Photovoltaic power generation adjustable control potential force modeling method fusing multi-modal environment perception

Abstract

This invention discloses a method for quantitative modeling of the controllable potential of photovoltaic power generation, integrating multimodal environmental perception. It relates to the fields of new energy power generation and intelligent power regulation technology. The method first constructs a photovoltaic output correction function that integrates temperature decay coefficient, illuminance, and nonlinear correction terms from a convolutional neural network to accurately calculate the actual output power. Second, it uses a lightweight generative adversarial network to generate extreme weather output scenarios, analyzes minute-level power change rates, and calculates the dynamically adjustable power range by incorporating equipment aging factors. When the predicted output fluctuation exceeds a threshold, it triggers collaborative predictive control between the photovoltaic cluster and the energy storage system to smooth the output power. Finally, it establishes a quantitative model of controllable potential, outputting real-time adjustable range, predicted fluctuation amplitude, ramp rate, and environmental impact sensitivity indicators. This invention achieves dynamic optimization of the entire chain from environment to output to regulation, significantly improving the predictability, adjustability, and grid support capabilities of photovoltaic power generation.

Description

Technical Field

[0001] This invention relates to the fields of new energy power generation and intelligent power regulation technology, and in particular to a method for quantitative modeling of the controllable potential of photovoltaic power generation that integrates multimodal environmental perception. Background Technology

[0002] As a crucial component of clean and renewable energy, photovoltaic (PV) power generation has seen its installed capacity and grid connection ratio continuously increase as the global energy structure transitions towards a low-carbon model. However, PV output is significantly affected by natural environmental factors, exhibiting characteristics such as intermittency, randomness, and volatility, posing substantial challenges to the safe, stable, and economical operation of the power system. Especially under complex meteorological conditions such as cloudy skies, drastic temperature changes, increased humidity, and dust accumulation, the output power of PV modules often exhibits nonlinear decay characteristics, making it difficult for traditional prediction models to accurately characterize its dynamic behavior.

[0003] Currently, photovoltaic power prediction and potential assessment methods mainly rely on historical meteorological data and empirical formulas, and generally employ linear or semi-empirical correction models based on irradiance and temperature. While these models can reflect the impact of major environmental factors to some extent, they often neglect the nonlinear effects caused by the coupling of multiple factors such as humidity, dust accumulation, and module aging, resulting in limited prediction accuracy in complex environments. Furthermore, existing methods lack the ability to generate and simulate minute-level rapid power fluctuations and extreme weather scenarios, making it difficult to accurately assess the actual controllable potential of photovoltaic systems to participate in grid frequency regulation and peak shaving.

[0004] In recent years, with the development of artificial intelligence technology, some studies have attempted to use data-driven methods such as convolutional neural networks (CNNs) and generative adversarial networks (GANs) for photovoltaic power output prediction. However, these methods mostly focus on optimizing a single prediction task and have not yet formed an integrated dynamic modeling framework that combines "environmental perception, power output modeling, and control response." In particular, in areas such as the coordinated operation of photovoltaic and energy storage systems and cluster coordinated control, there is still a lack of complete technical solutions that can quantify the system's adjustable potential in real time and support dynamic grid dispatch.

[0005] Therefore, how to integrate multimodal environmental perception information, construct a photovoltaic power output model that can accurately reflect the coupling effect of complex environment, and realize dynamic evaluation and optimization control of the controllability potential of photovoltaic system on this basis has become a key issue in improving the predictability, adjustability and grid friendliness of photovoltaic power generation, and has important theoretical research and engineering application value. Summary of the Invention

[0006] To address this, this invention provides a method for quantitative modeling of the controllable potential of photovoltaic power generation that integrates multimodal environmental perception. This method solves the problems in existing technologies where traditional models struggle to accurately characterize the nonlinear coupling effects of multiple environmental factors such as temperature, humidity, and dust accumulation, and are unable to dynamically assess the minute-level power ramp-up characteristics and controllable potential boundaries of photovoltaic systems under extreme weather conditions.

[0007] To address the aforementioned technical problems, this invention provides a method for quantitatively modeling the controllable potential of photovoltaic power generation by integrating multimodal environmental perception. This method includes the following steps: Step S1: Construct a photovoltaic output correction function. The correction function integrates the temperature decay coefficient, light intensity, and nonlinear correction terms output by a convolutional neural network. The nonlinear correction terms are generated based at least on real-time humidity and dust accumulation data to establish a correlation model between environmental factors and the electrical characteristics of photovoltaic modules, which is used to dynamically calculate the actual output power of photovoltaic modules. Step S2: Based on real-time collected multimodal environmental data, a lightweight generative adversarial network is used to generate photovoltaic output prediction scenarios under various extreme weather conditions, and the power change rate at the minute level is analyzed. Based on the analysis results, the maximum power ramp-up rate of the photovoltaic system is calculated, and a dynamically adjustable power range considering equipment aging factors is output. Step S3: When the predicted fluctuation of photovoltaic output exceeds the set threshold, the collaborative control mechanism of the edge-side photovoltaic cluster is triggered, and the energy storage system is linked to execute the predictive control optimization strategy to smooth the photovoltaic output power curve. Step S4: Based on the results of steps S1 to S3, establish a quantitative model of the controllable potential of the photovoltaic system under multimodal environmental conditions, and dynamically output the quantitative index of the photovoltaic system's ability to participate in grid frequency regulation and peak regulation.

[0008] Preferably, in step S1, the temperature decay coefficient Based on the material properties of photovoltaic modules and real-time temperature, the calculation formula is as follows: , in, The temperature power coefficient is related to the component material. The average effective temperature is obtained by measuring multiple temperature sensors located at different positions within the photovoltaic array. The reference temperature is set under standard test conditions. The convolutional neural network takes time-series data on humidity, dust accumulation, light intensity, and temperature as input, and learns the nonlinear mapping relationship between these data and photovoltaic power output deviation through training. It then outputs a correction value to correct the theoretical power output. .

[0009] Preferably, in step S1, the photovoltaic output correction function Represented as: , in, The nominal power under standard test conditions. To measure the actual light intensity, The light intensity under standard test conditions. The temperature decay coefficient is This is a correction value for the theoretical power.

[0010] Preferably, in step S2, the lightweight generative adversarial network adopts a dual discriminator structure, including a first discriminator and a second discriminator; the first discriminator is used to determine the authenticity of the generated photovoltaic power output sequence in terms of its time variation trend; the second discriminator is used to determine the physical rationality of the combination of environmental parameters corresponding to the generated sequence; the training objective of the generative adversarial network is to make the generated sample pass the authenticity test of both discriminators simultaneously.

[0011] Preferably, in step S2, the maximum power ramp rate The calculation method is as follows: for minute-level power sampling sequences Calculate the average power change rate during the continuous stable change phase, select the maximum value within the statistical period, and correct for equipment response delay and aging factor, i.e.: , in, The maximum power ramp rate, The sampling interval is... This is the equipment response efficiency factor. This is a decay factor based on the number of years of operation.

[0012] Preferably, in step S2, the dynamically adjustable power range The upper and lower limits are based on the current power. Maximum power ramp rate and prediction time window Confirmed, the calculation formula is: , , and Not lower than the minimum stable output power after considering aging.

[0013] Preferably, in step S3, the predictive control optimization strategy prioritizes power smoothing, and its objective function is: , in, Contribute to photovoltaic forecasting, For the charging and discharging power of the energy storage system, This is the smoothed reference power. For real-time state of charge of energy storage, This is a reference value for the state of charge. The weighting coefficients are used; the optimization process must simultaneously satisfy the energy storage power limit, capacity limit, and response time constraints.

[0014] Preferably, in step S3, the response time constraint requires the energy storage system to have a delay time from receiving a command to a change in power output. satisfy ,in The decision is determined by the type of energy storage; if the limit is exceeded, the backup fast response unit will be activated.

[0015] Preferably, in step S4, the quantitative indicators output by the adjustable potential quantification model include at least: the real-time adjustable power range and the predicted fluctuation amplitude within a specified future time window. Current power ramp rate And sensitivity indicators for environmental impacts such as temperature, humidity, light, and dust accumulation. Among them, the environmental impact sensitivity index Calculated using the following formula: , in, When fixing other environmental parameters, only environmental factors change The resulting change in output power and These are the baseline power and the baseline environmental factor values, respectively.

[0016] This invention also provides an electronic device, which includes a processor, a memory, and a bus system. The processor and the memory are connected through the bus system. The memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to realize the above-described method for quantitative modeling of the controllable potential of photovoltaic power generation by integrating multimodal environmental perception.

[0017] As can be seen from the above technical solutions, this invention application has the following beneficial effects: (1) This invention breaks through the limitations of traditional linear or semi-empirical models. By constructing a composite correction function that integrates physical mechanisms and data-driven approaches, it for the first time couples the nonlinear effects of humidity and dust accumulation learned by convolutional neural networks (CNNs) with dynamic temperature attenuation coefficients and light intensity based on material properties. This innovation solves the problem that existing technologies struggle to characterize the nonlinear interactions of multiple environmental factors (temperature, light, humidity, and dust). At the same time, a generative adversarial network (GAN) with a dual discriminator structure is used to generate physically plausible extreme weather scenarios, enabling the model to predict and adapt to sudden and rare weather conditions. Thus, it can achieve high-precision dynamic calculation of photovoltaic power output at the minute level in various complex and variable actual operating environments.

[0018] (2) This invention proposes for the first time a complete dynamic quantitative assessment chain for potential. By analyzing the minute-level power change rate and innovatively incorporating the equipment response delay factor and long-term aging degradation factor into the calculation, a maximum power ramp-up rate that is closer to the physical limit is obtained. Based on this, the real-time adjustable power range is dynamically calculated, and a series of quantitative results such as predicted fluctuation amplitude and environmental impact sensitivity index are output. This makes the regulation capability of photovoltaic power plants no longer a static or empirical estimate, but a precise quantitative indicator that is updated in real time with the environment and equipment status, providing unprecedented refined data support for grid dispatching and significantly improving the feasibility and economy of photovoltaic participation in ancillary service markets such as frequency regulation and peak shaving.

[0019] (2) This invention does not stop at the level of prediction and evaluation, but further constructs an active control closed loop. When the predicted power output fluctuation exceeds the threshold, the system can intelligently trigger the collaborative control mechanism of the edge-side photovoltaic cluster and link the energy storage system to execute the optimization strategy based on model predictive control (MPC). This strategy aims at power smoothing, and while meeting the energy and power constraints of the energy storage itself, it strictly considers the response time constraint and realizes the real-time optimal allocation of charging and discharging power. This mechanism transforms the traditional passive response to fluctuations into active smoothing control based on advanced prediction, effectively suppressing the impact of photovoltaic power mutations on the power grid, significantly improving the operational stability and security of the power system under high-proportion photovoltaic access, and enhancing the grid support capability of photovoltaic power generation as a friendly power source. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Referring to the drawings will make the features and advantages of the present invention clearer. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1This is a flowchart of a method for quantitative modeling of the controllable potential of photovoltaic power generation that integrates multimodal environmental perception, provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1: To address the shortcomings of existing technologies, traditional models struggle to accurately characterize the nonlinear coupling effects of multiple environmental factors such as temperature, humidity, and dust accumulation, and are unable to dynamically assess the minute-level power ramp-up characteristics and controllable potential boundaries of photovoltaic systems under extreme weather conditions. For example... Figure 1 As shown, this invention proposes a method for quantitative modeling of the controllable potential of photovoltaic power generation that integrates multimodal environmental perception. This method includes the following steps: Step S1: Construct a photovoltaic output correction function. The correction function integrates the temperature decay coefficient, light intensity, and nonlinear correction terms output by a convolutional neural network. The nonlinear correction terms are generated based at least on real-time humidity and dust accumulation data to establish a correlation model between environmental factors and the electrical characteristics of photovoltaic modules, which is used to dynamically calculate the actual output power of photovoltaic modules. Step S2: Based on real-time collected multimodal environmental data, a lightweight generative adversarial network is used to generate photovoltaic output prediction scenarios under various extreme weather conditions, and the power change rate at the minute level is analyzed. Based on the analysis results, the maximum power ramp-up rate of the photovoltaic system is calculated, and a dynamically adjustable power range considering equipment aging factors is output. Step S3: When the predicted fluctuation of photovoltaic output exceeds the set threshold, the collaborative control mechanism of the edge-side photovoltaic cluster is triggered, and the energy storage system is linked to execute the predictive control optimization strategy to smooth the photovoltaic output power curve. Step S4: Based on the results of steps S1 to S3, establish a quantitative model of the controllable potential of the photovoltaic system under multimodal environmental conditions, and dynamically output the quantitative index of the photovoltaic system's ability to participate in grid frequency regulation and peak regulation.

[0023] As can be seen from the above technical solution, this invention proposes a quantitative modeling method for the controllable potential of photovoltaic power generation that integrates multimodal environmental perception. Firstly, by constructing a photovoltaic output correction function that integrates temperature decay, illumination, and nonlinear correction terms from a convolutional neural network, this method accurately characterizes the coupled effects of multiple factors such as humidity and dust accumulation, significantly improving power prediction accuracy in complex environments. Secondly, by generating extreme weather scenarios based on a lightweight generative adversarial network and analyzing minute-level power change rates, it can dynamically assess the system's maximum power ramp-up capability and controllable potential, enhancing adaptability to sudden weather conditions and the real-time nature of potential assessment. Then, when predicted fluctuations exceed the threshold, it triggers collaborative predictive control between the photovoltaic cluster and energy storage, actively smoothing the output curve, effectively suppressing grid impacts and improving operational stability. Finally, by integrating the above steps to establish a quantitative model of controllable potential, it dynamically outputs quantitative indicators of the photovoltaic system's ability to participate in grid frequency regulation and peak shaving, providing a direct and scientific decision-making basis for grid optimization scheduling.

[0024] This invention provides a method for quantitative modeling of the controllable potential of photovoltaic power generation by integrating multimodal environmental perception. Its core lies in the integration of multi-dimensional environmental perception and intelligent algorithms to achieve accurate dynamic modeling of the output characteristics of photovoltaic systems, adaptability to extreme operating conditions, real-time quantitative assessment of future controllable potential, and coordinated optimization control with energy storage systems, thereby significantly improving the predictability, controllability, and grid support capabilities of photovoltaic power generation.

[0025] Furthermore, in step S1, the present invention first constructs a photovoltaic power output correction function that comprehensively considers multimodal environmental factors.

[0026] Specifically, the system uses the standard test conditions (STC) of photovoltaic modules as a benchmark to collect environmental parameters in real time, including illuminance (G), module temperature (T), ambient humidity (H), and dust accumulation (D). Illuminance is measured by a radiometer installed on the photovoltaic array plane, and the module temperature is obtained by averaging its values ​​using multiple temperature sensors placed at key locations in the array (such as the center and edges). Humidity is obtained through an ambient humidity sensor, while dust accumulation is obtained through a regularly calibrated optical sensor (such as a reflectance meter) or an image recognition-based dust coverage detection device.

[0027] Temperature decay coefficient ( The calculation of ) is crucial in this step. It does not use fixed empirical values, but is dynamically calculated based on the specific material properties of the photovoltaic module and the real-time temperature. The formula is as follows: , in, The temperature power coefficient of the module material (e.g., approximately -0.004 / °C for monocrystalline silicon, approximately -0.0045 / °C for polycrystalline silicon, and approximately -0.0025 / °C for thin-film modules). The reference temperature for standard test conditions (typically 25°C) is used. Multiple temperature sensors are employed to acquire the temperature. It effectively reflects the temperature field non-uniformity caused by installation location, wind speed, and obstruction, and significantly improves the correction accuracy compared to single-point measurement.

[0028] For factors such as humidity and dust accumulation that are difficult to characterize with linear models, this invention introduces a convolutional neural network (CNN) to extract their nonlinear effects. First, the system needs to collect historical operating data as a training set, with a data sampling period of 1 minute, continuously recording for at least 90 days, including time-series data of humidity (H), dust accumulation (D), light intensity (G), and temperature (T) and their corresponding actual output power (H). The training objective is to enable the CNN to learn environmental parameters and the theoretical power calculated by a traditional linear model (based only on G and T). Deviation between ) ).

[0029] When the trained CNN model is run in real time, it receives time-series sequences. The data is processed through its convolutional layers to extract local spatiotemporal correlations of environmental features, ultimately outputting a nonlinear correction term. Therefore, the actual output power of photovoltaic modules ( The following correction function is used for dynamic calculation: , in, and These represent the nominal power and irradiance of the component under standard test conditions, respectively. This formula breaks through the limitations of traditional linear or semi-empirical models, and for the first time integrates the nonlinear effects of humidity and dust accumulation learned by CNN with physically defined temperature and illumination correction terms, achieving high-precision modeling of environmental coupling effects.

[0030] Furthermore, in step S2, in order to evaluate the system's performance under sudden environmental changes and quantify its dynamic adjustment potential, the present invention utilizes a lightweight generative adversarial network (GAN) to generate extreme weather scenarios.

[0031] This GAN employs an innovative dual-discriminator structure, comprising a generator (G), a time-series discriminator (D1), and an environmental physics discriminator (D2). The generator takes a random noise vector and a current environmental data fragment as input to generate a simulated future photovoltaic (PV) output sequence and corresponding virtual environmental parameters. D1 focuses on determining whether the generated PV output sequence's temporal characteristics, such as minute-level fluctuations and autocorrelation, are consistent with real historical extreme weather data. D2 focuses on determining whether the combination of environmental parameters corresponding to the generated sequence (such as a sudden drop in sunlight accompanied by a slight increase in temperature) conforms to physical laws, avoiding the generation of physically impossible scenarios.

[0032] During training, the generator strives to produce data that can simultaneously "fool" both D1 and D2, while the two discriminators independently optimize their own discriminative abilities. This structure ensures that the generated extreme scenarios possess both realistic historical statistical characteristics and conform to environmental physical constraints, thereby significantly improving the model's adaptability to rare but possible sudden weather events (such as fast-moving clouds or sandstorms).

[0033] Based on various extreme scenarios generated by GAN, the system performs statistical analysis of power change rates on a minute-by-minute basis. For the power sequence P(t), its minute-by-minute differences are calculated, and a sliding window method is used for smoothing to eliminate noise. The system identifies all stable phases where power changes continuously in the same direction and calculates their average rate of change. The maximum power ramp rate (…) It is determined by the following formula: , in, The sampling interval is 1 minute. This indicates that the maximum value is taken within a statistical period (e.g., 24 hours). The efficiency factor (typically 0.95~0.99) is used to characterize the response hysteresis of devices such as inverters. An aging factor based on the component's service life and degradation curve (e.g., an average annual degradation of 0.5%, then after N years of operation). This calculation method is the first to incorporate the dynamic response characteristics of the equipment and long-term aging factors into the ramp-up capability assessment, making the results closer to the physical limits of the actual system.

[0034] Furthermore, the system can determine the future. Dynamically adjustable power range within a time period (e.g., 10 minutes) : , , in, This represents the current real-time power. The maximum possible output of the component at present (limited by lighting conditions). This represents the minimum stable operating power of the inverter. This range is dynamically updated, providing a clear indication of the safe power range that the photovoltaic system can adjust over a short period of time.

[0035] Furthermore, in step S3, when the photovoltaic power output fluctuation amplitude predicted based on the aforementioned model (such as the standard deviation of the power change rate within the next 5 minutes) exceeds a preset threshold (for example, twice the standard deviation of the historical fluctuation rate), the system triggers the edge-side photovoltaic cluster collaborative control mechanism.

[0036] This mechanism includes a cluster-level communication module and a local control module. Upon detecting excessive fluctuations, the communication module collects the current output, adjustable margin, and status of connected energy storage units within the cluster within milliseconds. The master control node calculates the total power adjustment required to smooth the overall output curve using a power balance algorithm and optimally allocates it to each unit, prioritizing units with large adjustable margins and low operating temperatures.

[0037] Simultaneously, the energy storage system executes a predictive control optimization strategy. This strategy constructs a rolling time-domain optimization problem whose objective function aims to minimize fluctuations in grid-connected photovoltaic power and maintain the energy storage's state of charge (SOC) within a healthy range: , in, Contribute to photovoltaic forecasting, The charging and discharging power of the energy storage system (discharging is positive). The smoothed target power (usually (moving average) The target state of charge (e.g., 50%). These are the weighting coefficients. Optimization must satisfy the following constraints: , and response time constraints (For example, lithium-ion batteries) (seconds). If the calculation finds that the response time constraint cannot be met, the system will instruct the activation of fast-response units such as supercapacitors as a supplement. By solving this optimization problem in real time, the energy storage system dynamically adjusts the charging and discharging power to achieve active smoothing of the photovoltaic output curve and effectively suppress grid impact.

[0038] Furthermore, in step S4, by integrating the results of the above-mentioned environmental coupling modeling, potential dynamic assessment and synergistic enhancement control, the system finally establishes and outputs a quantitative model of the controllable potential of the photovoltaic system.

[0039] This model not only outputs a dynamically adjustable power range It also outputs a series of quantitative indicators: 1. Predict the fluctuation range ( The standard deviation of the power sequence generated by GAN in multiple scenarios is calculated to reflect the uncertainty of power output in future periods.

[0040] 2. Current power ramp rate ( ): Real-time calculated minute-level power change rate, used to monitor the instantaneous dynamics of the system.

[0041] 3. Environmental impact sensitivity indicators ( The sensitivity of four key environmental factors—temperature, humidity, light intensity, and dust accumulation—was calculated using the controlled variable method. , This indicator quantifies the percentage change in power caused by a unit percentage change in each environmental factor, which helps maintenance personnel identify the most pressing environmental risks (such as high dust sensitivity indicating the need to clean components).

[0042] These structured and quantitative outputs provide precise data support for the power grid dispatching system, enabling it to dynamically assess the real-time capability boundaries of photovoltaic power plants in frequency regulation and peak shaving, and to make scientific dispatching decisions.

[0043] Example 2: This embodiment of the invention provides an electronic device, which includes a processor, a memory, and a bus system. The processor and the memory are connected through the bus system. The memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to realize the above-mentioned method for quantitative modeling of the controllable potential of photovoltaic power generation by integrating multimodal environmental perception.

[0044] In summary, this invention constructs a high-precision environment-output coupling model by integrating physical models and data-driven models (CNN, GAN); it achieves accurate quantification of the system's dynamic adjustment potential by introducing equipment response and aging factors; and it realizes proactive suppression of power fluctuations by designing a model predictive control-based energy storage collaborative optimization strategy. The entire method forms a closed loop from "multimodal perception" to "potential quantification" and then to "collaborative control," significantly improving the system friendliness and grid dispatchability of photovoltaic power generation, and possesses outstanding inventiveness and engineering application value.

[0045] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0046] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0047] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception, characterized in that, Includes the following steps: Step S1: Construct a photovoltaic output correction function. The correction function integrates the temperature decay coefficient, light intensity, and nonlinear correction terms output by a convolutional neural network. The nonlinear correction terms are generated based at least on real-time humidity and dust accumulation data to establish a correlation model between environmental factors and the electrical characteristics of photovoltaic modules, which is used to dynamically calculate the actual output power of photovoltaic modules. Step S2: Based on real-time collected multimodal environmental data, a lightweight generative adversarial network is used to generate photovoltaic output prediction scenarios under various extreme weather conditions, and the power change rate at the minute level is analyzed. Based on the analysis results, the maximum power ramp-up rate of the photovoltaic system is calculated, and a dynamically adjustable power range considering equipment aging factors is output. Step S3: When the predicted fluctuation of photovoltaic output exceeds the set threshold, the collaborative control mechanism of the edge-side photovoltaic cluster is triggered, and the energy storage system is linked to execute the predictive control optimization strategy to smooth the photovoltaic output power curve. Step S4: Based on the results of steps S1 to S3, establish a quantitative model of the controllable potential of the photovoltaic system under multimodal environmental conditions, and dynamically output the quantitative index of the photovoltaic system's ability to participate in grid frequency regulation and peak regulation.

2. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 1, characterized in that, In step S1, the temperature decay coefficient Based on the material properties of photovoltaic modules and real-time temperature, the calculation formula is as follows: ， in, The temperature power coefficient is related to the component material. The average effective temperature is obtained by measuring multiple temperature sensors located at different positions within the photovoltaic array. The reference temperature is set under standard test conditions. The convolutional neural network takes time-series data on humidity, dust accumulation, light intensity, and temperature as input, and learns the nonlinear mapping relationship between these data and photovoltaic power output deviation through training. It then outputs a correction value to correct the theoretical power output. .

3. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 1 or 2, characterized in that, In step S1, the photovoltaic output correction function Represented as: ， in, The nominal power under standard test conditions. To measure the actual light intensity, The light intensity under standard test conditions. The temperature decay coefficient is This is a correction value for the theoretical power.

4. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 1, characterized in that, In step S2, the lightweight generative adversarial network adopts a dual discriminator structure, including a first discriminator and a second discriminator; the first discriminator is used to determine the authenticity of the generated photovoltaic power output sequence in terms of its time variation trend; the second discriminator is used to determine the physical rationality of the combination of environmental parameters corresponding to the generated sequence; the training objective of the generative adversarial network is to make the generated sample pass the authenticity test of both discriminators at the same time.

5. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 1, characterized in that, In step S2, the maximum power ramp rate The calculation method is as follows: for minute-level power sampling sequences Calculate the average power change rate during the continuous stable change phase, select the maximum value within the statistical period, and correct for equipment response delay and aging factor, i.e.: ， in, The maximum power ramp rate, The sampling interval is... This is the equipment response efficiency factor. This is a decay factor based on the number of years of operation.

6. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 5, characterized in that, In step S2, the dynamically adjustable power range The upper and lower limits are based on the current power. Maximum power ramp rate and prediction time window Confirmed, the calculation formula is: ， ， and Not lower than the minimum stable output power after considering aging.

7. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 1, characterized in that, In step S3, the predictive control optimization strategy prioritizes power smoothing, and its objective function is: ， in, Contributing to photovoltaic forecasting For the charging and discharging power of the energy storage system, This is the smoothed reference power. For real-time state of charge of energy storage, This is a reference value for the state of charge. The weighting coefficients are used; the optimization process must simultaneously satisfy the energy storage power limit, capacity limit, and response time constraints.

8. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 7, characterized in that, In step S3, the response time constraint requires the energy storage system to have a delay time from receiving a command to a change in power output. satisfy ,in The decision is determined by the type of energy storage; if the limit is exceeded, the backup fast response unit will be activated.

9. The method for quantitative modeling of the controllable potential of photovoltaic power generation integrating multimodal environmental perception as described in claim 1, characterized in that, In step S4, the quantitative indicators output by the adjustable potential quantification model include at least: the real-time adjustable power range and the predicted fluctuation amplitude within a specified future time window. Current power ramp rate And sensitivity indicators for environmental impacts such as temperature, humidity, light, and dust accumulation. Among them, the environmental impact sensitivity index Calculated using the following formula: ， in, When fixing other environmental parameters, only environmental factors change The resulting change in output power and These are the baseline power and the baseline environmental factor values, respectively.

10. An electronic device, characterized in that, The electronic device includes a processor, a memory, and a bus system. The processor and the memory are connected through the bus system. The memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the photovoltaic power generation controllable potential quantitative modeling method integrating multimodal environmental perception as described in any one of claims 1 to 9.

Images