Photovoltaic power generation intelligent control system of photovoltaic panel in optical storage direct flexible system

Abstract

The invention discloses a photovoltaic power generation intelligent control system of photovoltaic panels in an optical storage direct flexible system, and relates to the technical field of photovoltaic power generation, the intelligent control system comprises a plurality of photovoltaic panel assemblies, the plurality of photovoltaic panel assemblies comprise main node assemblies, one main node assembly is arranged every 5-10 groups of photovoltaic panel assemblies, and the main node assemblies are connected with the photovoltaic panel assemblies; the master node assembly is configured with a high-precision double-axis sun sensor and an edge calculation module and is responsible for collecting a solar azimuth angle, a current inclination angle of the assembly and environment temperature and humidity data in real time, and the slave node assembly is configured with a single-axis inclination angle sensor and adjusts a tracking angle by receiving a master node instruction. Specifically, a master-slave node heterogeneous design is adopted, the slave node only needs a single-axis tilt angle sensor and a wireless communication module, the hardware configuration number is reduced, and cost control is achieved. According to the method for photovoltaic panel tracking optimization, the control instruction quantity of the system is reduced, the central control pressure is remarkably reduced, and the problems of cost and complexity during large-scale application of photovoltaic panel tracking are solved.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, specifically to an intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-DC-flexible system. Background Technology

[0002] The photovoltaic-storage-DC-flexible system is a new type of power system technology that deeply integrates photovoltaic, energy storage, DC power distribution, and flexible control technologies. Its core purpose is to improve the flexibility of load electricity consumption and the utilization rate of renewable energy. It connects distributed photovoltaic, energy storage, and load through a "DC bus" to achieve efficient power distribution and dynamic regulation. The photovoltaic power generation intelligent control system of the photovoltaic panel is a branch subsystem of the photovoltaic-storage-DC-flexible system. It is mainly used as a comprehensive management system for photovoltaic panels, integrating functions such as data acquisition, real-time monitoring, remote control, intelligent early warning, and data analysis. It can seamlessly connect to various sensors, inverters, combiner boxes, and other equipment in the photovoltaic power station to achieve comprehensive monitoring and management of the power station's operating status and environmental parameters.

[0003] To maximize energy extraction from photovoltaic (PV) panels, current intelligent control systems employ light tracking to enable these panels to perform solar tracking. This tracking is typically achieved using light sensors. However, PV panels are usually clustered together, requiring additional sensors, controllers, and communication links for each new cluster. Hardware costs increase linearly with the number of panels. Furthermore, each PV panel uses the same sensing, control, and tracking methods, leading to system complexity and significant control pressure. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-flexible system, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic power generation intelligent control system for photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system, the intelligent control system comprising multiple photovoltaic panel modules, the multiple photovoltaic panel modules comprising:

[0006] Master node component: One master node component is set for every 5-10 groups of photovoltaic panels. The master node component is equipped with a high-precision dual-axis solar sensor and an edge computing module, which is responsible for collecting solar azimuth angle, current tilt angle of the panel and ambient temperature and humidity data in real time.

[0007] Both the master node component and the slave node component are equipped with wireless communication modules for bidirectional communication between them.

[0008] Slave node component: The slave node component is equipped with a single-axis tilt sensor, which adjusts the tracking angle by receiving commands from the master node;

[0009] The edge computing module collects the real-time output power of each component every 15 minutes and uses the K-means algorithm to divide the components into 3 to 5 tracking subgroups according to power similarity.

[0010] Node monitoring module: used for online monitoring between the master node component and the slave node component. Utilizing the bidirectional communication of the wireless communication module, when a node in the master node component or the slave node component goes offline, an alarm message is sent to the management center after detection.

[0011] Fault-tolerant tracking module: When the master node component goes offline, during the fault period, the slave node component uses inertial navigation compensation to continue tracking and turning the light rays;

[0012] Time-sharing and zone-based sleep module: When the photovoltaic panel module is in the sleep state for one hour before or after sunrise or sunset, the steering drive component of the slave node module enters the sleep state, and only the tilt sensor and communication module remain working. At night, the self-calibration mode is activated, and the drive components of the master node module and slave node module rotate synchronously to the initial horizontal position. The mechanical zero point is calibrated by the built-in accelerometer to prevent errors in the steering drive component.

[0013] Preferably, the master node component calculates the optimal tilt angle of the photovoltaic panel component based on the NREL astronomical model, and corrects it by combining the data from the dual-axis solar sensor to generate a subgroup reference tracking curve.

[0014] Preferably, the calculation formula for the NREL-based astronomical model is as follows:

[0015] Solar declination angle δ:

[0016] δ=0.3723+23.2567sinθ+0.1149sin2θ-0.1712sin3θ-0.758cosθ+0.3656cos2θ+0.0201cos3θ

[0017] The solar angle θ is calculated from the accumulated days N over a year:

[0018]

[0019] Where N is the accumulated days of the year, calculated as January 1st = 1, December 31st in a leap year = 366, and 365.2422 is the average number of days in a year considering leap years;

[0020] Mean Time Difference (EoT):

[0021] EoT=0.0028-1.9857sinθ+9.9059sin2θ-7.0924cosθ-0.6882cos2θ

[0022] Among them, EoT is the difference between true solar time and mean solar time, with a maximum value of approximately ±15 minutes;

[0023] Solar hour angle ω:

[0024] ω=15(T s -12)

[0025] Among them, T s Local true solar time;

[0026] Solar altitude angle α:

[0027]

[0028] Where, ω rad The value is the hour angle in radians. The latitude is the local latitude, with positive numbers for the Northern Hemisphere.

[0029] Solar azimuth angle γ:

[0030]

[0031] Here, cosα is the cosine of the altitude angle, and it should not be divided by zero.

[0032] Preferably, the slave node component adopts a reference angle and deviation compensation control mode. The relative position deviation is detected by the RSSI signal strength between adjacent slave node components, and the angle difference caused by installation error is automatically compensated to ensure that the photovoltaic panel component turning error of the slave node component is reduced.

[0033] Preferably, the slave node component implements the reference angle and deviation compensation by reading single-axis tilt sensor data through the SPI interface and reporting the status to the master node every 200ms.

[0034] Preferably, the fault-tolerant tracking module includes:

[0035] Central node election module: When the slave node component detects that the communication of the master node component in the corresponding area is interrupted, during the navigation compensation process, it sends a communication request to the corresponding master node component at preset intervals, and repeatedly detects 3 to 5 times.

[0036] If the master node component is always offline, the ant colony reconstruction protocol is initiated. By releasing exploration signals, a temporary master node component collaborative core is dynamically elected based on the signal strength of adjacent master node components. The network is re-established after the master node components in the corresponding area resume communication.

[0037] Preferably, the specific calculation formula of the K-means algorithm of the edge computing module is as follows:

[0038] X for each data point i Assigned to the cluster C to which the nearest centroid belongs. j :

[0039]

[0040] Where, μ k Let ||k| be the centroid vector of the k-th cluster, and ||·|| be the Euclidean distance.

[0041] Recalculate the centroid based on the data points within the cluster:

[0042]

[0043] Among them, |C k |For cluster C k The number of data points within;

[0044] By iteratively optimizing the "assignment-update" process, the sum of squares within the cluster is minimized:

[0045]

[0046] Where K is the preset number of clusters.

[0047] Preferably, the master and slave nodes of the photovoltaic panel module automatically form a shadow adaptation group based on the power deviation of the modules that are shaded in front of it, which is >20%. A lag tracking strategy is adopted, and the tracking speed of the slave node modules is reduced by 50% to reduce mechanical losses. The unshaded modules in the center form a high-efficiency tracking group to perform real-time dynamic tracking. The number of groups of the photovoltaic panel modules is automatically adjusted with changes in light intensity.

[0048] Preferably, the intelligent control system further includes an energy consumption optimization module: collecting historical tracking data of each subgroup through the management and control center, training the optimal grouping strategy through a deep reinforcement learning model, and automatically updating the grouping parameters of the edge computing module every quarter to achieve continuous optimization of the tracking efficiency of the intelligent control system of the photovoltaic panel module.

[0049] Preferably, the steering drive components of the master node component and the slave node component are servo motors. The steering drive component of the master node component is connected to a dual-axis solar sensor, and a drive signal is sent to the steering drive component through the dual-axis solar sensor to start the steering drive component. The steering drive component is installed on the photovoltaic panel component.

[0050] The slave node component receives the steering information from the master node sensor via the wireless communication module, and then its drive component is activated.

[0051] This invention provides an intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system. It has the following beneficial effects:

[0052] (1) The present invention adopts a master-slave node heterogeneous design. Only one master node is configured for every 5-10 groups of photovoltaic panels. The remaining slave nodes only need a single-axis tilt sensor and a wireless communication module. The number of hardware configurations is reduced compared with the traditional homogeneous scheme, which directly reduces the repeated investment in sensors, controllers and communication links, thereby achieving cost control. Furthermore, the edge computing module is used to dynamically divide 3-5 tracking subgroups, replacing the traditional centralized architecture that uses a unified strategy to control all components. This reduces the number of system control commands, significantly reduces the central control pressure, and the number of communication network nodes and the complexity of control logic decrease exponentially, solving the cost and control complexity issues when photovoltaic panel tracking is applied on a large scale.

[0053] (2) A multi-fault tolerance and self-calibration mechanism is designed. The node monitoring module detects node offline and alarms in real time through bidirectional wireless communication, avoiding delays caused by manual inspection. When the master node fails, the slave node initiates 3 to 5 detection confirmations and then dynamically elects a temporary master node. Inertial navigation compensation is used during the fault period, which shortens the recovery time compared to the traditional centralized system. It can also ensure that the slave node can still perform regular tracking of the photovoltaic panel module during the master node failure, ensuring the tracking stability of the photovoltaic panel module and preventing the problem of the slave node being unable to track due to the master node being offline.

[0054] (3) By using the time-sharing and partitioning hibernation module, different time periods are adopted. The slave node drive components are put into hibernation within a preset time before and after sunrise / sunset, which reduces the power consumption of the slave node, thereby reducing the energy consumption of photovoltaic power generation and reducing operating costs. At the same time, the self-calibration mode is started at night, and the mechanical zero point is automatically calibrated by the accelerometer, reducing the cost of manual quarterly calibration and ensuring the tracking accuracy of photovoltaic panel components. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of a photovoltaic power generation intelligent control system for a photovoltaic panel in a photovoltaic-storage-direct-flexible system according to the present invention.

[0056] Figure 2 This is a schematic diagram of a node submodule of a photovoltaic power generation intelligent control system for photovoltaic panels in a photovoltaic-storage direct-drive flexible system according to the present invention.

[0057] Figure 3 This is a schematic diagram of the photovoltaic panel installation in a photovoltaic power generation intelligent control system for a photovoltaic-storage-direct-flexible system according to the present invention. Detailed Implementation

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying 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 efforts are within the scope of protection of the present invention.

[0059] Example 1

[0060] Please see Figure 1-3 This invention provides an intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system. To achieve the above objectives, this invention is implemented through the following technical solution: The intelligent control system includes multiple photovoltaic panel modules, and the multiple photovoltaic panel modules include:

[0061] Master node module: One master node module is set up for every 5-10 groups of photovoltaic modules. The master node module is equipped with a high-precision dual-axis solar sensor and an edge computing module, which is responsible for collecting solar azimuth angle, current tilt angle of the module and ambient temperature and humidity data in real time.

[0062] Both the master node component and the slave node component are equipped with wireless communication modules for bidirectional communication between them.

[0063] Slave node component: The slave node component is equipped with a single-axis tilt sensor, which adjusts the tracking angle by receiving commands from the master node;

[0064] Furthermore, the master node component is equipped with a backup capacitor power supply with a capacity that supports 30 minutes of offline operation, ensuring that fault node detection and temporary policy issuance can still be completed during brief power outages. The steering drive component of the slave node component is designed with dual windings. The main winding is used for normal tracking, while the secondary winding maintains low-power standby during sleep, balancing response speed and energy efficiency optimization.

[0065] Edge computing module: Collects the real-time output power of each component every 15 minutes, and uses the K-means algorithm to divide the components into 3 to 5 tracking subgroups according to power similarity;

[0066] Furthermore, based on setting one master node module for every nine photovoltaic panel modules, and using a typical photovoltaic array layout, nine groups are arranged in a 3×3 matrix as a basic distribution unit. The master node module is equipped with a high-precision dual-axis solar sensor, specifically a four-quadrant silicon photovoltaic cell array + MEMS tilt meter integrated module, with an azimuth angle detection accuracy of ±0.5° and an elevation angle accuracy of ±0.6°. It also includes an edge computing module, using an STM32H743VI chip, integrating 256KB RAM and an NPU neural network acceleration unit, supporting real-time data processing and edge-side algorithm operation. It is responsible for real-time acquisition of solar azimuth angle and current tilt angle of the module, and for feedback from the encoder of the driving module and environmental temperature and humidity data (external SHT30 sensor). The slave node module is equipped with a single-axis tilt sensor, using a MEMS tilt sensor. The wireless communication module uniformly adopts the LoRaWAN protocol, with a working frequency band of 470MHz and a communication distance of ≥1km. It adjusts the tracking angle by receiving commands from the master node, which include target values ​​for azimuth angle and elevation angle.

[0067] The master node broadcasts the reference tracking angle, including azimuth and pitch angles, to the slave node every minute. The command format follows a custom binary protocol, and the data frame is 16 bytes long, including a check bit. After receiving the command, the slave node first verifies the validity of the command through its built-in circuit, and then drives the servo motor to adjust the angle with an accuracy of 0.1° / step.

[0068] Node monitoring module: Used for online monitoring between master node components and slave node components. It utilizes bidirectional communication of the wireless communication module. When a node in the master node component or slave node component goes offline, an alarm message is sent to the management center after detection.

[0069] Using the two-way heartbeat mechanism of the wireless communication module, when no response is received from the target node for 3 to 5 consecutive times (3 times is preferred), the node is determined to be offline and an alarm message is sent to the management center through the wireless communication module.

[0070] Fault-tolerant tracking module: When the master node component goes offline, during the fault period, the slave node component uses inertial navigation compensation to continue ray tracking and steering;

[0071] Furthermore, when the master node component goes offline, the slave node component immediately activates its built-in inertial navigation unit. Using a MEMS gyroscope and accelerometer at a sampling frequency of 100Hz, it establishes a motion model based on the historical tracking data from the previous 5 minutes, including angle, velocity, and acceleration, and continues to adjust the tracking angle at a predicted speed of 5° / min.

[0072] Time-sharing and zone-based sleep module: Used for one hour before or after sunrise or sunset for photovoltaic modules. The drive components of the node components enter sleep mode, with only the tilt sensor and communication module remaining operational. At night, a self-calibration mode is activated, and the drive components of the master node components and slave node components rotate synchronously to the initial horizontal position. The mechanical zero point is calibrated through the built-in accelerometer to prevent errors in the steering drive components.

[0073] The steering drive components of the master node component and the slave node component are servo motors. The steering drive component of the master node component is connected to a dual-axis solar sensor. The dual-axis solar sensor sends a drive signal to the steering drive component to start it. The steering drive component is installed on the photovoltaic panel component.

[0074] After receiving the steering information from the master node's sensor via the wireless communication module, the slave node component starts its drive components.

[0075] In this embodiment, the cost of sensors and communication links for large-scale photovoltaic arrays is significantly reduced by differentiating the hardware configuration of master and slave nodes. The number of 9 unit nodes is reduced by 8 compared to the traditional solution. The time-sharing sleep and automatic calibration mechanism reduces system energy consumption while avoiding the time cost and operational risks of manual calibration, and improves the mechanical accuracy and stability of long-term operation.

[0076] Example 2

[0077] Specifically: The node component is based on NREL's astronomical model to calculate the optimal tilt angle of the photovoltaic panel component, and is corrected by combining dual-axis solar sensor data to generate a subgroup reference tracking curve.

[0078] Furthermore, based on the NREL astronomical model, the optimal tilt angle of the photovoltaic panel module is calculated by combining the local latitude and longitude. First, the real-time UTC time of the photovoltaic panel module area is obtained by GPS positioning and converted into local true solar time. The solar declination angle, hour angle, altitude angle and azimuth angle are calculated every 3 minutes. The dual-axis solar sensor collects the actual illumination angle every 15 seconds. The values ​​are fused with the astronomical model calculation values ​​by Kalman filtering to generate the subgroup reference tracking curve, thereby realizing the node tracking correction of the photovoltaic panel module.

[0079] The specific calculation formula for the astronomical model based on NREL is as follows:

[0080] Solar declination angle δ:

[0081] δ=0.3723+23.2567sinθ+0.1149sin2θ-0.1712sin3θ-0.758cosθ+0.3656cos2θ+0.0201cos3θ

[0082] The solar angle θ is calculated from the accumulated days N over a year:

[0083]

[0084] Where N is the accumulated days of the year, calculated as January 1st = 1, December 31st in a leap year = 366, and 365.2422 is the average number of days in a year considering leap years;

[0085] Mean Time Difference (EoT):

[0086] EoT=0.0028-1.9857sinθ+9.9059sin2θ-7.0924cosθ-0.6882cos2θ

[0087] Among them, EoT is the difference between true solar time and mean solar time, with a maximum value of approximately ±15 minutes;

[0088] Solar hour angle ω:

[0089] ω=15(T s -12)

[0090] Among them, T s Local true solar time;

[0091] Solar altitude angle α:

[0092]

[0093] Where, ω rad The value is the hour angle in radians. The latitude is the local latitude, with positive numbers for the Northern Hemisphere.

[0094] Solar azimuth angle γ:

[0095]

[0096] Where cosα is the cosine of the altitude angle, avoid dividing by zero;

[0097] In this embodiment, the astronomical model calculation error is within ±1°. After real-time correction by the sensor, the accuracy of the reference angle is improved to ±0.8°, which meets the requirements of high-precision tracking. For sudden changes in illumination caused by cloudy weather, the correction algorithm can complete the angle adjustment within two sampling periods. By fusing the astronomical model with real-time sensor data, the long-term predictive stability of the astronomical model is utilized, and the real-time errors such as atmospheric refraction and local occlusion are compensated by the sensor correction. This improves the accuracy of the reference tracking curve compared to a single model or single sensor solution, and significantly reduces over-tracking or tracking lag phenomena, especially under complex weather conditions.

[0098] Example 3

[0099] Specifically: The slave node component adopts a reference angle and deviation compensation control mode. It detects the relative position deviation by the RSSI signal strength between adjacent slave node components, automatically compensates for the angle difference caused by installation error, and ensures that the photovoltaic panel component turning error of the slave node component is reduced. The slave node component realizes reference angle and deviation compensation by reading single-axis tilt sensor data through the SPI interface and reporting the status to the master node every 200ms.

[0100] Furthermore, in the reference angle and deviation compensation control mode, the reference angle issued by the master node is the average optimal tilt angle of the subgroup. The slave node uses the RSSI signal strength of the LoRa module between adjacent components. The signal strength is negatively correlated with the distance, based on a preset empirical formula:

[0101]

[0102] Where A is the transmission power and n is the environmental attenuation factor, the relative position deviation is calculated. If the tilt angle difference between adjacent slave nodes is detected to be greater than 1.5°, and reasonable deviations caused by differences in illumination are excluded, it is determined to be an installation error. The drive motor is adjusted by the built-in PID controller, and the adjustment range is half of the deviation value each time.

[0103] The fault-tolerant tracing module includes:

[0104] Central node election module: When a slave node component detects that the communication of the corresponding area master node component is interrupted, during the navigation compensation process, it sends a communication request to the corresponding master node component at preset intervals. After repeated detection 3 to 5 times, if the master node component is always offline, the ant colony reconstruction protocol is started. By releasing exploration signals, a temporary master node component collaborative core is dynamically elected based on the signal strength of adjacent master node components. The network is re-established after the corresponding area master node component resumes communication.

[0105] Furthermore, when a slave node detects a communication interruption with the master node, resulting in the loss of 3-5 consecutive heartbeat packets over a period of 30 seconds, it first enters the inertial navigation compensation mode. Based on the last received reference angle and historical motion speed, it adjusts the angle at a conservative rate of 1° / min. Simultaneously, it releases detection signals to adjacent slave nodes, including its own ID, current angle, battery level, communication quality, and other parameters. After adjacent nodes receive more than 3 exploration signals, they are sorted according to signal strength. Nodes with a signal strength > -80dBm are considered valid. The master node with the strongest signal and a battery level > 50% is elected as the temporary master node, and the network is re-established after the original master node resumes communication.

[0106] In this embodiment, the RSSI deviation compensation mechanism can correct installation errors within ±2.5°, covering more than 95% of engineering installation deviation scenarios. The temporary master node election process is completed within 15 seconds. During the fault, the tracking angle error of the slave node is controlled within ±2.5°. Non-contact position deviation detection is achieved through the RSSI signal of the neighboring nodes, eliminating the need for additional ranging sensors, reducing hardware costs. The centerless node election mechanism avoids the single-point failure risk of traditional centralized systems, shortening the system reconstruction time to 15 seconds when the master node fails, which is a significant improvement over traditional solutions, ensuring the tracking continuity of photovoltaic modules under complex working conditions.

[0107] Example 4

[0108] Specifically: The master and slave nodes of the photovoltaic panel module automatically form a shadow adaptation group based on the power deviation of the modules that are shaded in front of it >20%, and adopt a lag tracking strategy. The tracking speed of the slave node modules is reduced by 50% to reduce mechanical losses. The unshaded modules in the center form a high-efficiency tracking group to perform real-time dynamic tracking. The number of photovoltaic panel modules in the group is automatically adjusted according to the change of light intensity.

[0109] Specifically, the high-efficiency tracking group (power deviation < 10%): Execution, real-time response mode, the master node updates the reference angle every 30 seconds, and the servo motor dynamically adjusts at a speed of 4° / minute to ensure a response delay of < 1 minute for rapid cloud movement. The shadow adaptation group (power deviation > 20%): Activation, energy-saving tracking mode, the reference angle update cycle is extended to 2 minutes, the tracking speed is reduced to 2° / minute, and the motor drive voltage is reduced by 15%, reducing wear on mechanical parts and effectively reducing the wear of the steering drive components.

[0110] The specific calculation formula for the K-means algorithm in the edge computing module is as follows:

[0111] X for each data point i Assigned to the cluster C to which the nearest centroid belongs. j :

[0112]

[0113] Where, μ k Let ||k| be the centroid vector of the k-th cluster, and ||·|| be the Euclidean distance.

[0114] Recalculate the centroid based on the data points within the cluster:

[0115]

[0116] Among them, |C k |For cluster C k The number of data points within;

[0117] By iteratively optimizing the "assignment-update" process, the sum of squares within the cluster is minimized:

[0118]

[0119] Where K is the preset number of clusters.

[0120] The K-means grouping algorithm employs an incremental update mechanism, real-time acquisition of the output power of each component via a Hall current sensor integrated into the inverter, and synchronous recording of the current solar altitude angle from the master node sensor and the tilt angle of the photovoltaic panel components. The initial centroid selection adopts a power ternary method, sorting the components by power from high to low, and selecting the median of the top 20%, the middle 60%, and the bottom 20% as the initial cluster center to avoid clustering bias caused by random initialization. After grouping, differentiated control is implemented for different subgroups.

[0121] The intelligent control system also includes an energy consumption optimization module: it collects historical tracking data of each subgroup through the management and control center, trains the optimal grouping strategy through a deep reinforcement learning model, and automatically updates the grouping parameters of the edge computing module every quarter to achieve continuous optimization of the tracking efficiency of the intelligent control system for photovoltaic modules.

[0122] Furthermore, the management and control center collects historical tracking data of photovoltaic modules through wireless communication. The tracking data includes daily power generation curves, grouping strategies, fault records, etc., with a storage period of 1 year. At the end of each quarter, the optimal grouping strategy is trained using a deep reinforcement learning model. The state space includes features such as light intensity, module spacing, and drive motor running time. The action space consists of cluster number adjustment and tracking speed parameters. The generated optimization parameters are remotely updated to the edge computing module via OTA.

[0123] Furthermore, the strategy iteration is achieved through the deep reinforcement learning model of the management center. The model input includes historical illumination data (from the weather station API), component aging degree (evaluated by power decay curve), driving unit runtime and other features. The output is the optimal number of groups and the tracking parameters of each subgroup (such as speed and accuracy thresholds). At the end of each quarter, the management center pushes the updated strategy file to the edge computing module through a secure encrypted channel. The edge nodes complete the firmware upgrade during the low load period at night.

[0124] In this embodiment, the specific implementation process of the K-means algorithm is as follows: For example, with a period of 15 minutes, the output power, tilt angle, and ambient temperature and humidity of each component are collected as feature vectors. The initial centroid is selected from the three photovoltaic panel components with the highest, middle, and lowest power in this period. The distance between each component and the centroid is calculated by Euclidean distance and assigned to the nearest cluster. Then, the mean value within the cluster is recalculated as the new centroid. The iteration continues until the change in the centroid position is <0.5° or the upper limit of 6 iterations is reached. After grouping, for the slave node components of the shadow adaptation group, the servo motor drive voltage is reduced from 24V to 12V by PWM signal, so that the tracking speed is reduced from 4° / minute to 2° / minute. The high-efficiency tracking group maintains the original voltage and the tracking speed is maintained at 4° / minute.

[0125] Furthermore, the power grouping threshold of the K-means algorithm is dynamically calibrated using field measurement data. The initial threshold is set at 20%, and subsequently adjusted by ±5% based on seasonal changes. This dynamic grouping strategy, based on power and environmental characteristics, reduces the mechanical losses of components in shaded areas by 40%. Simultaneously, the grouping is automated by using a deep reinforcement learning model for training. The deep reinforcement learning model has a training cycle of 72 hours, which can improve the average annual tracking efficiency by 8% to 12%. The deep reinforcement learning model training, driven by historical data, achieves self-optimization of the grouping strategy, avoiding the lag of manual experience adjustments. Especially in plateau and mountain power plants with complex lighting conditions, it can significantly improve the long-term power generation efficiency and equipment lifespan.

[0126] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A photovoltaic power generation intelligent control system for photovoltaic panels in a photovoltaic-storage-DC-flexible system, characterized in that, The intelligent control system includes multiple photovoltaic panel modules, which include: Master node component: One master node component is set for every 5-10 groups of photovoltaic panels. The master node component is equipped with a high-precision dual-axis solar sensor and an edge computing module, which is responsible for collecting solar azimuth angle, current tilt angle of the panel and ambient temperature and humidity data in real time. Both the master node component and the slave node component are equipped with wireless communication modules for bidirectional communication between them. Slave node component: The slave node component is equipped with a single-axis tilt sensor, which adjusts the tracking angle by receiving commands from the master node; The edge computing module collects the real-time output power of each component every 15 minutes and uses the K-means algorithm to divide the components into 3 to 5 tracking subgroups according to power similarity. Node monitoring module: used for online monitoring between the master node component and the slave node component. Utilizing the bidirectional communication of the wireless communication module, when a node in the master node component or the slave node component goes offline, an alarm message is sent to the management center after detection. Fault-tolerant tracking module: When the master node component goes offline, during the fault period, the slave node component uses inertial navigation compensation to continue tracking and turning the light rays; Time-sharing and zone-based sleep module: When the photovoltaic panel module is in the sleep state for one hour before or after sunrise or sunset, the steering drive component of the slave node module enters the sleep state, and only the tilt sensor and communication module remain working. At night, the self-calibration mode is activated, and the drive components of the master node module and slave node module rotate synchronously to the initial horizontal position. The mechanical zero point is calibrated by the built-in accelerometer to prevent errors in the steering drive component.

2. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 1, characterized in that: The master node component calculates the optimal tilt angle of the photovoltaic panel based on the NREL astronomical model, and then corrects it by combining the data from the dual-axis solar sensor to generate the subgroup reference tracking curve.

3. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 2, characterized in that, The specific calculation formula for the NREL-based astronomical model is as follows: Solar declination angle δ: δ=0.3723+23.2567sinθ+0.1149sin2θ-0.1712sin3θ-0.758cosθ+0.3656cos2θ+0.0201cos3θ The solar angle θ is calculated from the accumulated days N over a year: Where N is the accumulated days of the year, calculated as January 1st = 1, December 31st in a leap year = 366, and 365.2422 is the average number of days in a year considering leap years; Mean Time Difference (EoT): EoT=0.0028-1.9857sinθ+9.9059sin2θ-7.0924cosθ-0.6882cos2θ Among them, EoT is the difference between true solar time and mean solar time, with a maximum value of approximately ±15 minutes; Solar hour angle ω: ω=15(T s -12) Among them, T s Local true solar time; Solar altitude angle α: Where, ω rad The value is the hour angle in radians. The latitude is the local latitude, with positive numbers for the Northern Hemisphere. Solar azimuth angle γ: Here, cosα is the cosine of the altitude angle, and it should not be divided by zero.

4. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-DC-flexible system according to claim 1, characterized in that: The slave node component adopts a reference angle and deviation compensation control mode. It detects the relative position deviation by measuring the RSSI signal strength between adjacent slave node components, automatically compensates for the angle difference caused by installation error, and ensures that the photovoltaic panel component of the slave node component is turned less.

5. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 4, characterized in that: The slave node component implements the reference angle and deviation compensation by reading data from the single-axis tilt sensor through the SPI interface and reporting the status to the master node every 200ms.

6. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 1, characterized in that: The fault-tolerant tracking module includes: Central node election module: When the slave node component detects that the communication of the master node component in the corresponding area is interrupted, during the navigation compensation process, it sends a communication request to the corresponding master node component at preset intervals, and repeatedly detects 3 to 5 times. If the master node component is always offline, the ant colony reconstruction protocol is initiated. By releasing exploration signals, a temporary master node component collaborative core is dynamically elected based on the signal strength of adjacent master node components. The network is re-established after the master node components in the corresponding area resume communication.

7. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 1, characterized in that: The specific calculation formula of the K-means algorithm of the edge computing module is as follows: X for each data point i Assigned to the cluster C to which the nearest centroid belongs. j : Where, μ k Let ||k| be the centroid vector of the k-th cluster, and ||·|| be the Euclidean distance. Recalculate the centroid based on the data points within the cluster: Among them, |C k |For cluster C k The number of data points within; By iteratively optimizing the "assignment-update" process, the sum of squares within the cluster is minimized: Where K is the preset number of clusters.

8. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 1, characterized in that: The master and slave nodes of the photovoltaic panel module automatically form a shadow adaptation group based on the power deviation of the modules that are blocked by the front row, which is >20%. A lag tracking strategy is adopted, and the tracking speed of the slave node modules is reduced by 50% to reduce mechanical losses. The unblocked modules in the center form a high-efficiency tracking group to perform real-time dynamic tracking. The number of groups of the photovoltaic panel modules is automatically adjusted with the change of light intensity.

9. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-direct-drive-flexible system according to claim 1, characterized in that: The intelligent control system also includes an energy consumption optimization module: it collects historical tracking data of each subgroup through the management and control center, trains the optimal grouping strategy through a deep reinforcement learning model, and automatically updates the grouping parameters of the edge computing module every quarter to achieve continuous optimization of the tracking efficiency of the intelligent control system of the photovoltaic panel module.

10. The intelligent control system for photovoltaic power generation of photovoltaic panels in a photovoltaic-storage-DC-flexible system according to claim 1, characterized in that: The steering drive components of the master node component and the slave node component are servo motors. The steering drive component of the master node component is connected to a dual-axis solar sensor. The dual-axis solar sensor sends a drive signal to the steering drive component to start it. The steering drive component is installed on the photovoltaic panel component. The slave node component receives the steering information from the master node sensor via the wireless communication module, and then its drive component is activated.

Images