Solar generator speed regulation and control monitoring system

Abstract

A solar generator speed regulation and control monitoring system comprises a multi-source fusion sensing unit used for collecting multi-dimensional heterogeneous data in the operation process of a solar power generation system in real time; the self-adaptive collaborative decision-making unit is used for receiving the data from the multi-source fusion sensing unit and generating collaborative control instructions oriented to different execution mechanisms based on a preset collaborative optimization model and algorithm; the multi-mode execution driving unit is used for receiving and executing the cooperative control instruction issued by the self-adaptive cooperative decision-making unit and driving a corresponding physical execution mechanism to act; and the global state monitoring and diagnosis unit is used for gathering and analyzing state data from the multi-source fusion sensing unit and each unit in the system in real time, and realizing global visualization, performance evaluation and early fault early warning of the system operation state.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of solar power generation, and particularly relates to a solar power generator speed regulation and monitoring system. BACKGROUND

[0002] With the rapid development of renewable energy technology, solar power generation has become an important direction of global energy structure transformation due to its clean and sustainable characteristics. Photovoltaic power generation, as the main form of solar energy utilization, is based on the photovoltaic effect to directly convert solar radiation energy into electrical energy. The overall efficiency and operational stability of the power generation system are key factors in determining its economic and reliable performance.

[0003] Among them, the solar power generator speed regulation and monitoring system aims to comprehensively control the operating state of the power generation unit to maximize energy capture efficiency and maintain system stability. The system usually involves tracking control of photovoltaic arrays, maximum power point tracking, energy storage management, and grid-connected / off-grid operation mode switching, among other interrelated sub-processes. The goal is to dynamically match the output power of the power generation system with load demand or grid scheduling instructions.

[0004] In the prior art, there are many integrated challenges in the operation regulation and monitoring of solar power generators. Traditional photovoltaic array tracking mechanisms rely on single sensor feedback, with limited tracking accuracy and susceptibility to environmental interference, resulting in insufficient utilization of solar irradiance. Conventional maximum power point tracking algorithms have a lag in response when environmental parameters such as light and temperature change rapidly, causing power generation losses. The power regulation response speed of the energy storage system is slow, making it difficult to effectively smooth the inherent intermittency and volatility of solar power generation, affecting power supply quality. When operating in grid-connected mode, inverters lacking virtual synchronous machine characteristics cannot provide the necessary inertia and damping support, easily causing grid frequency and voltage fluctuations. At the same time, existing monitoring systems are mostly dispersed parameter collection, lacking coordinated analysis and intelligent regulation of the state of power generation, tracking, energy storage, and other links, and cannot achieve global optimization and fault warning of system operating state. Therefore, how to build a solar power generator speed regulation and monitoring system that can be synergistically optimized, respond in real time, and have high stability is a technical problem that needs to be solved. SUMMARY

[0005] A solar power generator speed regulation and monitoring system, characterized by comprising: A multi-source fusion perception unit for real-time acquisition of multi-dimensional heterogeneous data during the operation of a solar power generation system; An adaptive collaborative decision-making unit for receiving data from the multi-source fusion perception unit and generating collaborative control instructions for different execution mechanisms based on a pre-set collaborative optimization model and algorithm; The multimodal execution drive unit is used to receive and execute the collaborative control commands issued by the adaptive collaborative decision-making unit, and drive the corresponding physical actuators to move. The global status monitoring and diagnosis unit is used to collect and analyze status data from the multi-source fusion sensing unit and various units within the system in real time, so as to realize global visualization of system operation status, performance evaluation and early fault warning. The adaptive collaborative decision-making unit runs a hierarchical collaborative optimization control algorithm. The collaborative optimization control algorithm constructs a global optimization function at the top level; in the middle layer, the top-level global optimization objective is decomposed into several mutually coupled sub-optimization tasks; at the bottom level, a dedicated adaptive controller is designed for each sub-optimization task, and dynamic coupling constraints and information interaction protocols between sub-tasks are defined.

[0006] The sub-optimization tasks include: photovoltaic array dual-axis tracking control task, maximum power point tracking control task, energy storage system power smoothing control task, and grid-connected inverter virtual synchronous machine control task.

[0007] The execution process of the photovoltaic array dual-axis tracking control task is as follows: The adaptive collaborative decision-making unit first integrates data from a high-precision solar position sensor array and uses the weighted least squares method to calculate the precise azimuth and altitude angles of the sun in the sky coordinate system. The adaptive collaborative decision-making unit receives real-time power generation data from the distributed photovoltaic array sensor network; The adaptive collaborative decision-making unit has a built-in tracking strategy optimizer based on reinforcement learning. This optimizer uses maximizing the total output power of the photovoltaic array at the current moment as the immediate reward and the energy consumption of the tracking mechanism's actions as the penalty. It learns online and dynamically adjusts the motion trajectory of the tracking mechanism. The adaptive collaborative decision-making unit outputs the optimal tracking command, which includes azimuth and pitch angle adjustments, to the multimodal execution drive unit.

[0008] The execution process of the maximum power point tracking control task is as follows: The adaptive collaborative decision-making unit calculates the output power characteristic curve of the photovoltaic array in real time based on the voltage and current data collected by the distributed photovoltaic array sensor network. The adaptive collaborative decision-making unit employs an adaptive algorithm that combines a hybrid perturbation observation method with an admittance increment method. During periods of gradual changes in light intensity and temperature, the algorithm employs a variable step size admittance increment method for precise optimization. When a sudden change in light or temperature is detected, the algorithm automatically switches to the perturbation observation method mode with a large-area scan to quickly locate the new maximum power point region, and then switches back to the admittance increment method for fine adjustment. The step size adjustment coefficient in the algorithm is dynamically correlated with the real-time open-circuit voltage and short-circuit current of the photovoltaic array.

[0009] The execution process of the power smoothing control task of the energy storage system is as follows: The adaptive collaborative decision-making unit calculates the difference between the instantaneous power of photovoltaic power generation and the smoothed power reference value after processing by a low-pass filter in real time. This difference is the power command that the energy storage system needs to absorb or release. The adaptive collaborative decision-making unit adopts a power allocation strategy based on model predictive control; The strategy uses the future time period as the prediction time domain, maintains the state of charge of the energy storage system within a preset safe range, and aims to minimize the charge and discharge cycle losses and power fluctuations of the energy storage system. It optimizes the solution of the optimal charge and discharge power command of the energy storage converter in each control cycle.

[0010] The execution process of the grid-connected inverter virtual synchronous machine control task is as follows: The adaptive collaborative decision-making unit is used to simulate the rotor motion equation and electromagnetic transient characteristics of a synchronous generator in a grid-connected inverter. The adaptive collaborative decision-making unit dynamically calculates the angular velocity and power angle of the virtual rotor based on the power grid frequency and voltage information collected by the power grid interaction interface sensor. The adaptive collaborative decision-making unit adjusts the rotational inertia parameter and damping coefficient parameter of the virtual rotor, enabling the inverter to actively adjust its output active power according to the grid frequency deviation and actively adjust its output reactive power according to the grid voltage deviation. The virtual moment of inertia and damping coefficient are adaptively adjusted based on the real-time strength assessment results of the power grid.

[0011] The multimodal execution drive unit includes a high-precision dual-axis tracking servo driver, a bidirectional energy storage converter driver, and a full-bridge inverter driver. The high-precision dual-axis tracking servo driver receives tracking commands from the adaptive collaborative decision-making unit and drives the azimuth motor and pitch motor to move to the designated position with preset acceleration and velocity curves. The bidirectional energy storage converter driver receives power smoothing control commands and controls the switching timing of insulated gate bipolar transistors or metal oxide semiconductor field-effect transistors to realize bidirectional, fast, and precise energy flow between the energy storage battery pack and the DC bus. The full-bridge inverter driver receives virtual synchronous machine control commands, generates drive signals through space vector pulse width modulation technology, controls the power switching devices of the inverter bridge arm, realizes the conversion of DC to AC, and injects current with virtual synchronous machine characteristics into the power grid.

[0012] The global status monitoring and diagnosis unit includes a data fusion center, a performance evaluation engine, and a fault diagnosis inference engine; Receive and time-align all sensor data from the multi-source fusion sensing unit, as well as the internal state data from the adaptive collaborative decision-making unit and the multimodal execution drive unit; The performance evaluation engine calculates system-level performance indicators in real time based on the fused data, including daily cumulative power generation, real-time power generation efficiency, maximum power point tracking efficiency, energy storage system smoothing effect indicators, and grid connection point power quality qualification rate. The fault diagnosis inference engine has a built-in rule-based and case-based hybrid knowledge base, which defines the multi-parameter association patterns of each component of the system under normal and typical fault states. The fault diagnosis inference engine continuously compares the real-time data stream with the patterns in the knowledge base. When it detects that the parameters deviate from the normal association pattern and continue to exceed the preset sampling period, it generates early warning information that includes the location of the faulty component, the fault type, and the confidence level.

[0013] A two-way feedback channel is established between the adaptive collaborative decision-making unit and the global status monitoring and diagnosis unit; The global status monitoring and diagnosis unit feeds back the performance evaluation results and fault early warning information to the adaptive collaborative decision-making unit in real time. The adaptive collaborative decision-making unit dynamically adjusts the weight parameters and constraints of its internal optimization algorithm based on feedback information. The process by which the adaptive collaborative decision-making unit dynamically adjusts the weight parameters and constraints of its internal optimization algorithm based on feedback information is as follows: When a warning of performance degradation of a photovoltaic string is received, the adaptive collaborative decision-making unit reduces the power generation weight of the string in the global optimization function and adjusts the dependence of the maximum power point tracking algorithm on the string accordingly. When a decline in the health status of the energy storage system is detected, the adaptive collaborative decision-making unit tightens the safety range constraint of the energy storage system's state of charge.

[0014] The multi-source fusion sensing unit includes a high-precision solar position sensor array, a distributed photovoltaic array sensor network, an energy storage system status sensor group, and a grid interaction interface sensor. The distributed photovoltaic array sensor network is used to synchronously collect the output voltage, output current, surface temperature and backsheet temperature of each or each group of photovoltaic modules. The energy storage system status sensor group is used to monitor the state of charge, health status, charging and discharging current, terminal voltage and internal temperature distribution of the energy storage battery pack in real time. The power grid interaction interface sensor is used to collect the voltage, frequency, phase, and power factor of the grid connection point in real time.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The combination of a high-precision solar position sensor array and a reinforcement learning tracking strategy significantly improves tracking accuracy and energy capture efficiency under complex weather conditions. The hybrid adaptive maximum power point tracking algorithm effectively overcomes the power loss problem caused by sudden environmental changes. The energy storage power allocation strategy based on model predictive control smooths power fluctuations while optimizing the management of the energy storage system's operating status and lifespan. The introduction of virtual synchronous machine control technology enables the solar power generation system to actively support the stable operation of the power grid and improves the security of the power grid under a high proportion of new energy access. The global status monitoring and diagnostic unit realizes transparent, quantitative management and intelligent early warning of the system's operating status, providing a data foundation for preventive maintenance and continuous optimization of system performance. The bidirectional information flow between units forms a complete intelligent closed loop of perception, decision-making, execution, evaluation and optimization, enabling the entire system to have self-learning and self-adaptive capabilities and continuously cope with complex and ever-changing internal and external operating environments. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall technical architecture of the solar generator speed regulation and monitoring system proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the hierarchical collaborative optimization control algorithm in this invention; Figure 3 This is a logical flow diagram of the photovoltaic array dual-axis tracking control task in this invention; Figure 4 This is a schematic diagram of the multi-level interaction and data flow between the power smoothing control task of the energy storage system and the virtual synchronous machine control task of the grid-connected inverter in this invention. Detailed Implementation

[0017] Please refer to the attached document. Figures 1-2 A solar generator speed regulation and monitoring system includes: a multi-source fusion sensing unit, an adaptive collaborative decision-making unit, a multi-modal execution driving unit, and a global state monitoring and diagnosis unit.

[0018] The multi-source fusion sensing unit is used to collect multi-dimensional heterogeneous data in real time during the operation of the solar power generation system. The adaptive collaborative decision-making unit receives data from the multi-source fusion sensing unit and, based on a preset collaborative optimization model and algorithm, generates collaborative control commands for different actuators. The multimodal execution drive unit receives and executes the collaborative control commands issued by the adaptive collaborative decision-making unit, driving the corresponding physical actuators to move. The global status monitoring and diagnosis unit aggregates and analyzes status data from the multi-source fusion sensing unit and various units within the system in real time, enabling global visualization of the system's operating status, performance evaluation, and early fault warning.

[0019] The multi-source fusion sensing unit is responsible for real-time acquisition of multi-dimensional heterogeneous data during the operation of the solar power generation system. This unit comprises four core sensor subsystems; The first subsystem is a high-precision solar position sensor array, which consists of at least two solar position sensors with different physical principles. This ensures that the field of view coverage has no overlapping blind spots and can effectively eliminate measurement deviations caused by local cloud obstruction, lens contamination or instantaneous errors of a single sensor through data fusion algorithms. The second subsystem is a distributed photovoltaic array sensor network, which is deployed at each string combiner box of the photovoltaic array and on the backplane of a representative photovoltaic module; The third subsystem is the energy storage system status sensor group. This sensor group is integrated into the energy storage battery management system and monitors the core parameters of the energy storage battery pack in real time, including the state of charge estimated based on the fusion of the ampere-hour integral method and the open-circuit voltage method, the health status calculated based on the battery internal resistance and capacity decay model, the total charging and discharging current of the battery pack, the total terminal voltage of the battery pack, and the internal temperature distribution data collected by several temperature sensors distributed at key points of the battery module; The fourth subsystem is the grid interaction interface sensor. This sensor group is installed on the AC output side of the grid-connected inverter. It collects the instantaneous values ​​of the three-phase voltage and three-phase current at the grid connection point in real time using a sampling rate. Based on these instantaneous values, the grid frequency, the phase difference between each phase voltage and current, and the real-time power factor of the system are calculated in real time using digital signal processing algorithms.

[0020] Please refer to the attached document. Figure 2The adaptive collaborative decision-making unit runs a hierarchical collaborative optimization control algorithm. At the top level, this algorithm constructs a global optimization function aimed at maximizing the power generation benefits throughout the system's entire lifecycle. This function is a multi-objective weighted sum function, with decision variables encompassing the azimuth and elevation angles of the photovoltaic array, the operating voltage of the photovoltaic array, the charging and discharging power of the energy storage system, and the active and reactive power output of the grid-connected inverter. The objective function comprehensively considers real-time irradiance capture efficiency, maximum power point tracking efficiency, energy storage system charging and discharging efficiency, grid-connected power quality indicators, and equipment lifespan depreciation costs. Specifically, the equipment lifespan depreciation cost is converted into economic costs through an equivalent cycle number model, converting tracking motor operation, energy storage battery charging and discharging depth, and power device switching losses into economic costs. The weight coefficients of this global optimization function are refreshed at a preset period, and the adjustment of the weight coefficients is based on the system performance evaluation results and equipment health status fed back by the global status monitoring and diagnosis unit. In the intermediate layer, the algorithm decomposes the top-level global optimization objective into four mutually coupled sub-optimization tasks: photovoltaic array dual-axis tracking control task, maximum power point tracking control task, energy storage system power smoothing control task, and grid-connected inverter virtual synchronous machine control task. Intermediate variables include the predicted output power of the photovoltaic array, the expected value of the DC bus voltage, the real-time state of charge of the energy storage system, and the frequency and voltage deviation of the grid. Dynamic coupling constraints define the intersection of the feasible domains of the control instructions of each sub-task. For example, the maximum charging power of the energy storage system is constrained by the current maximum output power of the photovoltaic system and the maximum grid-connected power of the inverter. Information interaction between sub-tasks is achieved through a shared memory area, and data synchronization is performed at the beginning of each control cycle. At the bottom layer, the algorithm designs a dedicated adaptive controller for each sub-optimization task. This controller receives sub-task target instructions from the intermediate layer and real-time data from the multi-source fusion sensing unit. It then generates specific, executable control instructions through a dedicated internal control algorithm and sends them to the multi-modal execution drive unit. There are also rapid information exchange channels between the bottom-layer controllers. For example, when a sudden frequency drop occurs in the power grid, the virtual synchronous machine controller needs to immediately call upon the backup power of the energy storage system.

[0021] Please refer to the attached document. Figure 3The execution logic of the photovoltaic array dual-axis tracking control task is as follows: The adaptive collaborative decision-making unit first receives raw data from the high-precision solar position sensor array. Each sensor provides a set of measurement values ​​containing the coordinates of the solar mass center in the sensor's local coordinate system. The data preprocessing module inside the unit uses the weighted least squares method to fuse and solve multiple sets of measurement values. The weighting weights are dynamically allocated according to the confidence level of each sensor at the current moment. The confidence level is obtained by evaluating the consistency between the sensor's own state self-check data and historical measurements. The solution process finally outputs the precise azimuth angle α_s and altitude angle β_s of the sun in the sky coordinate system. At the same time, the unit continuously receives real-time power generation data P_pv_i of each photovoltaic string from the distributed photovoltaic array sensor network. The unit's built-in reinforcement learning-based tracking strategy optimizer uses the total output power ΣP_pv_i of the photovoltaic array in the current control cycle as the immediate reward R_t and the electrical energy E_move consumed by the tracking mechanism to move from the previous position to the new position as the penalty term C_t. The optimizer constructs a state-action value function Q, where the state S includes the sun's position, cloud cover index, and the current array angle, and the action A is the adjustment amount for the azimuth and pitch angles. The optimizer continuously updates the Q function through online learning, aiming to maximize the cumulative discount reward. Its update rule follows the principle of temporal difference learning, and the core formula is:

[0022] Here, η represents the learning rate, initially set to 0.1 and dynamically decaying according to the learning progress; γ represents the discount factor, set to 0.9, used to balance current rewards with future rewards. Finally, in each control cycle, the optimizer selects the action that maximizes the Q value, i.e., the optimal azimuth adjustment Δα and pitch adjustment Δβ, and generates the tracking command.

[0023] The maximum power point tracking (MPPT) control task is executed by an adaptive controller. Based on string-level voltage U_pv and current I_pv data collected by a distributed photovoltaic array sensor network, it calculates the output power P_pv = U_pv * I_pv of the photovoltaic array in real time and constructs a local model of the power-voltage characteristic curve near the current operating point. Under normal conditions, i.e., when the controller determines that the changes in illumination and temperature are gradual by calculating the power change rate dP / dU and the conductance change rate dI / dU in adjacent cycles, the algorithm operates in admittance increment mode. At this time, the control quantity, i.e., the update step size ΔU of the photovoltaic array operating voltage reference value U_ref and the admittance increment value dI / dU + ... The I / U dynamic correlation involves automatically reducing the step size for fine-tuning when the value approaches zero, preventing oscillations near the maximum power point. When the sensor network detects that the rate of change of a string current exceeds a preset value within a preset time interval, or when the temperature sensor detects a sudden temperature change in a component exceeding a preset value, the controller determines that an environmental change has occurred. The algorithm immediately switches to a large-range scanning perturbation observation method mode. In this mode, the controller rapidly scans the power-voltage curve with a large step size and unidirectional perturbation of the operating voltage to locate the new maximum power point region. Once the power change trend stabilizes, the algorithm switches back to the admittance increment method for final precise positioning. The reference voltage parameter in this algorithm is dynamically correlated with the real-time open-circuit voltage and short-circuit current of the photovoltaic array, and is updated once every preset time interval based on sensor data.

[0024] Please refer to the attached document. Figure 4The power smoothing control task of the energy storage system and the virtual synchronous machine control task of the grid-connected inverter have a hierarchical interaction relationship. The energy storage power smoothing controller first receives the instantaneous power P_pv_inst from the photovoltaic array. This power is processed by a low-pass digital filter to generate a smooth power reference value P_smooth_ref. The difference between the two is ΔP = P_pv_inst - P_smooth_ref, which is the power command that the energy storage system needs to absorb or release. ΔP is positive when the energy storage is charging and negative when it is discharging. This controller employs a model predictive control-based power allocation strategy. The strategy uses the next few minutes as the prediction time domain and the unit of seconds as the control time domain. In each control cycle, the controller uses historical irradiance data and weather forecast information to predict the trend of photovoltaic power P_pv_pred within the next few minutes. Simultaneously, a simplified mathematical model of the energy storage system is established, with the state of charge (SOC) as the state variable and the charge / discharge power P_batt as the control variable. The SOC is forced to remain within a preset safety range for each prediction step. The objective function J aims to minimize the charge / discharge cycle losses and power fluctuations of the energy storage system. Specifically, it minimizes the integral of the sum of squares of the charge / discharge power within the prediction time domain, incorporating a penalty term for deviations from the ideal SOC. The controller solves a constrained quadratic programming problem, using rolling optimization to obtain the optimal charge / discharge power command sequence for the current control cycle and the next few cycles, but only the first command, P_batt_cmd, is executed. The sign and magnitude of this command directly determine whether the energy storage system is charging or discharging, and the energy throughput rate.

[0025] The virtual synchronous machine control task of the grid-connected inverter is executed by the virtual synchronous machine controller, which simulates the rotor motion equation and electromagnetic transient characteristics of the synchronous generator for the grid-connected inverter. The rotor motion equation simulates mechanical inertia, calculates the angular velocity ω_v of the virtual rotor, and the controller receives the grid angular frequency ω_g from the grid interface sensor in real time. The motion of the virtual rotor follows a second-order differential equation:

[0026] Where J represents the virtual moment of inertia, an adjustable parameter with a typical value range corresponding to a time constant of 2 to 10 seconds; D represents the virtual damping coefficient; T_m represents the virtual mechanical torque, calculated and mapped from the DC power provided by photovoltaic and energy storage; T_e represents the virtual electromagnetic torque, proportional to the active power output of the inverter. The angular velocity ω_v of the virtual rotor is obtained by solving this equation through real-time numerical integration, and the virtual power angle δ is obtained by integrating it. The electromagnetic transient characteristics are realized by simulating the excitation system of a synchronous generator. Based on the deviation between the grid voltage amplitude U_g and the reference voltage U_ref collected by the grid interface sensor, the virtual excitation voltage is calculated and output through a proportional-integral regulator, thereby controlling the reactive power Q_cmd output by the inverter. The virtual moment of inertia J and the damping coefficient D are not fixed values. The controller has a built-in grid strength assessment module. This module calculates a grid strength coefficient between 0 and 1 based on real-time data such as the grid frequency change rate and voltage fluctuation amplitude. When the coefficient is high, it indicates a strong grid, so J and D are appropriately reduced to allow the inverter to respond to power commands more quickly. When the coefficient is low, it indicates a weak grid, so J and D are increased to make the inverter exhibit stronger inertia and damping, supporting the grid in a more gentle way. The active power command P_vsm_cmd and reactive power command Q_vsm_cmd generated by the virtual synchronous machine controller, together with P_batt_cmd output by the energy storage system power smoothing controller and the smoothed photovoltaic power P_smooth_ref, constitute the coordinated requirements for the final output power of the inverter.

[0027] The multimodal execution drive unit is the physical execution layer of the system, responsible for converting the digital control commands issued by the adaptive collaborative decision-making unit into high-precision physical actions or power conversions. This unit includes three types of drivers. The first type is a high-precision dual-axis tracking servo driver, which receives tracking commands containing the target azimuth angle α_target and the target pitch angle β_target. The driver internally employs a fully digital three-loop control structure for position, speed, and current. The position loop receives the target angle and compares it with the actual angle fed back by the absolute encoder mounted on the motor shaft. The difference is calculated by a proportional-integral-derivative controller to generate the speed command. The speed loop ensures that the motor moves smoothly according to a preset S-shaped acceleration curve, with the maximum angular velocity limited to 3 degrees per second to avoid mechanical shock. The current loop precisely controls the motor's torque output. Ultimately, the driver drives the azimuth rotary motor and the pitch motor. The second type is the bidirectional energy storage converter driver. This driver receives the power smoothing control command P_batt_cmd. The driver adopts a dual active bridge topology and achieves electrical isolation through a high-frequency transformer. Its core control algorithm is phase-shift control. Based on the sign and magnitude of P_batt_cmd, the driver calculates the phase difference φ between the corresponding switching transistor drive signals of the primary and secondary bridges. By precisely adjusting φ, it controls the energy flow from the DC bus to the battery pack or vice versa. The switching devices use metal-oxide-semiconductor field-effect transistors. The driver monitors the battery terminal voltage and the DC bus voltage in real time. The third type is the full-bridge inverter driver. The driver receives the final active power command P_inv_cmd and reactive power command Q_inv_cmd from the adaptive cooperative decision-making unit. The driver employs a voltage source inverter topology with a dual-loop control strategy: an outer voltage loop and an inner current loop. The inner current loop uses decoupled control based on a rotating coordinate system, decomposing the command into a direct-axis current component I_d_cmd and a quadrature-axis current component I_q_cmd. The space vector pulse width modulation module calculates and generates drive signals based on the current DC bus voltage and current commands, precisely controlling the on / off timing of the insulated-gate bipolar transistors on the three bridge arms. This synthesizes a three-phase sinusoidal AC current with the required amplitude, frequency, and phase at the inverter output, which is then injected into the grid. The driver strictly guarantees that the total harmonic distortion (THD) of the output current is below 3%.

[0028] The Global Status Monitoring and Diagnosis Unit serves as the system's health management and performance evaluation center. This unit comprises three core components: a data fusion center, a performance evaluation engine, and a fault diagnosis inference engine. The data fusion center receives raw data streams from all sensors in the multi-source fusion sensing unit, as well as status flags and intermediate variables of each controller in the adaptive collaborative decision-making unit, and real-time feedback data from the multi-modal execution drive unit, such as the actual position of the motor, the actual power of the converter, and the switching temperature of the inverter. The data fusion center adds a unified time stamp to all input data and performs validity verification, removes outliers that are significantly beyond the physical range, and uses a first-order hold algorithm to interpolate temporarily lost data to ensure that the output to the subsequent engine is a complete, time-aligned system panoramic state dataset. The performance evaluation engine, based on the panoramic dataset provided by the data fusion center, calculates a series of system-level and subsystem-level performance indicators in real time. These indicators include: daily cumulative power generation, obtained by integrating the total output power of the photovoltaic array; real-time power generation efficiency, which is the ratio of the actual output power of the photovoltaic array to the theoretical irradiance reaching the surface of the photovoltaic module during the same period; maximum power point tracking efficiency, which is the ratio of the actual output power of the photovoltaic array to the theoretical maximum possible power under the current environment, with the theoretical maximum power obtained by combining standard test condition data provided by the photovoltaic module manufacturer with real-time temperature and irradiance correction; energy storage system smoothing effect indicator, defined as the ratio of the original power fluctuation amplitude of the photovoltaic system to the grid-connected power fluctuation amplitude after energy storage smoothing; and grid-connected power quality qualification rate, which judges whether indicators such as voltage deviation, frequency deviation, harmonic content, and three-phase imbalance are qualified in real time according to national standards, and calculates the 24-hour rolling qualification rate. All performance indicators are visualized in the form of numbers and trend curves through a human-computer interaction interface. The fault diagnosis inference engine incorporates a hybrid knowledge base based on rules and cases. The rule base draws from technical specifications provided by equipment manufacturers, operation and maintenance manuals, and the experience of domain experts. It defines the relationships between key parameters of various system components under normal conditions in an "if-then" format, such as: "If the solar azimuth angle is X degrees and the cloudless index is high, then the output current of photovoltaic string A should be within the interval [I_min, I_max]." The case base stores typical fault events recorded in the system's historical operation and their corresponding multi-parameter abnormal mode feature vectors. For example, the feature vector for "overheating fault of a certain insulated gate bipolar transistor" includes data from multiple dimensions such as distortion of the device's drive waveform, abnormal heat sink temperature gradient, and increased output current harmonics. The fault diagnosis inference engine continuously receives real-time data streams from the data fusion center, compares the current data with the normal patterns in the knowledge base, and uses a fuzzy matching algorithm to calculate the deviation. When the deviation of a certain parameter or a set of parameters continuously exceeds the preset threshold and the duration reaches the preset number of diagnostic cycles, the inference engine determines that an anomaly has occurred. Subsequently, the case matching process is initiated, and the similarity between the current abnormal feature vector and the feature vector in the case library is calculated. Finally, a structured warning message is generated, which contains at least four fields: fault component location, accurate to the specific device or circuit; fault type.

[0029] A bidirectional, high-speed feedback closed-loop optimization link is established between the adaptive collaborative decision-making unit and the global state monitoring and diagnosis unit. The global state monitoring and diagnosis unit pushes the efficiency indicators calculated by the performance evaluation engine and the early warning information generated by the fault diagnosis inference engine to the internal parameter management module of the adaptive collaborative decision-making unit in real time. Based on this feedback information, this module dynamically adjusts the weight parameters and constraints in the collaborative optimization algorithm. For example, when it receives an early warning from the fault diagnosis inference engine about "performance degradation of photovoltaic string B," the parameter management module will reduce the power generation weight coefficient of the region where string B is located in the power generation benefit term of the global optimization function. At the same time, it will notify the maximum power point tracking controller to reduce the dependence weight of the current sampling value from string B when calculating the global maximum power point, so as to avoid the degraded string pulling the operating point of the entire array off track. As another example, when it is detected that the grid connection point voltage is near the upper limit for a long time, the parameter management module will increase the weight of the voltage regulation term in the virtual synchronous machine control task, prompting the inverter to output more capacitive reactive power to support the voltage. This dynamic parameter adjustment mechanism based on real-time state feedback makes the entire system no longer a controller with fixed parameters.

[0030] Working principle of the invention: The system is equipped with a multi-source fusion sensing unit, an adaptive collaborative decision-making unit, a multi-modal execution driving unit, and a global status monitoring and diagnosis unit. The multi-source fusion sensing unit is used to collect multi-dimensional heterogeneous data during the operation of the solar power generation system in real time. The adaptive collaborative decision-making unit is used to receive data from the multi-source fusion sensing unit and generate collaborative control commands for different actuators based on a preset collaborative optimization model and algorithm. The multi-modal execution driving unit is used to receive and execute the collaborative control commands issued by the adaptive collaborative decision-making unit to drive the corresponding physical actuators. The global status monitoring and diagnosis unit is used to collect and analyze the status data from the multi-source fusion sensing unit and various units within the system in real time, so as to realize global visualization of the system's operating status, performance evaluation, and early fault warning.

Claims

1. A solar generator speed regulation monitoring system, characterized by, The application relates to a solar power generation system adaptive cooperative control method and system. The application comprises: a multi-source fusion perception unit for collecting multi-dimensional heterogeneous data in the operation process of a solar power generation system in real time; an adaptive cooperative decision unit for receiving data from the multi-source fusion perception unit and generating cooperative control instructions for different execution mechanisms based on a preset cooperative optimization model and algorithm; a multi-modal execution driving unit for receiving and executing the cooperative control instructions issued by the adaptive cooperative decision unit to drive corresponding physical execution mechanisms to act; a global state monitoring and diagnosis unit for collecting and analyzing state data from the multi-source fusion perception unit and each unit in the system in real time, realizing global visualization, performance evaluation and early fault warning of the system operation state; the adaptive cooperative decision unit runs a hierarchical cooperative optimization control algorithm, and the cooperative optimization control algorithm has a global optimization function at the top layer; 2. A solar power generator speed regulation monitoring system according to claim 1, wherein, in the middle layer, the global optimization target at the top layer is decomposed into a plurality of mutually coupled sub-optimization tasks; and in the bottom layer, a special adaptive controller is designed for each sub-optimization task, and dynamic coupling constraints and information interaction protocols among the sub-tasks are defined.

3. A solar generator speed regulation monitoring system according to claim 2, wherein, The sub-optimization tasks include a photovoltaic array double-axis tracking control task, a maximum power point tracking control task, a power smoothing control task of an energy storage system and a virtual synchronous machine control task of a grid-connected inverter. The execution process of the photovoltaic array double-axis tracking control task is as follows: The adaptive cooperative decision unit first fuses data from a high-precision sun position sensor array, and solves the accurate azimuth angle and elevation angle of the sun in the sky coordinate system by using a weighted least square method; The adaptive cooperative decision unit receives real-time power generation data from a distributed photovoltaic array sensor network; The adaptive cooperative decision unit is internally provided with a tracking strategy optimizer based on reinforcement learning, which takes maximizing the total output power of the photovoltaic array at the current moment as the immediate reward and takes the motion energy consumption of the tracking mechanism as the penalty term, and dynamically adjusts the motion trajectory of the tracking mechanism through online learning; 4. A solar power generator speed regulation monitoring system according to claim 2, wherein, The adaptive cooperative decision unit outputs optimal tracking instructions containing azimuth angle and pitch angle adjustment amounts to the multi-modal execution driving unit. The execution process of the maximum power point tracking control task is as follows: The adaptive cooperative decision unit calculates the output power characteristic curve of the photovoltaic array in real time based on the voltage and current data collected by the distributed photovoltaic array sensor network; The adaptive cooperative decision unit adopts an adaptive algorithm combining a hybrid disturbance observation method and an admittance increment method; In the gentle stage of illumination and temperature change, the algorithm adopts a variable-step admittance increment method for accurate optimization; When a sudden change in illumination or temperature is sensed, the algorithm automatically switches to a large-range scanning disturbance observation method mode to quickly locate a new maximum power point region, and then switches back to the admittance increment method for fine adjustment; 5. A solar power generator speed regulation monitoring system according to claim 2, wherein, The step length adjustment coefficient in the algorithm is dynamically associated with the real-time open-circuit voltage and short-circuit current of the photovoltaic array. The execution process of the power smoothing control task of the energy storage system is as follows: The adaptive co-decision unit calculates the difference between the instantaneous power of photovoltaic power generation and the smooth power reference value processed by the low-pass filter in real time, which is the power instruction that needs to be absorbed or released by the energy storage system; The adaptive co-decision unit adopts a power distribution strategy based on model predictive control; The strategy takes the future period as the prediction time domain, maintains the state of charge of the energy storage system within the preset safe interval, minimizes the charge and discharge cycle loss and power fluctuation of the energy storage system, and optimizes the optimal charge and discharge power instruction of the energy storage converter in each control period.

6. A solar power generator speed regulation monitoring system according to claim 2, wherein, The execution process of the virtual synchronous machine control task of the grid-connected inverter is as follows: The adaptive co-decision unit simulates the rotor motion equation and electromagnetic transient characteristics of the synchronous generator for the grid-connected inverter; The adaptive co-decision unit dynamically calculates the angular velocity and power angle of the virtual rotor according to the grid frequency and voltage information collected by the grid interaction interface sensor; The adaptive co-decision unit adjusts the moment of inertia parameter and damping coefficient parameter of the virtual rotor, so that the inverter can actively adjust the active power output according to the grid frequency deviation and the reactive power output according to the grid voltage deviation; The virtual moment of inertia and damping coefficient are adaptively adjusted according to the real-time strength evaluation result of the grid.

7. A solar power generator speed regulation monitoring system as claimed in claim 2, wherein, The multi-modal execution driving unit includes a high-precision dual-axis tracking servo driver, a bidirectional energy storage converter driver, and a full-bridge inverter driver; The high-precision dual-axis tracking servo driver receives the tracking instruction issued by the adaptive co-decision unit, and drives the azimuth angle motor and the pitch angle motor to move to the specified position with a preset acceleration and speed curve; The bidirectional energy storage converter driver receives the power smoothing control instruction, controls the switching time sequence of the insulated gate bipolar transistor or the metal oxide semiconductor field effect transistor, and realizes the bidirectional, fast and accurate flow of energy between the energy storage battery pack and the DC bus; The full-bridge inverter driver receives the virtual synchronous machine control instruction, generates a driving signal through the space vector pulse width modulation technology, controls the power switching device of the inverter bridge arm, realizes the conversion from DC to AC, and injects a current with virtual synchronous machine characteristics into the grid.

8. A solar power generator speed regulation monitoring system according to claim 2, wherein, The global state monitoring and diagnosis unit includes a data fusion center, a performance evaluation engine, and a fault diagnosis inference machine; Receive and time-align all sensor data from the multi-source fusion perception unit, as well as the internal state data of the adaptive co-decision unit and the multi-modal execution driving unit; The performance evaluation engine calculates the system-level performance indicators in real time based on the fused data, including daily cumulative power generation, real-time power generation efficiency, maximum power point tracking efficiency, energy storage system smoothing effect indicators, and grid point power quality qualification rate; The fault diagnosis inference machine has a hybrid knowledge base based on rules and cases, which defines the multi-parameter association mode of each component of the system in normal and typical fault states; The fault diagnosis inference engine continuously compares real-time data stream with patterns in the knowledge base, and when it detects that a parameter deviates from the normal correlation pattern and continues to exceed the preset sampling period, it generates early warning information containing fault component positioning, fault type and confidence level.

9. A solar power generator speed regulation monitoring system according to claim 2, wherein, The adaptive collaborative decision-making unit and the global state monitoring and diagnosis unit have a two-way feedback channel; The global state monitoring and diagnosis unit feeds back performance evaluation results and fault warning information to the adaptive collaborative decision-making unit in real time; The adaptive collaborative decision-making unit dynamically adjusts the weight parameters and constraint conditions of its internal optimization algorithm according to the feedback information; The adaptive collaborative decision-making unit dynamically adjusts the weight parameters and constraint conditions of its internal optimization algorithm according to the feedback information as follows: When receiving a performance degradation warning of a certain photovoltaic string, the adaptive collaborative decision-making unit reduces the power generation weight of the string in the global optimization function, and adjusts the dependence of the maximum power point tracking algorithm on the string accordingly; When the health status of the energy storage system is monitored to decline, the adaptive collaborative decision-making unit tightens the safety interval constraint of the state of charge of the energy storage system.

10. A solar generator speed regulation monitoring system according to claim 9, wherein, The multi-source fusion perception unit includes a high-precision sun position sensor array, a distributed photovoltaic array sensor network, an energy storage system state sensor group, and a power grid interaction interface sensor; The distributed photovoltaic array sensor network is used to synchronously collect the output voltage, output current, surface temperature and backboard temperature of each photovoltaic component or group; The energy storage system state sensor group is used to monitor the state of charge, health status, charge and discharge current, terminal voltage and internal temperature distribution of the energy storage battery group in real time; The power grid interaction interface sensor is used to collect the voltage, frequency, phase and power factor of the grid-connected point in real time.

Images