camera to rotate to a theoretical angle so as to realize tracking of the sun; while tracking the sun, the tracking camera captures and captures an image of weather near the sun rich in weather characteristic information, and the captured image information, data acquired by the environmental data acquisition device and historical photovoltaic power generation power data recorded by the MPPT solar controller are sent to the photovoltaic tracking controller; and the photovoltaic tracking controller makes different control strategies according to the contents. The probability of invalid tracking of the system in a complex environment is reduced, so that the photovoltaic tracking system is based on the optimal state, and the power generation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic tracking technology, and in particular to a photovoltaic tracking method and system based on deep learning optimal scheduling. Background Art

[0002] The photovoltaic tracking system is a photovoltaic system that can automatically track the sun and increase the overall power generation. The solar photovoltaic array automatic tracking system tracks the movement of the sun in real time so that the sunlight directly hits the photovoltaic array, thereby increasing the amount of solar radiation received by the photovoltaic array and improving the overall power generation of the solar photovoltaic power generation system. There are four widely used solar photovoltaic automatic tracking systems, including horizontal single-axis tracking, double-column inclined single-axis tracking, vertical single-axis tracking and dual-axis tracking. Among them, horizontal single-axis tracking, inclined single-axis tracking and vertical single-axis tracking have only one degree of rotational freedom, and dual-axis tracking has two degrees of rotational freedom. At present, the main defects of conventional photovoltaic tracking systems are:

[0003] (1) Large size and high consumables. Some parts of the system are large in size and have a slightly redundant layout, which results in waste of materials and increases additional production costs.

[0004] (2) Simple model and extensive regulation. Conventional photovoltaic tracking systems have the disadvantages of simple control strategy models and extensive regulation, which can easily cause a large amount of invalid tracking in complex environments, indirectly shortening the service life of the system and increasing subsequent maintenance costs.

[0005] (3) Lack of environmental data monitoring devices. Conventional photovoltaic tracking systems lack environmental data monitoring devices. However, by real-time monitoring of environmental data, the operation of the photovoltaic system can be optimized, making it work in the best state and maximizing power generation efficiency.

[0006] (4) Lack of a monitoring cloud platform for human-computer interaction. Conventional photovoltaic tracking systems lack a monitoring cloud platform for human-computer interaction. Through the monitoring cloud platform, equipment failures or abnormal conditions can be discovered in a timely manner to avoid serious problems that lead to large-scale shutdowns. At the same time, it can also save manpower and material costs and improve maintenance efficiency. Summary of the invention

[0007] The present invention provides a photovoltaic tracking method based on deep learning optimal scheduling, comprising the following steps:

[0008] Step 1: Start the power supply, and the photovoltaic tracking device and environmental data acquisition device begin preparation work to obtain initial information;

[0009] Step 2: After obtaining the initial information, within the set operation time range, the tracking camera of the environmental data acquisition device uses the apparent solar motion trajectory tracking algorithm model to obtain the relative position of the sun, and then calculates the tracking angle based on the solar position information and drives the camera to rotate to the theoretical angle to achieve tracking of the sun;

[0010] Step 3: While tracking the sun, the tracking camera captures weather images near the sun that are rich in weather feature information. The captured image information, data collected by the environmental data acquisition device, and historical photovoltaic power generation data recorded by the MPPT solar controller are sent to the photovoltaic tracking controller;

[0011] Step 4: Based on the image information, historical photovoltaic power generation data, and data collected by the environmental data collection device, the photovoltaic tracking controller uses a cloud trajectory prediction model based on Kalman filtering and a photovoltaic power generation prediction model based on multi-information fusion to output prediction results that serve as a basis for determining whether to start and stop the photovoltaic tracking system in cloudy weather, and then makes different control strategies.

[0012] Furthermore, the initial information obtained in step 1 includes: the longitude and latitude coordinates, date and time of the device installation location obtained by the tracking camera using the GPS positioning system, and the initial angle position information obtained by the angle sensor embedded in the tracking camera.

[0013] Furthermore, the different control strategies made by the photovoltaic tracking controller in step 4 include:

[0014] The photovoltaic tracking controller receives the sent data and extracts the wind speed data. If the wind speed is less than the set safety threshold, the captured image information is used to build a weather recognition neural network to achieve high-precision recognition of complex weather. The output weather recognition results are fed back to the photovoltaic tracking controller for subsequent optimization of the control strategy of the control system; if it is greater than the set safety threshold, the photovoltaic tracking power generation is stopped and switched to fixed photovoltaic power generation;

[0015] If the photovoltaic tracking controller receives the results of cloudy and rainy weather or other extremely bad weather from the weather recognition network, it will stop outputting current to the DC motor, stop photovoltaic tracking power generation, and switch to fixed photovoltaic power generation;

[0016] If the photovoltaic tracking controller receives the sunny weather recognition result output from the weather recognition network, it obtains initial information based on the GPS positioning system and angle sensor embedded in the controller, and then calculates the altitude and azimuth of the sun based on the apparent sun motion trajectory tracking algorithm model. Then, it calculates the tracking angle based on these two angles and drives the motor to rotate to the calculated angle. Finally, the position tracking closed loop is realized through the angle sensor to achieve the photovoltaic panel's tracking function of the sun.

[0017] If the photovoltaic tracking controller receives the cloudy weather recognition result output from the weather recognition network, it will receive and read the weather image data near the sun collected by the tracking camera in the environmental data acquisition device, the light intensity data collected by the light sensor, the temperature and humidity data collected by the temperature and humidity sensor, and the historical photovoltaic power generation data recorded by the MPPT solar controller;

[0018] If the photovoltaic tracking controller reads the image video stream near the sun and processes it through the cloud trajectory prediction model, the output is the result that there will be clouds covering the sun in the future; at the same time, the received temperature, humidity, light intensity and historical photovoltaic power generation data are processed by the photovoltaic power generation prediction model, and the output is that the photovoltaic power generation will have a sharp decline and fluctuation in the future. At this time, the current output to the motor is stopped, the photovoltaic tracking power generation is stopped, and fixed photovoltaic power generation is switched.

[0019] Furthermore, if the photovoltaic tracking controller receives the read image and text data and processes them through the cloud trajectory prediction model and the photovoltaic power generation prediction model, it cannot simultaneously receive output signals indicating that there will be clouds blocking the sun and a sharp drop in photovoltaic power generation in the future. At this time, the initial information is obtained based on the GPS positioning system and the angle sensor in the photovoltaic tracking controller, and the tracking angle is calculated using the apparent solar motion trajectory tracking algorithm model to drive the motor to rotate to the theoretical calculation angle. Finally, the position tracking closed loop is realized through the angle sensor to achieve the photovoltaic panel's tracking function of the sun.

[0020] Furthermore, the environmental data collection device and the MPPT solar controller transmit the collected image text data to the monitoring cloud platform in real time through the MQTT protocol and store it through the database.

[0021] Furthermore, the prediction result value output by the photovoltaic power generation prediction model is transmitted to the cloud and stored, and the stored and recorded data is finally visualized through the monitoring cloud platform for human-computer interaction.

[0022] The present invention also proposes a photovoltaic tracking system based on deep learning optimal scheduling, including an environmental data acquisition device and a photovoltaic tracking device, wherein the environmental data acquisition device includes a wind speed sensor, a wind direction sensor, a temperature and humidity sensor, an eight-quadrant light sensor, a mobile tracking camera, a placement platform, a telescopic column, and a base; the base is connected to the telescopic column, the placement platform is installed on the telescopic column, and the placement platform is installed on the wind speed sensor, wind direction sensor, temperature and humidity sensor, eight-quadrant light sensor, and a mobile tracking camera;

[0023] The photovoltaic tracking device comprises a photovoltaic panel, a crossbeam bracket, a driving push rod, a supporting column, a fixed base, a push rod, an MPPT solar controller, and a photovoltaic tracking controller; there are two fixed bases, the two fixed bases are respectively connected to the supporting columns, the crossbeam bracket is installed on the supporting column, the push rod, the MPPT solar controller, the photovoltaic tracking controller and the photovoltaic panel are arranged on the crossbeam bracket, and a driving push rod is also connected between one of the fixed bases and the crossbeam bracket;

[0024] The environmental data acquisition device is communicatively connected with the MPPT solar energy controller and the photovoltaic tracking controller.

[0025] Furthermore, it also includes a monitoring cloud platform, which is communicatively connected to the MPPT solar controller and the photovoltaic tracking controller; wherein the monitoring cloud platform is a monitoring cloud platform for the photovoltaic tracking system based on digital twins.

[0026] At least one of the above technical solutions adopted in the embodiments of the present invention can achieve the following beneficial effects:

[0027] In summary, the present invention has the following technical advantages:

[0028] (1) The present invention proposes a photovoltaic tracking method that can be optimized based on deep learning in complex environments. The method can refine the control strategy of the photovoltaic tracking system, reduce the probability of invalid tracking of the system in complex environments, indirectly extend the service life of the system, and reduce the subsequent system maintenance costs. The method can enable the photovoltaic tracking system to be in the optimal state and maximize the power generation efficiency.

[0029] (2) Based on the mobile tracking camera on the environmental data monitoring device, images near the sun with multi-level weather features are captured, and then combined with the ultra-lightweight weather recognition algorithm to achieve high-precision weather recognition in complex environments. Through the multiple sensors on the environmental data monitoring device, data such as temperature, humidity, and light intensity that affect photovoltaic power generation are captured, and then historical photovoltaic power generation data is captured through MPPT. The captured and captured data are then combined with the constructed multi-information fusion photovoltaic power generation prediction model to achieve high-precision prediction of photovoltaic power generation in the medium and short term.

[0030] (3) Based on the tracking camera on the environmental data monitoring device, the image of the sky near the sun containing the cloud position information is captured. Then YOLOV5S is used to detect the cloud target, DeepSORT is used as a target tracker to achieve multi-target tracking of the cloud and record the historical movement trajectory position information, and Kalman filter is used as a trajectory predictor to predict the movement trajectory of the cloud at future moments by fitting the recorded historical cloud movement trajectory, thereby realizing the cloud trajectory prediction.

[0031] (4) A scalable portable environmental data acquisition device with a smaller size, lower consumables, and more compact structure was designed. It can simultaneously collect text data including temperature, humidity, light intensity, wind speed and direction, as well as complex weather environment images containing a large number of tiny features. It realizes the integration of multiple sensors and cameras, reduces data acquisition costs, and improves data acquisition efficiency.

[0032] (5) The designed monitoring cloud platform accurately simulates the behavior of the photovoltaic tracking system by establishing a digital twin model. It also accesses the environmental data collected by the environmental data acquisition device, the photovoltaic tracking system power generation data collected by the MPPT, and the photovoltaic power generation power prediction results, and draws and updates the real-time data curve on the human-computer interaction interface. It realizes effective monitoring of the photovoltaic tracking system, enhances the possibility of discovering abnormal situations, and ensures the stable operation of the photovoltaic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0034] Figure 1 This is a diagram of weather recognition results using the apparent sun motion trajectory tracking algorithm of the present invention.

[0035] Figure 2 This is a weather recognition accuracy diagram using the apparent sun motion trajectory tracking algorithm of the present invention.

[0036] Figure 3 This is a weather identification flow chart using the apparent sun motion trajectory tracking algorithm of the present invention.

[0037] Figure 4 This is a diagram showing the photovoltaic power generation prediction effect based on multi-sensor multi-feature information fusion according to the present invention.

[0038] Figure 5 This is a flow chart of the photovoltaic power prediction method based on multi-sensor multi-feature information fusion of the present invention.

[0039] Figure 6 The present invention is a flow chart of a cloud trajectory prediction method based on rapid capture of dynamic features of a tracking camera.

[0040] Figure 7 This is a flow chart of the control strategy model of the photovoltaic tracking system that can be optimally scheduled based on deep learning under complex weather conditions according to the present invention.

[0041] Figure 8 This is a diagram of the photovoltaic tracking system monitoring cloud platform interface of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] The technical solutions provided by various embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0044] The present invention provides a photovoltaic tracking method based on deep learning optimal scheduling, comprising the following steps:

[0045] Step 1: Start the power supply, and the photovoltaic tracking device and the environmental data acquisition device begin preparation work to obtain initial information; the initial information includes: the longitude and latitude coordinates, date and time of the device installation site obtained by the tracking camera using the GPS positioning system, and the initial angle position information obtained by the angle sensor embedded in the tracking camera.

[0046] Step 2: After obtaining the initial information, within the set operating time range, the tracking camera of the environmental data acquisition device uses the apparent sun motion trajectory tracking algorithm model to obtain the relative position of the sun, and then calculates the tracking angle based on the sun position information and drives the camera to rotate to the theoretical angle to track the sun; there are a large number of weather features near the sun, and the tracking camera uses the apparent sun motion trajectory tracking algorithm to complete the description of the relative position of the sun and drive the motor to track the sun. The recorded video stream is transmitted to the weather recognition network to achieve real-time and high-precision weather recognition. The weather recognition results are as follows: Figure 1 As shown in Figure 2, the weather recognition accuracy is Figure 2 As shown, the overall flow chart is as follows Figure 3 shown.

[0047] Step 3: While tracking the sun, the tracking camera captures weather images near the sun that are rich in weather feature information. The captured image information, data collected by the environmental data acquisition device, and historical photovoltaic power generation data recorded by the MPPT solar controller are sent to the photovoltaic tracking controller; medium- and short-term predictions for photovoltaic power generation can be achieved, which can provide a reference for optimizing the control and start-stop of the photovoltaic tracking system in cloudy weather. The temperature, humidity, and light intensity data that affect photovoltaic power generation are collected through the environmental data acquisition device, and the historical photovoltaic power generation data are collected using MPPT records. The collected data is processed by the BiLSTM bidirectional long short-term memory network to achieve a higher-precision photovoltaic power generation prediction. The photovoltaic power generation power prediction effect diagram is shown below. Figure 4 As shown, the overall flow chart is as follows Figure 5 shown.

[0048] Step 4. The photovoltaic tracking controller uses the image information, historical photovoltaic power data, and data collected by the environmental data collection device, and uses the prediction results of the cloud trajectory prediction model based on Kalman filtering and the photovoltaic power prediction model based on multi-information fusion as the basis for judging the start and stop of the photovoltaic tracking system in cloudy weather, and then makes different control strategies. Among them, predicting the movement trajectory of clouds near the sun can provide an effective reference for optimizing the control strategy of the photovoltaic tracking system and reducing invalid tracking. The tracking camera obtains real-time video stream data near the sun by tracking the sun. The captured data is processed by YOLOV5 (target monitoring model) to achieve target monitoring of clouds, and then DeepSORT (multi-target tracking model) is added to achieve multi-target tracking of clouds and record the historical cloud movement trajectory. Finally, Kalman filtering is added. The Kalman filter predicts the movement trajectory of clouds in the future based on the historical cloud movement trajectory, and finally realizes the trajectory prediction of clouds. The cloud trajectory prediction flow chart is as follows Figure 6 shown.

[0049] like Figure 7 As shown, the flow chart of different control strategies made by the photovoltaic tracking controller of the present invention includes:

[0050] The photovoltaic tracking controller receives the data sent and extracts the wind speed data. If the wind speed is less than the set safety threshold, the captured image information is used to build a weather recognition neural network to achieve high-precision recognition of complex weather. The output weather recognition results are fed back to the photovoltaic tracking controller for subsequent optimization of the control strategy of the control system. If it is greater than the set safety threshold, in order to avoid the photovoltaic panel being blown away, the photovoltaic tracking power generation is stopped and switched to fixed photovoltaic power generation.

[0051] If the photovoltaic tracking controller receives the results of rainy and other extremely bad weather from the weather recognition network, it will stop outputting current to the DC motor, stop photovoltaic tracking power generation, and switch to fixed photovoltaic power generation.

[0052] If the photovoltaic tracking controller receives the sunny weather recognition result output from the weather recognition network, it obtains initial information based on the GPS positioning system and angle sensor embedded in the controller, and then calculates the altitude and azimuth of the sun based on the apparent sun motion trajectory tracking algorithm model. It then calculates the tracking angle based on these two angles and drives the motor to rotate to the calculated angle. Finally, the position tracking closed loop is realized through the angle sensor to achieve the photovoltaic panel's tracking function of the sun.

[0053] If the photovoltaic tracking controller receives the cloudy weather identification result output from the weather identification network, the photovoltaic tracking controller will receive and read the weather image data near the sun collected by the tracking camera in the environmental data acquisition device, the light intensity data collected by the light sensor, the temperature and humidity data collected by the temperature and humidity sensor, and the historical photovoltaic power generation data recorded by the MPPT solar controller.

[0054] If the video stream of the image near the sun is processed by the cloud trajectory prediction model, the output is the result that there will be clouds covering the sun in the future. At the same time, the received temperature, humidity, light intensity and historical photovoltaic power generation data are processed by the photovoltaic power generation prediction model, and the output is that there will be a sharp decline and fluctuation in photovoltaic power generation in the future. At this time, it can be accurately determined that there will be clouds covering the sun in the high altitude in the short and medium term in the future. In order to avoid unfavorable factors such as invalid tracking and additional electricity waste, the controller stops outputting current to the motor, stops photovoltaic tracking power generation, and switches to fixed photovoltaic power generation.

[0055] In addition, if the photovoltaic tracking controller receives the read image and text data and processes it through the cloud trajectory prediction model and the photovoltaic power generation prediction model, it cannot simultaneously receive output signals that clouds will block the sun and that the photovoltaic power generation will experience a sharp drop in future power. At this time, the initial information is obtained based on the GPS positioning system and angle sensor in the photovoltaic tracking controller, and then the tracking angle is calculated using the apparent solar motion trajectory tracking algorithm model to drive the motor to rotate to the theoretical calculation angle. Finally, the position tracking closed loop is realized through the angle sensor to achieve the photovoltaic panel's tracking function of the sun.

[0056] While the photovoltaic tracking system uses the deep learning network model for optimal scheduling, the environmental data acquisition device and the MPPT solar controller also transmit the collected image text data to the monitoring cloud platform in real time through the MQTT protocol and store it in the database. At the same time, the predicted result value output by the photovoltaic power prediction model constructed by deep learning will also be transmitted to the cloud and stored. The above stored and recorded data is finally visualized through the monitoring cloud platform for human-computer interaction.

[0057] The present invention also proposes a photovoltaic tracking system based on deep learning optimal scheduling, including an environmental data acquisition device and a photovoltaic tracking device, the environmental data acquisition device including a wind speed sensor 1, a wind direction sensor 2, a temperature and humidity sensor 3, an eight-quadrant light sensor 4, a mobile tracking camera 5, a placement platform 6, a telescopic column 7, and a base 8; the base 8 is connected to the telescopic column 7, the placement platform 6 is installed on the telescopic column 7, and the placement platform 6 is installed on the wind speed sensor 1, the wind direction sensor 2, the temperature and humidity sensor 3, the eight-quadrant light sensor 4, and the mobile tracking camera 5;

[0058] The photovoltaic tracking device includes a photovoltaic panel 9, a beam support 10, a driving push rod 11, a supporting column 12, a fixed base 13, a push rod 14, an MPPT solar controller 15, and a photovoltaic tracking controller 16;

[0059] There are two fixed bases 13, and the two fixed bases 13 are respectively connected to the support columns 12, and the crossbeam bracket 10 is installed on the support column 12. The push rod 14, the MPPT solar controller 15, the photovoltaic tracking controller 16 and the photovoltaic panel 9 are arranged on the crossbeam bracket 10, and a driving push rod 11 is also connected between one of the fixed bases 13 and the crossbeam bracket 10;

[0060] The wind speed sensor 1 , the wind direction sensor 2 , the temperature and humidity sensor 3 , the eight-quadrant light sensor 4 , and the mobile tracking camera 5 are communicatively connected with the MPPT solar controller 15 and the photovoltaic tracking controller 16 .

[0061] The environmental data acquisition device of the present invention adopts multiple sensors + mobile tracking cameras, taking into account image data rich in multi-level features, to provide more reference basis for optimizing the operation of the photovoltaic tracking system. At the same time, the possibility of discovering abnormal situations is enhanced, so that timely measures can be taken to prevent equipment failures, extend equipment life, and ensure the stable operation of the photovoltaic system. A movable telescopic column + loading platform is adopted to optimize the layout of the device, adopt a more compact and streamlined structure, reduce material waste, and reduce production costs. At the same time, using telescopic columns instead of fixed columns increases the portability of the device and expands the application scenarios of the device.

[0062] At the same time, the system also includes a monitoring cloud platform, which is connected to the MPPT solar controller and the photovoltaic tracking controller; wherein the monitoring cloud platform is a monitoring cloud platform for the photovoltaic tracking system based on digital twins. The monitoring cloud platform realizes the following functions:

[0063] Use in photovoltaic power generation: By real-time monitoring of environmental data and photovoltaic power generation data, equipment failures or abnormal conditions can be discovered in a timely manner to avoid serious problems that lead to large-scale shutdowns, while saving manpower and material costs and improving maintenance efficiency.

[0064] Online fault diagnosis: The data recorded and stored by the environmental data monitoring device and the MPPT solar controller are analyzed and processed by the fault diagnosis model, and the output results are visualized on the cloud platform to quickly discover potential faults, avoid fault deterioration, and effectively reduce equipment downtime for maintenance.

[0065] Deep learning prediction model + prediction function area visualization: The data stored in the environmental data monitoring device and MPPT records are analyzed and processed by the deep learning prediction model, and the prediction results are output and visualized on the cloud platform. This can further enhance the possibility of discovering abnormal situations, and also provide an effective reference for optimizing the subsequent operation of the photovoltaic tracking system.

[0066] In summary, the present invention has the following technical advantages:

[0067] (1) The present invention proposes a photovoltaic tracking method that can be optimized based on deep learning in complex environments. The method can refine the control strategy of the photovoltaic tracking system, reduce the probability of invalid tracking of the system in complex environments, indirectly extend the service life of the system, and reduce the subsequent system maintenance costs. The method can enable the photovoltaic tracking system to be in the optimal state and maximize the power generation efficiency.

[0068] (2) Based on the mobile tracking camera on the environmental data monitoring device, images near the sun with multi-level weather features are captured, and then combined with the ultra-lightweight weather recognition algorithm to achieve high-precision weather recognition in complex environments. Through the multiple sensors on the environmental data monitoring device, data such as temperature, humidity, and light intensity that affect photovoltaic power generation are captured, and then historical photovoltaic power generation data is captured through MPPT. The captured and captured data are then combined with the constructed multi-information fusion photovoltaic power generation prediction model to achieve high-precision prediction of photovoltaic power generation in the medium and short term.

[0069] (3) Based on the tracking camera on the environmental data monitoring device, the image of the sky near the sun containing the cloud position information is captured. Then YOLOV5S is used to detect the cloud target, DeepSORT is used as a target tracker to achieve multi-target tracking of the cloud and record the historical movement trajectory position information, and Kalman filter is used as a trajectory predictor to predict the movement trajectory of the cloud at future moments by fitting the recorded historical cloud movement trajectory, thereby realizing the cloud trajectory prediction.

[0070] (4) A scalable portable environmental data acquisition device with a smaller size, lower consumables, and more compact structure was designed. It can simultaneously collect text data including temperature, humidity, light intensity, wind speed and direction, as well as complex weather environment images containing a large number of tiny features. It realizes the integration of multiple sensors and cameras, reduces data acquisition costs, and improves data acquisition efficiency.

[0071] (5) The designed monitoring cloud platform accurately simulates the behavior of the photovoltaic tracking system by establishing a digital twin model. It also accesses the environmental data collected by the environmental data acquisition device, the photovoltaic tracking system power generation data collected by the MPPT, and the photovoltaic power generation power prediction results, and draws and updates the real-time data curve on the human-computer interaction interface. It realizes effective monitoring of the photovoltaic tracking system, enhances the possibility of discovering abnormal situations, and ensures the stable operation of the photovoltaic system.

[0072] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A photovoltaic tracking method based on deep learning optimal scheduling, characterized in that: The following steps are involved: Step 1: Start the power supply, and the photovoltaic tracking device and the environmental data acquisition device begin preparation work to obtain initial information; Step 2: After obtaining the initial information, within the set operation time range, the tracking camera of the environmental data acquisition device uses the apparent sun motion trajectory tracking algorithm model to obtain the relative position of the sun, and then calculates the tracking angle based on the sun position information and drives the camera to rotate to the theoretical angle to achieve tracking of the sun; Step 3: While tracking the sun, the tracking camera captures weather images near the sun that are rich in weather feature information, and the captured image information, data collected by the environmental data acquisition device, and historical photovoltaic power generation data recorded by the MPPT solar controller are sent to the photovoltaic tracking controller; Step 4: Based on the image information, historical photovoltaic power generation data, and data collected by the environmental data collection device, the photovoltaic tracking controller uses a cloud trajectory prediction model based on Kalman filtering and a photovoltaic power generation prediction model based on multi-information fusion to output prediction results that serve as a basis for determining whether to start and stop the photovoltaic tracking system in cloudy weather, and then makes different control strategies.

2. A photovoltaic tracking method based on deep learning optimal scheduling according to claim 1, characterized in that: The initial information obtained in step 1 includes: the latitude and longitude coordinates, date and time of the device installation location obtained by the tracking camera using the GPS positioning system, and the initial angle position information obtained by the angle sensor embedded in the tracking camera.

3. A photovoltaic tracking method based on deep learning optimal scheduling according to claim 1, characterized in that: The different control strategies implemented by the photovoltaic tracking controller in step 4 include: The photovoltaic tracking controller receives the sent data and extracts the wind speed data. If the wind speed is less than the set safety threshold, the captured image information is used to build a weather recognition neural network to achieve high-precision recognition of complex weather. The output weather recognition results are fed back to the photovoltaic tracking controller for subsequent optimization of the control strategy of the control system; if it is greater than the set safety threshold, the photovoltaic tracking power generation is stopped and switched to fixed photovoltaic power generation; If the photovoltaic tracking controller receives the results of cloudy and rainy weather or other extremely bad weather from the weather recognition network, it will stop outputting current to the DC motor, stop photovoltaic tracking power generation, and switch to fixed photovoltaic power generation; If the photovoltaic tracking controller receives the sunny weather recognition result output from the weather recognition network, it obtains initial information based on the GPS positioning system and angle sensor embedded in the controller, and then calculates the altitude and azimuth of the sun based on the apparent sun motion trajectory tracking algorithm model. Then, it calculates the tracking angle based on these two angles and drives the motor to rotate to the calculated angle. Finally, the position tracking closed loop is realized through the angle sensor to achieve the photovoltaic panel's tracking function of the sun. If the photovoltaic tracking controller receives the cloudy weather recognition result output from the weather recognition network, it will receive and read the weather image data near the sun collected by the tracking camera in the environmental data acquisition device, the light intensity data collected by the light sensor, the temperature and humidity data collected by the temperature and humidity sensor, and the historical photovoltaic power generation data recorded by the MPPT solar controller; If the photovoltaic tracking controller reads the image video stream near the sun and processes it through the cloud trajectory prediction model, the output is the result that there will be clouds covering the sun in the future; at the same time, the received temperature, humidity, light intensity and historical photovoltaic power generation data are processed by the photovoltaic power generation prediction model, and the output is that the photovoltaic power generation will have a sharp decline and fluctuation in the future. At this time, the current output to the motor is stopped, the photovoltaic tracking power generation is stopped, and fixed photovoltaic power generation is switched.

4. A photovoltaic tracking method based on deep learning optimal scheduling according to claim 1, characterized in that: If the photovoltaic tracking controller receives the read image and text data and processes it through the cloud trajectory prediction model and the photovoltaic power generation prediction model, it cannot simultaneously receive output signals that clouds will block the sun and that the photovoltaic power generation will experience a sharp drop in future power. At this time, the initial information is obtained based on the GPS positioning system and angle sensor in the photovoltaic tracking controller, and the tracking angle is calculated using the apparent solar motion trajectory tracking algorithm model to drive the motor to rotate to the theoretically calculated angle. Finally, the position tracking closed loop is realized through the angle sensor to achieve the photovoltaic panel's tracking function of the sun.

5. The photovoltaic tracking method based on deep learning optimal scheduling according to claim 1 is characterized in that: The environmental data collection device and the MPPT solar controller transmit the collected image and text data to the monitoring cloud platform in real time through the MQTT protocol and store them through the database.

6. A photovoltaic tracking method based on deep learning optimal scheduling according to claim 1, characterized in that: The predicted result value output by the photovoltaic power generation prediction model is transmitted to the cloud and stored, and the stored and recorded data is finally visualized through the monitoring cloud platform for human-computer interaction.

7. A photovoltaic tracking system based on deep learning optimal scheduling, characterized in that: It includes an environmental data collection device and a photovoltaic tracking device, wherein the environmental data collection device includes a wind speed sensor, a wind direction sensor, a temperature and humidity sensor, an eight-quadrant light sensor, a mobile tracking camera, a placement platform, a telescopic column, and a base; the base is connected to the telescopic column, the placement platform is installed on the telescopic column, and the wind speed sensor, wind direction sensor, temperature and humidity sensor, eight-quadrant light sensor, and a mobile tracking camera are installed on the placement platform; The photovoltaic tracking device comprises a photovoltaic panel, a crossbeam bracket, a driving push rod, a supporting column, a fixed base, a push rod, an MPPT solar controller, and a photovoltaic tracking controller; there are two fixed bases, the two fixed bases are respectively connected to the supporting columns, the crossbeam bracket is installed on the supporting column, the push rod, the MPPT solar controller, the photovoltaic tracking controller and the photovoltaic panel are arranged on the crossbeam bracket, and a driving push rod is also connected between one of the fixed bases and the crossbeam bracket; The environmental data acquisition device is communicatively connected with the MPPT solar energy controller and the photovoltaic tracking controller.

8. A photovoltaic tracking system based on deep learning optimal scheduling according to claim 7, characterized in that: It also includes a monitoring cloud platform, which is communicatively connected to the MPPT solar controller and the photovoltaic tracking controller; wherein the monitoring cloud platform is a monitoring cloud platform for the photovoltaic tracking system based on digital twins.

Images