Intelligent photovoltaic power generation system control method

Abstract

The invention discloses an intelligent photovoltaic power generation system control method, and relates to the technical field of photovoltaic power generation management, and the method comprises the steps: collecting data in real time based on a sensor in a photovoltaic system, employing an edge computing technology, and transmitting important information to a cloud end; the future illumination change is predicted, so that the output of the inverter is adjusted in advance; based on an energy management system, the power demand and the generating capacity are monitored in real time, and interaction of photovoltaic power generation, energy storage and a power grid is coordinated; acquiring historical data, identifying a fault mode, and performing fault detection and prediction maintenance; based on the AR technology, the system state is displayed in real time. According to the method, environment data such as illumination, temperature and wind speed are collected in real time through a sensor, the edge computing technology is combined, the system state and the abnormal condition can be rapidly detected, the power generation efficiency is improved, illumination prediction based on the deep learning algorithm can dynamically adjust the working point of the inverter, the power generation capacity of the photovoltaic system is improved to the maximum extent, and the power generation efficiency is improved. And the method adapts to rapidly changing environmental conditions.

Description

Technical Field

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

[0002] With the rapid development of renewable energy, photovoltaic power generation has received more and more attention as a clean and efficient way of energy utilization. However, the performance of photovoltaic power generation systems is affected by many environmental factors, such as light, temperature, wind speed and equipment status. Therefore, how to monitor and optimize the operation of photovoltaic systems in real time, improve power generation efficiency, and reduce maintenance costs has become the focus of research and application.

[0003] Traditional photovoltaic power generation systems rely on static control and manual monitoring, lacking real-time and intelligence, resulting in energy loss and slow fault response. In addition, as the scale of photovoltaic systems expands, the amount of data also increases significantly. How to effectively process and analyze this data has become a technical challenge. In order to solve these problems, intelligent control methods based on edge computing and deep learning have emerged, realizing intelligent management of photovoltaic systems through real-time data collection, analysis and optimized control. Summary of the invention

[0004] In order to solve the above technical problems, a smart photovoltaic power generation system control method is provided. This technical solution solves the above problems.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A control method for an intelligent photovoltaic power generation system, comprising:

[0007] Based on the sensors in the photovoltaic system, environmental data such as light, temperature and wind speed as well as system status data are collected in real time, and important information is transmitted to the cloud using edge computing technology;

[0008] Based on the MPPT algorithm, the inverter's operating point is dynamically adjusted. Combined with the deep learning algorithm, future light changes are predicted, so as to adjust the inverter's output in advance.

[0009] Based on the energy management system, real-time monitoring of power demand and power generation is carried out to coordinate the interaction between photovoltaic power generation, energy storage and power grid;

[0010] Acquire historical data, identify failure modes, perform fault detection and predictive maintenance, and automatically generate fault reports and maintenance recommendations by combining natural language processing technology;

[0011] Based on AR technology, the system status is displayed in real time.

[0012] Preferably, the method of dynamically adjusting the working point of the inverter based on the MPPT algorithm and combining the deep learning algorithm to predict future illumination changes, thereby adjusting the output of the inverter in advance specifically includes:

[0013] Obtain real-time data from light sensors, temperature sensors, wind speed sensors, and voltage and current sensors. The sensors periodically collect environmental data and system status data.

[0014] Preprocessing the collected data in the sensor, wherein the preprocessing includes filtering, denoising and data normalization;

[0015] The collected environmental data and system status data are processed on the edge based on edge computing, and whether there are any abnormal conditions is detected based on the collected system status data;

[0016] Based on the analysis results, important information including abnormal conditions, power generation forecasts, and environmental changes are selected and prepared for sending to the cloud;

[0017] The important information is compressed and formatted, and the filtered and compressed important information is transmitted to the cloud platform through the wireless network. The cloud receives the data from the edge device and stores it in the database.

[0018] Preferably, the edge computing-based edge processing of the collected environmental data and system status data, and the detection of whether there is an abnormality based on the collected system status data specifically include:

[0019] Extract statistical features from the data, set thresholds based on historical data, and use anomaly detection algorithms to perform anomaly detection;

[0020] Based on the isolation forest algorithm, multiple random trees are constructed, anomaly scores are calculated, and thresholds are set. Once an abnormal situation is detected, an alarm is immediately sent to relevant personnel, and the abnormal event and related data are recorded in the database. The isolation forest algorithm formula is:

[0021]

[0022] Where AS is the anomaly score, E(x) is the average path length of all trees in the isolation forest for data point x, and e(n) is a constant. is the maximum anomaly score among all data points.

[0023] Preferably, the method of dynamically adjusting the working point of the inverter based on the MPPT algorithm and combining the deep learning algorithm to predict future illumination changes, thereby adjusting the output of the inverter in advance specifically includes:

[0024] A photovoltaic cell model is established, including the current-voltage characteristics of the photovoltaic module, and a single diode model is used to describe the photovoltaic cell. The single diode model expression is:

[0025]

[0026] Where, I is the output current, I ph is the photocurrent, I0 is the reverse saturation current, V is the output voltage, R S is the series resistance, R SH is the parallel resistance, n is the ideal factor, V t is the thermal voltage;

[0027] Collect real-time data of the photovoltaic system, including voltage, current, ambient temperature and light intensity, and use the convolutional neural network (CNN) model to predict future light changes;

[0028] Taking light data and weather forecast data as input, the output is the light intensity forecast for a period of time in the future;

[0029] Use the perturbation and observation method to dynamically adjust the inverter's operating point to ensure that the photovoltaic system always operates at the maximum power point;

[0030] Calculate current power, analyze power changes, and adjust operating voltage or current to achieve maximum power output;

[0031] Based on the prediction results of the deep learning model, the output of the inverter is adjusted in advance to adapt it to the upcoming light changes.

[0032] Preferably, the real-time data of the photovoltaic system, including voltage, current, ambient temperature and light intensity, is collected, and the future light changes are predicted using a convolutional neural network (CNN) model, specifically including:

[0033] The data set is divided into training set, validation set and test set. Based on the convolutional neural network, a lighting change prediction model is constructed. The lighting change prediction model is expressed as:

[0034]

[0035] In the formula, is the predicted value of light intensity at time t+1, f is the activation function, W fc is the weight matrix of the fully connected layer, pool is the pooling operation, W l is the weight matrix of the convolutional layer, x(t) is the input feature vector at time t, b t is the bias term of the convolutional layer, b fc is the bias term of the fully connected layer, and Relu is the activation function;

[0036] The illumination change prediction model based on convolutional neural network processes the input time series data through multiple convolutional layers and fully connected layers, and finally outputs the prediction of future illumination intensity.

[0037] Preferably, the energy management system is based on which real-time monitoring of power demand and power generation and coordination of the interaction between photovoltaic power generation, energy storage and power grid specifically includes:

[0038] Real-time monitoring of the power demand of each load, identification of peak and off-peak periods, and monitoring of the real-time power generation of the photovoltaic power generation system, including the impact of meteorological conditions on power generation efficiency;

[0039] By monitoring the voltage, frequency and load conditions of the power grid, the operation status of the power grid is evaluated, and the power status, health status and charging and discharging capacity of the energy storage system are monitored;

[0040] Use machine learning algorithms to predict future electricity demand and optimize power dispatch;

[0041] The output of photovoltaic power generation is dispatched according to the real-time power generation and predicted demand. When the power demand is higher than the power generation, the energy storage system is dispatched to discharge. When the power generation is in excess, the energy storage system is dispatched to charge.

[0042] Develop optimal control strategies based on real-time data and forecast results to maximize the use of renewable energy, reduce peak loads, and lower electricity costs;

[0043] Utilize feedback control mechanisms to adjust the operating strategies of photovoltaics, energy storage, and loads in real time to ensure efficient operation of the system.

[0044] Preferably, the use of a machine learning algorithm to predict future power demand and optimize power dispatch specifically includes:

[0045] Based on the historical power demand data and influencing factor data, the power demand forecasting model is developed, with the historical power demand data and influencing factor data as input and the power demand forecasting data as output. The power demand forecasting model expression is:

[0046] D(t)=β0+β1x1(t)+β2x2(t)+…+β n x n (t)+∈

[0047] Where D(t) is the power demand at time t, β1β2β n are model parameters, x1 is humidity, x2 is temperature, x n is the nth characteristic affecting electricity demand, β0 is the intercept term, and ∈ is the error.

[0048] Preferably, the output of photovoltaic power generation is dispatched according to the real-time power generation and the predicted demand, and when the power demand is higher than the power generation, the energy storage system is dispatched to discharge, and when the power generation is in excess, the energy storage system is dispatched to charge, specifically including:

[0049] Collect power demand forecast data obtained from the power demand forecast model, obtain the power generation capacity, cost, start and stop time and operating status information of various power generation equipment, and collect information on the status of energy storage equipment and adjustable load;

[0050] With the goal of minimizing power generation and dispatching costs, the start-stop strategy of the generator set is formulated based on integer programming, and thresholds are set to determine when the energy storage system needs to be dispatched;

[0051] When the power demand is higher than the power generation, the energy storage system is dispatched to discharge based on the required discharge amount;

[0052] When the power demand is lower than the power generation, the required charging capacity is calculated based on the excess power generation, and the energy storage system is dispatched for charging;

[0053] According to the dispatching strategy, control instructions for charging or discharging are issued to the energy storage system, and the charging and discharging status of the energy storage system is monitored to ensure that the operation is within a safe range.

[0054] Preferably, the acquisition of historical data, identification of fault modes, fault detection and predictive maintenance, and automatic generation of fault reports and maintenance suggestions in combination with natural language processing technology specifically include:

[0055] By analyzing historical data, different failure modes and corresponding characteristics can be identified, and by deploying sensors and data acquisition systems, the operating status of the equipment can be monitored in real time;

[0056] Integrated fault detection and power demand forecasting results, including fault type, severity, and recommended maintenance time;

[0057] Natural language processing technology is used to convert technical results into reports, which include fault description, detection results, recommended maintenance measures and expected impacts.

[0058] Preferably, the real-time display of system status based on AR technology specifically includes:

[0059] Display system status, including equipment operation status, performance indicators and fault information;

[0060] Collect real-time data through sensors and monitoring equipment, transmit the collected data to the processing system through the network in real time, analyze the data in real time, and identify key indicators and abnormal conditions;

[0061] Design how users interact with AR interfaces, including touch and gesture controls.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention proposes real-time monitoring and optimization. By collecting environmental data such as light, temperature and wind speed in real time through sensors, combined with edge computing technology, it can quickly detect system status and abnormal conditions, ensure that the system operates in the best working state, thereby improving power generation efficiency. The light prediction based on the deep learning algorithm can dynamically adjust the operating point of the inverter, maximize the power generation capacity of the photovoltaic system, and adapt to rapidly changing environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a step flow framework diagram of the present invention. DETAILED DESCRIPTION

[0065] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0066] Reference Figure 1 As shown, a smart photovoltaic power generation system control method is characterized by comprising:

[0067] Based on the sensors in the photovoltaic system, environmental data such as light, temperature and wind speed as well as system status data are collected in real time, and important information is transmitted to the cloud using edge computing technology;

[0068] Based on the MPPT algorithm, the inverter's operating point is dynamically adjusted. Combined with the deep learning algorithm, future light changes are predicted, so as to adjust the inverter's output in advance.

[0069] Based on the energy management system, real-time monitoring of power demand and power generation is carried out to coordinate the interaction between photovoltaic power generation, energy storage and power grid;

[0070] Acquire historical data, identify failure modes, perform fault detection and predictive maintenance, and automatically generate fault reports and maintenance recommendations by combining natural language processing technology;

[0071] Based on AR technology, the system status is displayed in real time.

[0072] The method of dynamically adjusting the inverter working point based on the MPPT algorithm and combining the deep learning algorithm to predict future light changes, thereby adjusting the inverter output in advance, specifically includes:

[0073] Obtain real-time data from light sensors, temperature sensors, wind speed sensors, and voltage and current sensors. The sensors periodically collect environmental data and system status data.

[0074] Preprocessing the collected data in the sensor, wherein the preprocessing includes filtering, denoising and data normalization;

[0075] The collected environmental data and system status data are processed on the edge based on edge computing, and whether there are any abnormal conditions is detected based on the collected system status data;

[0076] Based on the analysis results, important information including abnormal conditions, power generation forecasts, and environmental changes are selected and prepared for sending to the cloud;

[0077] The important information is compressed and formatted, and the filtered and compressed important information is transmitted to the cloud platform through the wireless network. The cloud receives the data from the edge device and stores it in the database.

[0078] The edge computing-based edge processing of the collected environmental data and system status data, and the detection of whether there is an abnormality based on the collected system status data specifically include:

[0079] Extract statistical features from the data, set thresholds based on historical data, and use anomaly detection algorithms to perform anomaly detection;

[0080] Based on the isolation forest algorithm, multiple random trees are constructed, anomaly scores are calculated, and thresholds are set. Once an abnormal situation is detected, an alarm is immediately sent to relevant personnel, and the abnormal event and related data are recorded in the database. The isolation forest algorithm formula is:

[0081]

[0082] Where AS is the anomaly score, E(x) is the average path length of all trees in the isolation forest for data point x, and e(n) is a constant. is the maximum anomaly score among all data points.

[0083] The method of dynamically adjusting the inverter working point based on the MPPT algorithm and combining the deep learning algorithm to predict future light changes, thereby adjusting the inverter output in advance, specifically includes:

[0084] A photovoltaic cell model is established, including the current-voltage characteristics of the photovoltaic module, and a single diode model is used to describe the photovoltaic cell. The single diode model expression is:

[0085]

[0086] Where, I is the output current, I ph is the photocurrent, I0 is the reverse saturation current, V is the output voltage, R S is the series resistance, R SH is the parallel resistance, n is the ideal factor, Vt is the thermal voltage;

[0087] Collect real-time data of the photovoltaic system, including voltage, current, ambient temperature and light intensity, and use the convolutional neural network (CNN) model to predict future light changes;

[0088] Taking light data and weather forecast data as input, the output is the light intensity forecast for a period of time in the future;

[0089] Use the perturbation and observation method to dynamically adjust the inverter's operating point to ensure that the photovoltaic system always operates at the maximum power point;

[0090] Calculate current power, analyze power changes, and adjust operating voltage or current to achieve maximum power output;

[0091] Based on the prediction results of the deep learning model, the output of the inverter is adjusted in advance to adapt it to the upcoming light changes.

[0092] The real-time data of the photovoltaic system, including voltage, current, ambient temperature and light intensity, is collected, and the convolutional neural network (CNN) model is used to predict future light changes, including:

[0093] The data set is divided into training set, validation set and test set. Based on the convolutional neural network, a lighting change prediction model is constructed. The lighting change prediction model is expressed as:

[0094]

[0095] In the formula, is the predicted value of light intensity at time t+1, f is the activation function, W fc is the weight matrix of the fully connected layer, pool is the pooling operation, W l is the weight matrix of the convolutional layer, x(t) is the input feature vector at time t, b t is the bias term of the convolutional layer, b fc is the bias term of the fully connected layer, and Relu is the activation function;

[0096] The illumination change prediction model based on convolutional neural network processes the input time series data through multiple convolutional layers and fully connected layers, and finally outputs the prediction of future illumination intensity.

[0097] The energy management system is based on real-time monitoring of power demand and power generation, and coordination of the interaction between photovoltaic power generation, energy storage and power grid, specifically including:

[0098] Real-time monitoring of the power demand of each load, identification of peak and off-peak periods, and monitoring of the real-time power generation of the photovoltaic power generation system, including the impact of meteorological conditions on power generation efficiency;

[0099] By monitoring the voltage, frequency and load conditions of the power grid, the operation status of the power grid is evaluated, and the power status, health status and charging and discharging capacity of the energy storage system are monitored;

[0100] Use machine learning algorithms to predict future electricity demand and optimize power dispatch;

[0101] The output of photovoltaic power generation is dispatched according to the real-time power generation and predicted demand. When the power demand is higher than the power generation, the energy storage system is dispatched to discharge. When the power generation is in excess, the energy storage system is dispatched to charge.

[0102] Develop optimal control strategies based on real-time data and forecast results to maximize the use of renewable energy, reduce peak loads, and lower electricity costs;

[0103] Utilize feedback control mechanisms to adjust the operating strategies of photovoltaics, energy storage, and loads in real time to ensure efficient operation of the system.

[0104] The use of machine learning algorithms to predict future power demand and optimize power dispatch specifically includes:

[0105] Based on the historical power demand data and influencing factor data, the power demand forecasting model is developed, with the historical power demand data and influencing factor data as input and the power demand forecasting data as output. The power demand forecasting model expression is:

[0106] D(t)=β0+β1x1(t)+β2x2(t)+…+β n x n (t)+∈

[0107] Where D(t) is the power demand at time t, β1β2β n are model parameters, x1 is humidity, x2 is temperature, x n is the nth characteristic affecting electricity demand, β0 is the intercept term, and ∈ is the error.

[0108] The method of dispatching the output of photovoltaic power generation according to the real-time power generation and the predicted demand, dispatching the energy storage system to discharge when the power demand is higher than the power generation, and dispatching the energy storage system to charge when the power generation is in excess specifically includes:

[0109] Collect power demand forecast data obtained from the power demand forecast model, obtain the power generation capacity, cost, start and stop time and operating status information of various power generation equipment, and collect information on the status of energy storage equipment and adjustable load;

[0110] With the goal of minimizing power generation and dispatching costs, the start-stop strategy of the generator set is formulated based on integer programming, and thresholds are set to determine when the energy storage system needs to be dispatched;

[0111] When the power demand is higher than the power generation, the energy storage system is dispatched to discharge based on the required discharge amount;

[0112] When the power demand is lower than the power generation, the required charging capacity is calculated based on the excess power generation, and the energy storage system is dispatched for charging;

[0113] According to the dispatching strategy, control instructions for charging or discharging are issued to the energy storage system, and the charging and discharging status of the energy storage system is monitored to ensure that the operation is within a safe range.

[0114] The acquisition of historical data, identification of failure modes, fault detection and predictive maintenance, and automatic generation of fault reports and maintenance suggestions in combination with natural language processing technology specifically include:

[0115] By analyzing historical data, different failure modes and corresponding characteristics can be identified, and by deploying sensors and data acquisition systems, the operating status of the equipment can be monitored in real time;

[0116] Integrated fault detection and power demand forecasting results, including fault type, severity, and recommended maintenance time;

[0117] Natural language processing technology is used to convert technical results into reports, which include fault description, detection results, recommended maintenance measures and expected impacts.

[0118] The real-time display of system status based on AR technology specifically includes:

[0119] Display system status, including equipment operation status, performance indicators and fault information;

[0120] Collect real-time data through sensors and monitoring equipment, transmit the collected data to the processing system through the network in real time, analyze the data in real time, and identify key indicators and abnormal conditions;

[0121] Design how users interact with AR interfaces, including touch and gesture controls.

[0122] In summary, the advantages of the present invention are:

[0123] By collecting environmental data such as light, temperature and wind speed in real time through sensors and combining them with edge computing technology, the system status and abnormal conditions can be quickly detected to ensure that the system operates in the best working state, thereby improving power generation efficiency.

[0124] Light prediction based on deep learning algorithms can dynamically adjust the inverter's operating point to maximize the photovoltaic system's power generation capacity and adapt to rapidly changing environmental conditions;

[0125] Combining historical data with machine learning algorithms, it can effectively identify and predict failure modes, automatically generate failure reports and maintenance recommendations, thereby reducing maintenance costs and improving system reliability;

[0126] The energy management system can monitor power demand and power generation in real time, achieve efficient coordination between photovoltaic power generation, energy storage and power grid, optimize power dispatch, reduce peak load and improve system economy;

[0127] Use AR technology to display system status in real time, improve user monitoring experience and operation convenience, and help users quickly understand system operation status and respond in a timely manner.

[0128] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A control method for an intelligent photovoltaic power generation system, characterized in that: include: Based on the sensors in the photovoltaic system, environmental data such as light, temperature and wind speed as well as system status data are collected in real time, and important information is transmitted to the cloud using edge computing technology; Based on the MPPT algorithm, the inverter's operating point is dynamically adjusted. Combined with the deep learning algorithm, future light changes are predicted, so as to adjust the inverter's output in advance. Based on the energy management system, real-time monitoring of power demand and power generation is carried out to coordinate the interaction between photovoltaic power generation, energy storage and power grid; Acquire historical data, identify failure modes, perform fault detection and predictive maintenance, and automatically generate fault reports and maintenance recommendations by combining natural language processing technology; Based on AR technology, the system status is displayed in real time.

2. The intelligent photovoltaic power generation system control method according to claim 1, characterized in that: The method of dynamically adjusting the inverter working point based on the MPPT algorithm and combining the deep learning algorithm to predict future light changes, thereby adjusting the inverter output in advance, specifically includes: Obtain real-time data from light sensors, temperature sensors, wind speed sensors, and voltage and current sensors. The sensors periodically collect environmental data and system status data; Preprocessing the collected data in the sensor, the preprocessing including filtering, denoising and data normalization; The collected environmental data and system status data are processed on the edge based on edge computing, and whether there are any abnormal conditions is detected based on the collected system status data; Based on the analysis results, important information including abnormal conditions, power generation forecasts, and environmental changes are screened out and prepared to be sent to the cloud; The important information is compressed and formatted, and the filtered and compressed important information is transmitted to the cloud platform through the wireless network. The cloud receives the data from the edge device and stores it in the database.

3. The intelligent photovoltaic power generation system control method according to claim 2, characterized in that: The edge computing-based edge processing of the collected environmental data and system status data, and the detection of whether there is an abnormality based on the collected system status data specifically include: Extract statistical features from the data, set thresholds based on historical data, and use anomaly detection algorithms to perform anomaly detection; Based on the isolation forest algorithm, multiple random trees are constructed, anomaly scores are calculated, and thresholds are set. Once an abnormal situation is detected, an alarm is immediately sent to relevant personnel, and the abnormal event and related data are recorded in the database. The isolation forest algorithm formula is: Where AS is the anomaly score, E(x) is the average path length of all trees in the isolation forest for data point x, and e(n) is a constant. is the maximum anomaly score among all data points.

4. The intelligent photovoltaic power generation system control method according to claim 3, characterized in that: The method of dynamically adjusting the inverter working point based on the MPPT algorithm and combining the deep learning algorithm to predict future light changes, thereby adjusting the inverter output in advance, specifically includes: A photovoltaic cell model is established, including the current-voltage characteristics of the photovoltaic module, and a single diode model is used to describe the photovoltaic cell. The single diode model expression is: Where, I is the output current, I ph is the photocurrent, I0 is the reverse saturation current, V is the output voltage, R S is the series resistance, R SH is the parallel resistance, n is the ideal factor, V t is the thermal voltage; Collect real-time data of the photovoltaic system, including voltage, current, ambient temperature and light intensity, and use the convolutional neural network (CNN) model to predict future light changes; Taking light data and weather forecast data as input, the output is the light intensity forecast for a period of time in the future; Use the perturbation-observation method to dynamically adjust the inverter's operating point to ensure that the PV system always operates at the maximum power point; Calculate current power, analyze power changes, and adjust operating voltage or current to achieve maximum power output; Based on the prediction results of the deep learning model, the output of the inverter is adjusted in advance to adapt it to the upcoming light changes.

5. The intelligent photovoltaic power generation system control method according to claim 4, characterized in that: The real-time data of the photovoltaic system, including voltage, current, ambient temperature and light intensity, is collected, and the convolutional neural network (CNN) model is used to predict future light changes, including: The data set is divided into training set, validation set and test set. Based on the convolutional neural network, a lighting change prediction model is constructed. The lighting change prediction model is expressed as: In the formula, is the predicted value of light intensity at time t+1, f is the activation function, W fc is the weight matrix of the fully connected layer, pool is the pooling operation, W l is the weight matrix of the convolutional layer, x(t) is the input feature vector at time t, b t is the bias term of the convolutional layer, b fc is the bias term of the fully connected layer, and Relu is the activation function; The illumination change prediction model based on convolutional neural network processes the input time series data through multiple convolutional layers and fully connected layers, and finally outputs the prediction of future illumination intensity.

6. The intelligent photovoltaic power generation system control method according to claim 5, characterized in that: The energy management system is based on real-time monitoring of power demand and power generation, and coordination of the interaction between photovoltaic power generation, energy storage and power grid, specifically including: Real-time monitoring of the power demand of each load, identification of peak and off-peak periods, and monitoring of the real-time power generation of the photovoltaic power generation system, including the impact of meteorological conditions on power generation efficiency; By monitoring the voltage, frequency and load conditions of the power grid, the operation status of the power grid is evaluated, and the power status, health status and charging and discharging capacity of the energy storage system are monitored; Use machine learning algorithms to predict future electricity demand and optimize power dispatch; The output of photovoltaic power generation is dispatched according to the real-time power generation and predicted demand. When the power demand is higher than the power generation, the energy storage system is dispatched to discharge. When the power generation is in excess, the energy storage system is dispatched to charge. Develop optimal control strategies based on real-time data and forecast results to maximize the use of renewable energy, reduce peak loads, and lower electricity costs; Utilize feedback control mechanisms to adjust the operating strategies of photovoltaics, energy storage, and loads in real time to ensure efficient operation of the system.

7. The intelligent photovoltaic power generation system control method according to claim 6, characterized in that: The use of machine learning algorithms to predict future power demand and optimize power dispatch specifically includes: Based on the historical power demand data and influencing factor data, the power demand forecasting model is developed, with the historical power demand data and influencing factor data as input and the power demand forecasting data as output. The power demand forecasting model expression is: D(t)=β0+β1x1(t)+β2x2(t)+…+β n x n (t)+∈ Where D(t) is the power demand at time t, β1β2β n are model parameters, x1 is humidity, x2 is temperature, x n is the nth characteristic affecting electricity demand, β0 is the intercept term, and ∈ is the error.

8. The intelligent photovoltaic power generation system control method according to claim 7, characterized in that: The method of dispatching the output of photovoltaic power generation according to the real-time power generation and the predicted demand, dispatching the energy storage system to discharge when the power demand is higher than the power generation, and dispatching the energy storage system to charge when the power generation is in excess specifically includes: Collect power demand forecast data obtained from the power demand forecast model, obtain the power generation capacity, cost, start and stop time and operating status information of various power generation equipment, and collect information on the status of energy storage equipment and adjustable load; With the goal of minimizing power generation and dispatching costs, the start-stop strategy of the generator set is formulated based on integer programming, and thresholds are set to determine when the energy storage system needs to be dispatched; When the power demand is higher than the power generation, the energy storage system is dispatched to discharge based on the required discharge amount; When the power demand is lower than the power generation, the required charging capacity is calculated based on the excess power generation, and the energy storage system is dispatched for charging; According to the dispatching strategy, control instructions for charging or discharging are issued to the energy storage system, and the charging and discharging status of the energy storage system is monitored to ensure that the operation is within a safe range.

9. The intelligent photovoltaic power generation system control method according to claim 8, characterized in that: The acquisition of historical data, identification of failure modes, fault detection and predictive maintenance, and automatic generation of fault reports and maintenance suggestions in combination with natural language processing technology specifically include: By analyzing historical data, different failure modes and corresponding characteristics can be identified, and by deploying sensors and data acquisition systems, the operating status of the equipment can be monitored in real time; Integrated fault detection and power demand forecasting results, including fault type, severity, and recommended maintenance time; Natural language processing technology is used to convert technical results into reports, which include fault description, detection results, recommended maintenance measures and expected impacts.

10. The intelligent photovoltaic power generation system control method according to claim 9, characterized in that: The real-time display of system status based on AR technology specifically includes: Display system status, including equipment operation status, performance indicators and fault information; Collect real-time data through sensors and monitoring equipment, transmit the collected data to the processing system in real time through the network, analyze the data in real time, identify key indicators and abnormal conditions; design the interaction method between users and AR interfaces, including touch and gesture control.

Images