Photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion

Abstract

A photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion, with a central control system used to coordinate and manage the photovoltaic output of each substation. Each substation is equipped with an independent monitoring center and sensor network to collect key data of photovoltaic panels and related equipment in real time; a deep learning architecture that integrates convolutional neural networks and long-short-term memory networks; a feature recognition system based on an attention mechanism to enhance the recognition and processing capabilities of key features; and a fuzzy logic control module that automatically adjusts the operating parameters of photovoltaic equipment when performance is detected to be below a preset threshold. The present invention improves data integrity and system response speed through multi-source information fusion, overcomes the data fragmentation problem in traditional monitoring systems, and can achieve functions such as accurate prediction of large-scale, distributed photovoltaics, real-time tracking after grid connection, dynamic consistency verification, coordinated control, and status early warning. It can effectively support the safe and efficient operation of low-inertia, multi-source new energy power grids.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic energy management and monitoring, and in particular to a photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion. Background Art

[0002] In traditional photovoltaic system monitoring technology, there are problems such as data fragmentation, slow response speed, insufficient early warning accuracy and low energy management efficiency. These systems are usually faced with the challenges of incomplete data integration, lack of real-time response mechanism, and insufficient timeliness of fault diagnosis and processing. In addition, due to the lack of efficient energy allocation and optimization strategies, traditional photovoltaic monitoring is often unable to effectively balance and utilize stored energy in the case of excess or shortage of electricity. In response to these limitations, modern photovoltaic monitoring technology has begun to adopt advanced methods such as multi-source information fusion, deep learning and fuzzy logic control to enhance the system's data processing capabilities, improve response speed and early warning accuracy, and optimize overall energy management efficiency. Based on this technical background, the present invention can effectively support the safe and efficient operation of low-inertia, multi-source new energy power grids by integrating innovative monitoring technologies and intelligent algorithms, providing strong technical support for modern energy management systems. Summary of the Invention

[0003] The technical problem addressed by this invention is to provide a real-time photovoltaic performance assessment and early warning system and method based on multi-source information fusion. This system incorporates a deep learning architecture and fuzzy logic control technology. By integrating image data and time series data processing, it not only monitors photovoltaic system performance in real time, but also provides efficient early warning of performance degradation and potential failures. Furthermore, the system utilizes an attention mechanism to enhance the identification and analysis of key features, thereby optimizing the accuracy of performance assessment.

[0004] To achieve the above objectives, the photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion of the present invention includes a central control system, a deep learning architecture, a feature recognition system based on an attention mechanism, and a fuzzy logic control module;

[0005] The central control system is used to coordinate and manage the photovoltaic output of each substation. Each substation has an independent monitoring center and sensor network to collect key data from photovoltaic panels and related equipment in real time.

[0006] A deep learning architecture that integrates convolutional neural networks (CNNs) and long short-term memory (LSTMs) to process and fuse image and time series data;

[0007] A feature recognition system based on the attention mechanism to enhance the recognition and processing capabilities of key features;

[0008] A fuzzy logic control module is used to automatically adjust the operating parameters of the photovoltaic equipment when performance is detected to be below a preset threshold.

[0009] The method of the photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion of the present invention includes the following method steps:

[0010] S1. Build a comprehensive central control system, optimize the layout to support photovoltaic output in multiple areas, achieve centralized management and data integration, and provide a foundation for large-scale photovoltaic monitoring;

[0011] S2. Develop a deep learning architecture for multi-source information fusion. Utilizing convolutional neural networks and long short-term memory networks, design a deep learning architecture to accurately predict the output efficiency of individual photovoltaic panels and track the performance of photovoltaic systems in real time after grid connection.

[0012] S3. Implement a dynamic performance monitoring and adjustment mechanism, set local photovoltaic output thresholds, and automatically adjust operating parameters through fuzzy logic control when equipment performance does not reach the predetermined thresholds, ensuring dynamic consistency and efficient energy management of the system;

[0013] S4. Configure a virtual power plant and coordinated control system. Through the central control system, establish a virtual power plant model to optimize the coordinated control of photovoltaic output in each substation, and achieve stable and efficient operation of the new low-inertia, multi-source power grid.

[0014] Furthermore, the temperature sensor calculation value in S1 is expressed as:

[0015] T=T0+K*I

[0016] Where T represents the temperature of the photovoltaic panel in degrees Celsius, T0 represents the ambient temperature, that is, the temperature of the photovoltaic panel when there is no light, in degrees Celsius, K represents the temperature coefficient, which represents the rate of increase of temperature when the unit light intensity increases, in degrees Celsius / watt ℃ / W, and I represents the light intensity, in watts / square meter W / m 2 , describes the relationship between the temperature of the photovoltaic panel and the ambient temperature and light intensity. As the light intensity increases, the temperature of the photovoltaic panel will also increase accordingly, but the rate of increase is determined by the temperature coefficient K.

[0017] Furthermore, the light sensor calculation value in S1 is expressed as:

[0018] I 光 =I0*e -k*d

[0019] Among them I 光 Indicates light intensity in W / m 2 , I0 represents the solar radiation intensity at the location of the light sensor, in W / m2 , k represents the atmospheric attenuation coefficient, which reflects the attenuation of solar radiation by the atmosphere; d is the solar altitude angle, which represents the angle between the sun's rays and the ground, and usually ranges from 0 to 90 degrees.

[0020] Furthermore, the feature fusion using the attention mechanism in S2 is expressed as:

[0021] e i =v T tanh(W c h c +W c h l +b)

[0022]

[0023] where h c Represents the prediction result of the CNN network, h l Represents the prediction result of LSTM network, W c W c Represent the weight parameters of CNN and LSTM networks respectively, b represents the bias parameter, v is the learnable vector of attention, e i represents the score of each attention weight, α i represents the attention weight, and n represents the number of features. f represents the fused feature representation, h i Denotes the feature representation of each network.

[0024] Furthermore, the fuzzy control operation in S3 is expressed as:

[0025] First, for each input variable (, the membership function is used to map the actual value to the membership value;

[0026]

[0027] Where x is the temperature, light intensity, voltage and current value, a and b are the left and right boundaries of the triangular membership function, the temperature is set to 10-50℃, and the light intensity is set to 50W / m 2 -200W / m 2 , for current and voltage, they are usually set to 200mA-800mA, 20V-80V;

[0028] Then, a set of fuzzy rules is defined, which states that when the temperature is “high” and the light is “low” or the temperature is “low” and the light is “high”, the lighting angle can be adjusted appropriately;

[0029] Furthermore, for each fuzzy rule, the intersection between its premise part and the membership degree of the input variable is calculated to obtain the inference result. Here, the minimum operator min is used as the intersection operation, and the illumination angle is adjusted according to the inference result.

[0030] β=min(u 温度 ,u 光照 ,u 电压电流 )

[0031] The above inference results are aggregated and the maximum operator Max is used to obtain the final fuzzy output of each data item. Finally, the averaging method is used to convert the fuzzy output into specific operating parameter adjustment values, which are then applied to the operating parameters of the photovoltaic equipment. This can intelligently adjust the operating status of the photovoltaic equipment, improve photovoltaic efficiency, and respond to environmental changes and internal state changes.

[0032] out = mean(max(β)).

[0033] Furthermore, the electric energy stored in S4 is expressed as:

[0034]

[0035] Among them, E 存储 is the stored electrical energy, E 产出 is the electricity produced by solar photovoltaic, E 用电 is the regional electricity demand, η is the efficiency of the energy storage system, and the above formula is used to calculate the amount of electricity that needs to be stored during the low electricity demand period.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] (1) This application proposes a deep learning architecture for photovoltaic prediction using convolutional neural networks and long short-term memory networks. This architecture can accurately predict the output efficiency of each photovoltaic panel and track the performance of the photovoltaic system after grid connection in real time, thereby supporting the efficient operation of low-inertia, multi-source power grids.

[0038] (2) This application proposes a dynamic performance monitoring and adjustment mechanism that can effectively monitor the abnormal status of the multi-source new power grid, issue status warnings, and make timely dispatching strategies based on the status warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0040] Figure 1This is a flow chart of a photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion proposed by the present invention;

[0041] Figure 2 This is a control center layout diagram of a photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion proposed by the present invention.

[0042] Figure 3 This is a photovoltaic efficiency prediction neural network structure diagram of a photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion proposed by the present invention. DETAILED DESCRIPTION

[0043] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. It should be noted that the terms "front," "rear," "left," "right," "up," and "down" used in the following description refer to directions in the accompanying drawings, and the terms "inward" and "outward" refer to directions toward or away from the geometric center of a particular component, respectively.

[0044] refer to Figure 1 , a photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion, including

[0045] A central control system is used to coordinate and manage the photovoltaic output of each substation. Each substation has an independent monitoring center and sensor network to collect key data from photovoltaic panels and related equipment in real time.

[0046] A deep learning architecture that integrates convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) for processing and fusing image and time series data;

[0047] A feature recognition system based on attention mechanism to enhance the recognition and processing capabilities of key features;

[0048] A fuzzy logic control module is used to automatically adjust the operating parameters of the photovoltaic equipment when performance is detected to be below a preset threshold.

[0049] In this embodiment, the photovoltaic efficiency prediction neural network structure specifically includes:

[0050] Use CNN network to extract photovoltaic panel image features;

[0051] Use LSTM network to capture the long-term dependencies among temperature, current, voltage and light intensity;

[0052] The attention mechanism is used for feature fusion, and the predicted contents of the CNN network and LSTM network are weighted and merged to predict the final result.

[0053] A photovoltaic real-time performance evaluation and early warning system and method based on multi-source information fusion includes the following steps:

[0054] S1. Establish a comprehensive central control system, optimize the layout to support the photovoltaic output of multiple areas, realize centralized management and data integration, and provide a foundation for large-scale photovoltaic monitoring. Figure 2 As shown, since the central control system needs to coordinate and manage the photovoltaic output of each substation, it is best located in the center of the county to facilitate communication and data exchange with each substation. At the same time, various sensors are required for each photovoltaic panel and related equipment, including temperature sensors, current and voltage sensors, and light sensors. These sensors need to be installed in key locations such as photovoltaic panels and inverters to continuously monitor key system indicators.

[0055] The S1 specifically comprises the following steps:

[0056] S11. Install a temperature sensor. Photovoltaic panels are exposed to sunlight during operation, which generates heat over long periods of time. Excessive temperatures may affect the performance and lifespan of the panels. Therefore, by installing a temperature sensor, you can monitor the temperature of the panels in real time and detect temperature anomalies in a timely manner so that you can take appropriate measures, such as adjusting the operating state or lowering the temperature, to protect the panels and related equipment and ensure their normal operation. The specific implementation method is as follows:

[0057] T=T0+K*I ⑴

[0058] Where T represents the temperature of the photovoltaic panel in degrees Celsius, T0 represents the ambient temperature, that is, the temperature of the photovoltaic panel when there is no light, in degrees Celsius, K represents the temperature coefficient, which represents the rate of increase of temperature when the unit light intensity increases, in degrees Celsius / watt (℃ / W), and I represents the light intensity, in watts / square meter (W / m 2 This formula describes the relationship between the temperature of a photovoltaic panel, the ambient temperature, and the light intensity. As light intensity increases, the temperature of the photovoltaic panel will also increase accordingly, but the rate of increase is determined by the temperature coefficient K.

[0059] This recording uses seasonality, recording data every two hours in spring, autumn, and winter, forming the data table [T1, T2, T3, ... T10]. In summer, data is recorded every 15 minutes from 9:00 to 17:00, and every three hours at other times.

[0060] S12. Install current and voltage sensors. These sensors can detect the voltage and current generated by the photovoltaic panels in real time, and record and form a data table [(A1, V1), (A2, V2), (A3, V3), (A4, V4), ... (An, Vn)].

[0061] S13. Install a light sensor. The light sensor can monitor light intensity in real time, helping to evaluate the performance of the photovoltaic system. Light intensity is one of the important factors affecting the power generated by photovoltaic panels. By monitoring changes in light intensity, we can understand how the photovoltaic system works under different lighting conditions. The implementation formula is:

[0062] I 光 =I0*e -k*d ⑵

[0063] Among them I 光 Indicates light intensity in W / m 2 ), I0 represents the solar radiation intensity at the location of the light sensor, the unit is (W / m 2 ), k is the atmospheric attenuation coefficient, which reflects the attenuation of solar radiation by the atmosphere, and d is the solar altitude angle, which represents the angle between the sun's rays and the ground, and usually ranges from 0 to 90 degrees.

[0064] The above data will be used to monitor the operating status indicators of photovoltaic panels in real time and provide data support for photovoltaic efficiency prediction.

[0065] S2. Using convolutional neural networks and long short-term memory networks, a deep learning architecture was designed to accurately predict the output efficiency of individual photovoltaic panels and track the performance of the grid-connected photovoltaic system in real time. During this phase, a CNN and LSTM network were innovatively combined. The CNN network was used to extract the operating status of the photovoltaic panels, while the LSTM network was used to capture the long-term dependencies between temperature, current, voltage, and light intensity. An attention mechanism was also implemented to perform a weighted summation of image and time series features, allowing the model to focus on important features and reduce the impact of irrelevant information, ultimately achieving photovoltaic panel efficiency prediction.

[0066] The S2 specifically comprises the following steps:

[0067] S21. Use CNN network to extract features of photovoltaic panel images. The images of photovoltaic panels contain rich visual information, such as light intensity, shadow distribution, damage, etc. This information can help understand the working status and performance of photovoltaic panels. By extracting image features through CNN, this visual information can be converted into machine-understandable feature vectors, thereby realizing the prediction of photovoltaic efficiency. The specific network structure is as follows: Figure 3As shown in the figure, the network consists of multiple convolutional layers, pooling layers, and fully connected layers to gradually extract and combine abstract features from the image, thereby achieving high-precision prediction of photovoltaic efficiency. First, the photovoltaic panel image (size 224×224×3, representing the RGB channels) passes through three convolutional layers, each using a 3×3 convolution kernel for feature extraction, outputting 32, 64, and 128 feature maps, respectively. Subsequently, the feature maps are downsampled through a max pooling layer to reduce the number of parameters and computational complexity. Next, two fully connected layers are used to further process and map the features, ultimately outputting a single numerical value, namely the predicted photovoltaic panel efficiency. In addition, techniques such as batch normalization and dropout are used to accelerate the training process and improve the model's generalization ability.

[0068] S22. Use LSTM network to capture the long-term dependency of temperature, current, voltage and light intensity. Figure 3 As shown in Figure 1, the network consists of multiple LSTM layers. It achieves accurate prediction of photovoltaic efficiency by progressively learning and integrating the temporal features of time series data. First, time series data such as temperature, current, voltage, and light intensity are used as input. These data are then passed through multiple LSTM layers for temporal feature extraction and encoding. Each LSTM layer contains a large number of LSTM units, effectively capturing long-term dependencies in the data. Subsequently, the outputs of multiple LSTM layers are integrated and weighted together to produce the final time series feature representation. These features capture the complex temporal relationships between data such as temperature, current, voltage, and light intensity, providing rich information for subsequent photovoltaic efficiency prediction.

[0069] S23. Use the attention mechanism for feature fusion. In the model design, the prediction results of the CNN and LSTM networks are combined through the attention mechanism, which can dynamically assign different weights to the prediction results of the two networks, so as to focus more on the features that are critical to photovoltaic efficiency prediction. The implementation formula is as follows:

[0070] e i =v T tanh(W c h c +W c h l +b) ⑶

[0071]

[0072] where h c Represents the prediction result of the CNN network, h l Represents the prediction result of the LSTM network, W c W cRepresent the weight parameters of CNN and LSTM networks respectively, b represents the bias parameter, v is the learnable vector of attention, e i represents the score of each attention weight, α i represents the attention weight, and n represents the number of features. f represents the fused feature representation, h i Denotes the feature representation of each network.

[0073] In this way, by introducing the attention mechanism, the features extracted by CNN and LSTM networks can be more effectively utilized to achieve accurate prediction of large-scale, distributed photovoltaics, and provide guarantees for the safe and efficient operation of low-inertia, multi-source new power grids.

[0074] S3. Set the local photovoltaic output threshold P 阈 , when the device does not reach the output threshold, that is, P 产 <P 阈 , which will trigger the alarm system of the control center of the substation. The central console will automatically adjust the operating parameters of the photovoltaic equipment through fuzzy logic control to cope with environmental changes and internal state changes, ensuring the dynamic consistency and efficient energy management of the system.

[0075] The specific fuzzy control operations are:

[0076] First, for each input variable (such as temperature, current and voltage, light intensity, etc.), a membership function is used to map the actual value to a membership value.

[0077]

[0078] Among them, x is the temperature value, light intensity, voltage and current value, a and b are the left and right boundaries of the triangle membership function. The temperature is usually set to 10-50℃, and the light intensity is usually set to 50W / m 2 -200W / m 2 , for current and voltage, they are usually set to 200mA-800mA, 20V-80V.

[0079] Then, a set of fuzzy rules is defined, which states that when the temperature is “high” and the light is “low” or the temperature is “low” and the light is “high”, the lighting angle can be adjusted appropriately.

[0080] Next, for each fuzzy rule, the intersection between its premise and the membership of the input variable is calculated to obtain the inference result. Here, the minimum operator (min) is used as the intersection operation. Based on the inference result, the lighting angle is adjusted.

[0081] β=min(u 温度 ,u 光照 ,u 电压电流 ) ⑺

[0082] The above inference results are aggregated and the maximum operator (Max) is used to obtain the final fuzzy output for each data item. Finally, the averaging method is used to convert the fuzzy output into specific operating parameter adjustment values, which are then applied to the operating parameters of the photovoltaic system. For example, adjusting the illumination angle to control temperature, light intensity, voltage and current can intelligently adjust the operating state of the photovoltaic system, improve photovoltaic efficiency, and respond to environmental and internal state changes, thereby achieving coordinated control and status warning functions for the photovoltaic system.

[0083] out=mean(max(β)) ⑻

[0084] S4. The total value of solar photovoltaic output in each region will be recorded by the central control system. During periods of low electricity demand, the excess solar energy production capacity in the region will be stored in the form of a virtual power plant, integrating the decentralized solar energy output and energy storage system into a coordinated power supply system. By optimizing the photovoltaic output of each substation through the central control system, it can effectively support the stable and efficient operation of the low-inertia, multi-source new power grid. During peak electricity demand, if it is detected that the electricity demand in a certain area is greater than its solar energy output, but does not reach the expected level, the electricity stored in the virtual power plant will be delivered to the area to alleviate the local power shortage. This process can be expressed by the following formula:

[0085]

[0086] Among them, E 存储 is the stored electrical energy, E 产出 is the electricity produced by solar photovoltaic, E 用电 is the regional electricity demand, and η is the efficiency of the energy storage system. The above formula can be used to calculate the amount of energy required to store during periods of low electricity demand. This formula can guide the virtual power plant system in adjusting energy storage and utilization during periods of low electricity demand, so as to provide sufficient power support during peak demand periods, thereby alleviating local power shortages.

[0087] It should be noted that the above content merely illustrates the technical idea of ​​the present invention and cannot be used to limit the scope of protection of the present invention. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications all fall within the scope of protection of the claims of the present invention.

Claims

1. Photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion, characterized by: It includes a central control system, a deep learning architecture, a feature recognition system based on the attention mechanism, and a fuzzy logic control module; The central control system is used to coordinate and manage the photovoltaic output of each substation. Each substation has an independent monitoring center and sensor network to collect key data from photovoltaic panels and related equipment in real time. A deep learning architecture that integrates convolutional neural networks (CNNs) and long short-term memory (LSTMs) to process and fuse image and time series data; A feature recognition system based on the attention mechanism to enhance the recognition and processing capabilities of key features; Feature fusion based on the attention mechanism is expressed as: ； ； ； in Represents the prediction result of the CNN network, Represents the prediction result of the LSTM network, represents the weight parameters of the CNN network, represents the weight parameter of the LSTM network, b represents the bias parameter, 𝑣 is the learnable vector of attention, represents the score of each attention weight, represents the attention weight, 𝑛 represents the number of features, represents the fused feature representation, Represent the feature representation of each network; a fuzzy logic control module for automatically adjusting the operating parameters of the photovoltaic equipment when performance is detected to be below a preset threshold; The fuzzy logic control module operation is expressed as: First, for each input variable, the membership function is used to map the actual value to the membership value; ； Where 𝑥 is the temperature, light intensity, voltage, and current value, 𝑎 and b are the left and right boundaries of the triangular membership function. For temperature, it is set to 10-50℃, for light intensity, it is set to 50 W / m²-200 W / m², and for current and voltage, it is usually set to 200mA-800mA and 20V-80V; Then, a set of fuzzy rules is defined, which states that when the temperature is "high" and the light is "low" or the temperature is "low" and the light is "high", the lighting angle can be adjusted appropriately; Furthermore, for each fuzzy rule, the intersection between its premise part and the membership degree of the input variable is calculated to obtain the inference result. Here, the minimum operator min is used as the intersection operation, and the illumination angle is adjusted according to the inference result. ； The above inference results are aggregated and the maximum operator Max is used to obtain the final fuzzy output of each data item. Finally, the averaging method is used to convert the fuzzy output into specific operating parameter adjustment values, which are then applied to the operating parameters of the photovoltaic equipment. This can intelligently adjust the operating status of the photovoltaic equipment, improve photovoltaic efficiency, and respond to environmental changes and internal state changes. 。 2. The photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion according to claim 1 is characterized by: The calculated value of the temperature sensor in the sensor network is expressed as: ； Where T represents the temperature of the photovoltaic panel in degrees Celsius. It represents the ambient temperature, that is, the temperature of the photovoltaic panel when there is no light, in degrees Celsius. K represents the temperature coefficient, which represents the rate of increase of temperature when the unit light intensity increases, in degrees Celsius / watt ℃ / W. I represents the light intensity, in watts / square meter W / m², which describes the relationship between the temperature of the photovoltaic panel and the ambient temperature and light intensity. As the light intensity increases, the temperature of the photovoltaic panel will also increase accordingly, but the rate of increase is determined by the temperature coefficient K.

3. The photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion according to claim 1 is characterized by: The calculated value of the light sensor in the sensor network is expressed as: ； in Indicates light intensity in W / m². Indicates the solar radiation intensity at the location of the light sensor, in W / m². k is the atmospheric attenuation coefficient, which reflects the atmospheric attenuation of solar radiation. d is the solar altitude angle, which indicates the angle between the sun's rays and the ground, and usually ranges from 0 to 90 degrees.

4. The method of a photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion according to any one of claims 1 to 3, characterized in that: The method comprises the following steps: S1. Build a comprehensive central control system, optimize the layout to support photovoltaic output in multiple areas, achieve centralized management and data integration, and provide a foundation for large-scale photovoltaic monitoring; S2. Develop a deep learning architecture for multi-source information fusion. Utilizing convolutional neural networks and long short-term memory networks, design a deep learning architecture to accurately predict the output efficiency of individual photovoltaic panels and track the performance of photovoltaic systems in real time after grid connection. S3. Implement a dynamic performance monitoring and adjustment mechanism, set local photovoltaic output thresholds, and automatically adjust operating parameters through fuzzy logic control when equipment performance does not reach the predetermined thresholds, ensuring dynamic consistency and efficient energy management of the system; S4. Configure a virtual power plant and coordinated control system. Through the central control system, establish a virtual power plant model to optimize the coordinated control of photovoltaic output in each substation, and achieve stable and efficient operation of the new low-inertia, multi-source power grid.

5. The method of photovoltaic real-time performance evaluation and early warning system based on multi-source information fusion according to claim 4 is characterized by: The amount of electrical energy stored in the S4 virtual power plant model is expressed as: ； in, is the stored electrical energy, is the electricity produced by solar photovoltaics. is the regional electricity demand, is the efficiency of the energy storage system. The above formula is used to calculate the amount of electrical energy that needs to be stored during the period of low electricity demand.

Images