Photovoltaic grid-connected box electric energy quality monitoring system based on Internet of Things data acquisition

Abstract

The invention relates to the technical field of intelligent monitoring and fault diagnosis of a photovoltaic power generation system, in particular to a photovoltaic grid-connected box electric energy quality monitoring system based on data acquisition of the Internet of Things. Comprising an environment perception and time domain synchronization unit which is used for triggering sampling by using a unified clock source and generating a full-dimensional synchronous original data stream; the ideal reference physical reconstruction unit is used for calculating a theoretical output amplitude envelope and generating an ideal state reference vector; the fault mechanism parameterization evolution unit is used for converting the fault mechanism into a fault operator, selecting a high-risk fault mode according to the current operation condition, and superposing the corresponding fault operator onto the ideal state reference vector to generate a theoretical damaged state waveform; the double-rail-difference manifold extraction unit is used for obtaining a real residual vector and a theoretical fault residual vector; and the isomorphic feature coupling judgment unit is used for calculating the similarity between the real residual vector and the theoretical fault residual vector. According to the invention, the robustness of power quality monitoring under complex meteorological conditions is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and fault diagnosis technology for photovoltaic power generation systems, specifically a power quality monitoring system for photovoltaic grid-connected boxes based on Internet of Things (IoT) data acquisition. Background Technology

[0002] In existing technologies, power quality monitoring of photovoltaic grid-connected boxes mainly relies on setting fixed upper and lower thresholds for electrical parameters such as voltage and current; when the monitored value exceeds the threshold, an alarm is triggered; however, photovoltaic power generation systems have significant environmental dependence, and their output power varies drastically with fluctuations in irradiance and temperature.

[0003] In actual operation, this monitoring method based on fixed thresholds has obvious limitations: when the light intensity fluctuates drastically, the output power of the system will change dramatically, which is very easy for traditional monitoring systems to misjudge as electrical faults, resulting in a high false alarm rate; under the masking of strong power grid background noise and normal power frequency signals, the early weak fault characteristics are often small in amplitude, making it difficult to identify them through simple threshold comparison, resulting in missed faults and potential safety hazards.

[0004] Therefore, how to provide a photovoltaic grid-connected box power quality monitoring system that can effectively eliminate environmental interference and accurately extract weak fault characteristics under strong background noise is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention discloses a photovoltaic grid-connected box power quality monitoring system based on Internet of Things (IoT) data acquisition. Specifically, the technical solution of this invention includes:

[0006] The environmental perception and time-domain synchronization unit is used to connect voltage sensors, current sensors, irradiance sensors and temperature sensors. It uses a unified clock source to trigger sampling, performs timestamp alignment processing on the collected voltage, current, irradiance and temperature data, and generates a full-dimensional synchronized raw data stream.

[0007] The ideal reference physical reconstruction unit is used to receive real-time irradiance and temperature data, extract the grid voltage fundamental phase and frequency from the full-dimensional synchronous raw data stream, call the photovoltaic module IV characteristic curve parameters, inverter topology parameters and line impedance parameters, perform forward deduction through the circuit physical law model, calculate the theoretical output amplitude envelope, and use the grid voltage fundamental phase and frequency to modulate the theoretical output amplitude envelope onto a sinusoidal carrier to generate an ideal state reference vector;

[0008] The fault mechanism parameterization evolution unit is used to transform the fault mechanism into a fault operator, select a high-risk fault mode according to the current operating conditions, and superimpose the corresponding fault operator onto the ideal state reference vector to generate a theoretical damaged state waveform.

[0009] The dual-track differential manifold extraction unit is used to calculate the difference between the full-dimensional synchronous original data stream and the ideal state reference vector to obtain the actual residual vector; and to calculate the difference between the theoretical damaged state waveform and the ideal state reference vector to obtain the theoretical fault residual vector.

[0010] The isomorphic feature coupling decision unit is used to map the actual residual vector and the theoretical fault residual vector to a high-dimensional feature space, calculate the similarity between the actual residual vector and the theoretical fault residual vector, compare the similarity with the fault diagnosis threshold, and output the diagnosed fault type signal if the similarity is higher than the fault diagnosis threshold.

[0011] Optionally, the environmental perception and time-domain synchronization unit is configured with a hardware-level clock synchronization mechanism; the timestamp alignment process ensures that voltage and current data frames and irradiance and temperature data frames establish a correspondence in the nanosecond time dimension.

[0012] Optionally, the ideal reference physical reconstruction unit uses software phase-locked loop logic to perform the modulation; the ideal state reference vector represents the theoretical upper limit of system performance under the current environmental conditions and the current power grid frequency state.

[0013] Optionally, the fault mechanism parameterization evolution unit is equipped with a parameterization factor conversion engine to convert the textual fault mechanism into a fault operator including virtual fault impedance and harmonic injection factor.

[0014] Optionally, the dual-track differential manifold extraction unit cancels out common-mode fluctuations driven by environmental parameters by calculating the difference between the full-dimensional synchronous raw data stream and the ideal state reference vector; and removes the fundamental wave component by calculating the difference between the theoretical damaged state waveform and the ideal state reference vector.

[0015] Optionally, the isomorphic feature coupling decision unit uses the wavelet packet energy spectrum space as the high-dimensional feature space; uses cosine similarity as the similarity; and the fault diagnosis threshold is statistically derived based on the environmental white noise baseline level measured during the system's fault-free operation.

[0016] Optionally, it also includes a feedback correction closed loop;

[0017] When the isomorphic feature coupling decision unit detects a constant small deviation in the actual residual vector that does not belong to a sudden failure mode, it generates a parameter drift correction signal.

[0018] The feedback correction closed loop feeds the parameter drift correction signal back to the ideal reference physical reconstruction unit;

[0019] The ideal reference physical reconstruction unit responds to the parameter drift correction signal and fine-tunes the photovoltaic module IV characteristic curve parameters or the line impedance parameters.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. This invention achieves real-time generation of dynamic benchmarks based on physical models by constructing an ideal benchmark physical reconstruction unit. Unlike traditional monitoring methods that rely on fixed thresholds, this system utilizes real-time sensed irradiance and temperature data, combined with the photovoltaic module IV characteristics and inverter topology parameters, to perform forward physical deduction and construct an accurate theoretical output amplitude envelope under the current operating conditions. This mechanism establishes an ideal state benchmark vector that adapts to environmental changes, effectively eliminating normal output fluctuations caused by drastic fluctuations in light intensity or temperature changes from a physical perspective. This allows for accurate differentiation between performance fluctuations caused by environmental factors and actual electrical faults within the equipment, significantly improving the robustness of power quality monitoring under complex meteorological conditions.

[0022] 2. This invention utilizes a dual-track differential manifold extraction unit and innovatively employs differential logic to extract the net signal of fault features. By calculating the difference between the full-dimensional synchronous original data stream and the ideal state reference vector, the system can automatically cancel common-mode fluctuations driven by environmental parameters and remove high-amplitude grid fundamental components. This processing method effectively suppresses interference from strong grid background signals and clearly preserves weak fault features that are often masked. This greatly improves the system's sensitivity in capturing hidden power quality problems such as early minor faults and high-impedance faults, and solves the technical problem of difficulty in identifying weak fault signals under strong background noise.

[0023] 3. This invention combines parameterized evolution of fault mechanisms with isomorphic feature-coupled decision units, achieving a leap from black-box data matching to white-box mechanism diagnosis. By transforming abstract fault mechanisms into specific fault operators and evolving theoretical damaged state waveforms, the system can compare the geometric similarity between real residuals and theoretical fault residuals in a high-dimensional feature space. This decision method based on isomorphic mapping can not only determine whether a system fault has occurred, but also directly infer the specific fault type and cause by comparing the theoretical model with the highest similarity. This effectively reduces the false alarm rate, provides maintenance personnel with physically interpretable diagnostic results, and significantly improves fault diagnosis efficiency.

[0024] 4. This invention integrates a high-precision time-domain synchronization mechanism and a parameter drift feedback correction closed loop, ensuring the stability of the monitoring system throughout its entire lifecycle. Hardware-level clock synchronization ensures strict alignment of voltage, current, and environmental data at the nanosecond level, eliminating analysis errors caused by data asynchrony. Simultaneously, the feedback correction closed loop can identify constant, minute deviations in non-sudden faults and dynamically fine-tune the impedance or IV characteristic parameters in the model accordingly. This adaptive calibration capability enables the system to distinguish between normal equipment aging and sudden faults, preventing false alarms caused by the natural degradation of component performance, thereby extending the effective service life of the monitoring system and maintaining long-term detection accuracy. Attached Figure Description

[0025] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0026] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0028] Example 1

[0029] Please see Figure 1 A photovoltaic grid-connected box power quality monitoring system based on Internet of Things (IoT) data acquisition includes:

[0030] The environmental perception and time-domain synchronization unit is used to connect voltage sensors, current sensors, irradiance sensors and temperature sensors. It uses a unified clock source to trigger sampling, performs timestamp alignment processing on the collected voltage, current, irradiance and temperature data, and generates a full-dimensional synchronized raw data stream.

[0031] The ideal reference physical reconstruction unit is used to receive real-time irradiance and temperature data, extract the fundamental phase and frequency of the grid voltage from the full-dimensional synchronous raw data stream, and call upon the photovoltaic modules. Characteristic curve parameters, inverter topology parameters, and line impedance parameters are derived forward using the circuit physics law model to calculate the theoretical output amplitude envelope. The theoretical output amplitude envelope is then modulated onto a sinusoidal carrier wave using the fundamental phase and frequency of the grid voltage to generate an ideal state reference vector.

[0032] The fault mechanism parameterization evolution unit is used to transform the fault mechanism into a fault operator, select a high-risk fault mode according to the current operating conditions, and superimpose the corresponding fault operator onto the ideal state reference vector to generate a theoretical damaged state waveform.

[0033] The dual-track differential manifold extraction unit is used to calculate the difference between the full-dimensional synchronous original data stream and the ideal state reference vector to obtain the actual residual vector; and to calculate the difference between the theoretical damaged state waveform and the ideal state reference vector to obtain the theoretical fault residual vector.

[0034] The isomorphic feature coupling decision unit is used to map the actual residual vector and the theoretical fault residual vector to a high-dimensional feature space, calculate the similarity between the actual residual vector and the theoretical fault residual vector, compare the similarity with the fault diagnosis threshold, and output the diagnosed fault type signal if the similarity is higher than the fault diagnosis threshold.

[0035] In the system architecture detailed in this embodiment, the environmental perception and time-domain synchronization unit constitutes the data foundation of the entire monitoring system. Its design aims to eliminate timing deviations between multi-source heterogeneous data. In actual engineering configuration, this unit is connected to the voltage and current sensors on the photovoltaic grid-connected box side via hardwired connections to obtain electrical response data, and simultaneously connected to the irradiance and temperature sensors deployed on the photovoltaic panel side to obtain environmental excitation data. To ensure the time-domain consistency of the data, this unit does not adopt the conventional mode of autonomous polling of the internal clocks of each sensor, but is controlled by a high-precision unified clock source. Whenever the sampling period arrives, the clock source simultaneously sends a trigger command to the data acquisition cards of all sensors to force concurrent sampling.

[0036] Specifically, to achieve the aforementioned high-precision concurrent sampling, the environmental perception and time-domain synchronization unit adopts a heterogeneous processing architecture based on an FPGA as the underlying control core. The FPGA is internally configured with parallel I / O logic blocks, each connected to the ADC of each sensor channel via an independent SPI high-speed bus.

[0037] The timestamp alignment process performed on this basis refers to indexing and matching the received data frames according to the global clock tag of their trigger time, strictly eliminating time slippage caused by transmission delay, thereby generating a full-dimensional synchronous original data stream; this data stream specifically refers to a time series set containing environmental excitation information and electrical response information at the same nanosecond level, and its core value lies in providing source data without time delay error for subsequent physical causal analysis, which is different from the phase alignment error caused by traditional asynchronous acquisition;

[0038] After the data stream is established, the ideal reference physical reconstruction unit initiates its core dynamic modeling process. This unit's operational logic breaks away from the traditional static threshold monitoring method based on historical data statistics, instead employing first principles of physics for real-time deduction. The unit reads irradiance and temperature values ​​from the full-dimensional synchronous raw data stream in real time and uses a phase-locked loop algorithm to extract the fundamental phase and frequency information of the current grid voltage. The system calls pre-stored static physical parameters, including those describing the photoelectric conversion efficiency of the photovoltaic modules. Characteristic curve parameters, topology parameters describing the inverter circuit structure, and line impedance parameters describing the loss characteristics of connecting cables; using the circuit physics law model, this unit calculates the theoretically achievable output voltage and current amplitude envelopes under the current instantaneous environmental conditions; the circuit physics law model adopts a single diode equivalent circuit model, and its current-voltage relationship equation is:

[0039]

[0040] in, For photocurrent, This is the reverse saturation current of the diode. The amount of electron charge. For diode ideality factor, Boltzmann's constant, This refers to the absolute temperature of the battery. For series resistance, The resistors are in parallel; based on this equation, the voltage at the maximum power point is solved using the Newton-Raphson iterative method. With current This is taken as the theoretical output amplitude envelope at the current moment. The baseline value;

[0041] It should be noted that, in order to further improve the realism of the physical reconstruction, the ideal reference physical reconstruction unit, based on the single-diode model described above, also introduces a correction sub-model for complex environmental effects; this correction sub-model is used to correct the photocurrent parameters according to real-time operating conditions. The calculation formula is as follows:

[0042]

[0043] in, The short-circuit current under standard test conditions. The temperature coefficient of short-circuit current. For reference temperature, This represents the current real-time irradiance. Reference irradiance (1000W / Through this correction, the model can dynamically compensate for the nonlinear degradation of photoelectric conversion efficiency caused by temperature rise;

[0044] This unit uses the amplitude envelope as the modulation signal and the real-time extracted fundamental phase and frequency of the power grid as the carrier signal to synthesize an ideal state reference vector. This vector is technically defined as the theoretical waveform of the system's healthy operation under the current environment and power grid conditions. It not only follows the fluctuations of environmental parameters but also maintains strict synchronization with the power grid frequency, serving as a dynamic subtraction to cancel environmental noise in subsequent differential operations.

[0045] Parallel to the above process, the fault mechanism parameterization evolution unit performs a knowledge-driven fault sample generation task. Addressing the challenge of scarce fault samples in actual operation, this unit is configured with a transformation logic from textual mechanisms to mathematical models. Based on the current system operating conditions, this unit selects fault modes with a high probability of occurrence in the current scenario from a pre-set fault mode library. The unit identifies several high-risk failure modes. For each mode, it transforms the corresponding textual failure mechanism into a mathematical failure operator and superimposes it onto the previously generated ideal state reference vector to synthesize the theoretical damaged state waveform; that is, the theoretical damaged state waveform. The calculation formula is:

[0046]

[0047] This formula clarifies that the fault waveform is not simply a superposition of signals, but a superposition of system response deviations caused by circuit parameter perturbations. This waveform is a digital simulation of the system output waveform when a specific fault is assumed to occur, providing the system with a fault comparison benchmark that does not require historical training data.

[0048] Following the aforementioned data stream, the dual-track differential manifold extraction unit performs crucial signal decoupling operations. This unit is designed with two parallel differential paths to achieve different signal processing objectives. In the first path, the difference between the full-dimensional synchronous original data stream and the ideal state reference vector is calculated. Since both are driven by the same real-time environmental parameters and have isomorphic physical response characteristics, this subtraction operation mathematically constitutes the complete cancellation of the environmental common-mode signal, thereby outputting a realistic residual vector that only contains the internal physical variation characteristics of the device and the white noise of the sensor.

[0049] In the second path, the difference between the theoretical damaged state waveform and the ideal state reference vector is calculated. This operation aims to remove the power frequency fundamental wave component that masks the weak fault characteristics, thereby extracting the pure theoretical fault residual vector. The innovation of this dual-track differential mechanism is that it does not use filters to passively suppress noise, but actively reconstructs and subtracts noise sources using physical models, thus achieving zero false alarm monitoring under drastic environmental conditions.

[0050] During the decision-making stage, the isomorphic feature coupling decision unit performs the final fault diagnosis. To overcome the limitation that it is difficult to distinguish small differences in time-domain signals, the unit simultaneously projects and maps the actual residual vector and the theoretical fault residual vector to a high-dimensional feature space. In this space, the unit calculates the geometric similarity between the two vectors and compares the calculation result with the preset fault diagnosis threshold.

[0051] The threshold set here is not a fixed constant, but is statistically derived based on the ambient white noise baseline level measured by the system during fault-free operation. The system determines that the real signal contains features isomorphic to the theoretical fault model if and only if the calculated similarity is significantly higher than the upper limit of the confidence interval corresponding to the noise baseline, and then outputs the confirmed fault type signal. This logic ensures that the judgment result has clear physical interpretability and effectively solves the technical pain point of low trust in black box AI models.

[0052] The environmental perception and time-domain synchronization unit is equipped with a hardware-level clock synchronization mechanism; timestamp alignment processing ensures that voltage and current data frames and irradiance and temperature data frames establish a correspondence in the nanosecond-level time dimension.

[0053] In the specific hardware configuration of this embodiment, in order to meet the stringent requirements of time synchronization accuracy for high-frequency transient analysis, the environmental perception and time-domain synchronization unit abandons the traditional soft clock calibration scheme and instead integrates a hardware-level clock synchronization mechanism. The core component of this mechanism is a master clock signal source generated by a high-precision crystal oscillator or a GPS / BeiDou dual-mode timing module. This signal source is directly connected to the external trigger pins of the analog-to-digital converters of each sensor through a low-latency physical bus. When the system is working, the master clock source sends a rising edge trigger signal to forcibly control the sample-and-hold circuits of all channels to freeze the analog quantity and start the conversion at the same physical moment.

[0054] The timestamp alignment process implemented on this basis generates time tags by parsing the header of each data packet, forcing a one-to-one mapping relationship between voltage and current data frames and irradiance and temperature data frames in the nanosecond time dimension. This nanosecond-level synchronization accuracy ensures that changes in environmental parameters and responses in electrical parameters exhibit strict physical causality at the data level, eliminating spurious phase differences introduced by sampling time differences, and laying an irreplaceable data foundation for subsequent high-precision physical reconstruction.

[0055] Example 2

[0056] The ideal reference physical reconfiguration unit uses software phase-locked loop logic to perform modulation; the ideal state reference vector represents the theoretical upper limit of system performance under the current environmental conditions and the current power grid frequency state;

[0057] In the reconstruction algorithm detailed in this embodiment, the ideal reference physical reconstruction unit uses software phase-locked loop logic to perform the key waveform modulation steps. This logic module includes an orthogonal signal generator, a Parker transform module, and a PI regulator. Its function is to track and lock the instantaneous phase angle and frequency of the grid voltage in the full-dimensional synchronous raw data stream in real time. After obtaining the amplitude envelope calculated based on the physical model, the unit multiplies it with the unit sine wave output by the phase-locked loop to complete the modulation from the DC envelope to the AC waveform.

[0058] The ideal state reference vector generated thus has a clear and strict physical meaning: it represents the theoretical upper limit of system performance under the current environmental conditions and the current power grid frequency state; the vector represents the output form of a perfect system that fully conforms to the design specifications, has no aging loss, and closely follows the fluctuations of the power grid frequency; by introducing this dynamic reference, the system can accurately extract the subtle distortions that deviate from the theoretical optimal value from the complex measured signals, thereby achieving keen detection of early weak faults.

[0059] The fault mechanism parameterization evolution unit is equipped with a parameterization factor conversion engine, which is used to convert the textual fault mechanism into fault operators including virtual fault impedance and harmonic injection factor.

[0060] In the fault evolution logic of this embodiment, the fault mechanism parameterization evolution unit has a built-in parameterization factor conversion engine. This engine internally stores a preset mechanism-parameter mapping rule table, which establishes a correspondence between fault text keywords and circuit parameter perturbation ranges. The engine extracts keywords from the input textual fault description. According to the mapping rule table: if a poor contact is matched, a parameter is added to the corresponding branch impedance. to Randomly generated impedance increments within the interval If a short circuit is detected, the impedance between the corresponding nodes will be forcibly set to [value]. ;

[0061] Specifically, the engine constructs the system node impedance matrix. For poor contact faults, increase the impedance increment in the corresponding branch. Let the fault impedance matrix For short-circuit faults, construct a fault impedance matrix. The impedance elements between corresponding nodes are replaced with At this point, the fault operator Defined as the time-domain difference function between the fault-state current response and the ideal-state current response:

[0062]

[0063] in Indicates the inverse Laplace transform. This represents finding the inverse of a matrix. As the excitation source; the fault involving nonlinear device failure is transformed into a harmonic injection factor operator, which simulates the open circuit or dead zone effect of the switch by superimposing harmonic components of specific frequency and amplitude on the fundamental wave;

[0064] The fault operator is defined here as a set of mathematical correction terms that can be directly applied to the ideal waveform equation. Its source is the perturbation processing of the standard circuit equation by the engine based on the physical cause of the fault. Through this parameterization transformation, the system can synthesize realistic fault waveforms through mathematical means without the need for real fault data, thus expanding the monitoring system's ability to identify rare faults.

[0065] The dual-track differential manifold extraction unit cancels common-mode fluctuations driven by environmental parameters by calculating the difference between the full-dimensional synchronous original data stream and the ideal state reference vector; and removes the fundamental wave component by calculating the difference between the theoretical damaged state waveform and the ideal state reference vector.

[0066] In the core signal processing stage of this embodiment, the dual-track differential manifold extraction unit uses two sets of independent differential operations to achieve deep signal cleaning. When processing real data, this unit calculates the full-dimensional synchronous original data stream minus the ideal state reference vector. Since the ideal reference vector is calculated based on real-time environmental parameters, the amplitude fluctuations it contains are completely consistent with the fluctuations caused by the environment in the original data stream. This subtraction operation can mathematically cancel the common-mode fluctuations driven by environmental parameters, so that the output results are no longer affected by weather changes and only retain the abnormal characteristics of the device itself.

[0067] When processing theoretical data, this unit calculates the theoretical damaged state waveform minus the ideal state reference vector. Since both contain the same power frequency fundamental wave component, this subtraction operation can completely strip away the fundamental wave component with the main energy concentration, thereby exposing the tiny fault feature vectors that are masked by the fundamental wave. This processing scheme not only solves the problem of false alarms caused by environmental interference, but also improves the signal-to-noise ratio and detection rate of weak fault signals by removing the high-energy fundamental wave background.

[0068] The isomorphic feature coupling decision unit uses wavelet packet energy spectrum space as high-dimensional feature space; cosine similarity is used as similarity; the fault diagnosis threshold is statistically derived based on the environmental white noise baseline level measured during the system's fault-free operation.

[0069] In the decision logic of this embodiment, in order to accurately quantify the similarities and differences of waveform features, the isomorphic feature coupling decision unit selects the wavelet packet energy spectrum space as the target space for feature mapping; the specific mapping process involves multi-layer wavelet packet decomposition of the time-domain residual signal. In this embodiment, Daubechies4 is selected as the wavelet basis function, and the number of decomposition layers is set to 3.

[0070] This decomposition process divides the original signal frequency band into... The signal energy is reconstructed by extracting wavelet coefficients from eight independent sub-band spaces. Constructing eigenvectors This allows the unit to capture the details of local abrupt changes in the signal in the time-frequency domain. In terms of similarity measurement, the unit uses a cosine similarity algorithm to calculate the cosine value of the angle between the actual residual feature vector and the theoretical fault feature vector in multi-dimensional space. The advantage of this algorithm is that it is not sensitive to changes in signal amplitude, but focuses on matching waveform fingerprint morphology, thus adapting to fault identification under different load levels.

[0071] Regarding the basis for the judgment, the setting of the fault diagnosis threshold follows strict statistical principles: during the background period when the system confirms no fault, it continuously collects and calculates the characteristic values ​​of the residual signal, statistically analyzes its distribution pattern, and calculates the mean and variance of the environmental white noise floor. Finally, the threshold is set above three times the standard deviation of the noise floor distribution. This dynamic threshold setting based on statistical characteristics ensures that the system can effectively shield the risk of false alarms caused by the inherent thermal noise of the sensor while maintaining extremely high sensitivity.

[0072] It also includes a feedback correction closed loop;

[0073] When the isomorphic feature coupling decision unit detects a constant small deviation in the real residual vector that does not belong to the sudden failure mode, it generates a parameter drift correction signal.

[0074] The feedback correction closed loop feeds the parameter drift correction signal back to the ideal reference physical reconstruction unit;

[0075] The ideal reference physical reconfiguration unit responds to the parameter drift correction signal to fine-tune the photovoltaic module. Characteristic curve parameters or line impedance parameters;

[0076] In the system maintenance mechanism of this embodiment, a feedback correction closed loop is added to address the aging problem throughout the entire life cycle of the equipment. During the long-term monitoring process, the isomorphic feature coupling decision unit continuously analyzes the long-term trend of the residual vector. When a non-sudden, non-transient constant small deviation is detected in the actual residual vector, and the similarity between the shape characteristics of this deviation and all known fault modes in the database is lower than the diagnostic threshold, the system determines that this phenomenon originates from the natural drift of the equipment's physical parameters and generates a parameter drift correction signal. This signal is essentially an error correction vector carrying the deviation direction and amplitude.

[0077] The feedback correction closed loop sends this signal back to the modeling end, driving the ideal reference physical reconstruction unit to initiate an adaptive calibration procedure; this procedure, based on the deviation characteristics, adjusts the photovoltaic modules in the internal physical model. The characteristic curve parameters or line impedance parameters are iteratively fine-tuned using a gradient descent algorithm. This fine-tuning process is based on the relationship between the physical model output and the parameters. The relationship exhibits a nonlinear implicit function, and this unit uses the numerical difference perturbation method to approximate the gradient; that is, in each iteration, the current parameters are adjusted. Apply small perturbation Calculate the change in the objective function to obtain approximate partial derivatives:

[0078]

[0079] Set the objective function The mean square error of the actual residual vector is given by the parameter update formula:

[0080]

[0081] in, Represents the physical parameters to be corrected. For the number of iterations, The learning rate is set to the preset value; the iteration termination condition is set to... Or until the maximum number of iterations is reached; until the calculated ideal waveform coincides with the actual waveform again;

[0082] Through this closed-loop mechanism, the system enables the physical model to evolve and update itself, ensuring that the monitoring benchmark remains anchored to the actual health status of the equipment throughout its decades-long operating cycle, thus avoiding false alarms caused by aging.

[0083] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention; any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A photovoltaic grid-connected box power quality monitoring system based on Internet of Things (IoT) data acquisition, characterized in that, include: The environmental perception and time-domain synchronization unit is used to connect voltage sensors, current sensors, irradiance sensors and temperature sensors. It uses a unified clock source to trigger sampling, performs timestamp alignment processing on the collected voltage, current, irradiance and temperature data, and generates a full-dimensional synchronized raw data stream. The ideal reference physical reconstruction unit is used to receive real-time irradiance and temperature data, extract the grid voltage fundamental phase and frequency from the full-dimensional synchronous raw data stream, call the photovoltaic module IV characteristic curve parameters, inverter topology parameters and line impedance parameters, perform forward deduction through the circuit physical law model, calculate the theoretical output amplitude envelope, and use the grid voltage fundamental phase and frequency to modulate the theoretical output amplitude envelope onto a sinusoidal carrier to generate an ideal state reference vector; The fault mechanism parameterization evolution unit is used to transform the fault mechanism into a fault operator, select a high-risk fault mode according to the current operating conditions, and superimpose the corresponding fault operator onto the ideal state reference vector to generate a theoretical damaged state waveform. The dual-track differential manifold extraction unit is used to calculate the difference between the full-dimensional synchronous original data stream and the ideal state reference vector to obtain the actual residual vector; and to calculate the difference between the theoretical damaged state waveform and the ideal state reference vector to obtain the theoretical fault residual vector. The isomorphic feature coupling decision unit is used to map the actual residual vector and the theoretical fault residual vector to a high-dimensional feature space, calculate the similarity between the actual residual vector and the theoretical fault residual vector, compare the similarity with the fault diagnosis threshold, and output the diagnosed fault type signal if the similarity is higher than the fault diagnosis threshold.

2. The photovoltaic grid-connected box power quality monitoring system based on Internet of Things data acquisition according to claim 1, characterized in that, The environmental perception and time-domain synchronization unit is equipped with a hardware-level clock synchronization mechanism; the timestamp alignment process ensures that voltage and current data frames and irradiance and temperature data frames establish a correspondence in the nanosecond time dimension.

3. The photovoltaic grid-connected box power quality monitoring system based on Internet of Things data acquisition according to claim 1, characterized in that, The ideal reference physical reconstruction unit uses software phase-locked loop logic to execute the modulation; the ideal state reference vector represents the theoretical upper limit of system performance under the current environmental conditions and the current power grid frequency state.

4. The photovoltaic grid-connected box power quality monitoring system based on Internet of Things data acquisition according to claim 1, characterized in that, The fault mechanism parameterization evolution unit is equipped with a parameterization factor conversion engine, which is used to convert the textual fault mechanism into fault operators including virtual fault impedance and harmonic injection factor.

5. A photovoltaic grid-connected box power quality monitoring system based on Internet of Things data acquisition according to claim 1, characterized in that, The dual-track differential manifold extraction unit cancels out common-mode fluctuations driven by environmental parameters by calculating the difference between the full-dimensional synchronous raw data stream and the ideal state reference vector; and removes the fundamental wave component by calculating the difference between the theoretical damaged state waveform and the ideal state reference vector.

6. The photovoltaic grid-connected box power quality monitoring system based on Internet of Things data acquisition according to claim 1, characterized in that, The isomorphic feature coupling decision unit uses the wavelet packet energy spectrum space as the high-dimensional feature space; Cosine similarity is used as the similarity; the fault diagnosis threshold is statistically derived based on the ambient white noise baseline level measured during fault-free operation of the system.

7. A photovoltaic grid-connected box power quality monitoring system based on Internet of Things data acquisition according to claim 1, characterized in that, It also includes a feedback correction closed loop; When the isomorphic feature coupling decision unit detects a constant small deviation in the actual residual vector that does not belong to the sudden failure mode, it generates a parameter drift correction signal. The feedback correction closed loop feeds the parameter drift correction signal back to the ideal reference physical reconstruction unit; The ideal reference physical reconstruction unit responds to the parameter drift correction signal and fine-tunes the photovoltaic module IV characteristic curve parameters or the line impedance parameters.

Images