Photovoltaic meteorological monitoring method and system based on multi-sensor cooperation

Abstract

The invention discloses a photovoltaic meteorological monitoring method and system based on multi-sensor collaboration, and the method comprises the steps: cooperatively collecting meteorological initial signals, such as irradiation, temperature and wind speed, through a data collection device, carrying out the deviation correction of the initial signals through a calibration model, obtaining a reliable parameter set, carrying out the multi-source data fusion based on the parameter set, and carrying out the data fusion. And if the deviation in the fusion process exceeds a threshold value, abnormal points are automatically eliminated to form a high-quality fusion data set, then the fusion data set is efficiently sent to an application layer from a sensing layer by using a transmission protocol to form a complete data stream, and the potential output power of the photovoltaic power station is calculated through analyzing a meteorological change mode by a prediction model in the application layer. When the predicted power is lower than a preset threshold value, an early warning signal is triggered immediately, finally, a reliability verification mechanism is integrated to carry out secondary confirmation on a prediction result, data consistency is ensured, an accurate monitoring report is output finally, and scientific decision support is provided for operation and maintenance of a power station. The power generation efficiency and the operation reliability of the photovoltaic power station are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic meteorological monitoring, and particularly discloses a photovoltaic meteorological monitoring method and system based on multi-sensor cooperation. BACKGROUND

[0002] Photovoltaic meteorological monitoring is a basic field to ensure the efficient power generation and stable operation of photovoltaic power stations, which directly determines the adaptability of the power station to complex natural environment and the reliability of long-term benefits. In the outdoor environment, the meteorological conditions change dramatically, and accurate real-time meteorological parameters are of great significance to power prediction, fault diagnosis and operation and maintenance scheduling of the power station.

[0003] Although the current photovoltaic meteorological monitoring method has deployed multiple sensors, the main defect is that the data acquisition and processing process is easily affected by environmental interference and equipment limitations, resulting in that the parameter measurement accuracy is difficult to maintain stable in actual operation. The sensors are exposed to high temperature, dust, strong wind or extreme weather for a long time, and the measurement values will gradually deviate or produce noise, and the existing method often lacks timely identification and compensation mechanism for these disturbances, so that the reliability of the collected meteorological data is reduced, and then the subsequent application judgment is affected. These problems further converge into the core technical difficulty when multiple sensors work together, that is, how to ensure the high consistency and complementarity between data in the process of collecting different meteorological parameters by multiple sensors at the same time. Different sensors have different response characteristics to the same environmental factor, for example, the response speed and sensitivity of the irradiance sensor and the module backboard temperature sensor are not completely synchronized when the light changes rapidly, and the wind speed and direction sensor will be affected by local turbulence to produce instantaneous fluctuations. These differences lead to internal conflicts in multi-source data before fusion. If the data inconsistency caused by the difference in response characteristics of the sensors cannot be effectively solved, the system will face a prominent contradiction in actual business: when the light rises rapidly in the morning or a sudden gust of wind comes, the measurement values of parameters such as irradiance, temperature and wind speed will appear a short but significant deviation, and the fused meteorological information cannot accurately reflect the real affected state of the power station, resulting in that the power generation prediction deviates from the actual situation or the fault warning is delayed, and the operation and maintenance personnel are difficult to respond correctly in a timely manner.

[0004] Therefore, how to realize effective calibration and data coordination and consistency of the response characteristic difference in the process of multi-sensor cooperative sensing, so as to obtain meteorological information with high reliability and truly reflecting the dynamic environmental changes of the power station, has become a key problem to improve the operation decision level of photovoltaic power stations.

[0004] Therefore, how to realize effective calibration and data coordination and consistency of the response characteristic difference in the process of multi-sensor cooperative sensing, so as to obtain meteorological information with high reliability and truly reflecting the dynamic environmental changes of the power station, has become a key problem to improve the operation decision level of photovoltaic power stations. SUMMARY

[0005] The present application provides a photovoltaic meteorological monitoring method and system based on multi-sensor cooperation, which aims to solve at least one of the defects in the prior art.

[0006] One aspect of the present application relates to a photovoltaic weather monitoring method based on multi-sensor cooperation, comprising the following steps: S100, acquiring initial collection signals by cooperatively collecting weather data through a data collection device, correcting deviations by processing the initial collection signals using a calibration model to obtain a calibrated parameter set, wherein the weather data includes irradiation parameters, temperature monitoring values, and wind speed parameters; S200, performing data fusion operations according to the calibrated parameter set, fusing the irradiation parameters, temperature monitoring values, and wind speed parameters, and if the deviation exceeds a preset threshold during the fusion process, removing the corresponding abnormal data points to determine a fused data group; S300, using the fused data group to execute a transmission protocol in a network architecture, sending the fused data group from a perception layer to an application layer through a transmission layer to obtain a transmission completed data stream; S400, analyzing weather change patterns through the transmission completed data stream, calculating potential output power of a photovoltaic power station using a prediction model, determining whether the potential output power of the photovoltaic power station is lower than a preset threshold, and if so, triggering a warning signal to obtain a prediction result set; S500, integrating a reliability improvement mechanism according to the prediction result set, performing secondary verification on the prediction result set to confirm data consistency, and outputting a final weather monitoring report to support photovoltaic power station operation and maintenance decisions.

[0007] Further, step S100 includes: S110, cooperatively collecting weather data through a data collection device, setting different sampling frequencies for irradiation parameters, temperature monitoring values, and wind speed parameters, obtaining initial collection signals from multiple sensor nodes and performing digital processing to generate an original data set; S120, using a calibration model to correct deviations of the original data set to generate a calibrated parameter set.

[0008] Further, step S200 includes: S210, according to the irradiation parameters, temperature monitoring values, and wind speed parameters in the calibrated parameter set, performing preliminary linear combination through a weighted fusion algorithm to generate an initial fusion numerical sequence; S220, using a sliding window mechanism to perform time series smoothing processing on the initial fusion numerical sequence to generate a smoothed fusion data stream; S230, detecting deviations of the smoothed fusion data stream through a preset rule and removing abnormal points to generate a purified fusion data set; S240, correcting the purified fusion data set to output a fused data group as a reliable fusion result of multi-source weather parameters.

[0009] Further, step S300 includes: S310, according to the characteristics of the fusion data set, the packet processing is carried out through the preset encoding mechanism in the perception layer, and the packet unit is generated; S320, according to the packet unit, the queue sorting is carried out in the transmission layer by using the preset transmission protocol, and the data priority is marked; S330, according to the sorted data queue, the dynamic routing selection mechanism is applied in the transmission layer, and the transmission path is planned; S340, according to the transmission path, the sorted data queue is distributed to the application layer and recombined to generate a complete data stream.

[0010] Further, step S400 includes: S410, the meteorological change key feature data is extracted through the transmission complete data stream, the meteorological change key feature data is analyzed in multiple dimensions by using a preset meteorological analysis model, a meteorological change dynamic trend graph is generated, and a meteorological change quantitative result set is obtained; S420, according to the meteorological change quantitative result set and the historical operation data of the photovoltaic power station, a preset power prediction model is used for multi-parameter fitting analysis to determine the potential output power prediction value interval of the photovoltaic power station; S430, the real-time environmental variable data of the photovoltaic power station is obtained for the prediction value interval, and the prediction value interval is dynamically corrected through a multi-source data fusion mechanism to obtain corrected power prediction data; S440, if the corrected power prediction data is lower than the preset threshold range, an abnormal state identifier is automatically triggered through a pre-warning signal generation module, a pre-warning signal data set is generated, and a pre-warning signal priority is determined; S450, according to the pre-warning signal priority, a hierarchical response mechanism is used to classify and process the pre-warning signal data set, a hierarchical response instruction set is generated, and a prediction result set is determined.

[0011] Further, step S500 includes: S510, according to the prediction result set, reliability verification key data is extracted, multi-level data comparison analysis is carried out on the prediction result set by using a preset consistency checking rule, and a consistency deviation quantitative result set is obtained; S520, the historical meteorological monitoring records of the photovoltaic power station are obtained for the consistency deviation quantitative result set, the deviation data is traced and located through a time sequence correlation analysis mechanism, and a deviation trace result set is obtained; S530, according to the deviation trace result set and the current meteorological real-time monitoring data, a preset calibration compensation model is used to automatically correct the deviation data, and a corrected reliability improvement data set is obtained; S540, if the corrected reliability improvement data set still has deviation beyond the preset threshold, the abnormal deviation part is recalculated through a secondary verification module, and a final unbiased reliability verification result set is obtained. S550, generating a structured weather monitoring report through the final unbiased reliability verification result set, and outputting an operation and maintenance decision support report containing complete reliability confirmation information.

[0012] Another aspect of the application relates to a photovoltaic weather monitoring system based on multi-sensor cooperation for executing the above-mentioned photovoltaic weather monitoring method based on multi-sensor cooperation, comprising: The calibrated parameter set acquisition module is configured to acquire initial acquisition signals by cooperatively collecting weather data through the data acquisition device, and correct the bias by processing the initial acquisition signals using the calibration model to obtain the calibrated parameter set, wherein the weather data includes irradiation parameters, temperature monitoring values, and wind speed parameters. The fusion data group determination module is configured to perform data fusion operation according to the calibrated parameter set, fuse the irradiation parameters, temperature monitoring values, and wind speed parameters, and if the bias exceeds the preset threshold during the fusion process, the corresponding abnormal data points are removed to determine the fusion data group. The transmission complete data stream acquisition module is configured to execute a transmission protocol in a network architecture using the fusion data group, send the fusion data group from the perception layer to the application layer through the transmission layer, and obtain the transmission complete data stream. The prediction result set acquisition module is configured to analyze the weather change mode through the transmission complete data stream, calculate the potential output power of the photovoltaic power station using the prediction model, determine whether the potential output power of the photovoltaic power station is lower than the preset threshold, and if so, trigger an early warning signal to obtain the prediction result set. The weather monitoring report output module is configured to integrate a reliability improvement mechanism according to the prediction result set, perform secondary verification on the prediction result set to confirm data consistency, and output a final weather monitoring report to support photovoltaic power station operation and maintenance decision.

[0013] The application has the following beneficial effects: This invention provides a photovoltaic meteorological monitoring method and system based on multi-sensor collaboration. It collaboratively collects initial meteorological signals such as irradiance, temperature, and wind speed using data acquisition equipment. A calibration model is then used to correct deviations in the initial signals, resulting in a reliable parameter set. Multi-source data is then fused based on this parameter set. If the deviation exceeds a threshold during fusion, outliers are automatically removed, forming a high-quality fused data set. Subsequently, a transmission protocol efficiently transmits the fused data set from the sensing layer to the application layer, forming a complete data stream. At the application layer, a prediction model analyzes meteorological change patterns to calculate the potential output power of the photovoltaic power station. When the predicted power is lower than a preset threshold, an early warning signal is immediately triggered. Finally, a reliability verification mechanism is integrated to reconfirm the prediction results, ensuring data consistency. Ultimately, an accurate monitoring report is output, providing scientific decision support for power station operation and maintenance. This invention effectively solves the problems of inaccurate power prediction and delayed early warning caused by meteorological data deviations, abnormal interference, and unstable transmission in photovoltaic power stations. It achieves precise end-to-end control of meteorological data from acquisition, calibration, fusion, transmission to prediction, significantly improving the power generation efficiency and operational reliability of photovoltaic power stations. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an embodiment of the photovoltaic meteorological monitoring method based on multi-sensor collaboration of the present invention. Figure 2 This is a functional block diagram of an embodiment of the photovoltaic meteorological monitoring system based on multi-sensor collaboration of the present invention.

[0015] Explanation of icon numbers: 10. Module for acquiring the set of parameters after calibration; 20. Module for determining the fused data set; 30. Module for acquiring the data stream after transmission completion; 40. Module for acquiring the prediction result set; 50. Module for outputting the meteorological monitoring report. Detailed Implementation

[0016] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0017] like Figure 1 As shown, the first embodiment of the present invention proposes a photovoltaic meteorological monitoring method and system based on multi-sensor collaboration, aiming to solve at least one defect existing in the above-mentioned prior art.

[0018] One aspect of the present invention relates to a photovoltaic meteorological monitoring method based on multi-sensor collaboration, comprising the following steps: Step S100: Collect meteorological data through data acquisition equipment to obtain initial acquisition signals, process the initial acquisition signals using a calibration model to correct deviations, and obtain a set of calibrated parameters, in which meteorological data includes irradiance parameters, temperature monitoring values, and wind speed parameters.

[0019] By using multiple types of data acquisition equipment to collaboratively collect meteorological data at photovoltaic power stations, initial acquisition signals including irradiance parameters, temperature monitoring values, and wind speed parameters are obtained. A calibration model is introduced to correct the deviation of the initial acquisition signals, eliminating equipment errors and environmental interference errors in the acquisition process, and obtaining a calibrated parameter set with data consistency and accuracy.

[0020] Specifically, the multi-type data acquisition equipment integrates irradiance sensors, temperature sensors, and wind speed sensors, with each sensor collecting data synchronously to ensure data consistency over time. Irradiance parameters include real-time irradiance and cumulative irradiance duration, directly related to the photovoltaic module's photoelectric conversion. Temperature monitoring values ​​include ambient temperature and photovoltaic module surface temperature, affecting module heat dissipation and conversion efficiency. Wind speed parameters include instantaneous wind speed and time-averaged wind speed, impacting module heat dissipation and power plant structural safety. The calibration model can employ a linear calibration model based on the least squares method or an adaptive calibration model based on machine learning. This effectively eliminates inherent sensor errors, irradiance value deviations caused by dust obstruction, temperature jumps due to electromagnetic interference, and wind speed measurement errors caused by installation angle misalignment, ensuring the accuracy of the calibrated parameter set. Simultaneously, timestamp alignment is performed on all calibrated parameters to ensure data consistency.

[0021] Step S200: Perform data fusion operation based on the calibrated parameter set, fusing irradiation parameters, temperature monitoring values, and wind speed parameters. If a deviation exceeding a preset threshold is detected during the fusion process, the corresponding abnormal data points are removed, and the fused data group is determined.

[0022] Multi-dimensional data fusion operations are carried out based on the calibrated parameter set. Data fusion processing is performed on irradiation parameters, temperature monitoring values ​​and wind speed parameters. During the fusion process, the data deviation is detected in real time to see if it exceeds the preset threshold. If the deviation exceeds the threshold, the corresponding abnormal data points are removed, and finally, the effective and abnormal fused data group is determined.

[0023] Specifically, the data fusion operation employs a weighted fusion algorithm. Based on the influence weights of various meteorological parameters on photovoltaic output power (e.g., irradiance parameters have the highest weight, followed by temperature parameters, and then wind speed parameters), a correlation mapping is established between various parameters. This integrates heterogeneous, calibrated parameters into a preliminary fusion data set with inherent logical connections. The preset threshold is calibrated based on the fluctuation range of historical normal meteorological data from photovoltaic power plants (e.g., using three times the standard deviation of historical data as the threshold). During the fusion process, the deviation of each data point from the preliminary fusion data set is compared in real time. If the deviation of a data point exceeds the preset threshold, it is determined to be an abnormal data point (e.g., jump values ​​caused by sensor failure or invalid collection values ​​under extreme weather conditions) and is directly removed. The dataset after removing abnormal data is supplemented with missing data using linear interpolation, ultimately forming a valid, abnormal-free, and highly correlated fusion data set.

[0024] Step S300: The fused data group is used to execute the transmission protocol in the network architecture, and the fused data group is sent from the perception layer through the transmission layer to the application layer to obtain the transmission completed data stream.

[0025] Based on the fused data group, the corresponding transmission protocol is executed in the preset network architecture. Relying on the transmission protocol, the fused data group is uploaded from the perception layer of the network architecture. After the data is relayed and transmitted through the transmission layer, it is pushed to the application layer of the network architecture to complete the end-to-end data transmission and form a transmission completion data stream.

[0026] Specifically, the pre-designed network architecture is a three-layer structure: "perception layer - transmission layer - application layer." The perception layer consists of various types of on-site data acquisition devices responsible for generating and outputting fused data sets. The transmission layer deploys wireless communication modules (such as LoRa (Long Range Radio) or 5G (5th Generation Mobile Communication Technology) modules) or wired communication modules, responsible for data relay and transmission. The application layer is the photovoltaic power station's backend monitoring platform, responsible for receiving and processing data. The transmission protocol adopts a lightweight transmission protocol (such as MQTT (Message Queuing Telemetry Transport) protocol) adapted to the real-time and integrity requirements of meteorological data. During transmission, mechanisms such as data compression, CRC (Cyclic Redundancy Check) verification, and breakpoint resumption are used to reduce bandwidth consumption, avoid data loss and distortion, and ensure the complete and efficient transmission of the fused data sets to the application layer. After the data completes the entire transmission chain, a transmission completion data stream is formed, containing fused data, timestamps, and sensor location information.

[0027] Step S400: Analyze the weather change pattern by transmitting the data stream, calculate the potential output power of the photovoltaic power station using the prediction model, determine whether the potential output power of the photovoltaic power station is lower than the preset threshold, and if it is lower, trigger an early warning signal to obtain the prediction result set.

[0028] Based on the completed data stream, the analysis of the on-site meteorological change patterns of the photovoltaic power station is carried out. The potential output power of the photovoltaic power station is calculated by substituting the meteorological data in the data stream into the preset prediction model. Then, it is determined whether the calculated potential output power is lower than the preset threshold. If it is determined to be lower than the preset threshold, an early warning signal is triggered immediately. Finally, the prediction results set including meteorological change patterns, potential output power values, and early warning status are integrated to form a prediction result set.

[0029] Specifically, meteorological change pattern analysis is conducted using time-series analysis algorithms (such as trend fitting and period decomposition) to uncover the temporal variation patterns of irradiance parameters, temperature monitoring values, and wind speed parameters (such as the intraday variation trend of irradiance intensity and the seasonal fluctuation characteristics of temperature) and the coupling relationships between parameters (such as the positive correlation between irradiance intensity and ambient temperature, and the negative correlation between wind speed and module surface temperature), thus forming a meteorological change pattern at the photovoltaic power station site; the preset prediction model is a time-series prediction model (such as LSTM (Long Short-Term Memory)). The model employs a Long Short-Term Memory (LSTM) neural network model. It uses historical meteorological data and corresponding photovoltaic (PV) output power data as training samples. After training, it inputs real-time meteorological data from the transmitted data stream to accurately calculate the potential output power of the PV power plant. A preset threshold is established based on the PV power plant's design output power or operational requirements (e.g., 80% of the design output power). If the calculated potential output power is lower than this preset threshold, the power plant is deemed to have insufficient power generation efficiency, triggering an immediate warning signal (including one or more of the following: platform pop-up warning, SMS warning for maintenance personnel, and on-site audible and visual warning). Finally, the model integrates meteorological change patterns, potential output power values, warning status, and warning trigger time to form a prediction result set.

[0030] Step S500: Based on the reliability enhancement mechanism of the prediction result set, perform secondary verification on the prediction result set to confirm data consistency, and output the final meteorological monitoring report to support the operation and maintenance decision of the photovoltaic power station.

[0031] Based on the pre-set reliability enhancement mechanism integrated with the prediction result set, a comprehensive secondary verification operation is performed on the prediction result set to verify and confirm the consistency and validity of all data in the prediction result set, and finally outputs a final meteorological monitoring report that can directly support the operation and maintenance decision of photovoltaic power plants.

[0032] Specifically, the pre-defined reliability enhancement mechanism includes multi-model cross-validation and historical data backtesting. Multi-model cross-validation involves introducing different types of prediction models (such as the BP (Backpropagation) neural network model) to recalculate the potential output power, comparing the calculation results of different models with the potential output power values ​​in the prediction result set to verify data consistency. Historical data backtesting compares the current meteorological data, potential output power values, and historical meteorological-power data for the same period to verify the rationality of the prediction results. Through the above two verification operations, the consistency and effectiveness of meteorological change patterns, potential output power values, and warning status within the prediction result set are verified and confirmed. Based on the verified prediction result set, a final meteorological monitoring report is generated. The report includes real-time meteorological parameters, meteorological change trend predictions (such as meteorological changes in the next 24 hours), potential output power values, warning information (including warning level and warning reason), and targeted operation and maintenance suggestions (such as "insufficient irradiance, it is recommended to clean the photovoltaic modules" and "excessive temperature, it is recommended to turn on the module heat dissipation system"). This report can directly support the operation and maintenance decisions of photovoltaic power plants.

[0033] Furthermore, in the photovoltaic meteorological monitoring method and system based on multi-sensor collaboration provided in this embodiment, step S100 includes: Step S110: Collect meteorological data from multiple points using data acquisition equipment. Set different sampling frequencies for irradiance parameters, temperature monitoring values, and wind speed parameters. Obtain initial acquisition signals from multiple sensor nodes and perform digital processing to generate the original data set.

[0034] The following formulas are used to set different sampling frequencies for irradiance parameters, temperature monitoring values, and wind speed parameters:

[0035] In formula (1), Indicates the first Sampling frequency of meteorological parameters Indicates the first The sampling period of the class parameter, Represents pi (π). Indicates the basic time interval. Indicates the first Frequency adjustment coefficient of class parameters. The control logic of formula (1) is based on the basic time interval. Based on the frequency adjustment coefficient To scale the sampling frequency of different meteorological parameters—assign corresponding parameters to different types of meteorological parameters (irradiance, temperature, wind speed). This allows us to calculate the sampling frequency corresponding to that parameter. (or sampling period) Its core function is to customize differentiated sampling frequencies for different types of meteorological parameters (irradiance, temperature, wind speed)—because different meteorological parameters have different rates of change and monitoring needs (e.g., rapid changes in irradiance may require higher sampling frequencies, while slow changes in temperature can require lower frequencies). Adjustment can balance the efficiency and cost of data acquisition while ensuring monitoring accuracy.

[0036] The initial acquired signal is digitized using the following formula to achieve the conversion from analog to digital signal: (2) In formula (2), Indicates the first The digital output value of each sampling point Indicates the first The analog voltage value at each sampling point This represents the minimum voltage value of the analog signal. This represents the maximum voltage value of the analog signal. This represents the number of bits used in the digitization process. The control logic of formula (2) is to first process the first... Analog voltage at each sampling point ,pass Normalize to the range of 0 to 1, then multiply by The maximum quantization value of a bit digital signal Finally passed (Rounding) yields the corresponding numerical output value. Its core function is to achieve the quantization conversion from analog to digital signals: mapping continuous analog voltage signals (such as the electrical signals corresponding to meteorological parameters) to digital signals. The discrete digital values ​​corresponding to the binary bits complete the core quantization process of an analog-to-digital converter (ADC), enabling analog signals to be processed and stored by digital systems.

[0037] In the meteorological monitoring system of a solar power plant, multi-point collaborative data acquisition is achieved through data acquisition devices deployed in different locations. These devices include multiple sensor nodes distributed around the photovoltaic panel array, capable of capturing environmental changes in real time. Specifically, these data acquisition devices employ wireless sensor network technology to ensure that data is synchronously transmitted from each node to the central processing unit, avoiding data loss due to the failure of a single node. In one embodiment, a high sampling frequency, such as once per minute, is set for the acquisition of irradiance parameters because irradiance intensity changes rapidly due to cloud cover. Temperature monitoring values ​​are sampled every 5 minutes to capture gradual trends, while wind speed parameters are sampled every 2 minutes based on wind fluctuation characteristics. This different frequency setting helps optimize data volume and reduce power consumption. In this way, the initial acquisition signals obtained from the sensor nodes are first digitized by an analog-to-digital converter, for example, converting analog voltage signals into digital values, forming a raw dataset containing timestamps and parameter values. This raw dataset includes hundreds of data points, covering monitoring records for one hour.

[0038] Step S120: Use the calibration model to correct the deviation of the original data set and generate a calibrated parameter set.

[0039] The basic transformation process for correcting bias in the original data using a calibration model is described by the following formula: (3) In formula (3), This represents the set of parameters after calibration. Represents the original data set. Represents the calibration model function. This represents the calibration coefficient vector. The control logic of formula (3) is based on the original data set. Based on this, by calibrating the model function Combined with calibration coefficient vector Calculate the deviation corresponding to the original data (i.e. Then subtract this deviation from the original data to obtain the calibrated data. Its core function is to correct systematic biases in the original data: the original collected data may have biases due to sensor errors, environmental interference, etc. By using a pre-determined calibration model and coefficients, the original data is adjusted compensatorily to improve the accuracy and reliability of the data, making the data closer to the real physical quantities.

[0040] For the generated raw data set, a calibration model is used to correct for biases, eliminating the influence of inherent sensor errors or environmental interference. Specifically, the calibration model is built based on the least squares method. First, standard reference data is collected, such as benchmark irradiance values ​​measured using precision instruments. Then, the difference between the raw data and the benchmark value is calculated. Parameters are adjusted through a fitting function, for example, applying a linear correction formula to temperature data, where the correction coefficients are derived from historical data training. In one embodiment, assuming a sensor node reports an irradiance value of 800 W / m², but the benchmark value is 820 W / m², the calibration model calculates the bias and applies a correction factor such as 0.975, making the calibrated value approximately 780 W / m², ensuring accuracy. This correction process not only improves data reliability but also provides a more accurate basis for subsequent power generation prediction. For example, in the correction of wind speed parameters, if the original wind speed reading is too low due to the installation height, the model introduces a height compensation factor and generates a calibrated parameter set through iterative calculation. This set can be directly input into the energy management system to achieve optimized prediction of photovoltaic output power.

[0041] Furthermore, in the photovoltaic meteorological monitoring method and system based on multi-sensor collaboration provided in this embodiment, step S200 includes: Step S210: Based on the irradiance parameters, temperature monitoring values, and wind speed parameters in the calibrated parameter set, perform a preliminary linear combination using a weighted fusion algorithm to generate an initial fused numerical sequence.

[0042] The initial fusion numerical sequence is obtained using the following formula: (4) In formula (4), Represents the first fused numerical sequence. Item value, The weighting coefficients representing the irradiation parameters. This indicates the irradiation parameters after calibration. This represents the weighting coefficient of the temperature monitoring values. This indicates the temperature monitoring value. This represents the weighting coefficient for the wind speed parameter. This represents the wind speed parameter. The control logic of formula (4) is based on the calibrated irradiance parameter. Temperature monitoring value Wind speed parameters These three types of meteorological parameters are each multiplied by their respective weighting coefficients. , , Then, the weighted results of these three parts are directly added together to obtain the first fused numerical sequence. Item value The core function of formula (4) is the weighted fusion of multiple meteorological parameters: the three independent meteorological parameters of irradiance, temperature and wind speed are linearly combined according to the preset weights (reflecting the importance of different parameters), and the scattered single-dimensional parameters are integrated into a comprehensive initial fusion sequence, so as to realize the initial integration of multi-source meteorological data and provide unified basic data for subsequent data analysis or model input.

[0043] For the irradiance parameters, temperature monitoring values, and wind speed parameters in the calibrated parameter set, the process of preliminary linear combination using a weighted fusion algorithm is a data integration method. The weighted fusion algorithm is essentially a calculation method based on weight allocation, which superimposes different parameters according to a preset importance ratio. Specifically, in the monitoring system of a solar power plant, these parameters come from multiple sensor nodes. For example, the irradiance parameter represents solar radiation intensity, the temperature monitoring value reflects changes in ambient heat, and the wind speed parameter captures airflow speed. The weighted fusion algorithm assigns a weight to each parameter, such as 0.5 for the irradiance parameter, 0.3 for the temperature monitoring value, and 0.2 for the wind speed parameter. Then, they are added together using a linear combination formula to generate an initial fused numerical sequence. This sequence is a continuous list of values, representing the preliminary quantitative result of the comprehensive meteorological impact. For example, in actual operation, if the irradiance parameter at a certain moment is 900 W / m², the temperature monitoring value is 25℃, and the wind speed parameter is 5 m / s, then the linear combination yields a fused value such as 620, which is used for subsequent processing. This method ensures the initial unification of multi-source data and avoids bias caused by a single parameter.

[0044] Step S220: Use a sliding window mechanism to perform time-series smoothing on the initial fused numerical sequence to generate a smoothed fused data stream.

[0045] The smoothed and fused data stream value is obtained using the following formula:

[0046] In formula (5), Indicates time After smoothing, the data stream values ​​are merged. This indicates the size of the sliding window. The control logic of formula (5) is based on the current time t as the endpoint, selecting a length of... Sliding window (i.e. containing arrive this Fusion data at each moment Calculate this The arithmetic mean of the data points is used as the smoothed fused data stream value at time t. Its core function is to smooth and reduce noise in the fused data stream: by calculating the mean of a sliding window, it weakens the impact of random fluctuations or sudden noise in the fused data, making the data stream more stable and continuous, and improving the stability of the data and the reliability of subsequent analysis.

[0047] In wind power applications, when using a sliding window mechanism to smooth the initial fused numerical data sequence, this technique reduces noise by defining a fixed-size window that moves across the sequence and calculating the average or median of the data within the window. Specifically, this sliding window mechanism sets the window size to 10 data points and slides across the sequence point by point. For example, if the initial fused numerical sequence contains a series of values ​​such as 620, 615, and 630, the window will take the average of the first 10 points as the first smoothing point, then move one step to take the average of the next 10 points, generating a smoothed fused data stream. This data stream more smoothly reflects meteorological trends and avoids interference from sudden fluctuations.

[0048] Step S230: Perform deviation detection and remove outliers on the smoothed and fused data stream according to preset rules to generate a cleaned and fused dataset.

[0049] The process of generating a cleaned fused dataset involves detecting deviations and removing outliers from a smoothed and fused data stream using pre-defined rules. These rules are based on thresholds or statistical methods, such as setting a deviation threshold as the average value of the data stream plus or minus two standard deviations. If a point exceeds this range, it is considered an anomaly and removed. Specifically, in an urban environmental monitoring project, the smoothed and fused data stream displayed a series of fused values ​​such as 618, 620, 850, and 622. Among these, 850 was significantly higher. After rule detection, it was removed, and the remaining points formed the cleaned fused dataset. This cleansing ensures the reliability of the data and provides a clean foundation for subsequent analysis.

[0050] Step S240: Perform state estimation correction on the purified fused dataset and output the fused data set as a reliable fusion result of multi-source meteorological parameters.

[0051] The following formula represents the optimal information fusion based on inverse covariance weighting, which corrects the state estimate of the purified fused dataset:

[0052] In formula (6), This represents the reliable fusion result after state estimation correction. Indicates the first State estimation of a purified data source Indicates the first The covariance matrix of each data source. Represents the fused covariance matrix. This indicates the number of data sources. The control logic of formula (6) is to first calculate the covariance matrix of all purified data sources. The sum of the inverse matrices, and then the inverse of those matrices, yields the fused covariance matrix. Then calculate the inverse covariance matrix for each data source. With its own state estimation The sum of the products, then multiplied by Multiplying these results yields a reliable fusion result after state estimation correction. Its core function is to perform high-precision weighted fusion of multi-source data: using the inverse of the covariance matrix (reflecting the uncertainty of the data) as the weight (the less uncertain the data source, the greater the weight), the state estimates of multiple purified data sources are weighted and fused, which not only integrates multi-source information, but also reduces the impact of data uncertainty, and improves the accuracy and reliability of state estimation.

[0053] When performing state estimation correction on the purified fused dataset to output the fused data set as a reliable fusion result of multi-source meteorological parameters, state estimation correction is an optimization technique that often uses principles such as Kalman filtering to estimate the true state. It corrects potential errors by iteratively adjusting the values ​​in the dataset. Specifically, in photovoltaic power generation optimization, correction is applied to the purified fused dataset. For example, if the initial dataset has values ​​such as 620, 618, and 622, the state estimation will consider historical states and measurement noise, gradually correcting it to a more accurate set such as 619, 617, and 621, which is then output as the fused data set. This correction improves the overall accuracy, resulting in more precise power generation forecasts and supporting energy management decisions.

[0054] Furthermore, in the photovoltaic meteorological monitoring method and system based on multi-sensor collaboration provided in this embodiment, step S300 includes: Step S310: Based on the characteristics of the fused data group, the perception layer performs packet processing through a preset encoding mechanism to generate packet units.

[0055] The following formula is used by the perception layer to perform packet processing based on the characteristics of the fused data set: (7) In formula (7), Indicates the generated first Each subcontracting unit This indicates a preset encoding mechanism. Indicates a merged data group. This indicates the number of sub-packages. The control logic of formula (7) is based on a preset coding mechanism. , will merge data groups According to the number of subcontracts The requirements are broken down into corresponding sub-package units. (That is, to encode a complete fused data set and then divide it into...) (Individual parts). Its core function is to adapt to the transmission / storage requirements of the perception layer: based on the characteristics of the fused data group, through encoding and packet processing, the data is divided into appropriately sized units, which facilitates efficient transmission, storage or subsequent processing by the perception layer, avoiding the transmission pressure or storage inconvenience caused by large-volume data.

[0056] Regarding the characteristics of fused data sets, when the perception layer performs packet processing through a preset encoding mechanism, this preset encoding mechanism is an encoding method for structured data packets. It divides the fused data set into smaller units according to size and type to adapt to network transmission requirements. Specifically, in the monitoring system of a solar power plant, the fused data set contains a comprehensive value sequence of irradiance, temperature, and wind speed. If a data set has hundreds of data points, the packet header is first defined through the encoding mechanism, including data type identifier, length field, and checksum. Then, the data set is evenly divided; for example, a 1000-byte data set is divided into 10 100-byte packet units, each carrying an independent sequence number to ensure reconstructability during transmission. This packet processing reduces the risk of packet loss caused by excessively large single packets and supports efficient data acquisition at the perception layer.

[0057] Step S320: Based on the packet unit, perform queue sorting at the transport layer using a preset transport protocol and mark data priority.

[0058] The data packets are sorted in the queue using the following formula: (8) In formula (8), This indicates the sorting result of the transport layer queue. Represents a set of data packets. Indicates the default transmission protocol. This indicates that data packets are queued and sorted according to the protocol rules of the preset transmission protocol. The control logic of formula (8) is based on the preset transmission protocol. The defined rules (such as priority, data type, timing rules, etc.) apply to the data packet set. The sorting operation is performed, and the final sorting result of the transport layer queue is obtained. Its core function is to regulate the transmission order of data packets: to sort data packets according to the rules of the transmission protocol (such as transmitting high-priority data first), so that the data packet queue of the transport layer conforms to the communication specifications and ensures the orderliness, efficiency or priority requirements of data transmission.

[0059] In wind power plant applications, when data is queued at the transport layer based on packet units using a preset transmission protocol, the preset transmission protocol is a custom protocol based on TCP (Transmission Control Protocol) or UDP (User Datagram Protocol). It achieves ordered data arrangement through queue management. Specifically, this process first places packet units into a buffer queue, then sorts them according to protocol rules such as FIFO (First In First Out) or a priority queue, while simultaneously marking data priorities. For example, real-time wind speed-related units are marked as high priority, and historical temperature units as medium priority. Priority fields in the protocol, such as values ​​from 0 to 3, ensure that high-priority data is processed first. This sorting mechanism prioritizes the transmission of critical meteorological data in operations, avoiding delays that could affect power generation control.

[0060] Step S330: Based on the sorted data queue, apply a dynamic routing mechanism at the transport layer to plan the transmission path.

[0061] The following formula is used to dynamically select a transmission path for each data item by minimizing the sum of the costs of all links along the path:

[0062] In formula (9), This indicates that the sorted data in the queue is... The optimal transmission path selected for each data item Represents the set of all paths. Indicates link The current dynamic cost, This means that the rule should cover the data queue. Each data item The control logic of formula (9) applies to each data item in the data queue. Iterate through the set of all possible transmission paths. Calculate the dynamic cost of all links on each path. The sum of all the links is used to select the path with the minimum total link cost as the optimal transmission path for that data item. Its core function is to optimize the cost of data transmission: dynamically matching the transmission path with the lowest link cost for each data item, minimizing resource consumption (such as bandwidth, latency, etc.) during transmission while ensuring data transmission, and improving the resource utilization efficiency of the transmission system.

[0063] Based on the sorted data queue, when planning transmission paths using a dynamic routing mechanism at the transport layer, the dynamic routing mechanism is a strategy that adjusts paths in real time according to network conditions. It selects routes by evaluating link load, latency, and bandwidth. Specifically, in a photovoltaic power generation optimization system, this mechanism is applied to the sorted queue. First, the network node status is monitored. If congestion is detected on the main path, it switches to an alternative path. For example, if there are high-priority packet units in the queue, the mechanism calculates the path cost, using a simple weighted formula considering a latency weight of 0.6 and a bandwidth weight of 0.4, and selects the path with the lowest cost. This planning ensures efficient data flow in complex networks and supports multi-site data synchronization in terms of business operations.

[0064] Step S340: Distribute the sorted data queue to the application layer and reassemble it according to the transmission path to generate a complete data stream for transmission.

[0065] The complete data stream is derived using the following formula: (10) In formula (10), This indicates the complete data stream generated by the reassembly. This represents the data queue distributed to the application layer. This indicates that the distributed data packets are reassembled at the application layer to restore the complete original data stream. The control logic of formula (10) is for the data queue distributed to the application layer. To carry out reorganization operations The previously fragmented and transmitted data packets are reassembled and integrated to obtain a complete data stream. Its core function is to restore the original complete data stream: to offset the data fragmentation caused by packet splitting at the perception layer and distribution at the transport layer, to restore the scattered data packets to the original complete data stream, and to ensure that the application layer can obtain continuous, complete and valid data.

[0066] When the sorted data queues are distributed to the application layer and reassembled to generate a complete data stream according to the transmission path, this process involves path-guided data forwarding and application layer reconstruction. Specifically, in urban energy monitoring projects, queues are distributed along planned paths, such as being forwarded to the application server via routers. Then, at the application layer, they are reassembled according to the packet sequence number and checksum. For example, after the original packet units 1-10 are received, they are sequentially spliced ​​back into the fused data group to form a complete data stream, such as a continuous irradiance temperature sequence. This reassembly ensures data integrity and provides reliable input for predictive models in business operations, supporting energy allocation decisions.

[0067] Furthermore, in the photovoltaic meteorological monitoring method and system based on multi-sensor collaboration provided in this embodiment, step S400 includes: Step S410: Extract key feature data of meteorological changes by transmitting the data stream, use a preset meteorological analysis model to perform multi-dimensional analysis of the key feature data of meteorological changes, generate a dynamic trend map of meteorological changes, and obtain a set of quantitative results of meteorological changes.

[0068] The quantitative results set of meteorological changes are obtained using the following formula:

[0069] In formula (11), Indicates the time span, This represents the time series of the analysis results. This represents the set of quantitative results of meteorological changes. Indicates cumulative trend, Indicates extreme value characteristics, This indicates volatility. The control logic of formula (11) is based on the time span. Within, the time series of meteorological analysis results Simultaneously, three key indicators—cumulative trend, extreme value characteristics, and volatility—are calculated, and these three indicators are ultimately integrated into a set of quantitative results for meteorological changes. Its core function is to quantify meteorological change characteristics in multiple dimensions: from the perspectives of cumulative trend, extreme values, and fluctuations, it comprehensively extracts key information from meteorological time series, transforming continuous meteorological changes into a quantifiable set of features, which facilitates subsequent analysis of the patterns or impacts of meteorological changes.

[0070] When extracting key meteorological change features from a transmitted data stream, the process first involves filtering out core meteorological indicators, such as irradiance fluctuation rate, temperature gradient, and wind speed variation coefficient. These features are obtained by parsing the time series data in the data stream. Specifically, the pre-defined meteorological analysis model is a hybrid model based on statistics and machine learning, including principal component analysis and time series decomposition components. The key feature data is first input into the model, and then subjected to multi-dimensional analysis, such as decomposition of time, space, and correlation dimensions. For example, when analyzing irradiance, the meteorological analysis model separates seasonal trends from random noise, and then generates a dynamic trend map of meteorological changes. This is a visual representation, such as a line chart combined with a heat map, ultimately yielding a quantitative set of meteorological change results, including numerical trend indicators such as average rate of change and peak deviation.

[0071] Step S420: Based on the quantitative results set of meteorological changes and the historical operation data of the photovoltaic power station, a multi-parameter fitting analysis is performed using a preset power prediction model to determine the range of potential output power prediction values ​​for the photovoltaic power station.

[0072] The following formula is used to determine the range of predicted potential output power of a photovoltaic power plant:

[0073] In formula (12), and This represents the lower and upper limits of the potential output power prediction range. This represents the predicted average power. The statistic represents the confidence level. This represents the standard deviation of the forecast based on historical data and meteorological quantification. The control logic of formula (12) is based on the mean of the predicted power. Centered on the confidence level, combined with the statistics corresponding to the confidence level With the predicted standard deviation Calculate the deviation range Subtracting this deviation from the mean yields the lower limit of the interval. The upper limit of the interval is obtained by adding the mean to the deviation. This allows for the determination of the potential output power prediction range. Its core function is to quantify the uncertainty range of power prediction: instead of providing a single predicted power value, it reflects the possible fluctuation range of photovoltaic power plant output power through confidence intervals, making the prediction results more realistic (considering the uncertainty of factors such as weather), and assisting power plants in making more reasonable power scheduling or risk assessments.

[0074] When combining the quantitative results set of meteorological changes with historical operating data of photovoltaic power plants, the preset power prediction model is a neural network-based model, such as a Long Short-Term Memory (LSTM) network. It integrates quantitative indicators from the results set with historical data, such as power generation and environmental records from the past year, through multi-parameter fitting analysis. Specifically, this fitting process first normalizes the data, then uses the input layer of the preset power prediction model to receive quantified values ​​of multiple parameters, such as irradiance and temperature. The hidden layer adjusts the weights to fit a nonlinear relationship, ultimately outputting a predicted range of potential output power for the photovoltaic power plant, such as 800-1200 kWh. This supports advance planning of energy output in business operations.

[0075] Step S430: Obtain real-time environmental variable data of photovoltaic power station for the predicted value range, and dynamically correct the predicted value range through multi-source data fusion mechanism to obtain corrected power prediction data.

[0076] The corrected power prediction data is obtained using the following formula: (13) In formula (13), This represents the corrected power prediction data. This represents the power prediction value corresponding to the original prediction range. This represents the dynamic correction amount calculated based on real-time environmental variables. The control logic of formula (13) uses the power prediction value corresponding to the original prediction value range. Based on this, plus dynamic corrections calculated based on real-time environmental variables. The corrected power prediction data were obtained. Its core function is to optimize power prediction accuracy by combining real-time environmental data: the original power prediction may be based on historical / predictive quantitative data, and by introducing dynamic correction values ​​corresponding to real-time environmental variables, the original prediction value is adjusted to offset the prediction deviation caused by changes in the real-time environment and improve the real-time accuracy of power prediction.

[0077] When acquiring real-time environmental variable data for photovoltaic power plants within the predicted range, this data includes current humidity, cloud cover, etc. Dynamic correction is performed through a multi-source data fusion mechanism. This mechanism is a strategy that combines weighted averaging and Kalman filtering. It first collects real-time data sources such as satellite imagery and ground sensors, and then applies correction to the predicted range. For example, if the real-time irradiance is lower than expected, the mechanism will adjust the lower limit of the range, ultimately obtaining corrected power prediction data accurate to 950 kWh.

[0078] Step S440: If the corrected power prediction data is lower than the preset threshold range, the abnormal state indicator is automatically triggered by the warning signal generation module to generate a warning signal dataset and determine the priority of the warning signal.

[0079] The following formula is used to define the triggering conditions for an abnormal status indicator: (14) In formula (14), Indicates an abnormal status indicator. This indicates the lower limit of the preset threshold range. As an indicator function, when the corrected power prediction data Below the lower limit of the preset threshold range hour, The value is 1 if it is set to 1, and 0 otherwise. The control logic of formula (14) is achieved through an indicator function. Determine the corrected power prediction data Is it below the lower limit of the preset threshold range? —If it is lower than that, an abnormal status is indicated. The value is 1; otherwise, it is 0. Its core function is to identify abnormally low power predictions: when the corrected power prediction is lower than the preset lower limit (which may correspond to risks such as abnormal power plant output or insufficient power supply), an anomaly indicator is triggered to facilitate timely warning and subsequent handling measures.

[0080] If the corrected power prediction data is lower than the preset threshold range, such as below 700 kWh, the warning signal generation module will automatically trigger the abnormal state indicator. This warning signal generation module is a rule engine that generates a warning signal dataset based on threshold comparison, including signal type such as low power alarm, and determines priority through severity scoring such as 1-5 levels.

[0081] Step S450: Based on the priority of the warning signals, a hierarchical response mechanism is used to classify the warning signal dataset, generate a hierarchical response instruction set, and determine the prediction result set.

[0082] The hierarchical response instruction set is derived using the following formula: (15) In formula (15), This indicates a hierarchical response instruction set. Indicates the priority of the warning signal. This represents the dataset of early warning signals. This represents a mapping function that classifies the dataset according to priority. The control logic of formula (15) is achieved through the mapping function. Combined with the priority of warning signals For the early warning signal dataset The system categorizes and processes warning signals, assigning them to corresponding response commands based on their priority, ultimately generating a tiered response command set. Its core function is to enable tiered response to early warnings: based on the priority differences of early warning signals, the early warning dataset is classified and processed, and corresponding response instructions are matched, so that early warnings of different severity correspond to different levels of response measures, thereby improving the pertinence and efficiency of the response.

[0083] The prediction result set is obtained using the following formula: (16) In formula (16), Represents the set of prediction results. This represents the function that determines the final prediction result based on the hierarchical response instruction set. The control logic of formula (16) is achieved through the function. hierarchical response instruction set As input, based on the hierarchical rules or processing logic in the instruction set, determine and output the corresponding final prediction result set. Its core function is to link response instructions with final predictions: it transforms the processing logic in the hierarchical response instruction set into specific prediction results, so that the decision of early warning response can be implemented into an output set of prediction results, ensuring the consistency between prediction results and response strategies.

[0084] When a graded response mechanism is adopted based on the priority of early warning signals, this graded response mechanism is a hierarchical processing framework. It classifies the dataset, such as high priority for immediate response and medium priority for scheduling and inspection, generates a graded response instruction set, such as notifying the maintenance team or adjusting the load, and finally determines the prediction result set. This ensures timely intervention in business to maintain the stable operation of the power plant.

[0085] Furthermore, in the photovoltaic meteorological monitoring method and system based on multi-sensor collaboration provided in this embodiment, step S500 includes: Step S510: Extract key data for reliability verification based on the prediction result set, and perform multi-level data comparison and analysis on the prediction result set using preset consistency verification rules to obtain a consistency deviation quantification result set.

[0086] The consistency deviation quantification result set is obtained using the following formula: (17) In formula (17), This represents the result set of consistency deviation quantification. Indicates the first The quantified value of the deviation of each prediction result. This indicates the preset consistency check threshold. This represents the total number of prediction result sets, and the consistency deviation quantification result set only includes deviation values ​​exceeding the threshold. The control logic of formula (17) is to traverse all... Quantitative value of the deviation of each prediction result Filter out those that meet the deviation quantification value Greater than the preset consistency check threshold Those values, those deviations exceeding the threshold are grouped into a set. The core function of formula (17) is to focus on significant deviations: from all the deviations of the prediction results, only the part that exceeds the preset threshold is extracted, and the acceptable small deviations are filtered out, so as to identify the consistency deviation items that need to be focused on and dealt with, which is convenient for subsequent targeted analysis or correction.

[0087] No. The quantified value of the deviation of each prediction result is obtained by the following formula: (18) In formula (18), Indicates the total number of comparison levels. Indicates the first Hierarchical weights Indicates the first The prediction result is in the first... Hierarchical predicted values, Indicates the first The prediction result is in the first... The reference value or key data value of the level. The control logic of formula (18) is for the first level. Each prediction result, first at each comparison level Next, calculate the predicted value. Compared with reference value The absolute difference (i.e.) Then, assign corresponding weights to the absolute differences at each level. Then multiply them, and finally sum the weighted results of all levels to obtain the quantified value of the prediction result's bias. The core function of formula (18) is to quantify the bias in a multi-dimensional weighted manner: considering that the importance of the prediction results is different at different levels (dimensions), by allocating weights, the degree of bias of multiple comparison levels is integrated to obtain a quantitative value that can reflect the overall bias, so that the bias assessment is more comprehensive and in line with the actual focus.

[0088] When extracting key data for reliability verification from the prediction result set, this process primarily focuses on filtering data indicators directly related to reliability assessment, such as predicted power values, meteorological condition parameters, and historical comparison data. Specifically, the extraction process uses data cleaning and feature selection mechanisms to remove redundant information from the original prediction result set, retaining core fields such as the daily average irradiance change rate and temperature fluctuation amplitude for subsequent consistency verification. For example, in the business scenario of a photovoltaic power plant, assuming the prediction result set contains power prediction data for the past 24 hours, the extraction will prioritize the matching degree between the predicted values ​​and the actual operating data to ensure that the data reflects the true operating status.

[0089] When using pre-defined consistency verification rules to perform multi-level data comparison and analysis on the prediction result set, it's important to note that these rules are typically a logical framework based on multi-dimensional comparisons, encompassing time, space, and parameter correlation dimensions. For example, in the time dimension, predicted data is compared with historical data from the same period to analyze for significant deviations. In the spatial dimension, predicted data from multiple photovoltaic power plants within the same region is compared to determine if regional anomalies exist. Specifically, in a particular analysis, if a power plant's predicted power output is higher than expected within a specific time period, while historical data and data from neighboring power plants show lower power output, this deviation is recorded and its degree is quantified as part of the consistency deviation quantification result set, such as a percentage or absolute difference.

[0090] Step S520: Obtain historical meteorological monitoring records of photovoltaic power plants for the consistency deviation quantification result set, and trace the source of deviation data through time series correlation analysis mechanism to obtain the deviation source tracing result set.

[0091] The deviation tracing result set is obtained using the following formula: (19) In formula (19), This represents the result set of deviation tracing. Indicates the relationship with the first One deviation Meteorological records with the highest temporal correlation, This represents the time series correlation analysis function. Indicates a point in time The meteorological characteristic vector. The control logic of formula (19) is for the first One deviation, analyzed by the time series correlation function Calculate this deviation at different time points meteorological feature vector The correlation between them is used to find the meteorological records corresponding to the time point with the highest correlation. All of these Composition of deviation source tracing result set The core function of formula (19) is to identify the meteorological factors associated with the positioning deviation: through time-series correlation analysis, the meteorological records most relevant to each deviation are found, thereby tracing the potential meteorological causes of the deviation and providing a basis for subsequent analysis of the source of the deviation and optimization of data quality.

[0092] To quantify consistency deviations, historical meteorological monitoring records of photovoltaic power plants are obtained and their sources are traced and located using a time-series correlation analysis (TSA) mechanism. TSA is an analytical tool based on time series data mining, designed to identify the root causes of deviations. For example, in the operation of a photovoltaic power plant, a large deviation in predicted power was found. By using TSA to trace back historical meteorological records, it was discovered that the deviation stemmed from a sudden change in cloud cover that had not been adequately considered within a certain period. This allowed for pinpointing the specific time point and meteorological factors, forming a deviation source tracing result set. This process helps the operation and maintenance team identify the root cause of the problem and provides a basis for subsequent calibration.

[0093] Step S530: Based on the deviation source tracing result set and the current real-time meteorological monitoring data, the deviation data is automatically corrected using a preset calibration compensation model to obtain the corrected reliability improvement dataset.

[0094] The revised reliability improvement dataset is derived using the following formula: (20) In formula (20), This represents the revised, improved reliability dataset. This represents the preset weights of the deviation tracing result set. This indicates the current real-time meteorological monitoring data. This indicates the preset weights of real-time meteorological monitoring data. Indicates deviation data, The preset weights represent the deviation data. The control logic of formula (20) is based on the deviation source tracing result set. Current real-time meteorological monitoring data Deviation data These three types of data are each multiplied by their respective preset weights. , , Sum the weighted results of the three factors, then divide by the sum of their weights. (By applying a weighted average), the final improved reliability dataset is obtained. The core function of formula (20) is to integrate multi-source data to optimize reliability: combining the correlation information of deviation tracing, real-time meteorological data, and original deviation data, and reflecting the importance of different data through weight allocation, the data is corrected by weighted fusion, which not only retains effective information but also weakens the impact of deviation, thereby improving the reliability of the final dataset.

[0095] When automatically correcting deviation data using a pre-set calibration compensation model based on the deviation source tracing results set and current real-time meteorological monitoring data, the calibration compensation model is typically a dynamic adjustment framework that combines historical and real-time data. Its aim is to reduce prediction deviations through parameter compensation. For example, in actual operations, if the source tracing results show that cloud cover is the main source of deviation, the calibration compensation model will adjust the predicted power value downwards based on the cloud cover ratio in the real-time meteorological monitoring data, generating a corrected reliability-enhanced dataset. This process ensures that the predicted data more closely reflects the actual operating environment.

[0096] Step S540: If the corrected reliability improvement dataset still has deviations exceeding the preset threshold, the abnormal deviation part is recalculated through the secondary verification module to obtain the final unbiased reliability verification result set.

[0097] The following formula is used to define the judgment criteria for abnormal deviations: (twenty one) In formula (21), Indicates the first Does each sample belong to the abnormal deviation portion? Indicates the first A revised reliability value, Indicates a reference reliability value. Indicates the preset threshold. A value of 1 indicates that this part is an abnormal deviation and requires secondary verification. The control logic of formula (21) is for the first... Calculate the corrected reliability value for each sample. Compared with reference reliability value The absolute difference, if the difference is greater than a preset threshold Then mark (Identified as an abnormal deviation); otherwise marked. (Judged as normal). The core function of formula (21) is to identify abnormal deviation samples: by comparing the difference between the corrected reliability value and the reference value, samples with deviations exceeding the threshold are selected and marked as abnormal items that need to be verified a second time, so as to facilitate subsequent targeted verification of data quality.

[0098] The final unbiased reliability verification result set is obtained through the following formula: (twenty two) In formula (22), This represents the final unbiased reliability verification result set. This indicates the result obtained by recalculating the abnormal deviation portion. The mask matrix representing the abnormal deviation portion. This indicates element-wise multiplication, and the formula achieves targeted replacement of abnormal deviations. The control logic of formula (22) utilizes the abnormal deviation mask matrix. (Where the position corresponding to the abnormal deviation is 1, and the normal position is 0), the two parts of data are combined point by point. Its core function is to accurately correct abnormal deviations: only the previously marked abnormal deviation part is replaced with the recalculated reliable result, while the normal part retains the original data. This not only fixes the abnormal problem, but also avoids unnecessary changes to the valid data, and efficiently obtains unbiased and reliable verification results.

[0099] The mask matrix for the abnormal deviation portion is obtained using the following formula: (twenty three) In formula (23), The mask matrix representing the abnormal deviation portion. This represents the bias distribution of the corrected dataset. Indicates the preset threshold. For indicator functions, This represents the absolute value of the deviation. Exceeding the preset threshold The corresponding position is 1 if the condition is met, and 0 otherwise; this formula is used to locate abnormal deviations that require secondary verification. The control logic of formula (23) is through an indicator function. The bias distribution of the corrected dataset The deviation value at each position is used to determine whether its absolute value exceeds a preset threshold. —If the value exceeds the limit, the corresponding mask matrix element at that position is set to 1; otherwise, it is set to 0, thus obtaining the final mask matrix. Its core function is to pinpoint the exact location of abnormal deviations: marking the locations where the deviation exceeds the standard in the corrected data as 1, clarifying which parts belong to abnormal deviations that require secondary verification, and providing accurate location markers for subsequent processing.

[0100] If the corrected reliability enhancement dataset still has deviations exceeding the preset threshold, a secondary verification module will be used for targeted recalculation. This secondary verification module is a specialized tool for handling anomalous data, typically combining finer-grained analysis rules and backup data sources for reassessment. For example, in a business operation, if the corrected power prediction value still exceeds the threshold range, the secondary verification module will call additional meteorological satellite data to recalculate the prediction values ​​for the anomalous time period, ultimately generating an unbiased reliability verification result set to ensure data accuracy.

[0101] Step S550: Generate a structured meteorological monitoring report using the final unbiased reliability verification result set, and output an operation and maintenance decision support report containing complete reliability confirmation information.

[0102] The operation and maintenance decision support report is calculated using the following formula through weighted summation:

[0103] In formula (24), Indicates the level of operational and maintenance decision support. Indicate the candidate decision-making options. This indicates the number of types of reliability confirmation information in the report. Indicate decision In the Compatibility score under class information Indicates the first The weights of class information. The control logic of formula (24) is for each candidate decision scheme. Calculate it Compatibility score under class reliability confirmation information Corresponding information weights The sum of the products (i.e., weighted summation) is then used to select the option with the largest weighted sum from all candidate options. The corresponding level is the optimal operation and maintenance decision support level. The core function of formula (24) is to select the optimal operation and maintenance decision scheme: combining the weights of different types of reliability information (reflecting the importance of the information), calculating the comprehensive compatibility score of each candidate decision, and selecting the operation and maintenance decision that best matches the current reliability information by maximizing the score, thereby improving the rationality and support of the decision.

[0104] When generating a structured meteorological monitoring report from the final unbiased reliability verification result set, it should be noted that this report is typically presented in a standardized format, including reliability confirmation information, deviation correction records, and an overview of meteorological conditions. For example, in the operation and maintenance scenario of a photovoltaic power plant, the generated structured meteorological monitoring report would detail the predicted power values ​​for each time period, the correction process, and the finally confirmed reliability indicators, along with a brief analysis of the meteorological conditions for the next 24 hours. The output is an operation and maintenance decision support report, providing data support for power plant management.

[0105] Please see Figure 2 This embodiment provides a photovoltaic meteorological monitoring system based on multi-sensor collaboration, used to execute the aforementioned photovoltaic meteorological monitoring method based on multi-sensor collaboration. It includes a calibration parameter set acquisition module 10, a fusion data group determination module 20, a transmission completion data stream acquisition module 30, and a prediction result set acquisition module. The system includes a parameter set acquisition module 40 and a meteorological monitoring report output module 50. The parameter set acquisition module 10 acquires initial acquisition signals by collaboratively collecting meteorological data through data acquisition equipment, processes the initial acquisition signals using a calibration model to correct deviations, and obtains a calibrated parameter set, where the meteorological data includes irradiance parameters, temperature monitoring values, and wind speed parameters. The fusion data set determination module 20 performs data fusion operations based on the calibrated parameter set, fusing irradiance parameters, temperature monitoring values, and wind speed parameters. If a deviation exceeding a preset threshold is detected during the fusion process, the corresponding abnormal data points are removed, and the fusion data set is determined. The transmission completion data stream acquisition module 30 executes the transmission protocol in the network architecture using the fusion data set, sending the fusion data set from the perception layer through the transmission layer to the application layer, obtaining the transmission completion data stream. The prediction result set acquisition module 30... Module 40 is used to analyze meteorological change patterns by transmitting data streams, calculate the potential output power of the photovoltaic power station using a prediction model, determine whether the potential output power of the photovoltaic power station is lower than a preset threshold, and trigger an early warning signal if it is lower, thus obtaining a prediction result set. Module 50 is used to integrate a reliability improvement mechanism based on the prediction result set, perform secondary verification on the prediction result set to confirm data consistency, and output a final meteorological monitoring report to support the operation and maintenance decisions of the photovoltaic power station.

[0106] The photovoltaic meteorological monitoring method and system based on multi-sensor collaboration provided in this embodiment, compared with existing technologies, collaboratively collects initial meteorological signals such as irradiance, temperature, and wind speed through data acquisition equipment. A calibration model is used to correct deviations in the initial signals to obtain a reliable parameter set. Multi-source data is then fused based on this parameter set. If the deviation exceeds a threshold during fusion, outliers are automatically removed to form a high-quality fused data set. Subsequently, the fused data set is efficiently transmitted from the sensing layer to the application layer using a transmission protocol to form a complete data stream. At the application layer, a prediction model analyzes meteorological change patterns to calculate the potential output power of the photovoltaic power station. When the predicted power is lower than a preset threshold, an early warning signal is immediately triggered. Finally, a reliability verification mechanism is integrated to reconfirm the prediction results, ensuring data consistency and ultimately outputting an accurate monitoring report, providing scientific decision support for power station operation and maintenance. This embodiment effectively solves the problems of inaccurate power prediction and delayed early warning caused by meteorological data deviations, abnormal interference, and unstable transmission in photovoltaic power stations. It achieves precise control of meteorological data throughout the entire chain from acquisition, calibration, fusion, transmission to prediction, significantly improving the power generation efficiency and operational reliability of photovoltaic power stations.

[0107] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A photovoltaic meteorological monitoring method based on multi-sensor collaboration, characterized in that, Includes the following steps: S100. Meteorological data is collected collaboratively by data acquisition equipment to obtain an initial acquisition signal. The initial acquisition signal is processed by a calibration model to correct the deviation and obtain a set of calibrated parameters. The meteorological data includes irradiance parameters, temperature monitoring values ​​and wind speed parameters. S200. Perform a data fusion operation based on the calibrated parameter set, fusing the irradiation parameter, the temperature monitoring value, and the wind speed parameter. If a deviation exceeding a preset threshold is detected during the fusion process, the corresponding abnormal data points are removed, and the fused data group is determined. S300: The fused data group is used to execute the transmission protocol in the network architecture, and the fused data group is sent from the perception layer through the transmission layer to the application layer to obtain the transmission completed data stream; S400. The data stream is used to analyze the weather change pattern through the transmission, and the potential output power of the photovoltaic power station is calculated using a prediction model. It is determined whether the potential output power of the photovoltaic power station is lower than a preset threshold. If it is lower, an early warning signal is triggered, and a prediction result set is obtained. S500. Based on the reliability enhancement mechanism of the prediction result set, perform secondary verification on the prediction result set to confirm data consistency, and output the final meteorological monitoring report to support the operation and maintenance decision of the photovoltaic power station.

2. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 1, characterized in that, Step S100 includes: S110. Meteorological data is collected from multiple points through data acquisition equipment. Different sampling frequencies are set for irradiance parameters, temperature monitoring values ​​and wind speed parameters. Initial acquisition signals are obtained from multiple sensor nodes and digitally processed to generate raw data sets. S120. The original data set is corrected for deviation using a calibration model to generate a calibrated parameter set.

3. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 1, characterized in that, Step S200 includes: S210. Based on the irradiance parameters, temperature monitoring values, and wind speed parameters in the calibrated parameter set, a preliminary linear combination is performed using a weighted fusion algorithm to generate an initial fused numerical sequence. S220. The initial fused numerical sequence is time-series smoothed using a sliding window mechanism to generate a smoothed fused data stream. S230. Perform deviation detection and remove outliers on the smoothed and fused data stream according to preset rules to generate a purified fused dataset. S240. Perform state estimation correction on the purified fused dataset and output the fused data set as a reliable fusion result of multi-source meteorological parameters.

4. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 1, characterized in that, Step S300 includes: S310. Based on the characteristics of the fused data group, the perception layer performs packet processing through a preset encoding mechanism to generate packet units. S320. According to the packet unit, the queue is sorted and the data priority is marked at the transport layer using a preset transport protocol; S330. Based on the sorted data queue, apply a dynamic routing mechanism at the transport layer to plan the transmission path; S340. According to the transmission path, the sorted data queue is distributed to the application layer and reassembled to generate a complete data stream for transmission.

5. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 1, characterized in that, Step S400 includes: S410. Extract key meteorological change feature data by transmitting the data stream, perform multi-dimensional analysis on the key meteorological change feature data using a preset meteorological analysis model, generate a dynamic trend map of meteorological changes, and obtain a set of quantitative results of meteorological changes. S420. Based on the meteorological change quantification result set and the historical operation data of the photovoltaic power station, a multi-parameter fitting analysis is performed using a preset power prediction model to determine the potential output power prediction range of the photovoltaic power station. S430. Obtain real-time environmental variable data of photovoltaic power station for the predicted value range, and dynamically correct the predicted value range through a multi-source data fusion mechanism to obtain corrected power prediction data. S440. If the corrected power prediction data is lower than the preset threshold range, the abnormal state indicator will be automatically triggered by the warning signal generation module to generate a warning signal dataset and determine the priority of the warning signal. S450. Based on the priority of the warning signals, a hierarchical response mechanism is used to classify the warning signal dataset, generate a hierarchical response instruction set, and determine the prediction result set.

6. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 1, characterized in that, Step S500 includes: S510. Extract key data for reliability verification from the prediction result set, and perform multi-level data comparison and analysis on the prediction result set using preset consistency verification rules to obtain a consistency deviation quantification result set. The consistency deviation quantification result set is obtained using the following formula: in, This represents the result set of consistency deviation quantification. Indicates the first The quantified value of the deviation of each prediction result. This indicates the preset consistency check threshold. This indicates the total number of prediction result sets, and the consistency deviation quantification result set only includes deviation values ​​that exceed the threshold; No. The quantified value of the deviation of each prediction result is obtained by the following formula: in, Indicates the total number of comparison levels. Indicates the first Hierarchical weights Indicates the first The prediction result is in the first... Hierarchical predicted values, Indicates the first The prediction result is in the first... Reference values ​​or key data values ​​for the hierarchy; S520. Obtain historical meteorological monitoring records of photovoltaic power stations for the consistency deviation quantification result set, and trace the source of deviation data through time series correlation analysis mechanism to obtain deviation source tracing result set; The deviation tracing result set is obtained using the following formula: in, This represents the result set of deviation tracing. Indicates the relationship with the first One deviation Meteorological records with the highest temporal correlation, This represents the time series correlation analysis function. Indicates a point in time Meteorological feature vectors; S530. Based on the deviation source tracing result set and the current real-time meteorological monitoring data, the deviation data is automatically corrected using a preset calibration compensation model to obtain the corrected reliability improvement dataset. S540. If the corrected reliability improvement dataset still has deviations exceeding the preset threshold, the abnormal deviation part is recalculated through the secondary verification module to obtain the final unbiased reliability verification result set. S550. Generate a structured meteorological monitoring report based on the final unbiased reliability verification result set, and output an operation and maintenance decision support report containing complete reliability confirmation information.

7. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 6, characterized in that, In step S530, the corrected reliability improvement dataset is obtained using the following formula: in, This represents the revised, improved reliability dataset. This represents the preset weights of the deviation tracing result set. This indicates the current real-time meteorological monitoring data. This indicates the preset weights of real-time meteorological monitoring data. Indicates deviation data, The preset weights represent the deviation data.

8. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 7, characterized in that, In step S540, the following formula is used to define the judgment criteria for the abnormal deviation portion: in, Indicates the first Does each sample belong to the abnormal deviation portion? Indicates the first A revised reliability value, Indicates a reference reliability value. Indicates the preset threshold. A value of 1 indicates that this part is an abnormal deviation and requires secondary verification; The final unbiased reliability verification result set is obtained through the following formula: in, This represents the final unbiased reliability verification result set. This indicates the result obtained by recalculating the abnormal deviation portion. The mask matrix representing the abnormal deviation portion. This formula represents element-wise multiplication, enabling targeted replacement of abnormal deviations. The mask matrix for the abnormal deviation portion is obtained using the following formula: in, The mask matrix representing the abnormal deviation portion. This represents the bias distribution of the corrected dataset. Indicates the preset threshold. For indicator functions, This represents the absolute value of the deviation. Exceeding the preset threshold The value is 1 if the condition is met, and 0 otherwise. This formula is used to locate abnormal deviations that require secondary verification.

9. The photovoltaic meteorological monitoring method based on multi-sensor collaboration according to claim 8, characterized in that, In step S550, the operation and maintenance decision support report is calculated by weighted summation using the following formula: in, Indicates the level of operational and maintenance decision support. Indicate the candidate decision-making options. This indicates the number of types of reliability confirmation information in the report. Indicate decision In the Compatibility score under class information Indicates the first Weights of class information.

10. A photovoltaic meteorological monitoring system based on multi-sensor collaboration, used to execute the photovoltaic meteorological monitoring method based on multi-sensor collaboration as described in any one of claims 1 to 9, characterized in that, include: The calibration parameter set acquisition module (10) is used to acquire initial acquisition signals by coordinating meteorological data acquisition through data acquisition equipment, process the initial acquisition signals with a calibration model to correct deviations, and obtain a calibration parameter set, wherein the meteorological data includes irradiance parameters, temperature monitoring values ​​and wind speed parameters. The data fusion group determination module (20) is used to perform data fusion operation based on the calibrated parameter set, fusing the irradiation parameter, the temperature monitoring value and the wind speed parameter. If a deviation exceeding a preset threshold is detected during the fusion process, the corresponding abnormal data point is removed and the data fusion group is determined. The transmission completion data stream acquisition module (30) is used to execute the transmission protocol in the network architecture using the fused data group, and send the fused data group from the perception layer through the transmission layer to the application layer to obtain the transmission completion data stream; The prediction result set acquisition module (40) is used to analyze the meteorological change pattern through the data stream of the transmission, calculate the potential output power of the photovoltaic power station using the prediction model, determine whether the potential output power of the photovoltaic power station is lower than the preset threshold, and if it is lower, trigger an early warning signal to obtain the prediction result set. The meteorological monitoring report output module (50) is used to integrate the reliability improvement mechanism according to the prediction result set, perform secondary verification on the prediction result set to confirm data consistency, and output the final meteorological monitoring report to support the operation and maintenance decision of the photovoltaic power station.

Images