A method and system for managing a photovoltaic four-can gateway

Abstract

The present application relates to photovoltaic gateway technical field, propose a kind of photovoltaic four can gateway management method and system, the management method includes: based on electrical parameter, equipment state parameter and environmental parameter, construct photovoltaic power generation power prediction model, and output future time period's photovoltaic power generation power prediction value;According to the fluctuation degree of photovoltaic power generation power prediction value, the fluctuation degree of photovoltaic power generation power prediction value is calculated, and based on fluctuation degree self-adapting adjustment MPPT algorithm's perturbation step length to track the maximum power point of photovoltaic module;Based on environmental parameter and electrical parameter, construct photovoltaic power generation power fitting model, by comparing the deviation of real-time power generation power and fitting power, determine photovoltaic panel dust accumulation state, and generate cleaning suggestion based on future weather forecast;Based on electrical parameter, equipment state parameter and environmental parameter, construct fault diagnosis model, and carry out real-time evaluation and fault early warning to photovoltaic panel state.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic gateway, in particular to a photovoltaic four-capable gateway management method and system. BACKGROUND

[0002] With the continuous growth of global energy demand and the increasing severity of environmental problems, photovoltaic power generation, as a clean and renewable energy solution, has ushered in an unprecedented development opportunity. However, there are still the following problems in photovoltaic applications:

[0003] MPPT (Maximum Power Point Tracking) technology is the core of photovoltaic power generation system, which is used to track the maximum power point of solar panels in real time, ensuring that the system operates at optimal efficiency. However, in reality, due to unreasonable disturbance step setting or rapid changes in external environment (such as sudden changes in irradiance), the MPPT algorithm often fails to adjust in time, resulting in overshooting of the system and a significant reduction in efficiency;

[0004] Dust accumulation on the surface of photovoltaic modules is one of the main factors affecting the energy efficiency of power stations. Dust accumulation not only directly reduces light transmittance and reduces power generation, but also hinders heat dissipation of the module, causing the working temperature to rise and further deteriorating the conversion efficiency. Long-term dust accumulation can even cause surface corrosion and accelerate the aging of the module, significantly shortening its service life. According to statistics, dust accumulation that is not cleaned up in time can cause a 2%-10% loss of power generation per month, which has a huge impact on operating income. At present, the industry urgently needs a simple and effective dust accumulation monitoring method to reasonably arrange cleaning time and find the optimal balance between operation and maintenance costs and power generation income;

[0005] Common faults of photovoltaic modules include cell aging, dust accumulation, hidden cracks, and diode failure. The main detection methods at present include light transmittance monitoring, EL (Electroluminescence) detection, and infrared thermal imaging method. However, these methods have many limitations, such as strict environmental requirements, complex operation, and high cost, making it difficult to meet the efficient operation and maintenance needs of large-scale photovoltaic power stations. SUMMARY

[0006] In view of the above defects, the present application aims to provide a photovoltaic four-capable gateway management method and system, which aims to solve the key problems of insufficient adaptability of MPPT algorithm, energy loss caused by dust accumulation, and limitations of fault detection methods in existing photovoltaic technology, thereby improving the operation efficiency of photovoltaic power stations and reducing operation and maintenance costs.

[0007] To achieve this purpose, the present application adopts the following technical solutions:

[0008] A photovoltaic four-capable gateway management method, the management method comprising the following steps:

[0009] S1: collecting electrical parameters, equipment state parameters and environmental parameters of the photovoltaic power station;

[0010] S2: constructing a photovoltaic power generation power prediction model based on the electrical parameters, equipment state parameters and environmental parameters, and outputting a photovoltaic power generation power prediction value of a future period;

[0011] S3: calculating a fluctuation degree of the photovoltaic power generation power prediction value according to the photovoltaic power generation power prediction value of the future period, and adaptively adjusting a perturbation step of an MPPT algorithm based on the fluctuation degree to track a maximum power point of the photovoltaic component;

[0012] S4: constructing a photovoltaic power generation power fitting model based on the environmental parameters and electrical parameters, determining a dust accumulation state of the photovoltaic panel by comparing a deviation between a real-time power generation power and a fitting power of the photovoltaic power generation power fitting model, and generating a cleaning suggestion based on a future weather forecast;

[0013] S5: constructing a fault diagnosis model based on the electrical parameters, equipment state parameters and environmental parameters, and performing real-time evaluation and fault warning on the photovoltaic panel state.

[0014] Preferably, step S1 comprises:

[0015] The electrical parameters include input and output voltages, input and output currents, input and output powers, active powers, reactive powers and power generation amounts of the photovoltaic component and the inverter;

[0016] The equipment state parameters include working modes and temperatures of the inverter, photovoltaic component states and photovoltaic switch states;

[0017] The environmental parameters include wind speeds, wind directions, light intensities, air pressures, humidities, outdoor temperatures, direct radiations, scattered radiations, solar elevation angles, solar azimuth angles and weather conditions.

[0018] Preferably, step S2 comprises:

[0019] S21: performing data preprocessing on the electrical parameters, equipment state parameters and environmental parameters, the data preprocessing including data cleaning, missing value and abnormal value processing and data normalization;

[0020] S22: performing feature extraction and construction based on the preprocessed data, the feature extraction and construction including extracting meteorological features from the environmental parameters, and constructing statistical quantities and signal analysis features including mean values, root mean square values, peak values, maximum difference values, peak value indexes, waveform indexes, pulse indexes, margin indexes, skewness and kurtosis from the electrical parameters and the meteorological features;

[0021] S23: Based on the results of feature extraction and construction, a long short-term memory network model is constructed and utilized for photovoltaic power prediction, and future photovoltaic power prediction values of several time granularities are output.

[0022] Preferably, step S3 comprises:

[0023] S31: Obtain photovoltaic power prediction values of future time periods, calculate the fluctuation degree of the photovoltaic power prediction values, and obtain the rolling standard deviation of photovoltaic power prediction , satisfying the relationship , wherein represents the size of the time window, represents the current time point, represents the photovoltaic power prediction value at a specific time point i, represents the average value of the photovoltaic power prediction values from t+1 to t+w+1.

[0024] S32: Compare the calculated fluctuation degree with the standard deviation set threshold , and adjust the perturbation voltage step of the MPPT algorithm according to the comparison result, wherein:

[0025] If the fluctuation degree is less than or equal to the standard deviation set threshold , a first perturbation voltage is used to adjust the output voltage of the photovoltaic panel;

[0026] If the fluctuation degree is greater than the standard deviation set threshold , a second perturbation voltage is used to adjust the output voltage of the photovoltaic panel, and the second perturbation voltage is less than the first perturbation voltage .

[0027] S33: After applying the perturbation voltage, measure the output power change of the photovoltaic system, and adjust the subsequent perturbation direction based on the output power change: if the output power increases, keep the current perturbation direction; if the output power decreases, reverse the perturbation direction.

[0028] Preferably, step S4 comprises:

[0029] S41: Select a historical period when the photovoltaic panel is in a clean state, and obtain the irradiance, outdoor temperature and photovoltaic power of the selected period as modeling data;

[0030] S42: Preprocess the modeling data, including outlier identification and screening, missing value filling and data normalization, fit the outdoor temperature, irradiance and photovoltaic power based on the preprocessed data, and construct a photovoltaic power fitting model;

[0031] S43: Obtain real-time irradiance and outdoor temperature, and calculate expected photovoltaic power generation power based on the photovoltaic power generation power fitting model;

[0032] S44: Compare the actual photovoltaic power generation power with the expected photovoltaic power generation power. If the actual photovoltaic power generation power is continuously less than the expected photovoltaic power generation power and the relative deviation exceeds the preset deviation threshold, it is determined that the photovoltaic panel is dusty, and the dust degree is calculated.

[0033] S45: Quantify the net income based on the dust degree, predicted power generation, predicted power generation attenuation, electricity price, and cleaning cost, and generate a cleaning suggestion. If the expected income increment after cleaning is greater than the cleaning cost, a cleaning suggestion of suggesting cleaning is generated, otherwise a cleaning suggestion of not suggesting cleaning is generated.

[0034] Preferably, the photovoltaic power generation power fitting model satisfies the relationship:

[0035] ;

[0036] wherein, , and represent fitting coefficients, represents the expected photovoltaic power generation power, represents irradiance, represents outdoor temperature, represents an irradiance correction coefficient.

[0037] Preferably, the dust degree satisfies the relationship:

[0038] ;

[0039] wherein, represents the expected photovoltaic power generation power, represents the actual photovoltaic power generation power;

[0040] power generation attenuation power .

[0041] Preferably, the calculation of the income satisfies the relationship:

[0042] ;

[0043] wherein, represents the power generation attenuation power, represents the electricity value, represents time.

[0044] Preferably, step S5 comprises:

[0045] S51: Collect historical fault label data corresponding to photovoltaic, and perform data preprocessing on the operation state data, including abnormal value identification and screening, missing value filling, time series synchronization and data normalization;

[0046] S52: Based on the preprocessed data, a derived feature reflecting the operation state of the photovoltaic panel is constructed, and based on the derived feature and the historical fault label data, a machine learning fault diagnosis model is constructed and trained;

[0047] S53: Using the trained machine learning fault diagnosis model, the state of the photovoltaic panel is evaluated and fault warning in real time.

[0048] A photovoltaic four-capable gateway management system, the photovoltaic four-capable gateway management system is applied to a photovoltaic four-capable gateway management method as described above, and the photovoltaic four-capable gateway management system comprises:

[0049] A data acquisition module for acquiring electrical parameters, equipment state parameters and environmental parameters of a photovoltaic power station;

[0050] A photovoltaic short-term prediction module for constructing a photovoltaic power generation power prediction model based on the electrical parameters, equipment state parameters and environmental parameters, and outputting a photovoltaic power generation power prediction value in a future period;

[0051] An MPPT adaptive adjustment module for calculating the fluctuation degree of the photovoltaic power generation power prediction value according to the photovoltaic power generation power prediction value in the future period, and adaptively adjusting the perturbation step of the MPPT algorithm based on the fluctuation degree to track the maximum power point of the photovoltaic component;

[0052] A photovoltaic panel operation and maintenance optimization module for constructing a photovoltaic power generation power fitting model based on the environmental parameters and electrical parameters, determining the dust accumulation state of the photovoltaic panel by comparing the deviation between the real-time power generation power and the fitting power of the photovoltaic power generation power fitting model, and generating a cleaning suggestion based on the future weather forecast;

[0053] A fault diagnosis module for constructing a fault diagnosis model based on the electrical parameters, equipment state parameters and environmental parameters, and evaluating and warning the state of the photovoltaic panel in real time.

[0054] One of the above technical solutions has the following advantages or beneficial effects:

[0055] The present application effectively solves the problems of poor adaptability of MPPT algorithm, insufficient dust monitoring and high cost of fault detection in existing photovoltaic systems. First, by collecting electrical parameters, equipment state parameters and environmental parameters of the photovoltaic power station, comprehensive and accurate data basis is provided for subsequent power prediction, MPPT adjustment, dust monitoring and fault diagnosis. Then, based on these parameters, a photovoltaic power generation power prediction model is constructed and the predicted value of the future period is output, so that the system can predict the trend of power generation power in advance, and provide a basis for subsequent adaptive MPPT adjustment. Then, the disturbance step of the MPPT algorithm is adaptively adjusted according to the predicted power generation power fluctuation degree, to ensure that the photovoltaic components always work near the maximum power point in complex environment, and improve the power generation efficiency. At the same time, by constructing a photovoltaic power generation power fitting model and comparing the deviation of the actual photovoltaic power generation power and the fitting power, the dust state of the photovoltaic panel is determined and the cleaning suggestion is generated based on the future weather forecast, to realize accurate monitoring and timely processing of the dust problem and reduce the power generation loss. Finally, based on the collected parameters, a fault diagnosis model is constructed to evaluate and warn the fault of the photovoltaic panel in real time, to reduce the cost of fault detection and improve the reliability and operation efficiency of the system. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only are the embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on the provided drawings.

[0057] Figure 1 is a flow chart of the management method of the photovoltaic four-gateway provided by the embodiment of the present application;

[0058] Figure 2 is a structural schematic diagram of the management system of the photovoltaic four-gateway provided by the embodiment of the present application. DETAILED DESCRIPTION

[0059] The embodiments of the present application will be described in detail below, and the examples of the embodiments are shown in the drawings, wherein the same or similar reference signs represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation of the present application.

[0060] In the present invention, the terms "comprising", "containing" or any other variant thereof are intended to cover non-exclusive inclusions, such that a process, method, article, or apparatus that comprises a list of elements not only includes those elements, but it is also possible that other elements that are not expressly listed are also included, or it is also possible that elements inherent to such a process, method, article, or apparatus are included. Without more limitations, the element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0061] A management method of a photovoltaic four-capable gateway, as shown in the figure, a preferred embodiment of the present invention, the method comprises the following steps: Figure 1

[0062] S1: Collecting electrical parameters, equipment state parameters and environmental parameters of the photovoltaic power station;

[0063] It should be noted that the electrical parameters refer to the input and output voltage, current, power, etc. of the photovoltaic module and the inverter, which are used to reflect the real-time running state of the photovoltaic system; the equipment state parameters include the working mode and temperature of the inverter, which are used to monitor the health condition and running mode of the equipment; the environmental parameters such as temperature and irradiance are used to provide external condition information for the operation of the photovoltaic system.

[0064] It can be understood that collecting electrical, equipment and environmental parameters of the photovoltaic power station can provide reliable data basis for subsequent power prediction, MPPT adjustment and fault diagnosis. The collection of the above parameters is realized through existing sensors and monitoring modules. The electrical parameters reflect the system running state, the equipment parameters reflect the health condition, and the environmental parameters provide external conditions, which constitute the input data source of the intelligent diagnosis system.

[0065] Specifically, the data can be collected in the following ways: 1) Through direct communication between the photovoltaic four-capable gateway and the inverter, the input and output current, voltage and power of the inverter and other system-level electrical parameters are obtained; 2) The environmental temperature, irradiance, wind speed and other environmental parameters are measured by using the weather station, and the weather station data is directly connected to the gateway. These collected data are summarized in the photovoltaic four-capable gateway through data lines or wireless transmission modules to form a complete data set. The combination of multiple sensors can improve the comprehensiveness and accuracy of data collection, and provide high-quality data support for subsequent intelligent diagnosis.

[0066] S2: Based on the electrical parameters, equipment state parameters and environmental parameters, a photovoltaic power generation power prediction model is constructed, and the photovoltaic power generation power prediction value of the future period is output;

[0067] ​It should be noted that the photovoltaic power prediction model is an algorithm model based on historical data and real-time data, which is used to predict the photovoltaic power in the future period. The model usually includes a data preprocessing module (for cleaning and normalizing data), a feature extraction module (for extracting features such as mean, variance, etc.), and a model training module (such as long short-term memory network LSTM). Its function is to provide a reference value for the future power for MPPT adjustment and dust accumulation determination.

[0068] It can be understood that the purpose of building a photovoltaic power prediction model is to know the future power in advance, so that the intelligent gateway can dynamically adjust the system parameters and optimize the power generation efficiency. The model learns the rules in the historical data to predict the future power change, thereby providing a basis for the intelligent adjustment of the system.

[0069] Specifically, an LSTM neural network model can be used for short-term power prediction, with inputs including historical power, irradiance, temperature, etc., and outputs including power prediction values for the next 30 minutes (5-minute granularity). Weather clustering analysis can also be combined to improve the adaptability of the model under different weather conditions.

[0070] S3: According to the photovoltaic power prediction value of the future period, calculate the fluctuation degree of the photovoltaic power prediction value, and adaptively adjust the perturbation step of the MPPT algorithm to track the maximum power point of the photovoltaic module based on the fluctuation degree;

[0071] It should be noted that the fluctuation degree of the photovoltaic power prediction value refers to the change amplitude of the power in the future period, which is usually quantified by calculating the rolling standard deviation. The MPPT algorithm is a control algorithm used to track the maximum power point of the photovoltaic module, and the perturbation step determines the accuracy and speed of tracking. The purpose of adaptively adjusting the perturbation step is to dynamically adjust the tracking strategy according to the power fluctuation, ensuring that the system can run efficiently in complex environments.

[0072] It can be understood that by calculating the fluctuation degree of the photovoltaic power prediction value, the power change situation faced by the system in the future period can be evaluated. When the fluctuation is small, a larger perturbation step can be used to quickly track the maximum power point; when the fluctuation is large, a smaller perturbation step can be used to avoid overshoot and oscillation, ensuring stable operation of the system. This adaptive adjustment mechanism can make the MPPT algorithm maintain efficient power tracking performance under different environments.

[0073] S4: Based on the environmental parameters and electrical parameters, a photovoltaic power fitting model is constructed, the deviation between the real-time power and the fitting power of the photovoltaic power fitting model is compared, the dust accumulation state of the photovoltaic panel is determined, and a cleaning suggestion is generated based on the future weather forecast;

[0074] It should be noted that the photovoltaic power generation power fitting model refers to a power-environment parameter relationship model established based on historical data under a clean state; the real-time power generation power refers to the power generation power of the photovoltaic panel in actual operation, and the fitted power refers to the expected photovoltaic power generation power, i.e., the ideal power generation power of the component, calculated according to the photovoltaic power generation power fitting model. The deviation between the two can reflect the dust accumulation degree of the photovoltaic panel. The cleaning suggestion is generated according to the dust accumulation degree and economic analysis, and is used to guide whether the photovoltaic panel needs to be cleaned.

[0075] It can be understood that, by comparing the expected output with the actual output through the model, the decrease in power generation efficiency caused by dust accumulation can be identified, thereby avoiding blind cleaning or delaying cleaning, and realizing the economy and scientificity of the cleaning decision.

[0076]

[0077] It should be noted that the fault diagnosis model refers to a classification or regression model trained based on historical fault data, which is used to identify whether the photovoltaic panel has a fault and the type of the fault; the real-time evaluation and fault warning refer to the judgment of whether a fault is likely to occur in the current operating state of the system, and the advance of the prompt.

[0078] It can be understood that, by fusing multi-source data such as electrical, environmental, and equipment data, the running state of the system can be comprehensively reflected, the accuracy and timeliness of fault identification can be improved, and the lag and high cost of traditional methods relying on manual inspection can be avoided.

[0079] Specifically, an XGBoost model can be constructed, the input includes voltage, current, temperature, power / irradiance ratio, etc., and the output is a fault type label; time series features can also be introduced to improve the sensitivity of the model to early faults.

[0080] Preferably, step S1 comprises:

[0081] The electrical parameters include the input and output voltages, input and output currents, input and output powers, active powers, reactive powers, and power generations of the photovoltaic components and the inverters;

[0082] The equipment state parameters include the working modes and temperatures of the inverters, the states of the photovoltaic components, and the states of the photovoltaic switches;

[0083] The environmental parameters include the wind speed, wind direction, light intensity, air pressure, humidity, outdoor temperature, direct radiation, scattered radiation, solar elevation angle, solar azimuth angle, and weather condition.

[0084] ​The electrical parameters are core data reflecting the power output characteristics of the photovoltaic power station. For example, the input and output voltages are used to determine whether the system is working in a reasonable voltage range, and the power data is used to evaluate the power generation efficiency. The equipment state parameters are used to monitor the system operation state, such as the inverter temperature which can be used to determine whether it is overheated. The environmental parameters are external conditions that affect photovoltaic power generation, such as the light intensity which directly determines the total amount of energy received by the photovoltaic panel, and the weather conditions such as sunny, rainy, cloudy, etc.

[0085] It can be understood that by refining the three types of parameter data, the subsequent prediction, fitting and diagnosis models have a unified, complete and comparable data base, avoiding model drift caused by missing dimensions or unknown states.

[0086] Specifically, the electrical parameters can be collected by the built-in sensors of the inverter, the environmental parameters can be obtained by the weather station or sensors, and the equipment state parameters can be read through the communication interface. The collection frequency can be set to once per minute to ensure data continuity. Edge computing devices can also be used for local preprocessing to reduce transmission delay.

[0087] Preferably, step S2 comprises:

[0088] S21: data preprocessing on the electrical parameters, equipment state parameters and environmental parameters, the data preprocessing comprising data cleaning, missing value and abnormal value processing and data normalization;

[0089] It should be noted that data preprocessing is a pre-process of data mining, data cleaning is a process of removing noise and error data, missing value processing is a method of filling data gaps, and data normalization is a technique of unifying data to a specific range.

[0090] It can be understood that the originally collected data often includes noise, missing values and scale differences. If directly used for analysis and modeling, it will seriously reduce the accuracy and reliability of the model. The cleaning operation filters out noise interference and error information, restores the pure nature of the data; the missing value processing fills in the data gaps with scientific interpolation, ensuring the integrity of the information; the normalization operation unifies the data scale, eliminates the dimensional differences, and makes the data more suitable for the operation characteristics of the model algorithm.

[0091] S22: based on the preprocessed data, feature extraction and construction, the feature extraction and construction comprising extracting meteorological features from the environmental parameters, and constructing statistical quantities and signal analysis features including mean, root mean square, peak value, maximum difference, peak value index, waveform index, pulse index, margin index, skewness, kurtosis from the electrical parameters and meteorological features;

[0092] It should be noted that feature extraction is a process of screening key features from massive data that have a significant impact on the prediction target, and feature construction is an operation of combining and transforming original features to generate new features to improve the expression ability of the model. The meteorological features in the environmental parameters include wind scale, wind speed, wind direction, and light intensity, which reflect the influence of the environment on power generation; statistical and signal analysis features such as mean, root mean square value, and peak value are quantitative indicators extracted from electrical parameters and meteorological features, which are used to capture the statistical characteristics and time series fluctuation rules of the data.

[0093] S23: Based on the results of feature extraction and construction, a long short-term memory network model is constructed and used for photovoltaic power generation power prediction, and future photovoltaic power generation power prediction values of several time granularities are output.

[0094] It should be noted that the long short-term memory network (LSTM) is a special recurrent neural network (RNN) architecture designed to handle long-distance dependency problems in time series data, and has the ability to remember long-term information. The photovoltaic power generation power prediction model is a machine learning model trained based on historical data, which is used to predict future power generation power. The results of feature extraction and construction provide input features for the model, including time series features and statistical features extracted from electrical parameters and environmental parameters, which provide key knowledge sources for the model.

[0095] Preferably, step S3 comprises:

[0096] S31: Obtain the photovoltaic power generation power prediction value of the future period, calculate the fluctuation degree of the photovoltaic power generation power prediction value, and obtain the rolling standard deviation of the photovoltaic prediction power , satisfying the relationship , wherein represents the size of the time window, represents the current time point, represents the photovoltaic power generation power prediction value at a specific time point i, represents the average value of the photovoltaic power generation power prediction value from t+1 to t+w+1.

[0097] S32: Compare the calculated fluctuation degree with the standard deviation set threshold , and adjust the perturbation voltage step size of the MPPT algorithm according to the comparison result, wherein:

[0098] If the fluctuation degree is less than or equal to the standard deviation set threshold , the first perturbation voltage is used to adjust the output voltage of the photovoltaic panel;

[0099] If the fluctuation degree is greater than the standard deviation set threshold , the second perturbation voltage Adjusting the output voltage of a photovoltaic panel, the second perturbation voltage less than the first perturbation voltage ;

[0100] S33: After applying the perturbation voltage, measure the output power change of the photovoltaic system, and adjust the subsequent perturbation direction based on the output power change: if the output power increases, keep the current perturbation direction; if the output power decreases, reverse the perturbation direction.

[0101] It should be noted that the photovoltaic predicted power rolling standard deviation is a statistical quantity that quantifies the degree of fluctuation of photovoltaic power generation in the future period. The size of the time window w determines the number of historical data points considered in the calculation, the current time point t indicates the fluctuation state at this moment, the predicted value is the power generation prediction value at time point i, and is predicted by a long short-term memory network (LSTM), is the average value of the predicted values within the time window, and these elements cooperatively define accurately depict the power fluctuation characteristics and provide a quantitative basis for subsequent perturbation step adjustment.

[0102] It can be understood that in order to adjust the perturbation step of the MPPT algorithm according to the fluctuation of the power generation prediction value, ensure that the photovoltaic component always works near the maximum power point, and improve the photovoltaic power generation efficiency. When the predicted power fluctuation is large, a smaller perturbation step is used to enable the system to more accurately track the maximum power point and avoid missing the optimal working point due to a too large step size; on the contrary, when the power fluctuation is small, the perturbation step can be appropriately increased to improve the tracking speed. In this way, the MPPT algorithm is adaptively adjusted, and the adaptability and stability of the photovoltaic system in complex environments are enhanced.

[0103] In an embodiment, the input features of the photovoltaic power generation prediction model PVModel are the actual photovoltaic power generation, the light intensity, the outdoor temperature, the direct / scattered radiation at the same time of the previous day, the day before yesterday and the prediction point, the average value / maximum value / minimum value of the photovoltaic power generation, the light intensity, the outdoor temperature, the direct / scattered radiation at the same time of the previous 7 days and the future prediction point, the actual photovoltaic power generation, the actual light intensity, the outdoor temperature, the direct / scattered radiation every 5 minutes at the current time and 1 hour ago, after normalization, the above features are used as input; the output is the photovoltaic power value r i+1 , r i+2 , ... r i+6The prediction model is a Long Short-Term Memory (LSTM) network model, with an overall architecture consisting of three layers: an input layer, an LSTM hidden layer, and an output layer. The input layer receives normalized input features.

[0104] LSTM hidden layers (2 layers in total): Layer 1 contains 64 hidden units, with dropout rate = 0.2 and return_sequences = True; Layer 2 contains 128 hidden units, with dropout rate = 0.2 and return_sequences = False.

[0105] Output layer: 6 nodes, outputting normalized 5-minute granular photovoltaic power generation values.

[0106] The model training parameters are: optimizer Adam (learning rate 0.001), loss function mean squared error (MSE), batch size = 64, and number of iterations = 100.

[0107] During forecasting, input features are extracted from historical data based on future forecast times. After normalization, these features are input into the PVModel. The forecast results are then inversely normalized to obtain the photovoltaic power generation forecast values ​​at a granularity of six 5-minute intervals, i.e., r. i+1 ,r i+2 ,...,r i+6 .

[0108] Set the time window size for the MPPT algorithm to w=6, and set a threshold for the standard deviation. =3. When the system obtains the sequence of 5-minute granular photovoltaic power generation prediction values ​​for 6 points over the next 30 minutes as 42, 45, 39, 55, 41, 50, it first calculates the average value of the prediction values ​​within the time window. Then, the rolling standard deviation is calculated based on the relationship. ≈6.09. Since 6.09 > 3 ( The system determines that the current fluctuation level is large, and therefore adopts a smaller second disturbance voltage. =0.2V to adjust the output voltage of the photovoltaic panel; after applying this disturbance voltage, the output power of the photovoltaic system was measured to increase from 45kW to 46kW. Based on this, the system determined that the current disturbance direction was correct and maintained this disturbance direction in subsequent control to continue tracking the maximum power point.

[0109] Preferably, step S4 includes:

[0110] S41: Select a historical period when the photovoltaic panels are in a clean state, and obtain the irradiance, outdoor temperature and photovoltaic power generation of the selected period as modeling data;

[0111] S42: data preprocessing is performed on the modeling data, including outlier identification and screening, missing value filling, and data normalization, fitting of outdoor temperature, irradiance, and photovoltaic power based on the preprocessed data, and construction of a photovoltaic power fitting model;

[0112] S43: real-time irradiance and outdoor temperature are obtained, and expected photovoltaic power is calculated based on the photovoltaic power fitting model;

[0113] S44: the actual photovoltaic power and the expected photovoltaic power are compared, and if the actual photovoltaic power is continuously less than the expected photovoltaic power and the relative deviation exceeds a preset deviation threshold, it is determined that the photovoltaic panel is dusty, and the dust degree is calculated;

[0114] S45: net income is quantified based on the dust degree, predicted power generation, predicted power generation attenuation, electricity price, and cleaning cost, and a cleaning suggestion is generated: if the expected income increment after cleaning is greater than the cleaning cost, a cleaning suggestion of suggesting cleaning is generated, otherwise a cleaning suggestion of not suggesting cleaning is generated.

[0115] It should be noted that outdoor temperature affects power generation by affecting the properties of semiconductor materials; the dust degree is quantified by comparing the deviation of real-time power generation and fitted power (expected photovoltaic power) to quantify the impact of dust on power generation efficiency; net income quantification is the process of economic benefit evaluation by comparing the power generation income increment after cleaning and the cleaning cost.

[0116] It can be understood that the purpose of step S4 is to accurately monitor the dusting state of the photovoltaic panel and make scientific decisions on the cleaning time. The principle is that by constructing a fitted power model, the power generation of the photovoltaic panel under the condition of no dust is simulated, and then compared with the actual power generation. If the actual power is continuously lower than the fitted power and the deviation exceeds the preset threshold, it is determined that the photovoltaic panel has dust problem, and further net income quantification is performed to calculate the difference between the power generation income increment after cleaning and the cleaning cost. The higher the dust degree, the smaller the predicted power generation and the larger the predicted power generation attenuation, which means the greater the potential power loss and the higher the income after cleaning. If the income increment is greater than the cost, cleaning is suggested, otherwise it is not suggested. This process not only accurately identifies the power loss caused by dust, but also optimizes the cleaning decision from an economic point of view, avoiding unnecessary cleaning expenses and ensuring the balance between photovoltaic power station operation and maintenance costs and power generation income.

[0117] For example, combined with weather forecast, the future 15-day hourly weather forecast is obtained through weather forecast. If the irradiance is continuously large in the next period, the expected income increment after cleaning is large, then cleaning is performed. If there are more rainy days in the next period, the number of hours with large irradiance is small, the income increment is less than the cleaning cost, or there is heavy rain to effectively clean the photovoltaic panel, then cleaning is not suggested.

[0118] Preferably, the photovoltaic power generation power fitting model satisfies the relationship:

[0119] ;

[0120] wherein, , and represent fitting coefficients, represents the expected photovoltaic power generation power, represents the irradiance, represents the outdoor temperature, represents the irradiance correction coefficient.

[0121] Specifically, the fitting coefficients are used to quantify the effects of irradiance and outdoor temperature on the power generation power, reflects the expected power generation capacity of the photovoltaic panel under given irradiance and outdoor temperature, and the irradiance is an environmental parameter that affects the power generation efficiency of the photovoltaic panel. The power generation power of the photovoltaic panel is affected by the irradiance and the outdoor temperature, and by constructing a fitting model, the influence relationship of these two factors on the power generation power can be quantified. Once the photovoltaic panel surface is dusty, the actual power generation power will be lower than the expected photovoltaic power generation power, and the deviation degree is related to the severity of dust accumulation. In addition, the irradiance correction coefficient is a coefficient that changes according to the weather type (such as sunny, cloudy, rainy, and snowy). The irradiance correction coefficient is used to improve the accuracy of the photovoltaic power generation power fitting model under different weather conditions. The photovoltaic power generation power fitting model provides a benchmark for dust accumulation monitoring and is the basis for realizing accurate dust judgment and cleaning decision. In an embodiment, the irradiance correction coefficient can be 0.3 for thick cloudy days, 0.25 for light rain, 0.15 for moderate rain, and 0.20 for snowy days.

[0122] Preferably, the dust degree satisfies the relationship:

[0123] ;

[0124] wherein, represents the expected photovoltaic power generation power, represents the actual photovoltaic power generation power;

[0125] power generation decay power .

[0126] It can be understood that the dust degree provides an index for quantitatively evaluating the influence of dust on the photovoltaic panel, and the principle is to reflect the degree of power generation capacity reduction caused by dust by comparing the difference between the actual power generation of the photovoltaic panel and the expected power generation of the photovoltaic panel. Specifically, when the photovoltaic panel is covered with dust, the light transmittance is reduced, and the power generation is reduced. By calculating the ratio of the difference between the actual power generation of the photovoltaic panel and the expected power generation of the photovoltaic panel to the actual power generation of the photovoltaic panel, the dust degree can be obtained, thereby directly quantifying the influence of dust on the power generation efficiency. The dust degree can help the operation and maintenance personnel to find the dust problem in time, and then take cleaning and other maintenance measures to optimize the power generation performance of the photovoltaic power station.

[0127] Preferably, the calculation of the benefit satisfies the relationship:

[0128] ;

[0129] wherein, represents the power generation attenuation power, represents the power value, represents the time.

[0130] It can be understood that the economic benefit of cleaning the photovoltaic panel is evaluated by analysis to determine whether it is necessary to perform cleaning operation, and the principle is to estimate the power generation loss caused by dust based on the dust degree, and to calculate the recoverable power generation and the corresponding economic benefit after cleaning by combining the light conditions and the electricity price factors. The recoverable power generation is related to the power generation attenuation power, and by calculating the power generation attenuation power as described above, combined with the time and the power unit price parameters, the increment of power generation benefit after cleaning can be obtained, that is, power generation attenuation power x time = power generation change value after cleaning. If the increment of benefit is greater than the cleaning cost, the cleaning operation is economically profitable and can bring net benefit to the photovoltaic power station; otherwise, it may not be recommended to clean, so as to avoid unnecessary economic expenditure. This process provides a scientific economic decision basis for the operation and maintenance management of the photovoltaic power station, and ensures the rationality and effectiveness of the cleaning operation.

[0131] Based on the above description, the following is part of the cleaning strategy of one of the embodiments: first, based on the real-time collected irradiance, outdoor temperature and weather type, the photovoltaic power generation power fitting model is used to calculate the expected power under the current environment, and compare it with the real-time collected actual power generation of the photovoltaic panel to calculate the current dust degree , based on the weather forecast data, the hourly irradiance , outdoor temperature and weather type (such as sunny, heavy rain, light rain, snow day, etc.) in the next 15 days are obtained, and the irradiance correction coefficient is determined according to the weather type.

[0132] Subsequently, the future hourly photovoltaic expected power generation is calculated , and the power generation attenuation power is obtained through the formula , wherein is the correction value considering the weather influence correction on the basis of the current dust degree , that is, =min(max( +the dust correction coefficient, 0), 100), for example, light rain (<10 mm / day): dust degree + 15% (forming a cohesive layer), heavy rain (>38 mm / day): dust degree - 80% (effective cleaning), sunny day: dust degree + 0%; and the correction value of the dust degree affected by the weather is continuous, if there is no large change in the weather, the last state is continued, for example, the future weather is light rain, light rain, light rain, heavy rain, heavy rain, sunny, sunny, and the dust degree is +15%, +15%, +15%, min(max( +15%-80%, 0), 100), min(max( +15%-80%, 0), 100), min(max( +15%-80%, 0), 100), min(max( +15%-80%, 0), 100).

[0133] The attenuation power at each time caused by dust is combined with the corresponding electricity price, and finally the estimated income is obtained , that is, the additional power generation income obtained by eliminating the dust attenuation through photovoltaic cleaning, for example, after obtaining the total value of the expected power generation loss caused by dust in the future 15 days of 85 yuan, the system compares the value with the preset single cleaning cost of 200 yuan: the calculated net income increment of cleaning is 85 yuan-200 yuan=-115 yuan; since the value is less than zero, it is determined that the cleaning behavior is not economically feasible, and the conclusion of "not recommended to clean" is drawn, and the monitoring is continued, if the dust intensifies or the weather forecast shows that the subsequent will appear continuous fine weather, the system will trigger the evaluation again.

[0134] Preferably, step S5 comprises:

[0135] S51: collecting historical fault label data corresponding to the photovoltaic, and performing data preprocessing including abnormal value identification screening, missing value filling, time series synchronization and data normalization on the running state data;

[0136] S52: Based on the preprocessed data, construct a derived feature reflecting the operating state of the photovoltaic panel, and based on the derived feature and the historical fault label data, construct and train a machine learning fault diagnosis model;

[0137] S53: Use the trained machine learning fault diagnosis model to perform real-time evaluation and fault warning of the photovoltaic panel state.

[0138] The fault diagnosis model refers to a model trained based on a large amount of data, which is used for real-time evaluation and fault warning of the operating state of the photovoltaic panel. The historical fault label data refers to the photovoltaic panel fault information recorded by the past operation and maintenance personnel, including fault type, occurrence time, etc. These data have clear labels and are used to train the model to identify different fault patterns. The derived feature is a new feature variable extracted from the operating data of the photovoltaic panel, such as the power-to-illumination ratio, which can more sensitively reflect the operating state of the device. Machine learning algorithms are technical methods for implementing fault diagnosis model construction and training, such as XGBoost, etc. These algorithms can learn the mapping relationship between features and faults from a large amount of labeled data, thereby achieving automatic fault diagnosis.

[0139] It can be understood that the purpose of step S5 is to realize real-time monitoring and early warning of photovoltaic panel state. The principle is to collect a large amount of historical fault data with labels and real-time photovoltaic panel operating data, and use machine learning algorithms to mine the correlation patterns between features and faults. The constructed fault diagnosis model can quickly and accurately judge whether the current state of the photovoltaic panel is normal when new operating data is input, and if there is an anomaly, further identify the specific fault type. This process can help operation and maintenance personnel to discover potential faults in time, take maintenance measures in advance, avoid fault expansion, reduce power generation loss, and improve the reliability and operating efficiency of the photovoltaic power station.

[0140] A photovoltaic four-capable gateway management system, as shown in Figure 2 The photovoltaic four-capable gateway management system is applied to the photovoltaic four-capable gateway management method described above, and the photovoltaic four-capable gateway management system comprises:

[0141] A data acquisition module for acquiring electrical parameters, device state parameters and environmental parameters of a photovoltaic power station;

[0142] A photovoltaic short-term prediction module for constructing a photovoltaic power generation power prediction model based on the electrical parameters, device state parameters and environmental parameters, and outputting a photovoltaic power generation power prediction value for a future period;

[0143] An MPPT adaptive adjustment module is configured to calculate a fluctuation degree of the predicted photovoltaic power according to the predicted photovoltaic power of the future period, and to adaptively adjust a perturbation step of an MPPT algorithm based on the fluctuation degree to track the maximum power point of the photovoltaic module;

[0144] A photovoltaic panel operation and maintenance optimization module is configured to construct a photovoltaic power fitting model based on the environmental parameters and the electrical parameters, to determine a dust accumulation state of the photovoltaic panel by comparing a deviation between a real-time power and a fitting power of the photovoltaic power fitting model, and to generate a cleaning suggestion based on a future weather forecast.

[0145] A fault diagnosis module is configured to construct a fault diagnosis model based on the electrical parameters, the equipment state parameters and the environmental parameters, and to perform real-time evaluation and fault warning on the photovoltaic panel state.

[0146] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the exemplary description of the above terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0147] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made thereto without departing from the principles and spirit of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A management method for a photovoltaic four-way gateway, characterized in that, The management method comprises the following steps: S1: collecting electrical parameters, equipment state parameters and environmental parameters of the photovoltaic power station; S2: based on the electrical parameters, equipment state parameters and environmental parameters, a photovoltaic power generation power prediction model is constructed, and a photovoltaic power generation power prediction value of a future period is output; S3: according to the photovoltaic power generation power prediction value of the future period, a fluctuation degree of the photovoltaic power generation power prediction value is calculated, and a perturbation step of an MPPT algorithm is adaptively adjusted based on the fluctuation degree to track a maximum power point of the photovoltaic component; S4: based on the environmental parameters and electrical parameters, a photovoltaic power generation power fitting model is constructed, by comparing a deviation of a real-time power generation power from a fitting power of the photovoltaic power generation power fitting model, a photovoltaic panel dusting state is determined, and a cleaning suggestion is generated based on a future weather forecast; S5: based on the electrical parameters, equipment state parameters and environmental parameters, a fault diagnosis model is constructed, and a photovoltaic panel state is evaluated and fault warned in real time; Step S4 comprises: S41: selecting a historical period in which the photovoltaic panel is in a clean state, and acquiring irradiance, outdoor temperature and photovoltaic power generation power of the selected period as modeling data; S42: data preprocessing is performed on the modeling data, including abnormal value identification and screening, missing value filling and data normalization, based on the preprocessed data, outdoor temperature, irradiance and photovoltaic power generation power are fitted to construct a photovoltaic power generation power fitting model; S43: real-time irradiance and outdoor temperature are acquired, and expected photovoltaic power generation power is calculated based on the photovoltaic power generation power fitting model; S44: the actual photovoltaic power generation power is compared with the expected photovoltaic power generation power, if the actual photovoltaic power generation power is continuously less than the expected photovoltaic power generation power and the relative deviation exceeds a preset deviation threshold, the photovoltaic panel is determined to be dusty, and the dusting degree is calculated; S45: based on the dusting degree, predicted power generation, predicted power generation attenuation, electricity price and cleaning cost, the net income is quantified, and a cleaning suggestion is generated: if the expected income increment after cleaning is greater than the cleaning cost, a cleaning suggestion of suggesting cleaning is generated, otherwise, a cleaning suggestion of not suggesting cleaning is generated; The photovoltaic power generation power fitting model satisfies the relationship: ； wherein, , and denote fitting coefficients, denotes the expected photovoltaic power, denotes the irradiance, denotes the outdoor temperature, denotes the irradiance correction coefficient.

2. The method of claim 1, wherein the photovoltaic four-can gateway is managed by a gateway management server. Step S1 comprises: The electrical parameters include input and output voltages, input and output currents, input and output powers, active powers, reactive powers and power generations of the photovoltaic component and the inverter; The equipment state parameters include working modes and temperatures of the inverter, photovoltaic component states and photovoltaic switch states; The environmental parameters include wind speed, wind direction, light intensity, air pressure, humidity, outdoor temperature, direct radiation, scattering radiation, solar elevation angle, solar azimuth angle and weather condition.

3. The method of claim 1, wherein the photovoltaic four-can gateway is managed by a gateway management server. Step S2 comprises: S21: data preprocessing is performed on the electrical parameters, equipment state parameters and environmental parameters, and the data preprocessing includes data cleaning, missing value and abnormal value processing and data normalization; S22: based on the pre-processed data, feature extraction and construction are performed, which include extracting meteorological features from the environmental parameters, and constructing statistical quantities and signal analysis features including mean, root mean square value, peak value, maximum difference value, peak value index, waveform index, pulse index, margin index, skewness, kurtosis from the electrical parameters and meteorological features; S23: based on the results of feature extraction and construction, a long short-term memory network model is constructed and used for photovoltaic power prediction, and future photovoltaic power prediction values of several time granularities are output.

4. The method of claim 1, wherein the photovoltaic four-can gateway is managed by a gateway management server. Step S3 includes: S31: Obtain the photovoltaic power prediction value of the future period, calculate the fluctuation degree of the photovoltaic power prediction value, and obtain the rolling standard deviation of the photovoltaic prediction power , the relationship formula is satisfied , wherein indicates the size of the time window, indicates the current time point, indicates the photovoltaic power prediction value at a specific time point i, indicates the average value of the photovoltaic power prediction value from t+1 to t+w+1. S32: set the calculated fluctuation degree with a standard deviation threshold value comparing, and adaptively adjusting the perturbation voltage step of the MPPT algorithm according to the comparison result, wherein: if the degree of fluctuation is less than or equal to a standard deviation set threshold then a first perturbation voltage is applied adjusting the output voltage of the photovoltaic panel If the degree of fluctuation is greater than a standard deviation set threshold then a second perturbation voltage is applied adjusting the output voltage of the photovoltaic panel, the second perturbation voltage is less than the first perturbation voltage ; S33: after applying the perturbation voltage, the output power change of the photovoltaic system is measured, and the subsequent perturbation direction is adjusted based on the output power change: if the output power increases, the current perturbation direction is maintained; if the output power decreases, the perturbation direction is reversed.

5. The method of claim 1, wherein the photovoltaic four-can gateway is managed by a gateway management server. The dust accumulation degree satisfies the relationship: ； wherein, represents the expected photovoltaic power generation power, represents the photovoltaic actual power generation power; Power generation decay power .

6. The method of claim 5, wherein the method further comprises: the calculation satisfies the relationship: ； wherein, represents the power of the decay of the electricity generation, represents the value of the electricity quantity, represents the time.

7. The method of claim 1, wherein the photovoltaic four-can gateway is managed by a gateway management server. Step S5 includes: S51: collect historical fault label data corresponding to the photovoltaic, and perform data preprocessing on the operating state data, including abnormal value identification and screening, missing value filling, time series synchronization, and data normalization; S52: based on the pre-processed data, a derived feature reflecting the operating state of the photovoltaic panel is constructed, and a machine learning fault diagnosis model is constructed and trained based on the derived feature and the historical fault label data; S53: use the trained machine learning fault diagnosis model to perform real-time evaluation and fault warning on the photovoltaic panel state.

8. A management system for photovoltaic four-can gateways, characterized by, The photovoltaic four-capable gateway management system is applied to the photovoltaic four-capable gateway management method of any one of claims 1-7, and the photovoltaic four-capable gateway management system comprises: a data acquisition module for acquiring electrical parameters, equipment state parameters and environmental parameters of a photovoltaic power station; a photovoltaic short-term prediction module for constructing a photovoltaic power prediction model based on the electrical parameters, equipment state parameters and environmental parameters, and outputting photovoltaic power prediction values of a future period; an MPPT adaptive adjustment module for calculating the fluctuation degree of the photovoltaic power prediction value according to the photovoltaic power prediction value of the future period, and adaptively adjusting the perturbation step of the MPPT algorithm based on the fluctuation degree to track the maximum power point of the photovoltaic component; a photovoltaic panel operation and maintenance optimization module for constructing a photovoltaic power fitting model based on the environmental parameters and electrical parameters, determining the dust accumulation state of the photovoltaic panel by comparing the deviation between the real-time power generation and the fitting power of the photovoltaic power fitting model, and generating a cleaning suggestion based on future weather forecasts; a fault diagnosis module for constructing a fault diagnosis model based on the electrical parameters, equipment state parameters and environmental parameters, and performing real-time evaluation and fault warning on the photovoltaic panel state.

Images