A method and system for monitoring and controlling micro-inverters based on Internet of Things (IoT) data.

Abstract

This invention relates to the technical field of microinverters and discloses a microinverter monitoring and control method and system based on Internet of Things (IoT) data. The method collects multi-dimensional environmental parameters such as temperature, light intensity, and grid load, performs feature extraction and pattern recognition to determine the operating scenario. It employs multivariate time series analysis and Kalman filtering to weightedly fuse heterogeneous data, and determines the data source credibility through information entropy calculation for anomaly detection. The data fusion weights are dynamically adjusted based on anomaly scores. A control decision matrix is ​​constructed based on the operating scenario and anomaly severity, and a fuzzy logic controller is used to intelligently adjust the inverter output. Feedback control continuously optimizes operating parameters, achieving efficient and stable operation of the microinverter. This invention effectively addresses the complex and ever-changing photovoltaic power generation environment, improving system reliability and power generation efficiency.

Description

Technical Field

[0001] This invention relates to the technical field of microinverters, and in particular to a microinverter monitoring and control method and system based on Internet of Things (IoT) data. Background Technology

[0002] In data acquisition systems, distributed photovoltaic (PV) power generation systems, as a core technology for clean energy conversion, play a crucial role in the global energy structure transformation. Microinverters, as key components of PV systems, directly impact the efficiency and stability of the entire power generation system. Therefore, establishing an effective monitoring and control mechanism is of great significance for ensuring the reliable operation of the system.

[0003] Current microinverter monitoring methods primarily rely on traditional centralized management models. This approach suffers from sluggish response times and struggles to achieve accurate status perception when faced with complex and ever-changing operating environments. Furthermore, existing monitoring systems generally suffer from insufficient data processing capabilities, failing to effectively integrate multi-source, heterogeneous operating data, thus limiting monitoring precision and decision-making accuracy.

[0004] Therefore, existing control systems cannot achieve intelligent control effects because they lack the ability to dynamically adjust according to the actual operating conditions. Summary of the Invention

[0005] This invention provides a micro-inverter monitoring and control method and system based on Internet of Things (IoT) data to achieve intelligent control.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a micro-inverter monitoring and control method based on Internet of Things (IoT) data, comprising:

[0007] Environmental parameters in the micro-inverter's operating environment are collected and a multi-dimensional environmental parameter matrix is ​​obtained. An operating scenario identification vector is obtained based on the multi-dimensional environmental parameter matrix. The heterogeneous data from different sensor nodes are weighted and fused based on the operating scenario identification vector to obtain a fused state feature vector.

[0008] Anomaly detection is performed on the state feature vector. If one of the sensor nodes is abnormal, the corresponding weight is reduced to obtain an updated state feature vector and perform time series analysis. If the feature vector deviation exceeds a preset deviation threshold for a preset number of consecutive time points, an anomaly score is obtained.

[0009] The control strategy parameters are determined based on the anomaly score and the operating scenario identifier vector; the micro-inverter operating parameters are obtained based on the control strategy parameters, and the inverter operating parameters are input into the system to obtain the updated multi-dimensional environmental parameter matrix, thereby obtaining the optimized micro-inverter operating parameters.

[0010] Preferably, the step of collecting environmental parameters in the operating environment of the micro-inverter and obtaining a multi-dimensional environmental parameter matrix includes:

[0011] The temperature, light intensity, and grid load power are obtained from the operating environment of the micro-inverter according to the preset sampling frequency, and the multi-dimensional environmental parameter matrix is ​​constructed.

[0012] A time-series analysis was performed on the multi-dimensional environmental parameter matrix to extract the changing trend characteristics of the temperature value, the light intensity, and the power grid load power, thus obtaining a time-series feature dataset.

[0013] If at least one parameter in the time-series feature dataset exceeds a preset parameter threshold, the deviation between the time-series feature dataset and historical data is compared to determine the abnormal state.

[0014] Increase the sampling frequency of the parameters corresponding to the abnormal state and obtain the updated environmental parameters.

[0015] Preferably, obtaining the operational scenario identifier vector based on the multi-dimensional environmental parameter matrix includes:

[0016] The temperature change rate is calculated based on the multi-dimensional environmental parameter matrix to obtain the feature dataset;

[0017] If the temperature change rate in the feature dataset is greater than a preset temperature change threshold and the light intensity is less than a preset light intensity threshold, then it is determined to be an inefficient operating mode, and a first scenario identifier vector is generated.

[0018] The fluctuation variance of the power grid load is calculated based on the feature dataset. If the fluctuation variance is greater than a preset variance threshold, it is determined to be an unstable operating scenario, and a second scenario identification vector is generated.

[0019] The first scenario identifier vector and the second scenario identifier vector are integrated to generate the running scenario identifier vector.

[0020] Preferably, the step of weighted fusion processing of heterogeneous data from different sensor nodes based on the operating scenario identifier vector to obtain a fused state feature vector includes:

[0021] Obtain heterogeneous data and denoise it to obtain a denoised data matrix;

[0022] The information entropy value H = -Σp(xi)log(p(xi)) for each data source is calculated based on the denoised data matrix, where p(xi) represents the probability distribution of data source i;

[0023] If the information entropy value of any data source is lower than the preset entropy threshold, the credibility weight of the corresponding data source is obtained, and a weight vector is obtained.

[0024] The weight vector and the denoised data matrix are weighted and fused to generate a fused state feature vector.

[0025] Preferably, the step of performing anomaly detection on the state feature vector, and reducing the corresponding weight if one of the sensor nodes is abnormal, to obtain an updated state feature vector, includes:

[0026] The abnormal score of each data source is calculated based on the denoised data matrix. If the abnormal score is higher than the preset score threshold, the corresponding data source is judged to be abnormal, and the abnormal data source set A(t) is obtained.

[0027] The information entropy value of each data source in A(t) is calculated using the information entropy formula H(t)=-Σp(xi)log(p(xi)), where p(xi) represents the probability distribution of data source i.

[0028] The weights of the abnormal data sources are updated based on the abnormal scores to obtain the updated weight vector;

[0029] The updated weight vector and the denoised data matrix are weighted and fused to generate an updated state feature vector.

[0030] Preferably, the step of obtaining an anomaly score if the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold includes:

[0031] The timing data sequence during the operation of the microinverter is obtained and segmented to obtain local data subsets within multiple time windows;

[0032] The deviation value of the state feature vector is calculated based on the local data subset. If the deviation value exceeds the preset deviation threshold, it is determined that there is a potential abnormal window, and a set of potential abnormal windows is obtained.

[0033] The state feature vectors in the potential anomaly window set are processed to calculate the anomaly score Score=2^{-E(h) / c(n)}, where E(h) represents the expected path length and c(n) represents the standardization factor.

[0034] Preferably, before determining the control strategy parameters based on the anomaly score and the operating scenario identifier vector, the method further includes:

[0035] If the abnormal score is higher than the preset score threshold, it is marked as a high abnormal node, and a set of high abnormal nodes is obtained.

[0036] Based on the set of highly abnormal nodes, the data weights of the corresponding nodes are adjusted to obtain the adjusted set of node weights. Combined with the node priority, a data source selection list is obtained.

[0037] Based on the data source selection list, process the time series data to generate optimized fused data input.

[0038] Preferably, determining the control strategy parameters based on the anomaly score and the operating scenario identifier vector includes:

[0039] Calculate the anomaly score and motion scenario identifier vector for each node to obtain the anomaly score set and scenario identifier set;

[0040] If the abnormal score of any node in the abnormal score set exceeds a preset threshold and the scenario identification vector indicates the inefficient operation mode, then a power adjustment instruction is generated, and a power adjustment parameter set is generated.

[0041] If the abnormal score of any node in the abnormal score set exceeds a preset threshold and the scenario identification vector indicates the unstable scenario, then a voltage stabilization command is generated, the voltage adjustment range is calculated, and a voltage stabilization parameter set is obtained.

[0042] A control decision matrix is ​​constructed based on the power regulation parameter set and the voltage stability parameter set to determine the control strategy parameters.

[0043] Preferably, obtaining the micro-inverter operating parameters based on the control strategy parameters includes:

[0044] Real-time output power and load change data are obtained from the microinverter, and fuzzy values ​​are calculated using a preset membership function μ(P), where P represents the output power, to obtain a set of fuzzy control variables.

[0045] If the power deviation indicated by the fuzzy control quantity set exceeds the preset power deviation threshold, a pulse width modulation signal adjustment instruction is generated to determine the duty cycle adjustment range.

[0046] The pulse width modulation signal is updated based on the duty cycle adjustment amplitude to obtain the optimized operating parameters of the micro-inverter.

[0047] Secondly, the present invention provides a micro-inverter monitoring and control system based on Internet of Things (IoT) data, comprising:

[0048] The detection end is used to collect environmental parameters in the operating environment of the micro-inverter and obtain a multi-dimensional environmental parameter matrix, and obtain an operating scenario identification vector based on the multi-dimensional environmental parameter matrix. The processing end is used to perform weighted fusion processing on heterogeneous data from different sensor nodes based on the operating scenario identification vector to obtain a fused state feature vector. Anomaly detection is performed on the state feature vector. If an anomaly exists in one of the sensor nodes, the corresponding weight is reduced to obtain an updated state feature vector and time series analysis is performed. If the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold, an anomaly score is obtained. Control strategy parameters are determined based on the anomaly score and the operating scenario identification vector. The optimization end is used to obtain the micro-inverter operating parameters based on the control strategy parameters, input the inverter operating parameters into the system, and obtain an updated multi-dimensional environmental parameter matrix to obtain optimized micro-inverter operating parameters.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) This invention collects environmental parameters in the operating environment of a micro inverter and obtains a multi-dimensional environmental parameter matrix. Based on the multi-dimensional environmental parameter matrix, it obtains an operating scenario identification vector. Through multi-dimensional data analysis, it ensures a comprehensive characterization of the operating status.

[0051] (2) The present invention performs weighted fusion processing on the heterogeneous data of different sensor nodes based on the operation scenario identification vector to obtain the fused state feature vector, ensuring that the state feature vector fully reflects the operation environment and enhancing the system's adaptability to complex scenarios.

[0052] (3) The present invention performs anomaly detection on the state feature vector. If one of the sensor nodes is abnormal, the corresponding weight is reduced to obtain the updated state feature vector and perform time series analysis. If the feature vector deviation at a consecutive preset number of time points exceeds the preset deviation threshold, an anomaly score is obtained. The combination of anomaly data source identification and weight adjustment can effectively filter unreliable data.

[0053] (4) The present invention determines the control strategy parameters based on the anomaly score and the operating scenario identification vector, thereby improving the system’s adaptability to environmental changes and providing a reliable basis for equipment operation.

[0054] (5) The present invention obtains the micro inverter operating parameters according to the control strategy parameters, inputs the micro inverter operating parameters into the system, obtains the updated multi-dimensional environmental parameter matrix, and obtains the optimized micro inverter operating parameters. The dynamic adjustment mechanism enables the inverter to maintain efficient operation under complex environment and load conditions. Attached Figure Description

[0055] Figure 1 This is a flowchart of a micro-inverter monitoring and control method based on Internet of Things data provided in an embodiment of the present invention;

[0056] Figure 2 This is a flowchart of another micro-inverter monitoring and control method based on Internet of Things data provided in this embodiment of the invention;

[0057] Figure 3 This is a schematic diagram of a micro-inverter monitoring and control system based on Internet of Things (IoT) data, provided in an embodiment of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Reference Figure 1 This invention provides a flowchart of a micro-inverter monitoring and control method based on Internet of Things (IoT) data, comprising the following steps:

[0060] S1. Collect environmental parameters in the operating environment of the micro-inverter and obtain a multi-dimensional environmental parameter matrix. Obtain an operating scenario identification vector based on the multi-dimensional environmental parameter matrix. S2. Perform weighted fusion processing on the heterogeneous data of different sensor nodes based on the operating scenario identification vector to obtain a fused state feature vector.

[0061] S3, perform anomaly detection on the state feature vector. If one of the sensor nodes is abnormal, reduce the corresponding weight to obtain the updated state feature vector and perform time series analysis. If the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold, obtain an anomaly score.

[0062] S4, determine the control strategy parameters based on the anomaly score and the operating scenario identifier vector; S5, obtain the micro-inverter operating parameters based on the control strategy parameters, input the inverter operating parameters into the system and obtain the updated multi-dimensional environmental parameter matrix to obtain the optimized micro-inverter operating parameters.

[0063] In step S1, environmental parameters in the micro-inverter operating environment are collected and a multi-dimensional environmental parameter matrix is ​​obtained. An operating scenario identifier vector is obtained based on the multi-dimensional environmental parameter matrix.

[0064] Preferably, the step of collecting environmental parameters in the operating environment of the micro-inverter and obtaining a multi-dimensional environmental parameter matrix includes:

[0065] The temperature, light intensity, and grid load power are obtained from the operating environment of the micro-inverter according to the preset sampling frequency, and the multi-dimensional environmental parameter matrix is ​​constructed.

[0066] A time-series analysis was performed on the multi-dimensional environmental parameter matrix to extract the changing trend characteristics of the temperature value, the light intensity, and the power grid load power, thus obtaining a time-series feature dataset.

[0067] If at least one parameter in the time-series feature dataset exceeds a preset parameter threshold, the deviation between the time-series feature dataset and historical data is compared to determine the abnormal state.

[0068] Increase the sampling frequency of the parameters corresponding to the abnormal state and obtain the updated environmental parameters.

[0069] Specifically, the distributed sensor network collects temperature T(t), light intensity L(t), and grid load power P(t) from the micro-inverter operating environment through multi-node collaborative work, forming an environmental parameter matrix E(t).

[0070] For example, in a distributed photovoltaic system, sensor nodes are deployed near the inverter, collecting data at a preset sampling frequency of once per minute. Assuming a certain moment T(t) = 35°C, L(t) = 800 W / m², and P(t) = 5 kW, these data form a matrix E(t), which is transmitted in real-time to the local processing unit via a wireless communication module. This acquisition method ensures high data timeliness, providing a reliable foundation for subsequent analysis.

[0071] Time-series analysis was performed on E(t) to extract trend characteristics. For example, by using a sliding window technique to analyze the mean and variance of T(t) over the past 10 minutes, it was found that the temperature slowly rose from 30°C to 35°C, indicating that the inverter's heat dissipation might be insufficient. The illuminance L(t) suddenly increased from 700W / m² to 800W / m², suggesting possible fluctuations in illuminance due to cloud movement. The grid load power P(t) increased from 4kW to 5kW, reflecting increased electricity demand. The extracted time-series feature dataset contains these trends and is used for subsequent anomaly detection. This analytical method can capture the dynamic characteristics of parameters and enhance the system's ability to perceive environmental changes.

[0072] For example, a temperature threshold of 40°C, a maximum light intensity of 1000W / m², and a maximum load power of 6kW are set. If T(t) = 42°C, exceeding the threshold, an anomaly detection is triggered. By comparing with historical data, assuming the historical average T(t) is 32°C, a deviation of 10°C is identified as a heat dissipation anomaly. After the anomaly is triggered, the system dynamically adjusts the sampling frequency, for example, increasing the temperature sampling frequency from once per minute to once every 10 seconds, to monitor the anomaly parameters more precisely. This dynamic adjustment mechanism improves the sensitivity of anomaly detection and reduces the risk of false positives. The adjusted sampling frequency generates a new E(t) containing higher frequency temperature data, such as T(t) = 42.5°C, 42.3°C, etc.

[0073] Assuming that the system detects temperature anomalies in advance and triggers the cooling fan to operate using the above method, reducing the temperature to 38°C and returning it to the normal range, this not only ensures the stable operation of the inverter but also optimizes the overall power generation efficiency of the photovoltaic system, reduces maintenance costs, and improves economic benefits.

[0074] Preferably, obtaining the operational scenario identifier vector based on the multi-dimensional environmental parameter matrix includes:

[0075] The temperature change rate is calculated based on the multi-dimensional environmental parameter matrix to obtain the feature dataset;

[0076] If the temperature change rate in the feature dataset is greater than a preset temperature change threshold and the light intensity is less than a preset light intensity threshold, then it is determined to be an inefficient operating mode, and a first scenario identifier vector is generated.

[0077] The fluctuation variance of the power grid load is calculated based on the feature dataset. If the fluctuation variance is greater than a preset variance threshold, it is determined to be an unstable operating scenario, and a second scenario identification vector is generated.

[0078] The first scenario identifier vector and the second scenario identifier vector are integrated to generate the running scenario identifier vector.

[0079] Specifically, distributed sensor networks collect environmental parameter matrices E(t)=[T(t), L(t), P(t)] through multi-node collaborative acquisition, providing a real-time data foundation for micro-inverter operation status analysis. The following sections provide detailed analysis and examples of technical topics such as temperature change rate calculation, pattern recognition algorithms, grid load power fluctuation analysis, and support vector machine integration, using photovoltaic system scenarios as examples, focusing on a single business area.

[0080] For example, the rate of temperature change, dT / dt, reflects the thermal dynamics of the inverter's operating environment. Sensor nodes collect temperature data T(t) at a frequency of once per minute, and dT / dt is calculated through time series analysis. Assuming that in a photovoltaic power station, at a certain time t1, T(t1) = 30°C, and at t2, T(t2) = 34°C, with a time interval of 2 minutes, then dT / dt = (34-30) / 2 = 2°C / min. If a preset threshold α = 1.5°C / min is set, and dT / dt > α, it indicates that the temperature rises too rapidly, possibly due to insufficient inverter heat dissipation or accumulation of ambient heat. This analysis helps to promptly identify potential overheating risks and optimize heat dissipation strategies.

[0081] Assuming a preset light intensity threshold β = 600 W / m², at a certain moment L(t) = 500 W / m², and dT / dt > α, the system analyzes historical data using a pattern recognition algorithm and finds that rapid temperature increases under low light conditions are often accompanied by a decrease in inverter efficiency. A first scenario identifier vector S1(t) is generated and marked as an inefficient operating mode.

[0082] In a cloudy day scenario, the light intensity L(t) remains below 600W / m², and the inverter enters an inefficient state due to insufficient input power. S1(t) records this state characteristic to facilitate subsequent optimization of power allocation.

[0083] In one embodiment, the variance Var(P) of the grid load power P(t) is used to assess operational stability. A sensor network collects P(t) data. Assuming the P(t) sequence over 10 minutes is 4.5kW, 5.2kW, 4.8kW, and 5.5kW, calculating Var(P) reveals it exceeds a preset threshold γ = 0.5kW², indicating severe load fluctuations. A data fusion algorithm analyzes the source of the fluctuations, which may be due to peak electricity consumption or grid instability, generating a second scenario identification vector S2(t) to mark it as an unstable operating scenario. This analysis helps the system dynamically adjust its output power and stabilize grid interaction.

[0084] Integrating S1(t) and S2(t), a final operating scenario identification vector S(t) is generated. In a photovoltaic system, S1(t) indicates inefficient operation, and S2(t) indicates load fluctuations. Combining the two vectors generates S(t), which describes the inverter as being in an "inefficient and unstable" state. The integration methods include logical integration, weighted linear combination, fuzzy logic integration, machine learning-based nonlinear fusion, and time-window dynamic integration. The appropriate method must be selected based on data characteristics and application requirements. For example, simple systems can use Boolean logic or weighted linear combination; complex and uncertain systems can use fuzzy logic or machine learning; and time-dependent systems can be combined with time-window dynamic verification. The ultimate goal is to accurately characterize the inverter's complex fault states, providing a basis for operation and maintenance decisions. The system adjusts operating parameters accordingly, such as reducing inverter output power to reduce overload risk. This integration improves the accuracy of state identification and enhances the system's adaptability to complex operating environments.

[0085] It should be noted that the above method uses multi-dimensional data analysis to ensure a comprehensive characterization of the operating status. S(t) can be used to guide the dynamic adjustment of the inverter, such as prioritizing the allocation of heat dissipation resources in inefficient modes and optimizing power output strategies in unstable scenarios, thereby improving the overall efficiency and reliability of the system.

[0086] In step S2, the heterogeneous data from different sensor nodes are weighted and fused according to the running scenario identifier vector to obtain the fused state feature vector.

[0087] In the Internet of Things (IoT) field, heterogeneous data refers to a collection of diverse data from different sources, in different formats, structures, and semantics. This data exhibits inherent inconsistencies due to differences in device type, communication protocols, and application scenarios, requiring special processing to achieve unified analysis and value extraction.

[0088] Preferably, the step of weighted fusion processing of heterogeneous data from different sensor nodes based on the operating scenario identifier vector to obtain a fused state feature vector includes:

[0089] Obtain heterogeneous data and denoise it to obtain a denoised data matrix;

[0090] Calculate the information entropy value of each data source based on the denoised data matrix;

[0091] If the information entropy value of any data source is lower than the preset entropy threshold, the credibility weight of the corresponding data source is obtained, and a weight vector is obtained.

[0092] The weight vector and the denoised data matrix are weighted and fused to generate a fused state feature vector.

[0093] For example, in a photovoltaic system, distributed sensor nodes collect environmental and operational data in real time, forming a heterogeneous data matrix D(t), which includes temperature T(t), illuminance L(t), and grid load power P(t). This data often contains interference due to environmental noise or sensor errors, affecting the accuracy of inverter condition analysis. The Kalman filter algorithm denoises D(t) through prediction and update steps, generating a first denoised data matrix D1(t).

[0094] A photovoltaic power station sensor collects a temperature of T(t) = 32°C, but due to noise fluctuations, it may display as 31.8°C to 32.2°C. After filtering, a smoothed value of T(t) = 32°C is obtained and included in D1(t). Based on D1(t), the information entropy value H of each data source is calculated. Information entropy reflects the uncertainty of the data; the lower the value, the more stable the data.

[0095] For example, if the probability distribution of T(t) at a certain moment is uniform and the information entropy H(T) is high, it indicates that the temperature data fluctuates greatly; if the L(t) data is concentrated in a certain range and H(L) is low, it indicates that the illumination data is stable. Assuming a preset threshold θ=0.8, if H(T)=0.6<θ, it indicates that the temperature data is highly stable and has high reliability. Next, the reliability weight W(t) of the data source is determined by weighted calculation.

[0096] For example, T(t) has a low H value and is given a higher weight W(T)=0.5, while L(t) and P(t) have higher H values ​​and are given weights W(L)=0.3 and W(P)=0.2 respectively, forming a weight vector W1(t).

[0097] For example, W1(t) is used to weight and fuse D1(t) to generate a state feature vector F(t). The fusion process integrates the credibility of each data source, highlighting the influence of reliable data.

[0098] For example, at a certain moment, D1(t) includes T(t) = 32°C, L(t) = 550W / m², and P(t) = 4.8kW. After weighting, F(t) is more likely to reflect the influence of temperature on the inverter's state because its data is more stable. This method ensures that F(t) accurately describes the inverter's operating state, such as its operating characteristics under high temperature and low light conditions.

[0099] F(t) can be used for subsequent inverter control optimization, such as adjusting power output to adapt to environmental changes. Assume a photovoltaic system operates on a cloudy day with L(t) = 500 W / m², below normal values, T(t) stable at 30°C, and P(t) fluctuating little. After Kalman filtering for noise reduction, the D1(t) data is smoother, and the information entropy calculation shows H(T) is the lowest, while the weight W(T) is the highest. The fused F(t) highlights temperature characteristics, indicating that the inverter needs to focus on heat dissipation management. This analysis improves the accuracy of the state description and provides a basis for dynamic adjustment. This method, through multi-dimensional data processing, ensures that the state feature vector comprehensively reflects the operating environment, enhancing the system's adaptability to complex scenarios.

[0100] In step S3, anomaly detection is performed on the state feature vector. If one of the sensor nodes is abnormal, the corresponding weight is reduced to obtain an updated state feature vector and perform time series analysis. If the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold, an anomaly score is obtained.

[0101] Preferably, the step of performing anomaly detection on the state feature vector, and reducing the corresponding weight if one of the sensor nodes is abnormal, to obtain an updated state feature vector, includes:

[0102] The anomaly score of each data source is calculated based on the denoised data matrix. If the anomaly score is higher than a preset score threshold, the corresponding data source is determined to be abnormal, and the abnormal data source is obtained.

[0103] The weights of the abnormal data source are updated based on the abnormal scores to obtain the updated weight vector;

[0104] The updated weight vector and the denoised data matrix are weighted and fused to generate an updated state feature vector.

[0105] In the field of Internet of Things (IoT) data, denoised data matrices and anomaly scores are two key concepts, typically used for data preprocessing and anomaly detection. A denoised data matrix is ​​a clean dataset obtained by removing noise (irrelevant information, random errors, or interference) from raw IoT data using algorithms or techniques. Noise may originate from sensor errors, communication interference, environmental fluctuations, etc. Anomaly scores are a quantitative metric used to measure the degree of anomaly at a specific data point (or time period) in IoT data. A higher score indicates a greater difference between that point and the normal pattern.

[0106] Dynamically updating the weights of anomalous data sources based on anomaly scores is a key step in optimizing system robustness and decision accuracy. The core idea is to reduce the impact of anomalous data sources on the overall analysis by quantifying the reliability of data sources (such as sensors, devices, or data streams).

[0107] A state feature vector is a numerical representation of the overall state of an IoT entity (such as a device, sensor, network node, or system) at a specific point in time. It quantifies the entity's operational status, environmental context, or behavioral patterns through a set of features (i.e., dimensions), and is fundamental to tasks such as data analysis, anomaly detection, and predictive modeling.

[0108] For example, a photovoltaic power station sensor collects a light intensity L(t) of 600 W / m², but due to cloud cover, the data fluctuates between 590 and 610 W / m². After filtering, a smoothed value L(t) of 600 W / m² is obtained and included in D1(t). This method ensures that the data is closer to the true value, providing a reliable basis for subsequent analysis. An anomaly score Score(t) is calculated for each data source based on D1(t). The anomaly score is determined by comparing the deviation between the current data and historical data.

[0109] For example, if the temperature T(t) = 30°C, and the historical average is 29°C with a standard deviation of 0.5°C, then the deviation of T(t) is small, and the score (T) is low, indicating that the data is normal. If the grid load power P(t) = 5kW, the historical average is 4.5kW, and the standard deviation is 0.8kW, then the score (P) is high, which may indicate an anomaly. Setting a threshold θ = 0.7, if the score (P) = 0.9 > θ, then P(t) is determined to be an anomaly and included in the anomaly data source set A(t). This anomaly detection method helps identify unreliable data and ensures the accuracy of subsequent data fusion.

[0110] For example, the probability distribution of T(t) is concentrated, with H(T) = 0.5, indicating stable temperature data; L(t) fluctuates significantly due to cloud cover, with H(L) = 0.9. For P(t) in the abnormal data source set A(t), its Score(P) = 0.9 and information entropy H(P) = 0.8, indicating low data reliability. Based on this, the weight of P(t) is reduced, for example, from 0.3 to 0.1, while the weights of T(t) and L(t) are adjusted to 0.5 and 0.4 respectively, forming the updated weight vector W2(t). This weight adjustment method emphasizes the influence of reliable data and improves the accuracy of the fusion results.

[0111] In one possible implementation, W2(t) is used to weight and fuse D1(t) to generate an optimized state feature vector F1(t).

[0112] For example, at a certain moment, D1(t) includes T(t) = 30°C, L(t) = 600W / m^2, and P(t) = 5kW. After weighting, F1(t) is more likely to reflect the effects of temperature and light intensity, as P(t) is considered abnormal and has a lower weight. F1(t) can be used to describe the operating status of a micro-inverter, for example, in high-temperature and high-light scenarios, indicating that the inverter needs to optimize heat dissipation and power distribution. This method ensures that F1(t) accurately reflects the system status by comprehensively considering the reliability of the data source, providing a basis for the dynamic adjustment of the inverter.

[0113] Understandably, the combination of identifying abnormal data sources and adjusting weights can effectively filter out unreliable data.

[0114] Preferably, the step of obtaining an anomaly score if the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold includes:

[0115] The timing data sequence during the operation of the microinverter is obtained and segmented to obtain local data subsets within multiple time windows;

[0116] The deviation value of the state feature vector is calculated based on the local data subset. If the deviation value exceeds the preset deviation threshold, it is determined that there is a potential abnormal window, and a set of potential abnormal windows is obtained.

[0117] The state feature vectors in the set of potential anomaly windows are processed to calculate the anomaly score.

[0118] In the Internet of Things (IoT) field, the potential anomaly window set is a time period that may contain anomalies, selected through preliminary detection (such as sliding window statistics and abrupt change detection). The state feature vector consists of numerical features extracted within each window (such as mean, variance, frequency domain energy, and temporal patterns). The anomaly score quantifies the degree to which the state deviates from normal within the window; a higher score indicates a greater anomaly.

[0119] Anomaly scores are calculated based on state feature vectors from a set of potential anomaly windows, and the appropriate method must be selected based on the data characteristics. Unsupervised methods (such as LOF and Isolation Forest) are suitable for unlabeled data; supervised methods are suitable for scenarios with historical anomaly annotations; and time-series methods (such as LSTM) are more effective for device status monitoring.

[0120] In photovoltaic systems, the time-series data sequence D(t) collected by distributed sensor nodes includes temperature T(t), irradiance L(t), and grid load power P(t), which reflect the system's operating status. To effectively handle data fluctuations, a sliding window mechanism divides D(t) into multiple local data subsets S(t) within a time window.

[0121] Specifically, assuming each window covers 10 minutes and the sampling frequency is 1 time / minute, each S(t) contains 10 sets of data, such as T(t) from 28°C to 30°C, L(t) from 550W / m^2 to 600W / m^2, and P(t) from 4.8kW to 5.2kW. This segmented processing facilitates the analysis of local data characteristics and the capture of short-term fluctuations.

[0122] In one possible implementation, the deviation value B(t) of the state feature vector F(t) is calculated for each S(t). F(t) consists of the statistical characteristics (such as mean and variance) of T(t), L(t), and P(t). The deviation value B(t) is obtained by comparing the difference between F(t) and historical data.

[0123] For example, if the average temperature (T(t)) within a certain window is 29°C, and the historical average is 28.5°C, the deviation is small; however, the average temperature (P(t)) is 5.1kW, and the historical average is 4.5kW, the deviation is large. If the preset threshold δ is 0.6, and the value of B(t) for P(t) exceeds δ, then the window is identified as a potential abnormal window and included in the set A(t). This method can quickly filter out abnormal data regions.

[0124] For example, if the path length is short and the volatile P(t) of a certain window is large, the score(t) will be high, such as 0.85; if the path length is long and the T(t) is stable, the score(t) will be low, such as 0.3. This algorithm is suitable for processing high-dimensional data and can efficiently distinguish abnormal features.

[0125] In one possible implementation, F2(t) is used to guide the operation of the photovoltaic inverter. For example, if F2(t) shows high temperature and high light intensity characteristics, it indicates that the inverter needs to strengthen its heat dissipation management; if L(t) fluctuates significantly due to cloud cover, F2(t) can guide the inverter to optimize power output. This method improves the system's adaptability to environmental changes by dynamically adjusting data weights, providing a reliable status basis for the inverter.

[0126] refer to Figure 2 Before step S4, the following is also included:

[0127] S6. If the abnormal score is higher than the preset score threshold, it is marked as a high abnormal node, and a set of high abnormal nodes is obtained.

[0128] Based on the set of highly abnormal nodes, the data weights of the corresponding nodes are adjusted to obtain the adjusted set of node weights. Combined with the node priority, a data source selection list is obtained.

[0129] Based on the data source selection list, process the time series data to generate optimized fused data input.

[0130] In the Internet of Things (IoT) field, a Data Source Selection List refers to the set of available data input sources and their attribute descriptions within an IoT system or data analysis process. It is typically presented as a structured list and used for dynamically configuring or optimizing data acquisition strategies. Fused Data Input refers to the integration, correlation, and comprehensive processing of heterogeneous data from different sensors, devices, or systems through multi-source data fusion technology (Data Fusion) to generate more accurate, complete, or high-dimensional data input to support decision-making or control. Its core objective is to eliminate the limitations of a single data source (such as noise and incompleteness) and improve the system's sensing capabilities and reliability.

[0131] Specifically, the time-series data collected by distributed sensor nodes is the core of photovoltaic system operation status analysis, covering key indicators such as temperature, irradiance, and grid load power. Regarding technical topics such as abnormal node identification, weight adjustment, and data fusion, the following explanations, through specific scenario analysis and examples, combined with the business logic of photovoltaic systems, highlight the implementation methods and beneficial effects of these technical topics.

[0132] For example, in a photovoltaic system, distributed sensor nodes collect data every minute, forming a time-series data sequence that includes temperature T(t), illuminance L(t), and grid load power P(t). The short-term mean of T(t), L(t), and P(t) for each node is calculated, and its deviation from the long-term mean is compared.

[0133] For example, a node has a temperature T(t) of 30°C within a 10-minute window, a long-term temperature of 28°C, and a deviation of 2°C; a temperature L(t) of 570 W / m², a long-term temperature of 550 W / m², and a deviation of 20 W / m²; and a temperature P(t) of 5.0 kW, a long-term temperature of 4.8 kW, and a deviation of 0.2 kW. With a preset threshold ε of 0.5, the deviations of P(t) and L(t) do not exceed the threshold, but the deviation of T(t) is large. This node is marked as a high-anomaly node and included in the high-anomaly node set. This method can quickly filter out nodes with abnormal data, ensuring the reliability of subsequent analysis. Regarding weight adjustment, the data weights of high-anomaly nodes need to be reduced to minimize their impact on system analysis.

[0134] For example, initially, each node has a weight of 0.1 (evenly distributed across 10 nodes). If node 1 and node 2 are marked as high-anomaly nodes, their weights are reduced from 0.1 to 0.05, while the weights of the remaining 8 nodes are correspondingly increased to 0.1125, and then redistributed using a weighted average method. This adjustment ensures that the impact of outlier data on the system is minimized while preserving the contribution of normal nodes.

[0135] It should be noted that the weight adjustment takes into account the node priority.

[0136] For example, nodes closer to the core area of ​​the photovoltaic array have higher priority, and their weights remain relatively advantageous even if reduced. For instance, node 1 has a higher priority, so its weight is adjusted to 0.07 instead of 0.05. This approach improves the targeting of data processing. The optimized data source selection list is generated based on the adjusted weight set and priorities.

[0137] For example, the system prioritizes nodes with a weight higher than 0.1 and ranking in the top 5 to form a data source selection list. Assuming nodes 3, 4, 5, 7, and 8 meet the conditions, their T(t), L(t), and P(t) data are selected first. This selection method ensures that the data source is reliable and representative.

[0138] Preferably, a data weighted fusion algorithm processes the data from selected nodes to generate fused data input. In the fused data, the weighted average of T(t) is 29.5°C, L(t) is 565 W / m², and P(t) is 4.9 kW. This fused data more accurately reflects the overall system state.

[0139] For example, fused data input can be used to guide the adjustment of operating parameters of photovoltaic inverters. When the fused data shows that T(t) is too high, the system prompts the inverter to increase heat dissipation power; when L(t) decreases due to cloud cover, the power output strategy is optimized. This method, through precise data filtering and fusion, improves the system's adaptability to environmental changes and provides a reliable basis for equipment operation.

[0140] In step S4, control strategy parameters are determined based on the anomaly score and the operating scenario identifier vector.

[0141] In the Internet of Things (IoT) field, control strategy parameters refer to a set of configurable variables or rules used to define, adjust, or optimize the control behavior of devices or processes in an IoT system. These parameters directly affect key performance indicators such as system response, energy efficiency, and stability.

[0142] The step of determining control strategy parameters based on the anomaly score and the operating scenario identifier vector includes:

[0143] Calculate the anomaly score and motion scenario identifier vector for each node to obtain the anomaly score set and scenario identifier set;

[0144] If the abnormal score of any node in the abnormal score set exceeds a preset threshold and the scenario identification vector indicates the inefficient operation mode, then a power adjustment instruction is generated, and a power adjustment parameter set is generated.

[0145] If the abnormal score of any node in the abnormal score set exceeds a preset threshold and the scenario identification vector indicates the unstable scenario, then a voltage stabilization command is generated, the voltage adjustment range is calculated, and a voltage stabilization parameter set is obtained.

[0146] A control decision matrix is ​​constructed based on the power regulation parameter set and the voltage stability parameter set to determine the control strategy parameters.

[0147] For example, a node might have a short-term temperature T(t) of 32°C and a long-term temperature of 28°C, showing a significant deviation; a short-term temperature L(t) of 540 W / m² and a long-term temperature of 550 W / m², showing a relatively small deviation; and a short-term temperature P(t) of 4.7 kW and a long-term temperature of 4.8 kW, with negligible deviation. Assuming a preset threshold of 0.5, if the T(t) deviation exceeds this threshold, the node is marked as an anomaly and included in the anomaly score set. This method can accurately identify data anomaly nodes, ensuring the reliability of subsequent decisions.

[0148] In one possible implementation, the context labeling vector is generated based on the node's operating state, reflecting efficient, inefficient, or unstable operating modes.

[0149] For example, if a node's T(t) is high and L(t) is low, it may be due to local overheating or shading, and the system generates an inefficient operation mode identifier vector. If P(t) fluctuates frequently, an unstable scenario identifier vector is generated. The scenario identifier set is generated using a vector clustering algorithm to ensure accurate classification of operating states. This method provides a precise basis for subsequent instruction generation. If a node's abnormal score exceeds a critical value and the scenario identifier indicates an inefficient operation mode, the system generates a power adjustment instruction.

[0150] For example, if node 1's T(t) is abnormal and identified as inefficient, the system determines a set of power regulation parameters through weighted calculation, reducing the output power of the inverter corresponding to that node, such as adjusting it from 5.0kW to 4.5kW. This adjustment reduces the risk of overheating and improves system stability. If a node's abnormal score exceeds the limit and the scenario is identified as unstable, the system generates a voltage stabilization command. For example, if node 2's P(t) fluctuates drastically, the system calculates the voltage adjustment range using an averaging algorithm, such as fine-tuning the voltage from 220V to 218V, and generates a set of voltage stabilization parameters. This method effectively smooths grid fluctuations and improves power supply quality.

[0151] In one embodiment, the control decision matrix is ​​generated using a linear regression algorithm based on power regulation parameters and voltage stability parameters. For example, the system integrates the parameter sets of 10 nodes to construct a matrix and determine the optimal control strategy parameters, such as setting the inverter output power to 4.8kW and stabilizing the voltage at 219V. This strategy ensures that the overall system operates efficiently and adapts to environmental changes.

[0152] For example, in a certain scenario, nodes in the core area have higher priority, and their parameters have greater weight in the matrix. For instance, the T(t) and L(t) data of node 3 dominate the decision-making process. This approach ensures that data from the core area has a greater impact on the control strategy, improving the targeting of decisions. Through matrix analysis, the system dynamically adjusts operating parameters to ensure the efficient and stable operation of the photovoltaic system.

[0153] In step S5, the micro-inverter operating parameters are obtained according to the control strategy parameters, and the inverter operating parameters are input into the system to obtain the updated multi-dimensional environmental parameter matrix, thereby obtaining the optimized micro-inverter operating parameters.

[0154] Preferably, obtaining the micro-inverter operating parameters based on the control strategy parameters includes:

[0155] Real-time output power and load change data are obtained from the microinverter, and fuzzy values ​​are calculated using a preset membership function μ(P), where P represents the output power, to obtain a set of fuzzy control variables.

[0156] If the power deviation indicated by the fuzzy control quantity set exceeds the preset power deviation threshold, a pulse width modulation signal adjustment instruction is generated to determine the duty cycle adjustment range.

[0157] The pulse width modulation signal is updated based on the duty cycle adjustment amplitude to obtain the optimized operating parameters of the micro-inverter.

[0158] In the field of Internet of Things (IoT) intelligent control systems, the fuzzified control variable set refers to the set of variables obtained by fuzzifying traditional control variables (such as temperature, humidity, speed, etc.) through fuzzy logic. These variables are no longer precise values ​​that are either black or white, but rather "fuzzy" linguistic values ​​(such as "higher" or "lower") described by membership functions and conforming to human thinking habits.

[0159] For example, real-time data acquisition of output power and load variations from microinverters provides the foundation for fuzzy control. Microinverters are typically deployed in photovoltaic systems, with each inverter independently controlling a single photovoltaic panel. The output power P dynamically fluctuates with changes in sunlight, temperature, and load. Real-time power data is acquired via sensors at second-level frequencies. For instance, an inverter might output 250W at a given moment, while the load demand is 260W, resulting in a deviation of 10W. A membership function μ(P) is used to fuzzify the power deviation, generating a set of control variables. The membership function is based on preset rules; for example, a deviation of 0-5W is considered "normal," 5-15W is "slight deviation," and over 15W is "severe deviation." For a 10W deviation, the fuzzified value might be "slight deviation," thus being included in the fuzzy control variable set. This fuzzification method transforms continuous power data into discrete control data.

[0160] For example, a deviation of 10W is judged as a "minor deviation". The system generates a duty cycle adjustment instruction through logical judgment, deciding to reduce the duty cycle from 50% to 48%. The determination of the duty cycle adjustment range is based on the deviation magnitude and load characteristics. For example, when the load power demand is stable, the adjustment range is small; if the load is frequent, it is understandable that the digital signal processor will quickly update the pulse width modulation signal after receiving the duty cycle adjustment instruction.

[0161] For example, the processor generates a new pulse width modulation signal based on a 48% duty cycle, adjusting the inverter's switching time to gradually increase the output power from 250W to close to 260W. This fast-response adjustment mechanism ensures that the power output matches the load demand, reducing energy waste.

[0162] Specifically, the optimized set of inverter operating parameters includes updated duty cycle, switching frequency, and power output values. The digital signal processor updates the pulse width modulation signal based on the duty cycle adjustment instructions.

[0163] For example, the processor generates a 48% duty cycle signal to adjust the inverter's switching time, bringing the output power closer to 300W from 290W. This rapid adjustment ensures that the power output matches the load.

[0164] Preferably, the processor continuously monitors the grid status; for example, if the voltage fluctuates to 228V but remains within ±5V, it confirms that the inverter has entered a dynamically optimized operating state. If the voltage drops to 225V, the processor may further fine-tune the duty cycle to 47% to stabilize power and voltage.

[0165] In one embodiment, multiple inverters work together, and the parameter matrix E(t) and control decision matrix D are integrated through a centralized controller.

[0166] For example, inverters in core areas have a higher weighting factor, prioritizing the power stability of critical loads. This collaborative optimization approach, supported by multi-faceted data, enhances the overall adaptability of the photovoltaic system. The dynamic adjustment mechanism ensures that inverters operate efficiently under complex environmental and load conditions.

[0167] In summary, this invention discloses a micro-inverter supervisory control method based on Internet of Things (IoT) data. It collects multi-dimensional environmental parameters such as temperature, light intensity, and grid load, performs feature extraction and pattern recognition to determine the operating scenario. Multivariate time series analysis and Kalman filtering are used to weighted fuse heterogeneous data, and information entropy calculation is used to determine the data source reliability. Anomaly detection is performed, and data fusion weights are dynamically adjusted based on anomaly scores. A control decision matrix is ​​constructed based on the operating scenario and anomaly severity, and a fuzzy logic controller is used to intelligently adjust the inverter output. Feedback control continuously optimizes operating parameters, achieving efficient and stable operation of the micro-inverter. This invention can effectively cope with complex and variable photovoltaic power generation environments, improving system reliability and power generation efficiency.

[0168] Reference Figure 3 This invention provides a schematic diagram of a micro-inverter monitoring and control system based on Internet of Things (IoT) data, comprising:

[0169] The detection end is used to collect environmental parameters in the operating environment of the micro-inverter and obtain a multi-dimensional environmental parameter matrix, and obtain an operating scenario identification vector based on the multi-dimensional environmental parameter matrix. The processing end is used to perform weighted fusion processing on heterogeneous data from different sensor nodes based on the operating scenario identification vector to obtain a fused state feature vector. Anomaly detection is performed on the state feature vector. If an anomaly exists in one of the sensor nodes, the corresponding weight is reduced to obtain an updated state feature vector and time series analysis is performed. If the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold, an anomaly score is obtained. Control strategy parameters are determined based on the anomaly score and the operating scenario identification vector. The optimization end is used to obtain the micro-inverter operating parameters based on the control strategy parameters, input the inverter operating parameters into the system, and obtain an updated multi-dimensional environmental parameter matrix to obtain optimized micro-inverter operating parameters.

[0170] It should be noted that the micro-inverter monitoring and control system based on Internet of Things (IoT) data provided in this embodiment of the invention is used to execute all the process steps of the micro-inverter monitoring and control method based on IoT data in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0171] This invention also provides a terminal device. The terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the various embodiments of the micro-inverter supervisory control method based on Internet of Things data described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments.

[0172] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0173] The terminal device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of terminal devices and do not constitute a limitation on the terminal device. It may include more or fewer components than described above, or a combination of certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0174] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0175] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0176] Wherein, if the modules / units integrated in the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0177] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0178] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that 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 for those skilled in the art.

Claims

1. A micro-inverter supervisory control method based on Internet of Things (IoT) data, characterized in that, include: Collect environmental parameters in the micro-inverter's operating environment and obtain a multi-dimensional environmental parameter matrix; obtain the operating scenario identification vector based on the multi-dimensional environmental parameter matrix. The heterogeneous data from different sensor nodes are weighted and fused based on the operation scenario identifier vector to obtain the fused state feature vector. Anomaly detection is performed on the state feature vector. If one of the sensor nodes is abnormal, the corresponding weight is reduced to obtain an updated state feature vector and perform time series analysis. If the feature vector deviation exceeds a preset deviation threshold for a preset number of consecutive time points, an anomaly score is obtained. The control strategy parameters are determined based on the anomaly score and the operational scenario identifier vector. The micro-inverter operating parameters are obtained based on the control strategy parameters, and the micro-inverter operating parameters are input into the system to obtain the updated multi-dimensional environmental parameter matrix, thereby obtaining the optimized micro-inverter operating parameters. The step of collecting environmental parameters from the microinverter's operating environment and obtaining a multi-dimensional environmental parameter matrix includes: acquiring temperature values, light intensity, and grid load power from the microinverter's operating environment according to a preset sampling frequency, and constructing the multi-dimensional environmental parameter matrix; performing time-series analysis on the multi-dimensional environmental parameter matrix to extract the changing trend characteristics of the temperature value, light intensity, and grid load power, and obtaining a time-series feature dataset; if at least one parameter in the time-series feature dataset exceeds a preset parameter threshold, comparing the deviation between the time-series feature dataset and historical data to determine an abnormal state; increasing the sampling frequency for the parameters corresponding to the abnormal state and obtaining updated environmental parameters; The step of obtaining the operating scenario identifier vector based on the multi-dimensional environmental parameter matrix includes: calculating the temperature change rate based on the multi-dimensional environmental parameter matrix to obtain a feature dataset; if the temperature change rate in the feature dataset is greater than a preset temperature change threshold and the light intensity is less than a preset light intensity threshold, then it is determined to be an inefficient operating mode, and a first scenario identifier vector is generated; calculating the fluctuation variance of the power grid load power based on the feature dataset, if the fluctuation variance is greater than a preset variance threshold, then it is determined to be an unstable operating scenario, and a second scenario identifier vector is generated; and integrating the first scenario identifier vector and the second scenario identifier vector to generate the operating scenario identifier vector.

2. The micro-inverter monitoring and control method based on Internet of Things data according to claim 1, characterized in that, The step of weighted fusion processing of heterogeneous data from different sensor nodes based on the operating scenario identifier vector to obtain a fused state feature vector includes: Obtain heterogeneous data and denoise it to obtain a denoised data matrix; Calculate the information entropy value of each data source based on the denoised data matrix; If the information entropy value of any data source is lower than the preset entropy threshold, the credibility weight of the corresponding data source is obtained, and a weight vector is obtained. The weight vector and the denoised data matrix are weighted and fused to generate a fused state feature vector.

3. The micro-inverter monitoring and control method based on IoT data according to claim 2, characterized in that, The step of performing anomaly detection on the state feature vector, whereby if an anomaly is found in one of the sensor nodes, the corresponding weight is reduced to obtain an updated state feature vector, includes: The anomaly score of each data source is calculated based on the denoised data matrix. If the anomaly score is higher than a preset score threshold, the corresponding data source is determined to be abnormal, and the abnormal data source is obtained. The weights of the abnormal data source are updated based on the abnormal scores to obtain the updated weight vector; The updated weight vector and the denoised data matrix are weighted and fused to generate an updated state feature vector.

4. The micro-inverter monitoring and control method based on Internet of Things data according to claim 3, characterized in that, If the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold, an anomaly score is obtained, including: The timing data sequence during the operation of the microinverter is obtained and segmented to obtain local data subsets within multiple time windows; The deviation value of the state feature vector is calculated based on the local data subset. If the deviation value exceeds the preset deviation threshold, it is determined that there is a potential abnormal window, and a set of potential abnormal windows is obtained. The state feature vectors in the set of potential anomaly windows are processed to calculate the anomaly score.

5. The micro-inverter monitoring and control method based on Internet of Things data according to any one of claims 1-4, characterized in that, Before determining the control strategy parameters based on the anomaly score and the operating scenario identifier vector, the method further includes: If the abnormal score is higher than the preset score threshold, it is marked as a high abnormal node, and a set of high abnormal nodes is obtained. Based on the set of highly abnormal nodes, the data weights of the corresponding nodes are adjusted to obtain the adjusted set of node weights. Combined with the node priority, a data source selection list is obtained. Based on the data source selection list, process the time series data to generate optimized fused data input.

6. The micro-inverter monitoring and control method based on Internet of Things data according to claim 1, characterized in that, The step of determining control strategy parameters based on the anomaly score and the operating scenario identifier vector includes: Calculate the anomaly score and scenario identification vector for each node to obtain the anomaly score set and scenario identification set; If the abnormal score of any node in the abnormal score set exceeds a preset threshold and the running scenario identifier vector indicates the inefficient running mode, then a power adjustment instruction is generated and a power adjustment parameter set is generated. If the abnormal score of any node in the abnormal score set exceeds the preset threshold and the operating scenario identifier vector indicates the unstable operating scenario, then a voltage stabilization command is generated, the voltage adjustment range is calculated, and a voltage stabilization parameter set is obtained. A control decision matrix is ​​constructed based on the power regulation parameter set and the voltage stability parameter set to determine the control strategy parameters.

7. The micro-inverter monitoring and control method based on Internet of Things data according to any one of claims 1-4, characterized in that, The process of obtaining the micro-inverter operating parameters based on the control strategy parameters includes: Real-time output power and load change data are obtained from the microinverter, and fuzzy values ​​are calculated using a preset membership function μ(P), where P represents the output power, to obtain a set of fuzzy control variables. If the power deviation indicated by the fuzzy control quantity set exceeds the preset power deviation threshold, a pulse width modulation signal adjustment instruction is generated to determine the duty cycle adjustment range. The pulse width modulation signal is updated based on the duty cycle adjustment amplitude to obtain the optimized operating parameters of the micro-inverter.

8. A micro-inverter monitoring and control system based on Internet of Things (IoT) data, characterized in that, For implementing the method as described in any one of claims 1-7, comprising: The detection end is used to collect environmental parameters in the operating environment of the micro-inverter and obtain a multi-dimensional environmental parameter matrix, and obtain the operating scenario identification vector based on the multi-dimensional environmental parameter matrix; The processing unit performs weighted fusion processing on heterogeneous data from different sensor nodes based on the operating scenario identifier vector to obtain a fused state feature vector; it performs anomaly detection on the state feature vector, and if one of the sensor nodes has an anomaly, it reduces the corresponding weight to obtain an updated state feature vector and performs time series analysis. If the feature vector deviation at a consecutive preset number of time points exceeds a preset deviation threshold, it obtains an anomaly score; and it determines control strategy parameters based on the anomaly score and the operating scenario identifier vector. The optimization end is used to obtain the micro-inverter operating parameters according to the control strategy parameters, input the inverter operating parameters into the system, obtain the updated multi-dimensional environmental parameter matrix, and obtain the optimized micro-inverter operating parameters.

Images