Distributed photovoltaic power monitoring system and control method thereof

Abstract

The invention relates to the technical field of power systems, and discloses a distributed photovoltaic power monitoring system and a control method thereof. The system comprises a data acquisition module used for collecting initial operation data of distributed photovoltaic in real time and preprocessing the initial operation data to obtain target operation data; the data transmission module is used for receiving the target operation data and transmitting the target operation data to the data processing module in real time; the data processing module is used for performing power prediction on the target operation data based on the improved PC-LSTM model to obtain a distributed photovoltaic power prediction result; and the data processing module is also used for carrying out operation processing on the power prediction result by adopting a self-tuning PID control algorithm to obtain distributed photovoltaic inverter reference electrical data, generating an inverter adjustment instruction based on the inverter reference electrical data, and sending the inverter adjustment instruction to the power prediction module. Therefore, the inverter adjustment instruction is sent to the distributed photovoltaic system through the data transmission module for real-time control. According to the invention, the distributed photovoltaic monitoring accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a distributed photovoltaic power monitoring system and a control method thereof. Background Art

[0002] Currently, distributed photovoltaic systems have become a key component of the new power system. Although distributed photovoltaic systems are developing well, existing monitoring systems have exposed numerous technical bottlenecks when faced with large-scale access and complex operating conditions. First, at the data perception layer, existing monitoring terminals suffer from insufficient sampling accuracy and high packet loss rates. This results in timing discontinuities and noise interference in the raw data received by the cloud platform, seriously affecting data quality. Second, at the communication transmission layer, traditional 4G / Wi-Fi transmission solutions have a high probability of channel collisions in densely populated device access scenarios and lack end-to-end encryption mechanisms, making it difficult to meet security requirements. Third, at the intelligent analysis layer, existing LSTM-based prediction models violate physical rules, with a high probability of predicted values exceeding the theoretical power limit and the prediction results violating the inverse temperature-efficiency law. The model is highly sensitive to data, and when the training sample size or signal-to-noise ratio is low, the prediction error increases nonlinearly. Prediction volatility is high under cloud shadow effects or component degradation conditions.

[0003] For the aggregator management platform, existing technologies make it difficult to achieve spatiotemporal alignment and collaborative analysis of multi-power station data, resulting in high regional power forecast errors, which seriously affects the accuracy of virtual power plants' day-ahead market declarations and the AGC regulation pass rate.

[0004] Therefore, it is urgent to design a new distributed photovoltaic power monitoring system to meet the development requirements of distributed photovoltaic system scale and efficiency. Summary of the Invention

[0005] In view of the limitations of the prior art, the present invention provides a distributed photovoltaic power monitoring system and a control method thereof.

[0006] In a first aspect, an embodiment of the present invention provides a distributed photovoltaic power monitoring system, comprising: a data acquisition module, a data transmission module, and a data processing module;

[0007] The data acquisition module is used to collect the initial operation data of the distributed photovoltaic system in real time and pre-process the initial operation data to obtain target operation data;

[0008] The data transmission module is configured to receive the target operation data and transmit the target operation data to the data processing module in real time;

[0009] The data processing module is configured to perform power prediction on the target operating data based on an improved PC-LSTM model to obtain a power prediction result of the distributed photovoltaic system, wherein the improved PC-LSTM model includes an energy conservation constraint, a semiconductor characteristic constraint, and an environmental coupling constraint. The energy conservation constraint includes limiting the power output range of the photovoltaic module based on a first thermodynamic equation. The semiconductor characteristic constraint includes establishing a relationship between the current and voltage of the photovoltaic module based on the Shockley equation. The environmental coupling constraint includes establishing a relationship between irradiance, temperature, and the output power of the photovoltaic module based on a second thermodynamic equation.

[0010] The data processing module is further used to use a self-tuning PID control algorithm to perform calculations on the power prediction results to obtain the inverter reference electrical data of the distributed photovoltaic, and generate inverter adjustment instructions based on the inverter reference electrical parameters, so that the inverter adjustment instructions are sent to the distributed photovoltaic through the data transmission module for real-time control.

[0011] Preferably, the data processing module is further used to perform fault diagnosis on the target operation data based on the improved PC-LSTM model to obtain a fault classification result of the distributed photovoltaic system, wherein the fault classification result includes a fault type and a fault probability.

[0012] Preferably, the improved PC-LSTM model includes an input layer, a feature selection layer, an LSTM layer, a physical constraint layer, a shared feature encoding layer and an output layer;

[0013] The input layer is configured to receive the target operating data and format the target operating data to obtain first target operating data in a tensor form;

[0014] The feature selection layer is used to perform feature screening on the first target operation data to obtain second target operation data without redundant variables;

[0015] The feature selection layer is further used to perform feature space compression on the second target operation data to obtain third target operation data;

[0016] The LSTM layer is used to extract time series features from the third target operation data to obtain fourth target operation data;

[0017] The physical constraint layer is used to fuse the fourth target operation data with the energy conservation constraint, the semiconductor characteristic constraint, and the environmental coupling constraint to obtain fifth target operation data;

[0018] The shared feature encoding layer is used to extract and encode features of the fifth target operation data to obtain shared operation feature data;

[0019] The output layer is used to generate corresponding distributed photovoltaic operation decision-related data based on the shared operation characteristic data according to task requirements, wherein the task requirements include power prediction tasks, fault diagnosis tasks and inverter control tasks, and the distributed photovoltaic operation decision-related data includes power prediction results, fault classification results and inverter adjustment instructions.

[0020] Preferably, the LSTM layer is composed of a plurality of dual-channel LSTM units, each of which includes a physical feature channel, an electrical feature channel and a feature crossover mechanism;

[0021] The physical feature channel is used to receive the first physical feature data in the third target operation data, and perform time series feature extraction on the first physical feature data to obtain second physical feature data;

[0022] The electrical characteristic channel is used to receive the first electrical characteristic data in the third target operation data, and perform time series feature extraction on the first electrical characteristic data to obtain second electrical characteristic data;

[0023] The feature crossover mechanism is used to dynamically weight the second physical feature data and the second electrical feature data according to a preset weight update frequency to obtain the output data of the dual-channel LSTM unit.

[0024] Preferably, the energy conservation constraint is characterized by the following formula:

[0025]

[0026] Among them, P pred Denotes the predicted power, G t represents irradiance, A represents the area of photovoltaic panels, η ttheory represents the theoretical conversion efficiency, β represents the temperature coefficient, T pv Indicates the temperature of the photovoltaic module, T STC Indicates photovoltaic standard test condition temperature;

[0027] The semiconductor characteristic constraints are characterized by the following formula:

[0028]

[0029] Among them, I pred Represents the predicted current, I ph represents the photocurrent, I0 represents the reverse saturation current of the diode, q represents the amount of electron charge, V represents the voltage applied across the pn junction, I represents the current passing through the pn junction, and R srepresents the series resistance, n represents the emission coefficient, k represents the Boltzmann constant, T represents the absolute temperature, R sh represents parallel resistance;

[0030] The following formula is used to characterize the environmental coupling constraint:

[0031]

[0032] Among them, G represents irradiance, α represents temperature coefficient, η represents photoelectric conversion efficiency, C th represents heat capacity, T amb Indicates the ambient temperature, R th Indicates thermal resistance.

[0033] Preferably, the power prediction result is calculated using the following formula:

[0034]

[0035] in, Represents the power prediction result for the future period, W p represents the weight matrix of the output layer, ELU represents the exponential linear unit activation function, W1 and W2 represent the weight matrices of the fully connected layer, b1 and b2 represent the bias terms of the fully connected layer, and h t Indicates shared operation characteristic data, b p Represents the bias term of the output layer.

[0036] Preferably, the data processing module includes:

[0037] An operation processing unit, configured to perform a linear operation on the power prediction result using a self-tuning PID control algorithm to obtain an inverter reference voltage and an inverter reference current of the distributed photovoltaic system;

[0038] an instruction generating unit, configured to generate an inverter adjustment instruction based on the inverter reference voltage and the inverter reference current;

[0039] An instruction sending unit is used to send the inverter adjustment instruction to the distributed photovoltaic system through the data transmission module for real-time control.

[0040] Preferably, the data processing module includes:

[0041] a spatiotemporal feature extraction unit, configured to extract spatiotemporal features from the shared operation feature data to obtain spatiotemporal feature data;

[0042] an attention weighting unit, configured to perform weighted processing on the spatiotemporal feature data through an attention mechanism to obtain weighted feature data;

[0043] A fault classification unit is used to perform time series feature extraction and classification mapping on the weighted feature data to obtain the fault probability of the distributed photovoltaic system in each fault type, and determine the corresponding fault type based on the fault probability.

[0044] Preferably, the data processing module is deployed in the cloud, and the data processing module performs remote real-time monitoring and management of the distributed photovoltaic system through cloud computing technology.

[0045] In a second aspect, an embodiment of the present invention provides a distributed photovoltaic power monitoring system control method, which is applied to the distributed photovoltaic power monitoring system as described above, and includes:

[0046] The data acquisition module collects the initial operation data of the distributed photovoltaic system in real time, and pre-processes the initial operation data to obtain target operation data;

[0047] receiving the target operation data through the data transmission module, and transmitting the target operation data to the data processing module in real time;

[0048] Performing power prediction on the target operating data based on an improved PC-LSTM model by the data processing module to obtain a power prediction result of the distributed photovoltaic system, wherein the improved PC-LSTM model includes an energy conservation constraint, a semiconductor characteristic constraint, and an environmental coupling constraint. The energy conservation constraint includes limiting the power output range of the photovoltaic module based on a first thermodynamic equation. The semiconductor characteristic constraint includes establishing a relationship between the current and voltage of the photovoltaic module based on the Shockley equation. The environmental coupling constraint includes establishing a relationship between irradiance, temperature, and the output power of the photovoltaic module based on a second thermodynamic equation.

[0049] The power prediction result is processed by the data processing module using a self-tuning PID control algorithm to obtain the inverter reference electrical data of the distributed photovoltaic, and an inverter adjustment instruction is generated based on the inverter reference electrical parameters, so that the inverter adjustment instruction is sent to the distributed photovoltaic through the data transmission module for real-time control.

[0050] Compared with the existing technology, the distributed photovoltaic power monitoring system and control method of the embodiment of the present invention have the following beneficial effects: through high-precision data acquisition and intelligent control, the power generation efficiency and stability of distributed photovoltaics are significantly improved; the modular design makes the monitoring system have good scalability and maintainability, reducing the subsequent operation and maintenance costs; the use of deep learning algorithms realizes accurate prediction and optimized control of distributed photovoltaic performance, improving power generation efficiency; based on cloud computing technology, remote monitoring and intelligent operation and maintenance of distributed photovoltaics are realized, reducing the difficulty and cost of operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic structural diagram of a distributed photovoltaic power monitoring system according to an embodiment of the present invention;

[0052] Figure 2 is a schematic structural diagram of a data processing module according to an embodiment of the present invention;

[0053] Figure 3 is another structural diagram of a data processing module according to an embodiment of the present invention;

[0054] Figure 4 The present invention is a flow chart of a distributed photovoltaic power monitoring system control method. DETAILED DESCRIPTION

[0055] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0056] In the description of the present invention, it should be understood that the terms "first" and "second" etc. are used in the present invention to distinguish different objects rather than to describe a specific order.

[0057] In describing the present invention, it should be noted that, unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. Those skilled in the art will understand the specific meanings of the above terms in the present invention in specific circumstances.

[0058] like Figure 1 As shown, an embodiment of the present invention provides a distributed photovoltaic power monitoring system, including: a data acquisition module 1, a data transmission module 2 and a data processing module 3.

[0059] The data acquisition module is used to collect the initial operation data of distributed photovoltaic in real time, and pre-process the initial operation data to obtain the target operation data.

[0060] Specifically, high-precision sensors are installed at every key location in the distributed photovoltaic system, such as photovoltaic panels, inverters, and combiner boxes, to collect key parameters such as voltage, current, power, and temperature. These data constitute the initial operational data for the distributed photovoltaic system. Furthermore, industrial-grade Hall sensors are used to collect voltage and current, digital irradiance sensors and temperature sensors are integrated, contact relays are used to monitor circuit breaker status, and LED optocouplers are used for isolation to achieve switch value acquisition.

[0061] Furthermore, the data acquisition module pre-processes the collected initial operation data, including data cleaning, format conversion, etc., to obtain target operation data, thereby ensuring the accuracy and consistency of the data.

[0062] The data transmission module is used to receive target operation data and transmit the target operation data to the data processing module in real time.

[0063] Specifically, the data transmission module compresses and encrypts the target operation data and transmits it to the data processing module through a stable and reliable communication channel. Furthermore, a stable data transmission network is built based on 4G technology to transmit the received target operation data to the data processing module in real time. The main channel uses a 4G DTU module with a built-in breakpoint resume function. In the event of a network interruption, 72 hours of data can be locally cached. The backup channel is equipped with an RS-485 interface, supports up to 31 nodes in cascade, and has a communication distance of 1200 meters. It uses a custom frame format (synchronization header + length field + CRC16 check) and a bit error rate of less than 10^-6.

[0064] The data processing module is used to perform power prediction on the target operating data based on the improved PC-LSTM model to obtain the power prediction results of distributed photovoltaics.

[0065] It should be noted that the LSTM network, as a special recurrent neural network, performs well in photovoltaic power generation forecasting. Its characteristics are that it can handle long-term dependencies in time series data and retain previous information through an internal self-loop structure, thereby establishing temporal correlation between time series data, which provides strong support for the accurate prediction of photovoltaic power generation.

[0066] The PC-LSTM model is an innovative model that incorporates physical constraints on the LSTM network. By extracting physical constraints from domain knowledge and physical laws in the photovoltaic field and integrating them into the LSTM model, the PC-LSTM model overcomes the limitations of machine learning algorithms trained solely on large amounts of data, improving prediction accuracy and robustness. Furthermore, through the design of modules such as data filtering, cropping, and loss function reconstruction, the rationality and reliability of the model output are further ensured.

[0067] Specifically, the improved PC-LSTM model includes energy conservation constraints, semiconductor property constraints, and environmental coupling constraints. The energy conservation constraint limits the power output range of the PV module based on the first thermodynamic equation. The semiconductor property constraint establishes the relationship between the current and voltage of the PV module based on the Shockley equation. The environmental coupling constraint establishes the relationship between irradiance, temperature, and the output power of the PV module based on the second thermodynamic equation.

[0068] The improved PC-LSTM model of the present invention is described in detail below:

[0069] The improved PC-LSTM model includes input layer, feature selection layer, LSTM layer, physical constraint layer, shared feature encoding layer and output layer.

[0070] An input layer, configured to receive target operating data and format the target operating data to obtain first target operating data in a tensor form;

[0071] Specifically, the target operation data of this embodiment is a time series feature vector with 12 dimensions, namely:

[0072]

[0073] Among them, X t Indicates target operation data, G t represents irradiance, T pv Indicates the temperature of the photovoltaic module, V oc Indicates the open circuit voltage, I sc Indicates short-circuit current, V mp Indicates the maximum power point voltage, I mp Represents the maximum power point current, RH represents the relative humidity, W s Indicates wind speed, T amb Indicates the ambient temperature, SoC indicates the state of charge of the energy storage system, P load represents the load power, and Δt represents the time interval.

[0074] Furthermore, the input layer formats the time series feature vector and converts it into a tensor form suitable for model processing.

[0075] A feature selection layer is used to perform feature screening on the first target operating data to obtain the second target operating data after eliminating redundant variables;

[0076] Specifically, a feature correlation matrix is constructed based on the photovoltaic system differential equations. Redundant variables with correlation coefficients less than a preset threshold are removed to obtain the second target operating data without redundant variables. It can be understood that the feature selection layer selects features based on the statistical properties of the data itself, which conforms to the characteristics of filter-based feature selection.

[0077] The feature selection layer is further used to perform feature space compression on the second target operation data to obtain third target operation data;

[0078] Specifically, a modified mutual information method combined with L1 regularization is used to achieve feature dimension compression, resulting in the third target operational data. It can be understood that the wrapping process combines the feature selection process with the learning algorithm, evaluating the quality of feature subsets based on model performance. The feature selection layer uses a modified mutual information method combined with L1 regularization to achieve feature dimension compression, aiming to improve model performance, consistent with the characteristics of wrapper feature selection.

[0079] It should be noted that the feature selection layer adopts a two-stage hybrid method combined with filtering and packaging processes, which can optimize highly correlated feature vectors and improve the prediction performance of the model.

[0080] The LSTM layer is used to extract time series features from the third target operation data to obtain the fourth target operation data;

[0081] Specifically, the LSTM layer consists of several dual-channel LSTM units, each of which includes a physical feature channel, an electrical feature channel, and a feature crossover mechanism.

[0082] a physical feature channel, configured to receive first physical feature data in the third target operation data, and perform time series feature extraction on the first physical feature data to obtain second physical feature data;

[0083] Specifically, the physical feature channel processes physical quantities such as irradiance and photovoltaic module temperature.

[0084] an electrical characteristic channel, configured to receive first electrical characteristic data in the third target operation data, and perform time series feature extraction on the first electrical characteristic data to obtain second electrical characteristic data;

[0085] Specifically, the electrical characteristic channel processes electrical parameters such as voltage and current.

[0086] The feature crossover mechanism is used to dynamically weight the second physical feature data and the second electrical feature data according to a preset weight update frequency to obtain the output data of the dual-channel LSTM unit.

[0087] Specifically, the feature cross-mechanism realizes the dynamic weighted fusion of dual-channel features through the attention layer. The preset weight update frequency in this embodiment is 10 seconds.

[0088] The physical constraint layer is used to fuse the fourth target operation data with energy conservation constraints, semiconductor characteristic constraints, and environmental coupling constraints to obtain the fifth target operation data;

[0089] The physical constraint layer fuses the pre-defined physical constraints with the feature information output by the LSTM layer, and adjusts the output of the model through the constraints to make it conform to the physical laws.

[0090] Specifically, the present invention constructs a photovoltaic system physical knowledge base that integrates three types of constraints:

[0091] 1) Energy conservation constraint: The power output range is limited by the thermodynamic equations of photovoltaic modules;

[0092] 2) Semiconductor characteristic constraints: Establish the current-voltage relationship boundary conditions based on the pn junction equation;

[0093] 3) Environmental coupling constraint: The irradiance-temperature-output power transfer function is introduced as a dynamic threshold.

[0094] Furthermore, the energy conservation constraint is characterized by the following formula:

[0095]

[0096] Among them, P pred Denotes the predicted power, G t represents irradiance, A represents the area of photovoltaic panels, η ttheory represents the theoretical conversion efficiency, β represents the temperature coefficient, T pv Indicates the temperature of the photovoltaic module, T STC Indicates photovoltaic standard test condition temperature.

[0097] It should be noted that the energy conservation constraint, on the one hand, limits the predicted power to not exceed the theoretical maximum value, and on the other hand, corrects the prediction deviation caused by the abnormality of the irradiance sensor.

[0098] Furthermore, the following formula is used to characterize the semiconductor characteristic constraints:

[0099]

[0100] Among them, I pred Represents the predicted current, I ph represents the photocurrent, I0 represents the reverse saturation current of the diode, q represents the amount of electron charge, V represents the voltage applied across the pn junction, I represents the current passing through the pn junction, and R s represents the series resistance, n represents the emission coefficient, k represents the Boltzmann constant, T represents the absolute temperature, R sh Represents parallel resistors.

[0101] It should be noted that semiconductor characteristic constraints ensure that the predicted current-voltage curve conforms to the pn junction characteristics while preventing physically infeasible electrical parameter combinations. The pn junction equation, also known as the Shockley equation, is the fundamental equation describing the current-voltage characteristics of a pn junction.

[0102] Furthermore, the following formula is used to characterize the environmental coupling constraint:

[0103]

[0104] Among them, G represents irradiance, α represents temperature coefficient, η represents photoelectric conversion efficiency, C th represents heat capacity, T amb Indicates the ambient temperature, R th Indicates thermal resistance.

[0105] It should be noted that the environmental coupling constraint, on the one hand, establishes a dynamic correlation between irradiance, temperature, and power, and on the other hand, improves the prediction stability under cloudy weather conditions.

[0106] a shared feature encoding layer, configured to extract and encode features of the fifth target operation data to obtain shared operation feature data;

[0107] The shared feature encoding layer extracts and encodes features from the input data, converting the selected data into a higher-level, more representative feature representation. Specifically, the shared feature encoding layer further integrates and abstracts the features learned by the previous LSTM layer, providing a unified feature foundation for subsequent specialized task heads with different task requirements. Each task head then processes these features to produce the output layer's results.

[0108] The output layer is used to generate corresponding distributed photovoltaic operation decision-making related data based on the shared operation characteristic data according to task requirements.

[0109] Specifically, the task requirements include power prediction, fault diagnosis, and inverter control. Based on the task requirements, the output layer generates the corresponding distributed photovoltaic operation decision-making data through corresponding activation functions and calculation methods.

[0110] The data processing module is used to perform power prediction on the target operating data based on the improved PC-LSTM model to obtain the power prediction results of distributed photovoltaics.

[0111] Specifically, the power prediction result is calculated using the following formula:

[0112]

[0113] in, Represents the power prediction result for the future period, Wp represents the weight matrix of the output layer, ELU represents the exponential linear unit activation function, W1 and W2 represent the weight matrices of the fully connected layer, b1 and b2 represent the bias terms of the fully connected layer, and h t Indicates shared operation characteristic data, b p Represents the bias term of the output layer.

[0114] Furthermore, the following formula is used to characterize the exponential linear unit activation function:

[0115]

[0116] Here, x represents the input value, and α represents the hyperparameter, which controls the curve shape of the activation function when x < 0 and affects the output value of the function.

[0117] The data processing module is also used to use the self-tuning PID control algorithm to calculate and process the power prediction results to obtain the inverter reference electrical data of the distributed photovoltaic, and generate inverter adjustment instructions based on the inverter reference electrical parameters, so that the inverter adjustment instructions are sent to the distributed photovoltaic through the data transmission module for real-time control.

[0118] Specifically, if Figure 2 As shown, the data processing module includes:

[0119] The processing unit 31 is used to perform linear operations on the power prediction results using a self-tuning PID control algorithm to obtain a reference voltage and a reference current of the inverter of the distributed photovoltaic system;

[0120] In the self-tuning PID control algorithm, the control quantity is determined by linearly combining the calculation results of the proportional, integral, and differential links, taking into account the current deviation, deviation accumulation, and deviation change trend, so as to achieve the adjustment of the inverter reference voltage and current.

[0121] Specifically, the control amount of the inverter reference voltage is calculated using the following formula:

[0122]

[0123] Where, ΔV ref Indicates the control quantity of the inverter reference voltage, K p Represents the proportionality coefficient, K i Indicates the integral coefficient, K d represents the differential coefficient, and e(t) represents the power deviation.

[0124] The instruction generating unit 32 is used to generate an inverter adjustment instruction based on the inverter reference voltage and the inverter reference current;

[0125] The instruction sending unit 33 is used to send the inverter adjustment instruction to the distributed photovoltaic system through the data transmission module for real-time control.

[0126] By controlling the inverter's output voltage and current in real time, power generation efficiency can be optimized or potential failures can be avoided.

[0127] The data processing module is also used to perform fault diagnosis on the target operation data based on the improved PC-LSTM model to obtain the fault classification results of distributed photovoltaics.

[0128] Specifically, if Figure 3 As shown, the data processing module includes:

[0129] The spatiotemporal feature extraction unit 34 is used to extract spatiotemporal features from the shared operation feature data to obtain spatiotemporal feature data;

[0130] One-dimensional convolution is used to extract features from shared running feature data, in which a nonlinear activation function is introduced to enhance the model's ability to express data features. The maximum pooling operation further extracts the main features and reduces the data dimension, reducing the amount of calculation to obtain spatiotemporal feature data.

[0131] an attention weighting unit 35 for performing weighted processing on the spatiotemporal feature data through an attention mechanism to obtain weighted feature data;

[0132] The spatiotemporal feature data is weighted through the attention mechanism to obtain weighted feature data, which enables the model to focus on important features and improve its sensitivity to key fault features.

[0133] The fault classification unit 36 is used to perform time series feature extraction and classification mapping on the weighted feature data to obtain the fault probability of the distributed photovoltaic system in each fault type, and determine the corresponding fault type based on the fault probability.

[0134] Specifically, the fault classification results include fault type and fault probability. This embodiment inputs weighted feature data into an LSTM network to further learn the temporal dependencies between features. After processing through a fully connected layer, the output is converted into a probability distribution for each of the 14 fault categories using a Softmax function. This determines the fault type and implements the classification of the 14 typical fault categories.

[0135] The data processing module is deployed in the cloud, and uses cloud computing technology to remotely monitor and manage distributed photovoltaics in real time.

[0136] Specifically, the data processing module combines cloud computing and edge computing technologies to enable remote, real-time monitoring and intelligent management of distributed photovoltaic systems. It's understandable that the data processing module is essentially a cloud computing platform. Operations and maintenance personnel can not only view real-time information such as the operating status and power generation efficiency of distributed photovoltaic systems through the platform, but also gain an intuitive understanding of the on-site environment of the photovoltaic power station through 3D visualization technology. Furthermore, the platform integrates deep learning algorithms, enabling immediate warning notifications when faults occur and automatically performing fault diagnosis and location, significantly reducing operation and maintenance costs and workload. Furthermore, the platform also features automated operation and preventive maintenance capabilities, intelligently generating maintenance plans and recommendations based on the equipment's operating status and historical data, further extending equipment life and improving power generation efficiency.

[0137] The embodiment of the present invention provides a distributed photovoltaic power monitoring system, which significantly improves the power generation efficiency and stability of distributed photovoltaics through high-precision data acquisition and intelligent control. The modular design makes the monitoring system scalable and maintainable, reducing the subsequent operation and maintenance costs. The deep learning algorithm is used to achieve accurate prediction and optimized control of distributed photovoltaic performance, thereby improving power generation efficiency. Cloud computing technology is used to achieve remote monitoring and intelligent operation and maintenance of distributed photovoltaics, reducing the difficulty and cost of operation and maintenance.

[0138] like Figure 4 As shown, an embodiment of the present invention provides a distributed photovoltaic power monitoring system control method, which is applied to the distributed photovoltaic power monitoring system as described above, and includes the following steps:

[0139] S1. Collect the initial operating data of the distributed photovoltaic system in real time through the data acquisition module, and pre-process the initial operating data to obtain the target operating data;

[0140] S2. Receive target operation data through the data transmission module and transmit the target operation data to the data processing module in real time;

[0141] S3. Power prediction of target operation data is performed based on the improved PC-LSTM model through the data processing module to obtain the power prediction result of distributed photovoltaics;

[0142] The improved PC-LSTM model includes energy conservation constraints, semiconductor characteristic constraints and environmental coupling constraints. The energy conservation constraint includes limiting the power output range of the photovoltaic module based on the first thermodynamic equation. The semiconductor characteristic constraint includes establishing the relationship between the current and voltage of the photovoltaic module based on the Shockley equation. The environmental coupling constraint includes establishing the relationship between irradiance, temperature and the output power of the photovoltaic module based on the second thermodynamic equation.

[0143] S4. The power prediction result is processed by the data processing module using the self-tuning PID control algorithm to obtain the inverter reference electrical data of the distributed photovoltaic, and the inverter adjustment instruction is generated based on the inverter reference electrical parameters, so that the inverter adjustment instruction is sent to the distributed photovoltaic through the data transmission module for real-time control.

[0144] It should be noted that, for the specific definition of a distributed photovoltaic power monitoring system control method, please refer to the above definition of a distributed photovoltaic power monitoring system. The two have the same functions and effects and will not be repeated here.

[0145] In summary, the embodiments of the present invention provide a distributed photovoltaic power monitoring system and a control method thereof, which significantly improve the power generation efficiency and stability of distributed photovoltaics through high-precision data acquisition and intelligent control; the modular design makes the monitoring system have good scalability and maintainability, reducing the subsequent operation and maintenance costs; the use of deep learning algorithms realizes accurate prediction and optimized control of distributed photovoltaic performance, thereby improving power generation efficiency; based on cloud computing technology, remote monitoring and intelligent operation and maintenance of distributed photovoltaics are realized, reducing the difficulty and cost of operation and maintenance.

[0146] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the method embodiment, since it is basically similar to the system embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the system embodiment. It should be noted that the various technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0147] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention. These improvements and substitutions should also be regarded as the scope of protection of the present invention.

Claims

1. A distributed photovoltaic power monitoring system, characterized in that: include: Data acquisition module, data transmission module and data processing module; The data acquisition module is used to collect the initial operation data of the distributed photovoltaic system in real time and pre-process the initial operation data to obtain target operation data; The data transmission module is used to receive the target operation data and transmit the target operation data to the data processing module in real time; The data processing module is configured to perform power prediction on the target operating data based on an improved PC-LSTM model to obtain a power prediction result of the distributed photovoltaic system, wherein the improved PC-LSTM model includes an energy conservation constraint, a semiconductor characteristic constraint, and an environmental coupling constraint. The energy conservation constraint includes limiting the power output range of the photovoltaic module based on a first thermodynamic equation. The semiconductor characteristic constraint includes establishing a relationship between the current and voltage of the photovoltaic module based on the Shockley equation. The environmental coupling constraint includes establishing a relationship between irradiance, temperature, and the output power of the photovoltaic module based on a second thermodynamic equation. The data processing module is further used to use a self-tuning PID control algorithm to perform calculations on the power prediction results to obtain the inverter reference electrical data of the distributed photovoltaic, and generate inverter adjustment instructions based on the inverter reference electrical parameters, so that the inverter adjustment instructions are sent to the distributed photovoltaic through the data transmission module for real-time control.

2. The distributed photovoltaic power monitoring system according to claim 1, characterized in that: The data processing module is further used to perform fault diagnosis on the target operation data based on the improved PC-LSTM model to obtain a fault classification result of the distributed photovoltaic system, wherein the fault classification result includes a fault type and a fault probability.

3. The distributed photovoltaic power monitoring system according to claim 2, characterized in that: The improved PC-LSTM model includes an input layer, a feature selection layer, an LSTM layer, a physical constraint layer, a shared feature encoding layer and an output layer; The input layer is configured to receive the target operating data and format the target operating data to obtain first target operating data in a tensor form; The feature selection layer is used to perform feature screening on the first target operation data to obtain second target operation data without redundant variables; The feature selection layer is further used to perform feature space compression on the second target operation data to obtain third target operation data; The LSTM layer is used to extract time series features from the third target operation data to obtain fourth target operation data; The physical constraint layer is used to fuse the fourth target operation data with the energy conservation constraint, the semiconductor characteristic constraint, and the environmental coupling constraint to obtain fifth target operation data; The shared feature encoding layer is used to extract and encode features of the fifth target operation data to obtain shared operation feature data; The output layer is used to generate corresponding distributed photovoltaic operation decision-related data based on the shared operation characteristic data according to task requirements, wherein the task requirements include power prediction tasks, fault diagnosis tasks and inverter control tasks, and the distributed photovoltaic operation decision-related data includes power prediction results, fault classification results and inverter adjustment instructions.

4. The distributed photovoltaic power monitoring system according to claim 3, characterized in that: The LSTM layer is composed of a plurality of dual-channel LSTM units, each of which includes a physical feature channel, an electrical feature channel and a feature crossover mechanism; The physical feature channel is used to receive the first physical feature data in the third target operation data, and perform time series feature extraction on the first physical feature data to obtain second physical feature data; The electrical characteristic channel is used to receive the first electrical characteristic data in the third target operation data, and perform time series feature extraction on the first electrical characteristic data to obtain second electrical characteristic data; The feature crossover mechanism is used to dynamically weight the second physical feature data and the second electrical feature data according to a preset weight update frequency to obtain the output data of the dual-channel LSTM unit.

5. The distributed photovoltaic power monitoring system according to claim 1, characterized in that: The energy conservation constraint is characterized by the following formula: P pred ≤G t ×A×η theory ×(1-β(T pv -T STC ))P pred ≤G t ×A×η theory ×(1-β(T pv -T STC )) Among them, P pred Denotes the predicted power, G t represents irradiance, A represents the area of photovoltaic panels, η ttheory represents the theoretical conversion efficiency, β represents the temperature coefficient, T pv Indicates the temperature of the photovoltaic module, T STC Indicates photovoltaic standard test condition temperature; The semiconductor characteristic constraints are characterized by the following formula: Among them, I pred Represents the predicted current, I ph represents the photocurrent, I0 represents the reverse saturation current of the diode, q represents the amount of electron charge, V represents the voltage applied across the pn junction, I represents the current passing through the pn junction, and R s represents the series resistance, n represents the emission coefficient, k represents the Boltzmann constant, T represents the absolute temperature, R sh represents parallel resistance; The following formula is used to characterize the environmental coupling constraint: Among them, G represents irradiance, α represents temperature coefficient, η represents photoelectric conversion efficiency, C th represents heat capacity, T amb Indicates the ambient temperature, R th Indicates thermal resistance.

6. The distributed photovoltaic power monitoring system according to claim 3, characterized in that: The power prediction result is calculated using the following formula: in, Represents the power prediction result for the future period, W p represents the weight matrix of the output layer, ELU represents the exponential linear unit activation function, W1 and W2 represent the weight matrices of the fully connected layer, b1 and b2 represent the bias terms of the fully connected layer, and h t Indicates shared operation characteristic data, b p Represents the bias term of the output layer.

7. The distributed photovoltaic power monitoring system according to claim 1, characterized in that: The data processing module includes: An operation processing unit, configured to perform a linear operation on the power prediction result using a self-tuning PID control algorithm to obtain an inverter reference voltage and an inverter reference current of the distributed photovoltaic system; an instruction generating unit, configured to generate an inverter adjustment instruction based on the inverter reference voltage and the inverter reference current; An instruction sending unit is used to send the inverter adjustment instruction to the distributed photovoltaic system through the data transmission module for real-time control.

8. The distributed photovoltaic power monitoring system according to claim 3, characterized in that: The data processing module includes: a spatiotemporal feature extraction unit, configured to extract spatiotemporal features from the shared operation feature data to obtain spatiotemporal feature data; an attention weighting unit, configured to perform weighted processing on the spatiotemporal feature data through an attention mechanism to obtain weighted feature data; A fault classification unit is used to perform time series feature extraction and classification mapping on the weighted feature data to obtain the fault probability of the distributed photovoltaic system in each fault type, and determine the corresponding fault type based on the fault probability.

9. The distributed photovoltaic power monitoring system according to claim 1, characterized in that: The data processing module is deployed in the cloud, and the data processing module performs remote real-time monitoring and management of the distributed photovoltaic through cloud computing technology.

10. A distributed photovoltaic power monitoring system control method, characterized in that: The distributed photovoltaic power monitoring system control method is applied to the distributed photovoltaic power monitoring system according to any one of claims 1 to 9, and the distributed photovoltaic power monitoring system control method includes: The data acquisition module collects the initial operation data of the distributed photovoltaic system in real time, and pre-processes the initial operation data to obtain target operation data; receiving the target operation data through the data transmission module, and transmitting the target operation data to the data processing module in real time; Performing power prediction on the target operating data based on an improved PC-LSTM model by the data processing module to obtain a power prediction result of the distributed photovoltaic system, wherein the improved PC-LSTM model includes an energy conservation constraint, a semiconductor characteristic constraint, and an environmental coupling constraint. The energy conservation constraint includes limiting the power output range of the photovoltaic module based on a first thermodynamic equation. The semiconductor characteristic constraint includes establishing a relationship between the current and voltage of the photovoltaic module based on the Shockley equation. The environmental coupling constraint includes establishing a relationship between irradiance, temperature, and the output power of the photovoltaic module based on a second thermodynamic equation. The power prediction result is processed by the data processing module using a self-tuning PID control algorithm to obtain the inverter reference electrical data of the distributed photovoltaic, and an inverter adjustment instruction is generated based on the inverter reference electrical parameters, so that the inverter adjustment instruction is sent to the distributed photovoltaic through the data transmission module for real-time control.

Images