Remote intelligent control method and system for photovoltaic power station

Abstract

The invention discloses a photovoltaic power station remote intelligent management and control method and system, and belongs to the technical field of photovoltaic power station management and control. According to the method, historical operation data of key parameters of a photovoltaic power station system are collected, a prediction model of the photovoltaic power station system is established based on a neural network, and PID controller parameters are optimized by using a photovoltaic power station system adaptive control algorithm optimized by ITAE; historical data are mined through a deep learning model, and a self-adaptive PID control strategy is combined, so that intelligent remote control of the photovoltaic power station is realized, and the power generation efficiency and the operation and maintenance management level are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power station control, and in particular to a photovoltaic power station remote intelligent control method and system. Background Art

[0002] At present, the proportion of large-scale photovoltaic power stations in the power system is constantly increasing, and their safe and stable operation has an increasingly greater impact on the power grid. However, traditional photovoltaic power station control systems mainly use fixed PID control parameters, which cannot adjust the control strategy in time according to the dynamic changes in environmental conditions, resulting in delayed system response and affecting power generation efficiency.

[0003] Photovoltaic power generation is intermittent and volatile, and its output power varies with external conditions such as light intensity and ambient temperature. This poses a challenge to the precise control of power stations. At the same time, photovoltaic power stations need to match the load characteristics of the power grid to achieve grid-friendly grid-connected operation.

[0004] Therefore, how to improve the control accuracy and intelligence level of photovoltaic power stations and achieve efficient operation of power stations has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to provide a remote intelligent control method and system for photovoltaic power plants, which is used to solve the problems of delayed response and low control accuracy of traditional photovoltaic power plant control systems, and to achieve the purpose of intelligent remote control and efficient operation of photovoltaic power plants.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a remote intelligent control method for a photovoltaic power station, which is used to improve the power generation efficiency and operation and maintenance management level of the photovoltaic power station. The photovoltaic power station uses a PID controller to regulate the inverter power output and the photovoltaic array angle to achieve matching adjustment of the power output of the photovoltaic power station with the grid load.

[0007] The method comprises the following steps: Step 1: Collect historical operating data of key parameters of the photovoltaic power station system, where the key parameters include: inverter power output, photovoltaic array angle, light intensity, ambient temperature and grid load.

[0008] Step 2: Establish a prediction model for the photovoltaic power station system based on a neural network, and provide inverter power settings and optimal angles of the photovoltaic array under current illumination and load response according to the prediction model.

[0009] Step 3: Use the ITAE-optimized PV power station system adaptive control algorithm to optimize the proportional, integral, and differential coefficients of the PID controller. Based on the established prediction model and the optimized PID controller, adjust the inverter power output and PV array angle in real time.

[0010] Furthermore, the prediction model is composed of an input layer, a one-dimensional convolutional layer, a pooling layer, a bidirectional memory loop layer, a Dropout layer, a fully connected layer and an output layer which are connected in sequence.

[0011] The one-dimensional convolution layer extracts the time series characteristics of the historical operation data of the photovoltaic power station, including the periodic change characteristics of light intensity, ambient temperature and the dynamic change characteristics of the grid load; the pooling layer simplifies the data dimension after the one-dimensional convolution and passes the data to the bidirectional gated loop layer; the bidirectional gated loop layer uses the bidirectional GRU to model the long-term and short-term dynamic characteristics of the photovoltaic power generation system, extracts the time series correlation characteristics of power output and photovoltaic array angle adjustment, and performs nonlinear transformation to determine the intrinsic logical relationship between light, temperature, load input and inverter power output and photovoltaic array angle.

[0012] After the data is subjected to secondary feature extraction by the bidirectional memory loop layer, some nodes of the data are randomly discarded in the Dropout layer to prevent overfitting and the processed data is passed to the fully connected layer. The fully connected layer maps the output features to the label space of the sample through nonlinear combination, and the output layer outputs the preset power output adapted to the current state of the photovoltaic power station through linear regression operation. and preset PV array angle The predicted value of .

[0013] Further, the historical operation data is divided into a training set and a feature set. time, use represents the state of the input layer, represents the output layer state, represents the hidden layer state during the bidirectional GRU layer transmission process, then: ; In the formula, Represents the feature state learned by the forward layer during one-way transmission. It is the feature state learned by the backward layer during the one-way transmission process; and It is the weight of the forward layer, the backward layer and the output layer in sequence; represents the bias vector added in the output layer; the training set is imported into the neural network model to obtain the parameters of the network model.

[0014] Furthermore, the prediction results output by the prediction model of the photovoltaic power station system are compared with the measured data in the test set, and the goodness of fit index is used. To describe the predictive performance of the model, ,when When , the model training stop condition is met. At this time, the model parameters are the optimal parameters of the neural network model structure. If not, the back propagation algorithm is used to adjust the model deviation and continue the cycle.

[0015] Goodness of fit index The calculation formula is: ; In the formula, Representative The actual power output value of the samples, For the The power output prediction value corresponding to the samples is: is the average value of power output; Representative The actual PV array angle value of samples, For the The predicted value of the photovoltaic array angle corresponding to the sample, is the average value of the photovoltaic array angle; is the total number of samples; and is the weight coefficient and satisfies , used to balance the importance of power output and angle prediction, the default value is: , .

[0016] Furthermore, the control transfer function of the PID controller is: ; in, is the inverter power output after Laplace transformation, is the Laplace transform value of the PV array angle, and are the Laplace transform values ​​of power deviation and angle deviation respectively, is the time variable; , and are the proportional, integral and differential coefficients of the power control channel PID controller, , , ;in, is the relative deviation of inverter power output, It is the relative deviation of the PID controller stroke of the power control channel; , and They are the proportional, integral, and derivative coefficients of the PID controller for the angle control channel, respectively; , , ; where is the relative deviation of angle control, is the relative deviation of the stroke of the PID controller for the angle control channel.

[0017] Furthermore, select index as the fitness function of the optimization algorithm , where is the simulation time.

[0018] Furthermore, compare the ITAE index values calculated from the system states at the previous moment S-1 and the current moment S and make a comparison. The ITAE index value at the previous moment is denoted as , and the ITAE index value at the current moment is denoted as .

[0019] When ITAE(S-1) ≥ ITAE(S), it indicates that the system performance is improving. At this time, keep the current PID controller parameters unchanged; according to the relative deviations of the controlled variables in the power and angle channels and the relative deviation of the PID controller stroke at the current moment; adjust the power output of the inverter and the angle of the photovoltaic array to make them reach the preset power output and the preset angle of the photovoltaic array .

[0020] When ITAE(S-1) < ITAE(S), it indicates that the system performance is deteriorating. At this time, restore the relative deviations of the controlled variables in the power and angle channels and the relative deviation of the PID controller stroke to the values at the previous moment, and recalculate the PID controller parameters; adjust the power output of the inverter and the angle of the photovoltaic array to make them reach the preset power output and the preset angle of the photovoltaic array .

[0021] Furthermore, the method further includes the following steps: Step 4, as the operation data accumulates continuously, regularly update the prediction model parameters and PID controller parameters; the frequency of the regular update is maintained at 15 - 30 days.

[0022] Furthermore, step 1 also includes preprocessing the historical operation data through a mean filtering method to remove noise and outliers; for each point in the historical operation data, take the average of this point and its two adjacent points as the output.

[0023] The specific method of the mean filtering is as follows:

[0024] in is the number of points in the filter window, N=5, the original value The first The value of a point.

[0025] In a second aspect, the present invention provides a remote intelligent management and control system for a photovoltaic power station, which is used to execute the method of the first aspect. The system includes: a data acquisition module, a data processing module and a parameter adjustment module connected in sequence.

[0026] The data acquisition module is used to collect historical operating data of key parameters of the photovoltaic power station system, and the key parameters include: inverter power output, photovoltaic array angle, light intensity, ambient temperature and grid load.

[0027] The data processing module is used to establish a prediction model for the photovoltaic power station system based on a neural network, and to provide inverter power settings and optimal angles of the photovoltaic array under current illumination and load response according to the prediction model.

[0028] The parameter adjustment module is used to optimize the proportional, integral and differential coefficients of the PID controller using the ITAE optimized photovoltaic power station system adaptive control algorithm, and adjust the inverter power output and photovoltaic array angle in real time based on the established prediction model and the optimized PID controller.

[0029] Furthermore, the system also includes: The remote communication unit is used to realize the remote transmission of system data, remote monitoring of system status and remote alarm of abnormal situations.

[0030] The data storage unit is used to store the historical operation data of the photovoltaic power station system, the neural network model parameters and the PID controller parameters.

[0031] The human-computer interaction unit includes a display module and a control module; wherein the display module is used to display the power generation power, array angle, environmental data and system operation status of the photovoltaic power station in real time; and the control module is used to receive the operator's control instructions and adjust the system operation parameters.

[0032] The monitoring execution unit is used to monitor the key parameters of the system and perform exception processing at a preset frequency, and the monitoring frequency shall not be less than once every 5 minutes; when an abnormality is detected, an alarm message is sent to the remote communication unit.

[0033] Compared with the prior art, the present invention has the following beneficial effects: The present invention adopts a prediction model based on a neural network and an adaptive PID control strategy optimized by ITAE, which can dynamically optimize system control parameters according to real-time conditions such as light intensity and ambient temperature, thereby improving the power generation efficiency and control accuracy of the photovoltaic power station; realizing intelligent remote monitoring and management of the photovoltaic power station, improving operation and maintenance efficiency, and reducing operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of a remote intelligent control method for a photovoltaic power station of the present invention; Figure 2 The figure is a schematic diagram of the composition of a remote intelligent management and control system for a photovoltaic power station of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] It should be noted that the present invention deploys a distributed data acquisition system on-site in the photovoltaic power station. In terms of inverter power output acquisition, a power meter with an accuracy of 0.2 and a sampling frequency of 100Hz is selected to ensure that the rapid fluctuation characteristics of power can be captured. The photovoltaic array angle adopts a high-precision angle sensor with a resolution of 0.1° to monitor the inclination changes of the array in real time. The light intensity is measured using a professional photovoltaic meteorological station, which includes 12 light meters distributed in different areas of the power station with a sampling frequency of 1Hz to form light intensity spatial distribution data; in addition to collecting air temperature, the ambient temperature is also equipped with an infrared thermometer to monitor the surface temperature of the photovoltaic module to fully reflect the temperature field distribution; the grid load data is obtained through a real-time communication interface with the grid dispatching system, including electrical parameters such as active power and reactive power.

[0037] Example 1 like Figure 1 As shown, it is a flow chart of a remote intelligent control method of a photovoltaic power station of the present invention, which is used to improve the power generation efficiency and operation and maintenance management level of the photovoltaic power station. The photovoltaic power station uses a PID controller to regulate the inverter power output and the photovoltaic array angle to achieve matching adjustment of the power output of the photovoltaic power station and the grid load.

[0038] The method comprises the following steps: Step 1: Collect historical operating data of key parameters of the photovoltaic power station system, where the key parameters include: inverter power output, photovoltaic array angle, light intensity, ambient temperature and grid load.

[0039] By configuring the data acquisition module, the key parameters of the photovoltaic power station operation process are recorded in real time at a sampling frequency of no less than once every 5 minutes; among them, the inverter power output is measured by a power meter, the photovoltaic array angle is obtained by an angle sensor, the light intensity is measured by a light meter, the ambient temperature is collected by a temperature sensor, and the grid load data is obtained from the grid dispatching system.

[0040] The present invention adopts a hierarchical and distributed data acquisition architecture. At the field level, each photovoltaic array is equipped with an independent data acquisition unit, including a high-precision light meter (accuracy ±1%, range 0-2000W / m²), a PT100 temperature sensor (accuracy ±0.1℃, range -30℃~+80℃), an angle sensor (accuracy ±0.1°, range 0-90°), etc. The inverter power output is measured using a 0.5S-level high-precision electric energy meter with a sampling frequency of 1Hz. All acquisition devices are connected to the local controller via the RS485 bus and communicate using the Modbus-RTU protocol to ensure the real-time and reliability of data transmission.

[0041] Before mean filtering, first perform outlier detection on the original data, and set the The data validity judgment criteria of the principle are: 1. Effective range of light intensity: 0-1200W / m²; 2. Effective range of ambient temperature: (μ is the average value of historical data, σ is the standard deviation); 3. Valid range of power data: 0-1.2 times installed capacity.

[0042] Data points that are out of range are marked as outliers. For continuous outliers, if the duration is less than 1 minute, they are corrected using linear interpolation; if the duration exceeds 1 minute, an alarm signal is triggered and manual confirmation is required.

[0043] The historical operation data is preprocessed by the mean filtering method to remove noise and outliers; for each point in the historical operation data, the average value of the point and its neighboring points, that is, the points within the filter window, is taken as the output.

[0044] The method of mean filtering is specifically as follows:

[0045] in is the number of points in the filter window, the original value The first The value of a point.

[0046] Select the filter window size N=5, that is, for each data point, take the average of the 5 points before and after it as the filtered value of the point. This method can effectively eliminate data noise caused by factors such as equipment jitter and electromagnetic interference. For example, the light intensity data collected 5 times at a certain moment is: 850 W / m², 890 W / m², 870 W / m², 860 W / m², 880 W / m². After the mean filter, the light intensity value at that moment is 870 W / m², which avoids the deviation caused by a single sampling.

[0047] In the mean filtering process, different sizes of filter windows are used for different types of data: for the light intensity data with faster changes, a small window of N=5 is selected to retain the fast-changing characteristics; for parameters with slower changes such as temperature, a large window of N=9 is used to further suppress noise; for the grid load data, a medium window of N=7 is used to balance the response speed and filtering effect; at the same time, an adaptive window mechanism is introduced. When a data mutation is detected (the deviation between two adjacent points exceeds the set threshold), the filter window size is automatically reduced to improve the system's response to the mutation; for example, when the light intensity suddenly drops from 1000 W / m² to 200 W / m², the window size is automatically adjusted from N=5 to N=3, allowing the system to quickly track this change.

[0048] Step 2: Establish a prediction model for the photovoltaic power station system based on a neural network, and provide inverter power settings and optimal angles of the photovoltaic array under current illumination and load response according to the prediction model.

[0049] The prediction model consists of an input layer, a one-dimensional convolutional layer, a pooling layer, a bidirectional memory loop layer, a Dropout layer, a fully connected layer, and an output layer connected in sequence.

[0050] The one-dimensional convolution layer extracts the time series characteristics of the historical operation data of the photovoltaic power station, including the periodic change characteristics of light intensity, ambient temperature and the dynamic change characteristics of the grid load; the pooling layer simplifies the data dimension after the one-dimensional convolution and passes the data to the bidirectional gated loop layer; the bidirectional gated loop layer uses the bidirectional GRU to model the long-term and short-term dynamic characteristics of the photovoltaic power generation system, extracts the time series correlation characteristics of power output and photovoltaic array angle adjustment, and performs nonlinear transformation to determine the intrinsic logical relationship between light, temperature, load input and inverter power output and photovoltaic array angle.

[0051] After the data is subjected to secondary feature extraction by the bidirectional memory loop layer, some nodes of the data are randomly discarded in the Dropout layer to prevent overfitting and the processed data is passed to the fully connected layer. The fully connected layer maps the output features to the label space of the sample through nonlinear combination, and the output layer outputs the preset power output adapted to the current state of the photovoltaic power station through linear regression operation. and preset PV array angle The predicted value of .

[0052] The specific structural parameters of the neural network model are as follows: Input layer: receives 5×24 dimensional time series data (5 parameters, 24 historical moments); One-dimensional convolution layer: contains three layers of cascade structure: First layer: 64 convolution kernels, kernel size 3, step size 1, ReLU activation; Second layer: 128 convolution kernels, kernel size 3, step size 1, ReLU activation; Third layer: 256 convolution kernels, kernel size 3, step size 1, ReLU activation; Pooling layer: uses maximum pooling, window size 2 to achieve feature dimensionality reduction; Bidirectional GRU layer: adopts a two-layer structure: the first layer: 256 hidden units, tanh activation function, the second layer: 128 hidden units, tanh activation function; Dropout layer: adopts a dropout rate of 0.3 to prevent overfitting; Fully connected layer: contains a three-layer structure: the first layer: 256 neurons, ReLU activation; the second layer: 128 neurons, ReLU activation; the third layer: 64 neurons, ReLU activation; Output layer: 2 neurons, outputting power and angle prediction values ​​respectively.

[0053] The historical operation data is divided into a training set and a feature set. time, use represents the state of the input layer, represents the output layer state, represents the hidden layer state during the bidirectional GRU layer transmission process, then: ; In the formula, Represents the feature state learned by the forward layer during one-way transmission. It is the feature state learned by the backward layer during the one-way transmission process; and It is the weight of the forward layer, the backward layer and the output layer in sequence; represents the bias vector added in the output layer; the training set is imported into the neural network model to obtain the parameters of the network model.

[0054] At time t, the input data xt contains the values ​​of the five key parameters at the current moment; through the forward propagation process, the GRU layer models the past time series data; through the back propagation process, the changing trend of the future time series data is captured. The final output layer obtains the prediction result through linear combination; for example, for a certain sample, the input parameters such as the current light intensity of 900 W / m², the ambient temperature of 25℃, and the grid load of 80% can predict that the optimal inverter power output should be 450kW and the photovoltaic array angle should be adjusted to 32°.

[0055] The specific settings of the network structure of the prediction model are as follows: the input layer receives 5 key parameters as input features; the one-dimensional convolution layer uses 64 convolution kernels, the kernel size is 3, and the step size is 1; the pooling layer adopts maximum pooling, and the pooling window size is 2; the bidirectional GRU layer contains 128 hidden units; the dropout rate of the Dropout layer is set to 0.3; the fully connected layer contains 64 neurons and uses the ReLU activation function; the output layer outputs two prediction values ​​of power and angle; it can effectively capture the temporal characteristics and nonlinear relationships of the data.

[0056] During model training, the data set is divided into: 70% training set, 15% validation set, and 15% test set; the Adam optimizer is used, the initial learning rate is 0.001, and it decays by 10% every 50 epochs; the batch size is set to 64, and the maximum number of training rounds is 1000; the early stopping strategy is used, and training is stopped when the validation set loss has not improved for 10 consecutive epochs; L2 regularization (coefficient 0.0001) is used to suppress overfitting.

[0057] The prediction results output by the prediction model of the photovoltaic power station system are compared with the measured data in the test set, and the fit index is used To describe the predictive performance of the model, ,when When , the model training stop condition is met. At this time, the model parameters are the optimal parameters of the neural network model structure. If not, the back propagation algorithm is used to adjust the model deviation and continue the cycle.

[0058] Goodness of fit index The calculation formula is: ; In the formula, Representative The actual power output value of the samples, For the The power output prediction value corresponding to the samples is: is the average value of power output; Representative The actual PV array angle value of samples, For the The predicted value of the photovoltaic array angle corresponding to the sample, is the average value of the photovoltaic array angle; is the total number of samples; and is the weight coefficient and satisfies , used to balance the importance of power output and angle prediction, the default value is: , .

[0059] The setting of weight coefficients w1 and w2 takes into account that power output has a greater impact on system efficiency, so a higher weight of 0.55 is given, and the weight of angle prediction is set to 0.45; when R²≥0.95, it indicates that the deviation between the model prediction result and the actual value is within an acceptable range; if the condition is not met, the Adam optimizer is used for back propagation, the learning rate is set to 0.001, the batch size is 64, and the maximum number of iterations is 1000.

[0060] Step 3: Use the ITAE-optimized PV power station system adaptive control algorithm to optimize the proportional, integral, and differential coefficients of the PID controller. Based on the established prediction model and the optimized PID controller, adjust the inverter power output and PV array angle in real time.

[0061] In the process of tuning the PID controller parameters, the initial parameter values ​​are first determined according to the Ziegler-Nichols method, and then fine-tuned through the ITAE optimization algorithm; the PID parameters of the power control channel and the angle control channel are optimized independently, but the coupling relationship between the two is established through the ITAE indicator to achieve coordinated control.

[0062] Initial parameters of PID controller: power control channel: Kp1=0.5, Ti1=2.0s, Td1=0.1s; angle control channel: Kp2=0.3, Ti2=5.0s, Td2=0.2s; ITAE optimization process: set simulation time st=100s, use particle swarm algorithm for optimization, population size 50, maximum number of iterations 100; parameter search range: Kp1∈[0.1,1.0], Ti1∈[1.0,5.0], Td1∈[0.05,0.5]; Kp2∈[0.1,0.8], Ti2∈[2.0,8.0], Td2∈[0.1,0.8].

[0063] The control transfer function of the PID controller is: ; in, is the inverter power output after Laplace transformation, is the Laplace transform value of the PV array angle, and are the Laplace transform values of the power deviation and the angle deviation respectively, is the time variable; , and are the proportional, integral and derivative coefficients of the PID controller in the power control channel respectively, , , ; where, is the relative deviation of the inverter power output, is the relative deviation of the stroke of the PID controller in the power control channel; , and are the proportional, integral and derivative coefficients of the PID controller in the angle control channel respectively; , , ; where, is the relative angle control deviation, is the relative deviation of the stroke of the PID controller in the angle control channel.

[0064] Select the index as the fitness function of the optimization algorithm , where, is the simulation time.

[0065] Compare the ITAE index values by calculating the system states at the previous moment S-1 and the current moment S and make a comparison. The ITAE index value at the previous moment is denoted as , and the ITAE index value at the current moment is denoted as .

[0066] When ITAE(S-1)≥ITAE(S), it indicates that the system performance is improving. At this time, keep the current PID controller parameters unchanged; according to the relative deviations of the controlled variables in the power and angle channels at the current moment and the relative deviation of the PID controller stroke; adjust the inverter power output and the photovoltaic array angle so that they reach the preset power output and the preset photovoltaic array angle .

[0067] When ITAE(S-1)<ITAE(S), it indicates that the system performance is deteriorating. At this time, restore the relative deviations of the controlled variables in the power and angle channels and the relative deviation of the PID controller stroke to the values at the previous moment, recalculate the PID controller parameters; adjust the inverter power output and the photovoltaic array angle so that they reach the preset power output and the preset photovoltaic array angle .

[0068] Taking a 100MW photovoltaic power station as an example, after adopting the remote intelligent control method of the present invention, under typical sunny conditions, the power generation efficiency of the power station is improved by 3.2%; when there is a sudden change in light intensity (such as a sudden drop from 1000 W / m² to 600 W / m²), the system can complete power regulation within 2 seconds, and the response time is shortened by 60% compared with the traditional fixed parameter PID control; at the same time, the average power prediction error within 24 hours a day is reduced to 3.5%, which is a significant improvement over the 7.8% of the traditional method; this shows that the present invention can effectively improve the operating efficiency and control performance of photovoltaic power stations.

[0069] The method further comprises the following steps: Step 4: As the operation data continues to accumulate, the prediction model parameters and PID controller parameters are regularly updated; the frequency of the regular update is maintained at 15 to 30 days.

[0070] For example, the system parameters can be updated every 20 days. The update process is performed during the low-load period in the early morning. The neural network model parameters are updated through incremental learning, and the PID controller parameters are re-optimized based on the latest operating data. This regular update mechanism ensures that the system can adapt to the impact of seasonal changes and equipment performance degradation.

[0071] Example 2 like Figure 2 As shown, it is a schematic diagram of the composition of a remote intelligent management and control system for a photovoltaic power station of the present invention. The system includes: a data acquisition module, a data processing module and a parameter adjustment module connected in sequence; data is exchanged between the modules of the system through industrial Ethernet, and the TCP / IP protocol stack is used to ensure the real-time and reliability of data transmission.

[0072] The data acquisition module is used to collect historical operating data of key parameters of the photovoltaic power station system, and the key parameters include: inverter power output, photovoltaic array angle, light intensity, ambient temperature and grid load.

[0073] The data acquisition module adopts a hierarchical distributed architecture, and deploys multiple data acquisition units at the field level to collect historical operating data of key parameters of the photovoltaic power station system; power data acquisition uses a 0.5S-level high-precision smart energy meter with a sampling frequency of 1Hz and a measurement range of 0-120% rated power, and transmits data through an RS485 interface; angle data acquisition uses a high-precision encoder with a resolution of 0.1°, a measurement range of 0-90°, and a temperature compensation function; light intensity acquisition uses 12 distributed light meters with an accuracy of ±1% and a range of 0-2000W / m² to form a light intensity spatial distribution map; ambient temperature acquisition is configured with a PT100 temperature sensor (accuracy of ±0.1℃) and an infrared thermometer to comprehensively monitor the ambient temperature and component surface temperature; grid load data communicates with the grid dispatching system in real time through the standard IEC 60870-5-104 protocol.

[0074] The data processing module is used to establish a prediction model for the photovoltaic power station system based on a neural network, and to provide inverter power settings and optimal angles of the photovoltaic array under current illumination and load response according to the prediction model.

[0075] The data processing module is built on an industrial-grade server, equipped with a dual-core Intel Xeon processor, 32GB memory, and a real-time operating system, and is used to perform the following functions: data preprocessing uses a mean filtering algorithm to remove data noise, the window size is configurable (5-9 points), outlier detection and processing is based on the 3σ principle, data normalization is performed, and various types of data are uniformly mapped to the [0,1] interval; the neural network model running model framework is implemented using PyTorch, supports GPU accelerated computing, real-time computing response time <100ms, and model parameters are updated regularly (15-30 days); the output prediction power resolution is 0.1kW, and the prediction angle resolution is 0.1°.

[0076] The parameter adjustment module is used to optimize the proportional, integral and differential coefficients of the PID controller using the ITAE optimized photovoltaic power station system adaptive control algorithm, and to adjust the inverter power output and photovoltaic array angle in real time based on the established prediction model and the optimized PID controller.

[0077] The parameter adjustment module is implemented with a real-time controller, and the control cycle is configurable, typically set to 100ms. The PID parameter self-tuning range is reasonably designed, and the power control channel parameter range is: Kp1 (0.1-1.0), Ti1 (1.0-5.0s), Td1 (0.05-0.5s); the angle control channel parameter range is: Kp2 (0.1-0.8), Ti2 (2.0-8.0s), Td2 (0.1-0.8s). The control accuracy requirements are strict, the power control accuracy reaches ±0.5%, and the angle control accuracy reaches ±0.2°. The system responds quickly, the power adjustment response time is less than 2s, and the angle adjustment response time is less than 5s, which meets the needs of rapid adjustment of power stations.

[0078] The system further comprises: The remote communication unit is used to achieve remote transmission of system data, remote monitoring of system status, and remote alarm of abnormal situations. The remote communication unit adopts a redundant design and supports multiple interfaces such as 4G / 5G wireless communication, Gigabit Ethernet, and fiber optic communication. The communication protocol covers mainstream industrial protocols such as Modbus TCP / IP and IEC 60870-5-104, and supports data encryption transmission. The communication performance is excellent, the data transmission delay is controlled within 100ms, and the system reliability exceeds 99.99%. The unit has functions such as remote parameter configuration, remote program update, and remote fault diagnosis, providing strong support for remote operation and maintenance of the system.

[0079] The data storage unit is used to store the historical operation data of the photovoltaic power station system, the parameters of the neural network model, and the parameters of the PID controller; the data storage unit adopts a distributed storage architecture, with a local 256GB SSD for data cache and a remote database capacity of 10TB. The storage content includes real-time operation data, historical trend data, alarm event records, and model parameter backup, etc. The unit implements automatic data backup, data compression storage, and hierarchical storage strategies, uses RAID5 disk arrays to ensure data security, and implements strict data encryption storage and access control mechanisms.

[0080] The human-computer interaction unit includes a display module and a control module; wherein the display module is used to display the power generation power, array angle, environmental data and system operation status of the photovoltaic power station in real time; and the control module is used to receive the operator's control instructions and adjust the system operation parameters.

[0081] The human-machine interaction unit adopts B / S architecture design. The display module supports multi-screen real-time display, trend curve analysis, automatic report generation and 3D visualization display. The control module realizes online parameter adjustment, operation mode switching, remote operation authorization and emergency control. The system interface supports multi-resolution adaptation and mobile terminal access, and provides multi-language switching function. The operation authority adopts hierarchical management, supports operation record tracing, and implements confirmation mechanism for key operations.

[0082] The monitoring execution unit is used to monitor the key parameters of the system and perform abnormal processing according to a preset frequency, and the monitoring frequency is not less than once every 5 minutes; when an abnormality is detected, an alarm message is sent to the remote communication unit.

[0083] The monitoring execution unit has complete real-time monitoring and fault handling capabilities. The monitoring functions include equipment operation status monitoring, performance index calculation, fault warning analysis and system diagnosis and evaluation; the execution functions cover automatic adjustment control, fault protection action, emergency response and optimized control execution. The system sets the monitoring frequency according to different time scales: 5 minutes / time for conventional parameters, 10 seconds / time for key parameters, and 100ms / time for protection parameters. The abnormality handling mechanism adopts a four-level alarm system to realize functions such as automatic fault isolation, backup plan switching and automatic fault recovery.

[0084] The overall performance indicators of the system are excellent, with screen refresh time less than 1s, control response time less than 2s, and alarm response time less than 1s. The annual operating reliability of the system exceeds 99.9%, and the average trouble-free operation time exceeds 5,000 hours. The system has good scalability and supports capacity expansion, function upgrades, and interface expansion; the economic benefits are significant, and the power generation efficiency can be increased by 3-5%, and the operation and maintenance costs can be reduced by 30%.

[0085] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A remote intelligent control method for a photovoltaic power station, used to improve the power generation efficiency and operation and maintenance management level of the photovoltaic power station, wherein the photovoltaic power station uses a PID controller to control the inverter power output and the photovoltaic array angle to achieve matching adjustment of the power output of the photovoltaic power station with the grid load; characterized in that: The method comprises the following steps: Step 1, collecting historical operating data of key parameters of the photovoltaic power station system, wherein the key parameters include: inverter power output, photovoltaic array angle, light intensity, ambient temperature and grid load; Step 2, establishing a prediction model of the photovoltaic power station system based on a neural network, and providing the inverter power setting and the optimal angle of the photovoltaic array under the current light and load response according to the prediction model; Step 3: Use the ITAE-optimized PV power station system adaptive control algorithm to optimize the proportional, integral, and differential coefficients of the PID controller. Based on the established prediction model and the optimized PID controller, adjust the inverter power output and PV array angle in real time.

2. A photovoltaic power station remote intelligent management and control method according to claim 1, characterized in that: The prediction model consists of an input layer, a one-dimensional convolutional layer, a pooling layer, a bidirectional memory loop layer, a Dropout layer, a fully connected layer and an output layer connected in sequence; The one-dimensional convolution layer extracts the time series characteristics of the historical operation data of the photovoltaic power station, including the periodic change characteristics of the light intensity and the ambient temperature and the dynamic change characteristics of the grid load; the pooling layer simplifies the data dimension after the one-dimensional convolution and passes the data to the bidirectional gated loop layer; the bidirectional gated loop layer uses the bidirectional GRU to model the long-term and short-term dynamic characteristics of the photovoltaic power generation system, extracts the time series correlation characteristics of the power output and the photovoltaic array angle adjustment, and performs nonlinear transformation to determine the intrinsic logical relationship between the light, temperature, load input and the inverter power output and the photovoltaic array angle; After the data is subjected to secondary feature extraction by the bidirectional memory loop layer, some nodes of the data are randomly discarded in the Dropout layer to prevent overfitting and the processed data is passed to the fully connected layer. The fully connected layer maps the output features to the label space of the sample through nonlinear combination, and the output layer outputs the preset power output adapted to the current state of the photovoltaic power station through linear regression operation. and preset PV array angle The predicted value of The historical operation data is divided into a training set and a feature set. time, use represents the state of the input layer, represents the output layer state, represents the hidden layer state during the bidirectional GRU layer transmission process, then: ; In the formula, Represents the feature state learned by the forward layer during one-way transmission. It is the feature state learned by the backward layer during the one-way transmission process; and It is the weight of the forward layer, the backward layer and the output layer in sequence; represents the bias vector added in the output layer; the training set is imported into the neural network model to obtain the parameters of the network model.

3. A photovoltaic power station remote intelligent management and control method according to claim 2, characterized in that: The prediction results output by the prediction model of the photovoltaic power station system are compared with the measured data in the test set, and the fit index is used To describe the predictive performance of the model, ,when When , the model training stop condition is met. At this time, the model parameters are the optimal parameters of the neural network model structure. If not, the back propagation algorithm is used to adjust the model deviation and continue the cycle; Goodness of fit index The calculation formula is: ; In the formula, Representative The actual power output value of the samples, For the The power output prediction value corresponding to the samples is: is the average value of power output; Representative The actual PV array angle value of samples, For the The predicted value of the photovoltaic array angle corresponding to the sample, is the average value of the photovoltaic array angle; is the total number of samples; and is the weight coefficient and satisfies , used to balance the importance of power output and angle prediction, the default value is: , .

4. A method for remote intelligent control of a photovoltaic power station according to claim 3, characterized in that: The control transfer function of the PID controller is: ； in, is the inverter power output after Laplace transformation, is the Laplace transform value of the PV array angle, and are the Laplace transform values ​​of power deviation and angle deviation respectively, is the time variable; , and are the proportional, integral and differential coefficients of the power control channel PID controller, , , ;in, is the relative deviation of inverter power output, It is the relative deviation of the PID controller stroke of the power control channel; , and They are the proportional, integral and differential coefficients of the PID controller of the angle control channel respectively; , , ;in, To control the relative deviation for the angle, It is the relative deviation of the PID controller stroke of the angle control channel.

5. A method for remote intelligent control of a photovoltaic power station according to claim 4, characterized in that: choose The indicator is used as the fitness function of the optimization algorithm , where is the simulation time.

6. A method for remote intelligent management and control of a photovoltaic power station according to claim 5, characterized in that: Compare the system status at the previous moment S-1 and the current moment S to calculate the ITAE index value and compare them. The ITAE index value at the previous moment is recorded as , the ITAE index value at the current moment is recorded as ; When ITAE(S-1)≥ITAE(S), it indicates that the system performance is improving. At this time, keep the current PID controller parameters unchanged; according to the relative deviation of the controlled parameters of the power and angle channels at the current moment and the relative deviation of the PID controller stroke; adjust the inverter power output and the PV array angle to achieve the preset power output given by the prediction model. and preset PV array angle ; When ITAE(S - 1) < ITAE(S), it indicates that the system performance is deteriorating. At this time, the relative deviation of the controlled parameters in the power and angle channels and the relative deviation of the PID controller stroke are restored to the values at the previous moment, and the PID controller parameters are recalculated; the inverter power output and the photovoltaic array angle are adjusted to reach the preset power output given by the prediction model and the preset photovoltaic array angle .

7. A photovoltaic power station remote intelligent management and control method according to claim 1, characterized in that: The method further comprises the following steps: Step 4: As the operation data continues to accumulate, the prediction model parameters and PID controller parameters are regularly updated; The frequency of regular updates is maintained at 15 to 30 days.

8. A method for remote intelligent management and control of a photovoltaic power station according to claim 1, characterized in that: The step 1 also includes preprocessing the historical operation data by a mean filtering method to remove noise and outliers; for each point in the historical operation data, the average value of the point and the two points before and after it is taken as the output; The method of mean filtering is specifically as follows: in is the number of points in the filter window, N=5, the original value The first The value of a point.

9. A remote intelligent management and control system for a photovoltaic power station, used to execute the method according to any one of claims 1 to 8, characterized in that: The system comprises: a data acquisition module, a data processing module and a parameter adjustment module connected in sequence; The data acquisition module is used to collect historical operating data of key parameters of the photovoltaic power station system, and the key parameters include: inverter power output, photovoltaic array angle, light intensity, ambient temperature and grid load; The data processing module is used to establish a prediction model of the photovoltaic power station system based on a neural network, and to provide the inverter power setting and the optimal angle of the photovoltaic array under the current light and load response according to the prediction model; The parameter adjustment module is used to optimize the proportional, integral and differential coefficients of the PID controller using the ITAE optimized photovoltaic power station system adaptive control algorithm, and to adjust the inverter power output and photovoltaic array angle in real time based on the established prediction model and the optimized PID controller.

10. A photovoltaic power station remote intelligent management and control system according to claim 9, characterized in that: The system further comprises: Remote communication unit, used to achieve remote transmission of system data, remote monitoring of system status, and remote alarm of abnormal situations; A data storage unit, used to store historical operation data of the photovoltaic power station system, neural network model parameters and PID controller parameters; The human-computer interaction unit includes a display module and a control module; wherein the display module is used to display the power generation, array angle, environmental data and system operation status of the photovoltaic power station in real time; and the control module is used to receive the operator's control instructions and adjust the system operation parameters; The monitoring execution unit is used to monitor the key parameters of the system and perform exception processing at a preset frequency, and the monitoring frequency shall not be less than once every 5 minutes; when an abnormality is detected, an alarm message is sent to the remote communication unit.

Images