Photovoltaic control and regulation method and system applied to 5g integrated power supply

Abstract

The present application relates to photovoltaic and energy storage collaborative control technical field, especially to be applied to 5G integrated power supply photovoltaic control adjustment method and system, the present application proposes the following scheme, the photovoltaic characteristic parameter and corresponding time mark data are collected, three-dimensional curved surface model is constructed to carry out differential geometry curvature analysis, according to the curvature analysis result, the light change event type is determined and the photovoltaic maximum power point is positioned, according to the photovoltaic maximum power point, the photovoltaic output parameter adjustment and energy storage charging and discharging strategy is generated, the present application realizes the quick response to the quick fluctuation of light through the fusion of multidimensional characteristic parameters, adopts the positioning strategy of event-driven, improves the maximum power point tracking accuracy and dynamic stability, reduces power oscillation and energy loss, realizes the collaborative optimization of photovoltaic, energy storage and load.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic and energy storage collaborative control, and particularly relates to a photovoltaic control and regulation method and system applied to a 5G integrated power supply. BACKGROUND

[0002] Under the background of large-scale deployment of 5G communication networks, integrated power supply as a core power supply guaranteeing device needs to consider both energy efficient utilization and power supply stability. The control accuracy and response efficiency of photovoltaic components directly affect the system performance. However, the current technology still faces core challenges such as low multi-parameter fusion degree, poor dynamic adaptability, and rigid collaborative scheduling. First, photovoltaic systems involve multi-dimensional parameters. Due to differences in collection frequency and feature dimension, existing technologies cannot efficiently fuse and mine, resulting in incomplete characterization of the running state and inability to accurately capture the dynamic correlation between parameters. Second, in complex environments, light changes are diverse, and it is difficult to distinguish the nature of the event by only using power change rate, resulting in insufficient pertinence of power prediction and control strategy, which easily leads to output fluctuations. In addition, existing maximum power point tracking mostly uses fixed strategies, which cannot adjust tracking accuracy and response speed according to dynamic scenarios.

[0003] For example, a Chinese patent with the publication number CN116388279B discloses a grid-connected control method and system for a solar photovoltaic power generation system. The control method collects photovoltaic power generation historical data and grid-connected historical data, trains a neural network model based on historical environmental data, estimates future power generation power curve and power consumption curve, and trains a reinforcement learning model to decide grid-connected time and energy storage power by constructing a four-tuple set. The disclosed embodiment is driven by historical time series data, which improves grid-connected stability. For example, a Chinese patent with the publication number CN117498517B discloses a photovoltaic inverter power supply control system and method. The photovoltaic inverter power supply control method adjusts the voltage according to the preset voltage range by monitoring the output voltage of the photovoltaic power generation component, and controls the component to alternate start and stop by monitoring the temperature of the inverter component through the temperature detection module. The disclosed embodiment realizes the basic protection of the inverter power supply by linking the fixed voltage reduction range and the temperature threshold.

[0004] The above existing technologies all have the problems raised in the background: low multi-parameter fusion degree, poor dynamic adaptability, and weak collaborative control correlation, so it is difficult to meet the core needs of photovoltaic control accuracy, response speed, and power supply stability in the deployment scenario of 5G integrated power supply. To solve the above problems, the present application provides a photovoltaic control and regulation method and system applied to a 5G integrated power supply. SUMMARY

[0005] The technical problem to be solved by the present application is to provide a photovoltaic control and adjustment method and system applied to a 5G integrated power supply, which determines the light change event type and locates the photovoltaic maximum power point through differential geometry curvature analysis of a three-dimensional curved surface model, generates photovoltaic output parameter adjustment and energy storage charging and discharging strategies according to the photovoltaic maximum power point, and realizes fast response to rapid light fluctuation through deep fusion of multi-dimensional photovoltaic characteristic parameters, improves the maximum power point tracking accuracy and dynamic stability by using an event-driven positioning strategy, reduces power oscillation and energy loss, and realizes the collaborative optimization of photovoltaic, energy storage and load.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0007] The photovoltaic control and adjustment method applied to the 5G integrated power supply comprises:

[0008] Collecting photovoltaic characteristic parameters and corresponding time mark data, wherein the photovoltaic characteristic parameters include photovoltaic real-time parameters, energy storage state parameters and load parameters;

[0009] According to the photovoltaic real-time parameters, a three-dimensional curved surface model is constructed for differential geometry curvature analysis, the Gaussian curvature is calculated, and the light intensity change rate is obtained according to the Gaussian curvature;

[0010] According to the load parameters, a load demand matrix is constructed, and the light change event type is determined in combination with the photovoltaic real-time parameters, the Gaussian curvature and the light intensity change rate;

[0011] According to the Gaussian curvature, a time series prediction model is constructed to generate a future photovoltaic supply matrix, the virtual shading feature is extracted according to the Gaussian curvature, the future photovoltaic supply matrix is corrected, and a photovoltaic output matrix is obtained;

[0012] When the light change event type meets the preset triggering condition, according to the light change event type, a corresponding positioning strategy is adopted to locate the photovoltaic maximum power point within a preset time threshold, and the preset time threshold is determined according to the load priority in the load parameters and the light intensity change rate;

[0013] According to the photovoltaic maximum power point, the energy storage state parameters, the load parameters and the photovoltaic output matrix, the photovoltaic output parameters are dynamically adjusted and the energy storage charging and discharging strategy is generated.

[0014] The photovoltaic characteristic parameters are collected, which comprises:

[0015] The photovoltaic real-time parameters, including photovoltaic array voltage, current and light intensity, are collected;

[0016] The energy storage state parameter includes a state of charge value, voltage, current and battery pack temperature of the energy storage battery;

[0017] The load parameter includes core load power, flexible load power, real-time power demand and load priority;

[0018] The time mark data corresponds to each parameter acquisition time.

[0019] According to the photovoltaic real-time parameter, a three-dimensional curved surface model is constructed for differential geometry curvature analysis, and Gaussian curvature is calculated, including:

[0020] According to the photovoltaic array voltage, current and time mark data in the photovoltaic real-time parameter, the photovoltaic power at each sampling time is calculated, and a photovoltaic power data sequence is formed in a sliding time window;

[0021] Through data filtering, outlier elimination and data standardization processing, the photovoltaic power data sequence, photovoltaic array voltage and time mark data are preprocessed to obtain standardized photovoltaic power data sequence, standardized photovoltaic array voltage and standardized time mark data;

[0022] A three-dimensional curved surface model is constructed by using a polynomial regression method to fit the standardized photovoltaic power data sequence, standardized photovoltaic array voltage and standardized time mark data;

[0023] According to the three-dimensional curved surface model, the first-order partial derivative and the second-order partial derivative are calculated, and the first fundamental form coefficient and the second fundamental form coefficient are calculated according to the principle of differential geometry, and the Gaussian curvature is calculated according to the first fundamental form coefficient, the second fundamental form coefficient and the Gaussian curvature formula.

[0024] According to the Gaussian curvature, the illumination intensity change rate is obtained, including:

[0025] According to the Gaussian curvature, a mapping relationship model of Gaussian curvature and illumination intensity change rate is constructed;

[0026] According to the illumination intensity in the photovoltaic real-time parameter, the illumination intensity change characteristic is calculated, and according to the value range of the Gaussian curvature and the illumination intensity change characteristic, the segmented intervals of the mapping relationship model are divided;

[0027] According to the Gaussian curvature in each segmented interval and its corresponding illumination intensity change range, the correlation law is determined, and the linear conversion equation of the mapping relationship model is constructed according to the correlation law, and the coefficients of the linear conversion equation are determined according to theoretical derivation verification;

[0028] Based on the segmented intervals and linear transformation equations of the mapping relationship model, the initial rate of change of light intensity is obtained. The initial rate of change of light intensity is then dynamically verified using the light intensity in the real-time photovoltaic parameters to obtain the final rate of change of light intensity.

[0029] Based on the aforementioned real-time photovoltaic parameters, Gaussian curvature, rate of change of illuminance, and load demand matrix, the types of illuminance change events are determined, including:

[0030] Based on the real-time photovoltaic parameters, Gaussian curvature, rate of change of irradiance, and load demand matrix, the third derivative, spatial curvature characteristics, irradiance dynamic characteristics, and load power fluctuation coefficient of the photovoltaic power curve are calculated respectively.

[0031] The third derivative, spatial curvature feature, illumination dynamic feature, and load power fluctuation coefficient of the photovoltaic power curve are fused using a weighted fusion algorithm to obtain a multidimensional feature vector. The weight of the spatial curvature feature is determined based on the area of ​​the abnormal region of the Gaussian curvature, the weight of the illumination dynamic feature is determined based on the amplitude of the change rate of illumination intensity, and the weight of the load power fluctuation coefficient is dynamically adjusted based on the load priority in the load parameters.

[0032] Based on the numerical range and weight ratio of each feature component in the multidimensional feature vector, a discrimination criterion for illumination change events is constructed.

[0033] The type of illumination change event is determined according to the discrimination criterion, and the type of illumination change event includes local shadow events, rapid cloud movement events, and steady-state events.

[0034] Based on the Gaussian curvature, virtual shading features are extracted to correct the future photovoltaic supply matrix. This correction includes power attenuation coefficient superposition correction and time phase calibration correction. The virtual shading features include virtual shading area and virtual shading duration, wherein:

[0035] The power attenuation coefficient superposition correction is based on the virtual shading area to construct a piecewise power attenuation model, determine the attenuation function type through an adaptive function selection mechanism, and combine a multi-layer superposition algorithm to attenuate the photovoltaic power value of the future photovoltaic supply matrix. The virtual shading area is calculated based on the area of ​​abnormal regions exceeding a preset threshold in the spatial distribution of the Gaussian curvature.

[0036] The time phase calibration correction is constructed based on the virtual shading duration. The time phase correction model rearranges the time series of the future photovoltaic supply matrix using a sliding window technique and combines an interpolation algorithm to ensure the data continuity before and after the time series rearrangement. The virtual shading duration is determined based on the continuous span of the abnormal region area on the time axis.

[0037] The correction results of the power attenuation coefficient superposition correction and the time phase calibration correction are coupled and calculated to obtain the corrected future photovoltaic supply matrix, which is used as the photovoltaic output matrix.

[0038] The preset time threshold is determined based on the load priority and the rate of change of light intensity in the load parameters, including:

[0039] A threshold reference range is determined based on the rate of change of light intensity, and the lower limit of the threshold reference range increases with the rate of change of light intensity.

[0040] Based on the load priority in the load parameters, the threshold reference interval is dynamically corrected to obtain the corrected threshold reference interval. The upper limit of the corrected threshold reference interval decreases as the load priority increases, and the lower limit is greater than or equal to the minimum response time of the system hardware.

[0041] The median of the corrected threshold reference interval is used as the preset time threshold.

[0042] Based on the type of illumination change event, a corresponding positioning strategy is adopted to locate the new photovoltaic maximum power point. The positioning strategy includes a first positioning strategy, a second positioning strategy, and a third positioning strategy, wherein:

[0043] The first positioning strategy corresponds to the illumination change event type of local shadow event. The voltage sub-interval is divided according to the photovoltaic output matrix. By increasing the division density and scanning frequency of the voltage sub-interval, the photovoltaic maximum power point is located. The division density and scanning frequency of the voltage sub-interval increase with the increase of the virtual shading area.

[0044] The second positioning strategy corresponds to the light change event type of cloud rapid movement event. A photovoltaic power trend prediction curve is constructed based on the future photovoltaic supply matrix. Based on the photovoltaic power trend prediction curve, Gaussian curvature and virtual shading duration, the maximum photovoltaic power point is located by dynamically adjusting the frequency of the scanning step and the response speed. The frequency of the scanning step increases as the virtual shading duration decreases, and the response speed increases as the absolute value of the Gaussian curvature increases.

[0045] The third positioning strategy corresponds to a steady-state event type of light change event. The current photovoltaic maximum power point is taken as the new photovoltaic maximum power point and periodically corrected. The correction frequency of the periodic correction decreases as the rate of change of light intensity decreases.

[0046] The dynamic adjustment of photovoltaic output parameters and generation of energy storage charging and discharging strategies include:

[0047] The energy storage charging and discharging power is determined based on the state of charge value, battery temperature, and maximum allowable charging and discharging power in the energy storage state parameters. The maximum allowable charging and discharging power is calculated based on the voltage, current, and battery rated parameters in the energy storage state parameters.

[0048] Based on the core load power in the load parameters and the state of charge value in the energy storage state parameters, the flexible load power in the load parameters is corrected to obtain the corrected flexible load power.

[0049] The target output power of the photovoltaic system is calculated based on the core load power, energy storage charging and discharging power, corrected flexible load power, photovoltaic maximum power point, and photovoltaic output matrix in the load parameters.

[0050] The photovoltaic output parameters are adjusted according to the photovoltaic target output power. When there is a deviation between the actual photovoltaic output power and the photovoltaic target output power, the power prediction value of the sliding time window is obtained according to the photovoltaic output matrix. The energy storage charging and discharging power adjustment amount is calculated according to the battery pack temperature and power prediction value, and the energy storage charging and discharging power is dynamically adjusted according to the adjustment amount.

[0051] A photovoltaic control and regulation system for 5G integrated power supplies includes a parameter acquisition module, a differential calculation module, an analysis and judgment module, a power positioning module, and a coordinated regulation module, wherein:

[0052] The parameter acquisition module includes a voltage sensor, a current sensor, a light sensor, and a temperature sensor, and is used to acquire real-time photovoltaic parameters, energy storage status parameters, and load parameters.

[0053] The differential calculation module is used to construct a three-dimensional surface model, perform differential geometric curvature analysis, calculate Gaussian curvature, and obtain the rate of change of light intensity based on the Gaussian curvature.

[0054] The analysis and judgment module is used to analyze and determine the type of light change event, and to perform preset trigger condition judgment and preset time threshold calculation.

[0055] The power positioning module is used to locate the photovoltaic maximum power point by adopting the corresponding positioning strategy according to the type of light change event.

[0056] The coordinated adjustment module is used to perform photovoltaic output parameter adjustment and energy storage charging and discharging strategy generation, and output control signals to the photovoltaic inverter and energy storage battery management system.

[0057] The power positioning module and the collaborative adjustment module interact directly through an internal high-speed data bus that supports multi-priority scheduling. The output of the differential calculation module is fed into the analysis and judgment module and the collaborative adjustment module in parallel.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] This invention constructs a three-dimensional surface model using multi-dimensional photovoltaic characteristic parameters for differential geometric curvature analysis, achieving full-dimensional perception and intelligent prediction of the photovoltaic power generation environment. Through virtual shading feature extraction and future photovoltaic supply matrix correction, it effectively compensates for the impact of shading, significantly improving the accuracy of irradiance change prediction and the system's intelligence level. By determining the type of irradiance change event, this invention adopts an event-driven positioning strategy to improve the accuracy and dynamic stability of photovoltaic maximum power point tracking, reduce power oscillation and energy loss, effectively balance power generation supply and load demand, and significantly improve the real-time responsiveness, power generation efficiency, and operational stability of the 5G integrated power system. Attached Figure Description

[0060] Figure 1 This is a flowchart of the photovoltaic control and regulation method applied to 5G integrated power supply in this invention;

[0061] Figure 2 This is a flowchart illustrating the steps of dynamically adjusting photovoltaic output parameters and generating an energy storage charging and discharging strategy according to the present invention.

[0062] Figure 3 This is a schematic diagram of a three-dimensional surface model in an embodiment of the present invention;

[0063] Figure 4 A flowchart of the steps for determining the type of illumination change event provided by the present invention;

[0064] Figure 5 This is a schematic diagram illustrating the method for determining the type of illumination change event according to the present invention;

[0065] Figure 6 This is an overall framework diagram of the photovoltaic control and regulation system applied to the 5G integrated power supply in this invention. Detailed Implementation

[0066] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof.

[0067] The term "and / or" in the following text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0068] This application provides a photovoltaic control method, including:

[0069] Collect photovoltaic characteristic parameters and their corresponding time-stamped data. The photovoltaic characteristic parameters include real-time parameters, status parameters and load parameters.

[0070] Based on real-time parameters, a three-dimensional surface model is constructed for differential geometric curvature analysis, Gaussian curvature is calculated, and intensity change rate is obtained based on Gaussian curvature.

[0071] Based on the load parameters, a demand matrix is ​​constructed, and the types of change events are determined by combining real-time parameters, Gaussian curvature, and intensity change rate.

[0072] A time series prediction model is constructed based on Gaussian curvature to generate a future supply matrix. Virtual occlusion features are extracted based on the Gaussian curvature to correct the future supply matrix and obtain the output matrix.

[0073] When the type of change event meets the preset triggering conditions, the maximum power point is located within the preset time threshold, which is determined based on the load parameters and the intensity change rate.

[0074] Based on the maximum power point, state parameters, load parameters, and output matrix, the output parameters are dynamically adjusted and a charging / discharging strategy is generated.

[0075] For example, this embodiment uses a 5G integrated power system as an example to specifically illustrate the above photovoltaic control method.

[0076] In existing technologies, maximum power point tracking (MPPT) of photovoltaic (PV) systems typically employs two methods. The first method uses a fixed-step perturbation observation approach, periodically subjecting the PV array's operating voltage to minute perturbations and observing changes in output power to determine the direction of MPPT movement. This method ensures basic MPPT functionality with low computational complexity under relatively stable illumination conditions. However, this method suffers from a lag in response to changes in illumination, and the system's tracking accuracy is primarily affected by the perturbation step size and the rate of environmental change. This lag impacts the system's tracking performance in dynamic environments.

[0077] Method two involves using the incremental conductance method for maximum power point tracking (MPPT). This method determines the location of the maximum power point by comparing the instantaneous conductance and incremental conductance of the photovoltaic array. While this method theoretically offers high tracking accuracy, in practical applications, it suffers from high computational complexity due to the need for numerous differential operations. Furthermore, it remains prone to slow tracking speed and getting trapped in local optima when illumination conditions vary significantly. Therefore, existing technologies often struggle to achieve a perfect balance between improving tracking accuracy, reducing response latency, and minimizing computational complexity. This is particularly true in application scenarios like 5G base stations, where extremely high power supply stability requirements exist, making it difficult for current photovoltaic control methods to meet the dual demands of rapid response and efficient power generation.

[0078] To achieve efficient and stable operation of photovoltaic power generation in integrated power supply systems for 5G base stations, and to improve the adaptability of the control system to complex lighting environments, while also maximizing power generation efficiency through rapid response to changes in lighting, this application provides a photovoltaic control and regulation method for integrated 5G power supplies based on the aforementioned photovoltaic control method. Figure 1 As shown, the execution flow of this method includes:

[0079] S1: Collect photovoltaic characteristic parameters and their corresponding time stamp data. The photovoltaic characteristic parameters include real-time photovoltaic parameters, energy storage status parameters, and load parameters.

[0080] In this step, during the operation of the integrated power system for the 5G base station, photovoltaic characteristic parameters and their corresponding time-stamped data are collected, including real-time photovoltaic parameters, energy storage status parameters, and load parameters. Compared to traditional single-parameter monitoring methods, photovoltaic characteristic parameter acquisition covers parameter monitoring from three dimensions: the generation side, the energy storage side, and the load side. This approach significantly improves data integrity and the time synchronization accuracy of multi-source data, while also enhancing the system's environmental awareness and anomaly identification capabilities. A high-quality, highly consistent data foundation provides reliable support for subsequent control steps such as differential curvature geometry analysis, identification of illumination change event types, and photovoltaic maximum power point location, thereby ensuring the reliability and accuracy of the overall control system.

[0081] Specifically, the real-time photovoltaic parameters include photovoltaic array voltage, current, and illuminance. First, a voltage sensor monitors the photovoltaic array output, converting the DC voltage signal into a standard analog signal using electromagnetic induction. This signal is then digitally processed by a high-precision converter to obtain the accurate photovoltaic array voltage. Next, a current sensor monitors the photovoltaic array output circuit, converting the current flowing through the conductor into a resistance change based on electromagnetic effects. This resistance change is then converted back into a voltage signal via a bridge circuit, amplified and filtered by a signal conditioning circuit, and finally converted to an A / D converter to obtain the accurate photovoltaic array current. Finally, an illuminance sensor monitors the photovoltaic array plane. The photosensitive element generates a photocurrent under illumination, which is proportional to the incident light intensity. This photocurrent is converted into a voltage signal by a signal amplifier, corrected by a temperature compensation circuit, and outputs a standard signal linearly related to the illuminance. The illuminance is then obtained based on this standard signal.

[0082] Specifically, the energy storage state parameters include the state of charge (SOC) value, voltage, current, and battery pack temperature of the energy storage battery. First, the SOC value is estimated using a combination of charge measurement and voltage detection methods. A high-precision current sensor continuously monitors the charging and discharging current, and the battery capacity change is calculated through integration. Simultaneously, the battery open-circuit voltage is periodically measured for SOC calibration. A filtering algorithm is used to fuse the results of the two methods to obtain the SOC value. Next, a voltage sensor directly measures the terminal voltage of the energy storage battery pack. A voltage divider circuit and isolation amplifier convert the battery voltage into a standard signal range, which is then converted to the battery voltage via an A / D converter. Bidirectional current sensors are installed in the positive and negative terminal circuits of the energy storage battery. Based on the principle of electromagnetic induction, the direction and magnitude of the current are detected; positive current indicates charging, and negative current indicates discharging. The battery current is output through differential amplification and signal conditioning circuits. Finally, multiple temperature sensors are placed at key locations inside the energy storage battery pack. Utilizing the characteristic that the resistance of the temperature-sensing element changes with temperature, temperature data is obtained through a voltage divider circuit and an A / D converter. A multi-point averaging algorithm is used to calculate the battery pack temperature.

[0083] Specifically, the load parameters include core load power, flexible load power, real-time power demand, and load priority. First, in the core load power supply circuit, a voltage sensor acquires the load-side voltage signal using the voltage divider principle, and a current sensor detects the load current using the electromagnetic induction principle. A digital signal processor performs real-time calculations on the voltage and current signals to obtain the core load power. Then, in the flexible load power supply circuit, a combination of voltage and current sensors uses high-frequency synchronous sampling technology to acquire voltage and current waveform data. Combined with Fast Fourier Transform analysis of harmonic components, the active power, reactive power, and apparent power of the flexible load are calculated based on the product of voltage and current. Furthermore, the real-time power demand is calculated by summing the core load power and flexible load power in real time to obtain the total power demand. Based on this total power demand and a load prediction algorithm, the power demand trend within a short-term time window is calculated, and the real-time power demand is obtained from this trend. Finally, the priority configuration parameters of each load are read through the communication interface to establish a load priority database. Based on the current operating status and business importance, the priority weight coefficients of each load are dynamically updated to obtain the load priority.

[0084] S2: Based on the real-time photovoltaic parameters, a three-dimensional surface model is constructed to perform differential geometric curvature analysis, Gaussian curvature is calculated, and the rate of change of light intensity is obtained based on the Gaussian curvature.

[0085] In this step, the voltage, current, and power calculated from the real-time photovoltaic parameters are used as spatial coordinates to construct a three-dimensional surface model. Curvature analysis is performed on the surface using differential geometry theory, and the dynamic changes of photovoltaic characteristic parameters are quantified by calculating the Gaussian curvature at each point on the surface. Simultaneously, the rate of change of light intensity is extracted based on the time derivative of the Gaussian curvature, achieving a geometric description and mathematical quantification of the dynamic changes in illumination conditions. This method, by integrating the geometric features of multi-dimensional parameters, significantly improves the sensitivity and recognition accuracy of illumination changes, providing more accurate and stable basic data support for photovoltaic control systems.

[0086] S3: Based on the load parameters, construct a load demand matrix, and combine the real-time photovoltaic parameters, Gaussian curvature, and rate of change of light intensity to determine the type of light change event;

[0087] In this step, a load demand matrix is ​​first constructed based on the load parameters, serving as a quantitative expression of the load characteristics. Subsequently, multi-dimensional data, including real-time photovoltaic parameters, Gaussian curvature, and rate of change of illuminance, are integrated. Through feature correlation analysis and pattern matching, the types of illuminance change events are identified and classified. This process avoids the control strategy failure problems caused by misjudgment or over-response in traditional methods by distinguishing different illuminance change event patterns. Accurate identification of illuminance change event types provides a basis for the selection of subsequent positioning strategies, ensuring the targeting and effectiveness of the control strategy. It also plays a crucial role as the intelligent decision-making center in the entire scheme, significantly improving the system's dynamic response capability and stability.

[0088] S4: Construct a time series prediction model based on the Gaussian curvature to generate a future photovoltaic supply matrix. Extract virtual shading features based on the Gaussian curvature to correct the future photovoltaic supply matrix and obtain a photovoltaic output matrix.

[0089] In this step, a time-series forecasting model is first constructed using Gaussian curvature to generate a future photovoltaic (PV) supply matrix, predicting the PV system's supply capacity at different time periods. Subsequently, virtual shading features are extracted using Gaussian curvature to quantify the dynamic impact of shading on PV supply. The future supply matrix is ​​then corrected based on these shading features, forming a PV output matrix reflecting the impact of shading. This method achieves accurate prediction of PV supply and dynamic correction of shading effects through Gaussian curvature, effectively improving the reliability and accuracy of the PV output matrix. It provides a power benchmark consistent with actual operating conditions for subsequent power positioning and coordinated regulation, significantly reducing prediction errors caused by uncertainties such as shading, and ensuring the stability and efficiency of the PV control system.

[0090] S5: When the type of light change event meets the preset triggering condition, locate the photovoltaic maximum power point within the preset time threshold. The preset time threshold is determined according to the load priority and light intensity change rate in the load parameters.

[0091] In this step, the system first determines whether preset triggering conditions are met based on the type of light change event. If the conditions are met, a targeted photovoltaic maximum power point (MPP) positioning strategy is adopted within a dynamic time threshold. The preset triggering conditions are determined by a combination of the light change event type, load demand matrix, and Gaussian curvature, while the time threshold is dynamically adjusted based on load priority and the rate of change of light intensity in the load parameters. Through precise triggering of light change events and dynamic constraints on the time threshold, the PMP is rapidly located, adapting to different load priorities and light intensity changes, ensuring that the photovoltaic output approaches its maximum capacity. This process effectively improves energy utilization efficiency, enhances the timeliness and adaptability of the system response, and provides a reliable foundation for subsequent power regulation.

[0092] S6: Based on the photovoltaic maximum power point, energy storage state parameters, load parameters and photovoltaic output matrix, dynamically adjust the photovoltaic output parameters and generate an energy storage charging and discharging strategy.

[0093] In this step, by integrating the photovoltaic maximum power point, energy storage state parameters, load parameters, and photovoltaic output matrix, the dynamic adjustment of photovoltaic output parameters and the formulation of energy storage charging and discharging strategies are collaboratively completed to achieve power balance among photovoltaics, energy storage, and the load. The coordinated application of multi-dimensional parameters ensures dynamic adaptation between photovoltaic output and load demand. Simultaneously, the energy storage charging and discharging strategy enables flexible energy dispatch, guaranteeing stable power supply to core loads, improving the utilization efficiency of photovoltaic energy, enhancing the system's adaptability and reliability under conditions of fluctuating sunlight and changing loads, and reducing operational risks caused by supply-demand imbalances.

[0094] Specifically, the energy storage charge / discharge power is calculated based on the state of charge (SOC) value of the energy storage battery, the battery pack temperature, and the maximum allowable charge / discharge power in the energy storage state parameters. The maximum allowable charge / discharge power is calculated based on the voltage and current data in the energy storage state parameters, combined with the battery's rated parameters. The flexible load power is dynamically adjusted based on the core load power in the load parameters and the SOC value in the energy storage state parameters. If the energy storage SOC value is below the safe discharge limit, the flexible load power is appropriately reduced to lower the total demand and avoid over-discharge of the energy storage; if the energy storage battery's SOC value is sufficient, the flexible load power can be retained, forming the adjusted flexible load power. Subsequently, the target photovoltaic output power is calculated using the power balance formula based on the core load power, energy storage charge / discharge power, adjusted flexible load power, photovoltaic maximum power point, and photovoltaic output matrix in the load parameters.

[0095] Based on this, the photovoltaic output parameters are adjusted according to the maximum photovoltaic power point, and dynamic adaptation is achieved by combining load demand and energy storage status parameters. The load demand is calculated based on the real-time power demand and load priority in the load parameters. The photovoltaic output parameters include the output voltage and current of the photovoltaic array. The voltage and current corresponding to the maximum photovoltaic power point are set as the initial output reference to ensure that the photovoltaic operates within the high-efficiency conversion range under the current irradiance conditions. The actual photovoltaic output power is the maximum conversion efficiency point under the current irradiance. The voltage and current parameters are dynamically adjusted according to the calculated photovoltaic target output power, and the magnitudes of the photovoltaic target output power and the actual photovoltaic output power are compared, such as... Figure 2 As shown, if the target output power of the photovoltaic system is equal to the actual output power of the photovoltaic system, the voltage and current output of the initial reference are maintained, and the power output is stabilized by the inverter. If the target output power of the photovoltaic system is less than the actual output power of the photovoltaic system, the pulse width modulation duty cycle is adjusted by the inverter to appropriately reduce the output voltage or current, so that the actual output power of the photovoltaic system approaches the target output power of the photovoltaic system. During the adjustment process, the matching of voltage and current is maintained to avoid deviation from the safe operating range. At the same time, the power prediction trend of the photovoltaic output matrix is ​​referenced in real time to predict the impact of changes in light intensity on voltage and current, and the adjustment range is corrected in advance.

[0096] After adjusting the photovoltaic (PV) output parameters according to the target output power, it is necessary to monitor the deviation between the actual PV output power and the target PV output power in real time. If a deviation exists, the predicted power value within the sliding time window is extracted from the PV output matrix, and the battery pack temperature is obtained from the energy storage state parameters. The energy storage charging and discharging power adjustment amount is then calculated, and the energy storage charging and discharging power is dynamically adjusted according to the adjustment amount to compensate for the deviation. If the actual PV output power is less than the target PV output power, the energy storage discharging power is increased; if the actual PV output power is greater than the target PV output power, the energy storage charging power is increased.

[0097] The specific steps of S2 are as follows:

[0098] S2.1: Construct a three-dimensional curved surface model based on the real-time photovoltaic parameters;

[0099] In this embodiment, the real-time photovoltaic parameters include photovoltaic array voltage and current. Specifically, firstly, based on the photovoltaic array voltage and current, the photovoltaic power at each sampling time is calculated, and a continuous photovoltaic power data sequence is formed within a sliding time window. The length of the sliding time window is set according to the sampling step size and system response requirements. Then, the photovoltaic power data sequence, photovoltaic array voltage, and time-stamped data undergo three levels of preprocessing. The first level of processing is data filtering, which uses an adaptive Kalman filter algorithm to eliminate high-frequency noise. The filter coefficient is dynamically adjusted according to the fluctuation amplitude of the photovoltaic power; the larger the fluctuation, the smaller the filter coefficient. The second level of processing is outlier removal, which determines the outlier threshold by analyzing the central tendency and dispersion of the data sequence, and identifies and marks significant deviations. For outliers in the overall distribution trend of the sequence and sampling points exceeding the normal fluctuation range, linear interpolation using nearby valid data is used to correct the marked outliers. The third stage of processing is data standardization, which involves scaling the photovoltaic power data sequence, photovoltaic array voltage, and time-stamped data separately. First, the extreme value features of each parameter data sequence are extracted to determine the dynamic range of the data distribution as a conversion benchmark. Then, a linear transformation is used to map the original data values ​​to a unified dimensionless scale, ensuring that the numerical distributions of different parameters are on the same order of magnitude. This transformation eliminates the differences in parameter numerical ranges caused by differences in physical dimensions, outputting standardized photovoltaic power data sequences, standardized photovoltaic array voltage, and standardized time-stamped data. These processed data serve as inputs for subsequent 3D surface model fitting, ensuring a balanced weighting of the influence of each parameter on the 3D surface model fitting during the modeling process, and improving the fitting accuracy and feature extraction reliability of the 3D surface model.

[0100] Specifically, the core inputs are standardized photovoltaic power data sequences, standardized photovoltaic array voltages, and standardized time-stamped data. First, a three-dimensional coordinate system is defined, such as... Figure 3 As shown, with standardized time as the X-axis, standardized photovoltaic array voltage as the Y-axis, and standardized photovoltaic power as the Z-axis, a bivariate functional relationship between Z and X, Y is established; for example, Figure 3The median values ​​range from 0.0 to 1.2, using a grayscale gradient to represent different power intensity levels. Darker colors correspond to higher power output values, and lighter colors correspond to lower power output values. This color mapping allows for intuitive identification of the power output level of the photovoltaic array under different operating conditions. Then, a polynomial regression method is used to fit the input data to a surface, constructing a polynomial expression containing powers of X and Y and their cross terms. The polynomial coefficients are solved using the least squares method to optimize the fit of the expression to the input data. Finally, cross-validation is used for iterative testing to determine the optimal order of the polynomial. The input data is divided into training and validation sets. For polynomials of different orders, the model is fitted on the training set and the fitting effect is evaluated on the validation set. The order that minimizes the fitting error on the validation set is selected as the optimal order, ensuring that the model accurately represents the data characteristics while avoiding overfitting. The final output is a three-dimensional surface model constructed based on the optimal order. This model presents the mapping relationship between standardized time, standardized photovoltaic array voltage and standardized photovoltaic power in the form of a continuous surface. It can intuitively characterize the photovoltaic power variation characteristics at different times and with different photovoltaic array voltages, and provide a basic model for subsequent differential geometric curvature analysis.

[0101] S2.2: Perform differential geometric curvature analysis to calculate the Gaussian curvature;

[0102] In this embodiment, the first and second partial derivatives are first calculated based on the three-dimensional surface model, and the first and second fundamental form coefficients are calculated based on the principle of differential geometry. Then, the Gaussian curvature is calculated based on the first and second fundamental form coefficients and the Gaussian curvature formula.

[0103] like Figure 3 As shown, in this embodiment, the sampling points are first solved based on the three-dimensional surface model. The partial derivative of , where n is the number of sampling points. For example, for sampling points... By analyzing its adjacent sampling points and coordinate difference and and power difference Solve for the first-order partial derivatives and The height variation of the surface in the three-dimensional model reflects the tilt trend of the surface in the time and voltage dimensions. The height variation of the surface visually displays the distribution of power output strength, with higher regions corresponding to higher power output and lower regions corresponding to lower power output. The second-order partial derivative is then calculated using the rate of change of the first-order partial derivative to describe the increase or decrease in the tilt slope.

[0104] Based on this, the first and second fundamental form coefficients are derived. The first fundamental form coefficient is calculated by multiplying the squares of the first-order partial derivatives and is used to characterize the local scale properties of the surface. The second fundamental form coefficient is calculated by normalizing the combination of the second-order and first-order partial derivatives and is used to quantify the curvature properties of the surface. According to the principles of differential geometry, the first and second fundamental form coefficients are substituted into the Gaussian curvature formula for correlation calculation. Through the specific combination relationship of the two types of coefficients, the Gaussian curvature value of each sampling point is obtained. The positive and negative characteristics of the Gaussian curvature correspond to the curvature type of the surface; a positive value reflects a convex curvature, and a negative value reflects a concave curvature. The magnitude of the value indicates the degree of curvature of the surface.

[0105] S2.3: The rate of change of light intensity is obtained based on the Gaussian curvature;

[0106] In this embodiment, Gaussian curvature is used as the core feature parameter. By establishing a quantitative correlation between Gaussian curvature and changes in light intensity, a mapping relationship model between Gaussian curvature and the rate of change of light intensity is constructed. The rate of change of light intensity is determined based on the mapping relationship model. Specifically, firstly, based on the light intensity in the real-time photovoltaic parameters, light intensity change features are extracted. These features include light intensity change trend, light intensity change level, and light intensity change stability. The light intensity change trend is determined by calculating the difference in light intensity over continuous monitoring times. The light intensity change level is classified according to the relative change amplitude per unit time. The light intensity change stability is assessed by evaluating the degree of fluctuation of the change value over a continuous period.

[0107] Then, based on the Gaussian curvature and the characteristics of light intensity variation, the segmented intervals of the mapping relationship model are determined using a two-dimensional cross-division method. The horizontal division is determined based on the bending characteristics of the Gaussian curvature, including convex bending intervals, concave bending intervals, and smooth intervals. The vertical division is based on the characteristics of light intensity variation, including enhancement trend intervals, attenuation trend intervals, and stable trend intervals. Through cross-matching of the horizontal and vertical divisions, several segmented intervals are formed, each corresponding to a characteristic correlation pattern between the Gaussian curvature and the rate of change of light intensity.

[0108] For the aforementioned segmented intervals, based on the physical characteristics of photovoltaic cell light response, a linear transformation equation between Gaussian curvature and the rate of change of light intensity is derived. The equation is a linear function of the rate of change of light intensity and Gaussian curvature. Based on the inherent law of photovoltaic module power variation with light intensity, a theoretical range of values ​​for the equation coefficients is set. The Gaussian curvature and the rate of change of light intensity are then fitted and optimized to ensure that the equation accurately reflects the quantitative correlation between the two within the interval.

[0109] Based on the range of Gaussian curvature values, a corresponding linear transformation equation is determined to obtain the initial rate of change of illumination intensity. Based on the illumination intensity in the real-time photovoltaic parameters, the real-time rate of change of illumination intensity is calculated using the adjacent time-difference method. Dynamic verification is performed by comparing the deviation between the initial rate of change of illumination intensity and the real-time rate of change of illumination intensity. If the deviation is within a preset range, the initial result is directly output as the final rate of change of illumination intensity. If the deviation exceeds the range, the equation coefficients for the current interval are corrected according to the direction and magnitude of the deviation. After correction, the rate of change of illumination intensity is recalculated and output to ensure that the result is consistent with the actual illumination dynamics.

[0110] Please see Figure 4 The figure is a flowchart of the steps for determining the type of illumination change event provided in an embodiment of this application. The specific steps of S3 are as follows:

[0111] S3.1: Construct a load demand matrix based on the load parameters;

[0112] In this embodiment, a load demand matrix is ​​constructed based on the core load power, flexible load power, real-time power demand, and load priority among the load parameters. This matrix is ​​used to quantitatively characterize the dynamic demand features of the electricity consumer. The row dimension of the matrix corresponds to the time series, reflecting the load demand at different times. The column dimension corresponds to demand characteristic indicators, including instantaneous power value, power change trend, demand stability, and priority weight. The instantaneous power value is calculated based on the sum of the core load power and the flexible load power. The power change trend is calculated by the difference between the core load power and the flexible load power at adjacent times; a positive value indicates an increase in demand, and a negative value indicates a decrease in demand. The demand stability is calculated based on the standard deviation of the real-time power demand fluctuation over a continuous period; a smaller standard deviation indicates more stable demand. The priority weight is assigned a quantified value based on the load priority, with the core load having a higher weight than the flexible load. The above characteristic indicators are integrated into a load demand matrix according to the time series. The matrix elements are standardized to eliminate dimensional differences, comprehensively reflecting the dynamic characteristics of load demand.

[0113] S3.2: Determine the type of light change event based on the real-time photovoltaic parameters, Gaussian curvature, rate of change of light intensity, and load demand matrix;

[0114] In this embodiment, based on real-time photovoltaic parameters, Gaussian curvature, rate of change of illuminance, and load demand matrix as inputs, a feature extraction algorithm generates key feature indicators, including the third derivative of the photovoltaic power curve reflecting dynamic power changes, spatial curvature characteristics characterizing spatial distribution, dynamic illuminance characteristics quantifying the evolution of illuminance, and a load power fluctuation coefficient that comprehensively assesses load stability. Subsequently, a weighted fusion method is used to integrate these multi-dimensional features. The fusion ratio is dynamically allocated according to the physical meaning and influence weight of each key feature indicator. The weight of the spatial curvature feature is determined based on the area of ​​the abnormal region of the Gaussian curvature, the weight of the dynamic illuminance feature is determined based on the rate of change of illuminance, and the weight of the load power fluctuation coefficient is dynamically adjusted according to load priority, forming a comprehensive feature vector containing multi-dimensional information. Based on the fused feature vector, a discrimination criterion is constructed to classify event types, and the event type results are output through feature matching and logical verification. Determining the type of illuminance change events provides accurate scenario awareness for the photovoltaic system, supports targeted regulation strategy optimization, and improves system operational stability and energy utilization efficiency. Simultaneously, it provides a decision-making basis for the coordinated scheduling of photovoltaic, energy storage, and the power grid, enhancing the adaptability and overall reliability of the energy system to complex illuminance environments.

[0115] S3.2.1: Based on the real-time photovoltaic parameters, Gaussian curvature, rate of change of illuminance, and load demand matrix, the third derivative, spatial curvature characteristics, illuminance dynamic characteristics, and load power fluctuation coefficient of the photovoltaic power curve are calculated respectively.

[0116] In this embodiment, based on the real-time photovoltaic parameters, Gaussian curvature, rate of change of irradiance, and load demand matrix, a multi-dimensional feature vector is extracted, including the third derivative of the photovoltaic power curve, spatial curvature characteristics, dynamic irradiance characteristics, and load power fluctuation coefficient.

[0117] The third derivative of the photovoltaic power curve is calculated based on the real-time photovoltaic parameters and is used to quantify the instantaneous rate of change and trend reversal characteristics of the photovoltaic power curve. Specifically, the photovoltaic power is first calculated based on the voltage and current in the real-time photovoltaic parameters; then, the photovoltaic power calculation results at consecutive time points are arranged in chronological order to obtain power time-series data; finally, the derivative of the power time-series data is calculated to obtain the third derivative of the photovoltaic power curve.

[0118] The spatial curvature characteristics are obtained by analyzing the spatial distribution pattern of the Gaussian curvature and the morphological characteristics of abnormal regions, and are used to comprehensively characterize the degree of spatial curvature and non-uniformity of the photovoltaic power surface. Specifically, firstly, the spatial distribution data of the Gaussian curvature is analyzed. When the Gaussian curvature is positive, it indicates that the power distribution in that region is convex, that is, the local power is higher than the surrounding area. When the Gaussian curvature is negative, it indicates that the power distribution is concave, that is, the local power is lower than the surrounding area. The absolute value of the Gaussian curvature reflects the degree of curvature of the photovoltaic power surface. Then, the spatial distribution pattern of the Gaussian curvature is analyzed based on the spatial distribution data. The spatial distribution data includes the continuity of convex and concave regions and the distribution of extreme points. The continuity of convex and concave regions is determined by statistically analyzing whether the convex or concave regions with Gaussian curvature are continuous, that is, whether there are large continuous convex regions or scattered concave regions. The higher the continuity, the stronger the spatial consistency of the power distribution. The distribution of extreme points is determined by identifying the points with the largest absolute values ​​of the Gaussian curvature and analyzing the spatial location and density of these points. The more concentrated the extreme points, the more significant the protrusion or concavity characteristics of the power surface. Finally, based on the Gaussian curvature, the morphological characteristics of the abnormal regions are identified, and combined with the spatial distribution pattern, spatial curvature characteristics are obtained. The morphological characteristics of the abnormal regions are obtained by filtering out areas where the absolute value of the Gaussian curvature exceeds the normal range and analyzing their area, shape, and distribution. The spatial distribution pattern and the morphological characteristics of the abnormal regions are normalized and mapped to a specified interval, weights are assigned, and the normalized quantified indicators are multiplied by the corresponding weights, then integrated into a multi-dimensional feature vector, which serves as the spatial curvature characteristic.

[0119] The dynamic characteristics of illumination are determined based on the trend persistence, consistency of change direction, and fluctuation frequency of the illumination intensity change rate, and are used to quantitatively reflect the dynamic evolution law of illumination intensity. Specifically, firstly, the trend persistence of the illumination intensity change rate is analyzed based on the power time-series data. The duration of continuous enhancement and continuous decay is statistically analyzed from the power time-series data. The duration of continuous enhancement is calculated based on the cumulative time during which the photovoltaic power change rate is continuously positive; the duration of continuous decay is calculated based on the cumulative time during which the photovoltaic power change rate is continuously negative. Then, the consistency of the direction of illumination intensity change and the fluctuation frequency are analyzed. The consistency of direction is determined by statistically analyzing the number of times the direction of illumination intensity change rate changes per unit time; fewer changes indicate a more uniform direction of illumination change. The fluctuation frequency is obtained by calculating the number of times the peak value of the absolute value of the illumination intensity change rate occurs per unit time; a higher fluctuation frequency indicates more unstable illumination and more frequent fluctuations. Finally, the dynamic characteristics of illumination are obtained based on the trend persistence, consistency of direction, and fluctuation frequency. The trend persistence, directional consistency, and fluctuation frequency are normalized and uniformly mapped to a specified interval to eliminate dimensional differences. Weights are assigned according to the importance of each feature to the dynamic evolution of illumination. Finally, the three are integrated into a multi-dimensional feature vector through weighted fusion, which serves as the dynamic feature of illumination.

[0120] The load power fluctuation coefficient is calculated based on the load demand matrix and is used to comprehensively characterize the stability and fluctuation amplitude of load power. Specifically, firstly, based on the load demand matrix, time-series data of real-time power demand is extracted. The time-series data includes the total load power at different times, calculated by adding the core load power and the flexible load power. Then, the fluctuation standard deviation is calculated based on the time-series data to measure the dispersion of the time-series data. The smaller the fluctuation standard deviation, the closer the load power at different times is, indicating that the overall load is more stable. Finally, based on the fluctuation standard deviation, a weighted calculation is performed using the core load power ratio and the flexible load power ratio to obtain the load power fluctuation coefficient. The core load power ratio is calculated based on the proportion of core load power to total load power, and the flexible load power ratio is calculated based on the proportion of flexible load power to total load power. Since the core load has high priority and cannot be interrupted, a higher power ratio of the core load has a greater actual impact on system stability, even if the fluctuation standard deviation is the same. In this embodiment, this impact is strengthened through weight allocation. Conversely, a higher flexible load ratio results in a relatively weaker actual impact of fluctuation, so the corresponding weight is reduced.

[0121] S3.2.2: The third derivative, spatial curvature feature, illumination dynamic feature, and load power fluctuation coefficient of the photovoltaic power curve are fused using a weighted fusion algorithm to obtain a multidimensional feature vector. The weight of the spatial curvature feature is determined based on the area of ​​the abnormal region of the Gaussian curvature, the weight of the illumination dynamic feature is determined based on the amplitude of the change rate of illumination intensity, and the weight of the load power fluctuation coefficient is dynamically adjusted based on the load priority in the load parameters.

[0122] In this embodiment, firstly, the third derivative, spatial curvature characteristics, illumination dynamic characteristics, and load power fluctuation coefficient of the photovoltaic power curve are normalized and mapped to the [0,1] interval to eliminate the influence of dimensional differences on the fusion result. Then, weights are dynamically assigned based on the characteristic features. The weight of the spatial curvature characteristic is determined based on the area of ​​the abnormal region of the Gaussian curvature; the higher the proportion of the abnormal region area to the total area of ​​the photovoltaic panel, the wider the area where the Gaussian curvature deviates significantly from the normal range, and thus the larger the weight value. The weight of the illumination dynamic characteristics is determined based on the amplitude of the rate of change of illumination intensity; the amplitude is determined based on the peak absolute value of the rate of change of illumination intensity, and the larger the amplitude, the higher the weight. The weight of the load power fluctuation coefficient is dynamically adjusted according to the load priority in the load parameters; the higher the proportion of the core load power, the greater the weight. The weight of the third derivative of the photovoltaic power curve is set to a fixed benchmark value to ensure the fundamental influence of the power dynamic change characteristics. Finally, a weighted summation algorithm is used to fuse the four normalized features. Each feature component is multiplied by its corresponding weight and then summed to form a multidimensional feature vector that includes comprehensive information on power dynamics, spatial curvature, illumination evolution, and load fluctuation. This vector fully covers the dynamic correlation characteristics of illumination, power, and load, providing a unified feature input for the subsequent determination of illumination change event types.

[0123] S3.2.3: Based on the numerical range and weight ratio of each feature component in the multidimensional feature vector, construct a discrimination criterion for illumination change events;

[0124] In this embodiment, firstly, by analyzing the multi-dimensional feature vector samples corresponding to known illumination change event types, the numerical distribution patterns of the third derivative of the photovoltaic power curve, spatial curvature characteristics, illumination dynamic characteristics, and load power fluctuation coefficient under each illumination change event type are extracted. The mean, standard deviation, and extreme value range of each feature component are statistically analyzed to clarify the characteristic differences of different illumination change event types. For example, local shading events are characterized by a high mean of spatial anomaly characteristics and negative extreme values ​​of the power derivative. Among them, the spatial curvature characteristics are dominated by concave anomalies, and the mean of its concave component is significantly higher and the standard deviation is small. The third derivative of the photovoltaic power curve shows a negative mean due to the sudden drop in photovoltaic power, and the standard deviation reflects the fluctuation intensity of the local sudden drop, with the extreme value range concentrated in the negative large value range. The illumination dynamic characteristics are mainly characterized by local unidirectional attenuation, while the load power fluctuation shows the characteristics of increasing mean and expanding extreme value range as the proportion of core load increases. The rapid cloud movement event is characterized by high-frequency mean and wide-range extreme values ​​in its dynamic features. Specifically, the dynamic characteristics of illumination are characterized by prominent high-frequency fluctuations, with the mean frequency falling within the high-frequency range and the standard deviation large, while the extreme value range covers typical high-frequency fluctuations. The third derivative of the photovoltaic power curve exhibits oscillating characteristics with a mean close to zero but a large standard deviation due to rapid changes in illumination, and the extreme value range alternates between positive and negative values ​​with a wide span. The convex-concave switching component of the spatial curvature feature has a high mean, and the extreme value range covers the frequently switching feature range. The load power fluctuation shows a pattern of rising mean and widening standard deviation in sync with unstable illumination. Steady-state events are characterized by low mean, small standard deviation, and narrow extreme value ranges across all features. All features exhibit a mean close to a stable baseline value and a very small standard deviation due to low fluctuations, with the extreme value range narrow and concentrated in the low-fluctuation range, overall reflecting a distribution pattern of stable feature values ​​and weak fluctuations.

[0125] Building upon this foundation, the distribution patterns of features are further transformed into quantifiable decision boundaries, and the decision-making role of key features is strengthened through weighting to construct a discrimination criterion. Specifically, considering the characteristic differences of different types of illumination change events, the effective decision intervals for each feature component under each event type are defined using the mean of the feature components as the benchmark, the standard deviation as the fluctuation reference, and the extreme value range as the boundary constraint. For local shadow events, the high mean interval of the concave component of the spatial curvature feature is used as the core threshold, while the negative extreme value range of the third derivative of the photovoltaic power curve is used as an auxiliary constraint to ensure the coordinated judgment of spatial anomalies and power drop characteristics. For rapid cloud movement events, the high mean interval of the dynamic characteristic fluctuation frequency of illumination is used as the primary threshold, combined with the wide oscillation extreme value range of the third derivative of the photovoltaic power curve and the convex-concave switching interval of the spatial curvature feature to form a joint judgment condition for dynamic fluctuation characteristics. For steady-state events, the low mean interval and narrow extreme value range of all feature components are used as a unified threshold, and cross-validation is performed through the low fluctuation characteristics of each feature. At the same time, the weighting ratio is integrated into the discrimination criterion logic, giving higher priority to high-weight features. When a high-weight feature component falls within the threshold range of its corresponding event, the event type can be initially identified. A second verification is then performed using threshold matching of low-weight features, reducing the risk of misjudgment due to fluctuations in a single feature. Furthermore, a dynamic calibration mechanism for the discrimination criteria is established. A tolerance range is set based on the stability of the feature distribution pattern. When a feature component deviates slightly from the mean due to minor disturbances but remains within the tolerance range, it is still classified as the original event type. If a systematic shift occurs in the feature distribution, the feature extraction stage is traced back, and the threshold boundary is optimized by adjusting the statistical logic of the mean and standard deviation.

[0126] S3.2.4: Determine the type of illumination change event according to the discrimination criteria. The type of illumination change event includes local shadow events, rapid cloud movement events, and steady-state events.

[0127] like Figure 5As shown, the type of illumination change event is determined according to the constructed discrimination criterion. First, the third derivative of the illumination power curve, spatial curvature characteristics, illumination dynamic characteristics, and load power fluctuation coefficient are used as vectors as input data for event type determination. Subsequently, the input data is matched with the discrimination criterion in multiple dimensions. For the local shadow event, the core criterion is whether the high-weighted component of the spatial curvature feature falls within the core threshold range. If the high-weighted feature component falls within the corresponding core threshold range, it is further verified by auxiliary conditions such as the negative fluctuation characteristics of the third derivative of the photovoltaic power curve and the unidirectional attenuation characteristics of the dynamic characteristics of illumination. For the cloud rapid movement event, the core criterion is whether the dynamic characteristics of illumination meet the high-frequency fluctuation threshold. If the dynamic characteristics of illumination meet the high-frequency fluctuation threshold, it is jointly judged by auxiliary conditions such as the oscillation characteristics of the third derivative of the photovoltaic power curve and the convex-concave switching range of the spatial curvature feature to ensure the consistency of the dynamic fluctuation characteristics. For the steady-state event, the core criterion is whether all feature components fall within the low fluctuation threshold range. The accuracy of the judgment is ensured by verifying whether the power derivative approaches a stable value, whether there are significant anomalies in the space, and whether the illumination fluctuation is weak.

[0128] Finally, the event type of illumination change is output based on the matching results: if the feature vector fully satisfies the core criteria and auxiliary conditions of a certain type of event, it is directly determined as that event type; if feature cross-matching occurs, that is, it meets the criteria of two types of events, the matching degree score of the feature components and each type of criterion is calculated according to the weight ratio, the event type with the highest score is selected as the result, and it is marked as fuzzy judgment to prompt subsequent verification; if all features do not fall into any event criterion range, the anomaly handling mechanism is triggered, it is temporarily set as an unknown event and the original feature vector data is retained.

[0129] The specific steps for S4 are as follows:

[0130] S4.1: Construct a time series prediction model based on the Gaussian curvature to generate a future photovoltaic supply matrix;

[0131] In this embodiment, the Gaussian curvature is first preprocessed temporally to obtain structured Gaussian curvature time-series data. This preprocessing includes time-dimensional alignment, noise reduction, and missing value imputation. Time-dimensional alignment maps all Gaussian curvatures to a unified time axis using a timestamp synchronization algorithm, ensuring spatial correspondence of Gaussian curvatures in different regions at the same time, forming a spatiotemporally correlated data structure. Noise reduction and missing value imputation address outliers caused by sensor errors or environmental interference in the Gaussian curvature by employing a spatial correlation-based filtering algorithm. This algorithm calculates the deviation between the Gaussian curvature of a region and its surrounding regions, replacing noise values ​​with the weighted average of neighboring curvatures. For missing data, if the missing duration is short, temporal interpolation is used to fit and imput the data based on the trend of Gaussian curvature changes before and after the missing time. If the missing duration is long, a spatial interpolation algorithm is used to estimate the missing data based on the Gaussian curvature distribution of surrounding regions at the same time, and this is combined with historical Gaussian curvature change patterns and photovoltaic output characteristics for joint imputation, ensuring the integrity and physical rationality of the Gaussian curvature time-series data.

[0132] Subsequently, based on the Gaussian curvature time-series data, a time-series prediction model integrating spatial features and temporal patterns is constructed. This time-series prediction model uses multi-scale convolutional kernels in the convolutional layers to extract features from the spatial distribution of Gaussian curvature. These multi-scale convolutional kernels include small-scale and large-scale kernels. The small-scale kernels capture curvature abrupt changes in local regions, while the large-scale kernels capture the curvature distribution trend in the global region. Through multi-layer convolution and pooling operations, the two-dimensional spatial curvature distribution is transformed into a high-dimensional spatial feature vector, preserving the correlation between regions. The recurrent layer of the time-series prediction model connects to a bidirectional recurrent layer based on the output of the convolutional layers. A gating mechanism is used to memorize the spatial feature evolution patterns at different time steps. The forward loop captures the progressive change trend of the Gaussian curvature distribution, while the reverse loop strengthens the feature memorization of historical key nodes, solving the problem of long-term temporal dependence. Simultaneously, an attention mechanism is introduced to assign dynamic weights to features at different time steps, improving the accuracy of temporal capture. By fusing the spatial features output by the convolutional layer with the temporal features output by the recurrent layer through a fully connected layer, a joint mapping relationship between spatial patterns and temporal trends is constructed, enabling the model to simultaneously identify static spatial distributions and dynamic temporal evolutions.

[0133] During the model training phase, a sliding window method is used to divide the samples. Gaussian curvature time-series data for m consecutive time steps are used as input samples, and the corresponding actual photovoltaic output power for the next n time steps is used as the label. The model is divided into training, validation, and test sets according to time sequence. A loss function suitable for continuous value prediction is selected, and the model parameters are iteratively adjusted through an optimizer to bring the prediction error on the validation set to a stable range, ensuring that the model learns the intrinsic correlation between the spatial distribution of Gaussian curvature over time and photovoltaic power output. After training, Gaussian curvature time-series data is input. The model extracts the current spatial distribution features through convolutional layers and extrapolates the curvature evolution over a preset time period through recurrent layers, outputting the predicted spatial distribution of Gaussian curvature at each future time. Then, according to the preset mapping rules between Gaussian curvature and photovoltaic power, the predicted Gaussian curvature values ​​for each region are converted into corresponding photovoltaic power values. These are organized by time and region dimensions to form a preliminary future photovoltaic supply matrix. Each element in the matrix represents the predicted photovoltaic power value for a specific region at a specific time, reflecting the overall output trend of the future photovoltaic system.

[0134] S4.2: Extract virtual shading features based on the Gaussian curvature, and correct the future photovoltaic supply matrix to obtain the photovoltaic output matrix;

[0135] In this embodiment, virtual occlusion features are extracted based on Gaussian curvature. These features include virtual occlusion area and virtual occlusion duration. The virtual occlusion area is determined by setting a Gaussian curvature anomaly threshold based on a concave region judgment standard determined by physical mechanisms. A spatial region segmentation algorithm identifies all concave anomaly regions in the Gaussian curvature spatial distribution that exceed the anomaly Gaussian curvature threshold. The areas of each anomaly region are statistically analyzed and accumulated to obtain the virtual occlusion area, which quantifies the spatial coverage of the occlusion. The virtual occlusion duration is determined by continuous interval detection using the Gaussian curvature time-series data. The duration is tracked by beginning timing from the first appearance of the anomaly region until its area falls below a preset effective threshold. This time span represents the virtual occlusion duration, characterizing the duration of occlusion.

[0136] Furthermore, the future photovoltaic supply matrix is ​​corrected based on the virtual shading characteristics. This correction includes power attenuation coefficient superposition correction and time phase calibration correction. Specifically, the power attenuation coefficient superposition correction involves constructing a piecewise power attenuation model to attenuate the photovoltaic power values ​​of the future photovoltaic supply matrix. Several attenuation intervals are divided based on the numerical range of the virtual shading area. A matching attenuation function type is preset for each attenuation interval; for example, a linear attenuation function is used for low-shading intervals, and a nonlinear attenuation function is used for medium-to-high-shading intervals, forming an attenuation function library. An adaptive function selection mechanism is constructed based on the matching degree between the interval to which the virtual shading area belongs and historical attenuation patterns. A target attenuation function is selected from the attenuation function library, and a multi-layer superposition algorithm is used to perform regional correction of the future photovoltaic supply matrix. For regions with virtual shading, the predicted power value of that region is superimposed with the attenuation coefficient calculated by the corresponding attenuation function. Unshaded regions retain their original predicted power values. During the superposition process, the attenuation coefficients of adjacent shading regions are adjusted collaboratively to ensure that the attenuation correction conforms to the spatial distribution logic.

[0137] The time phase calibration correction involves constructing a time phase correction model based on the virtual shading duration to dynamically calibrate the time series of the future photovoltaic supply matrix. A sliding window size is set according to the virtual shading duration; the longer the virtual shading duration, the larger the corresponding sliding window size. This sliding window is used to capture the dynamic offset characteristics of the power time series. The time phase correction model uses a time series rearrangement mechanism as its core algorithm. Based on the time deviation between the shading start time, shading end time, and power prediction time series, it dynamically adjusts the time phase parameters of the power change trend. The sliding window technique is used to segment and rearrange the time series of the future photovoltaic supply matrix, ensuring precise alignment between power fluctuation nodes and the time series characteristics of the virtual shading duration. Interpolation algorithms are used to fill the data gaps generated after rearrangement, ensuring that the time series maintains numerical continuity and consistency with the power change trend before and after correction, avoiding power prediction distortion caused by time series breaks.

[0138] The correction results of the power attenuation coefficient superposition correction and the time phase calibration correction are coupled and processed. Based on the time-series data after time phase calibration correction, the regional power values ​​after power attenuation coefficient superposition correction are re-mapped and distributed according to the calibrated time-series structure. Through a weight allocation mechanism, the influence weights of the power attenuation coefficient superposition correction and the time phase calibration correction are balanced to obtain the corrected future photovoltaic supply matrix, which serves as the photovoltaic output matrix.

[0139] The specific steps for S5 are as follows:

[0140] S5.1: Determine whether the type of the illumination change event meets the preset triggering conditions;

[0141] In this embodiment, the preset triggering conditions are determined comprehensively based on the type of illumination change event, the load demand matrix, and the Gaussian curvature. Specifically, firstly, basic triggering conditions are determined based on the type of illumination change event. When the illumination change event type is a local shading event or a rapid cloud movement event, it is initially determined that the basic triggering conditions are met. Then, further screening is performed based on the power gap of the core load in the load demand matrix. If the power supply of the core load is lower than the demand threshold, or if the load demand matrix predicts a surge in future core load power demand, it is still determined that the triggering conditions are met even if the illumination change event does not reach an extreme level. If the overall load demand is stable and there is no core gap, the triggering threshold can be appropriately increased. Finally, the necessity of triggering is verified through the spatial distribution characteristics of the Gaussian curvature. When the illumination change event type is a local shading event, the actual existence and accuracy of the degree of influence of the shading area are verified based on the area ratio of the abnormal region and the curvature gradient of the abnormal region. When the illumination change event type is a rapid cloud movement event, the authenticity of the dynamic fluctuation of photovoltaic power output is verified based on the convex-concave switching frequency of the Gaussian curvature and the rate of change of light intensity.

[0142] S5.2: Determine the preset time threshold based on the load priority and the rate of change of light intensity in the load parameters;

[0143] In this embodiment, the rate of change of light intensity is divided into several levels according to the data range, and a corresponding threshold reference interval is matched for each level. The threshold reference interval includes a lower limit and an upper limit. The lower limit of the threshold reference interval increases monotonically with the rate of change of light intensity; the upper limit of the threshold reference interval is set according to the upper limit of the system computing power. The threshold reference interval is compressed and corrected according to the load priority. For high-priority loads, the upper limit of the threshold reference interval is lowered by a priority correction coefficient. The higher the load priority, the larger the correction coefficient and the lower the upper limit. For medium and low-priority core loads, the correction magnitude is reduced according to the priority level, and the reduction magnitude of the upper limit decreases sequentially. The lower limit of the corrected threshold reference interval is always greater than or equal to the minimum response time of the system hardware. The median of the corrected threshold reference interval is calculated and used as the final preset time threshold. The median value selection method can balance the tightness of the constraints within the interval, avoiding insufficient positioning accuracy due to an overly tight threshold and power loss due to an overly loose threshold, ensuring the adaptability of the time threshold to the scene requirements.

[0144] S5.3: Based on the type of light change event, adopt the corresponding positioning strategy to locate the new photovoltaic maximum power point. The positioning strategy includes a first positioning strategy, a second positioning strategy, and a third positioning strategy.

[0145] In this embodiment, based on the differences in the types of illumination change events, a dynamic positioning logic that adapts the positioning strategy to the event type is adopted. The first, second, and third positioning strategies are used to deal with local shadow events, rapid cloud movement events, and steady-state events in the illumination change event types, respectively.

[0146] When the illumination change event type is a local shadow event, a first positioning strategy is adopted. This first positioning strategy implements refined scanning positioning based on the power distribution characteristics of the photovoltaic output matrix, improving positioning accuracy in multi-peak scenarios through dynamic optimization of voltage sub-intervals. Specifically, firstly, based on the photovoltaic output matrix, the effective range of the voltage dimension is extracted, and the effective range is divided into several voltage sub-intervals according to the power gradient characteristics. The division of voltage sub-intervals becomes more refined as the virtual shading area increases. As the virtual shading area increases, the scanning frequency increases accordingly, capturing power peak position changes caused by shadow movement in real time through high-frequency scanning. During the scanning process, the photovoltaic power value within each voltage sub-interval is sampled and compared, and the voltage point corresponding to the maximum photovoltaic power is selected. This is further verified in conjunction with the photovoltaic output matrix to obtain the new photovoltaic maximum power point.

[0147] When the light change event type is a rapid cloud movement event, a second positioning strategy is adopted. This second positioning strategy achieves high-frequency response positioning based on power trend prediction and dynamic parameter adjustment. Specifically, firstly, based on the photovoltaic power time-series data of the future photovoltaic supply matrix, a sliding window fitting method is used to construct a photovoltaic power trend prediction curve. This photovoltaic power trend prediction curve includes at least the short-term power change slope and fluctuation frequency. Based on the photovoltaic power trend prediction curve, the spatial fluctuation characteristics of Gaussian curvature, and the duration of virtual shading, a dynamic adjustment model is constructed to adjust the frequency of the scanning step, the response speed, and the scanning direction in real time. The frequency of the scanning step is determined based on the duration of virtual shading; the response speed is determined based on the absolute value of the Gaussian curvature, with a larger absolute value resulting in a faster response speed for the positioning algorithm. The scanning direction is determined based on the photovoltaic power trend prediction curve; if the prediction curve shows an upward power trend, the scanning direction is adjusted towards the direction of increasing voltage; if the prediction curve shows a downward power trend, the scanning direction is adjusted towards the direction of decreasing voltage. Through the above dynamic adjustment, the frequency of the scanning step size, response speed and scanning direction are adjusted in real time, and the maximum power point of the new photovoltaic is finally located to adapt to the high-frequency power fluctuations caused by cloud movement.

[0148] When the illumination change event is a steady-state event, a third positioning strategy is employed. This third positioning strategy combines low-power continuous tracking with periodic correction, optimizing balance accuracy and energy consumption through periodic adjustments. Specifically, the current maximum photovoltaic power point is taken as the new maximum photovoltaic power point, and tracking of this new maximum photovoltaic power point is maintained without initiating high-frequency scanning; positioning accuracy is maintained solely through a periodic correction mechanism. The correction frequency of this periodic correction decreases as the rate of change in illumination intensity decreases. In terms of correction period, a smaller rate of change in illumination intensity results in a longer correction period; conversely, a slight increase in the rate of change in illumination intensity leads to a corresponding increase in correction frequency and a shorter correction period, rapidly compensating for peak position shift.

[0149] Furthermore, such as Figure 6 As shown, this application provides a photovoltaic control and regulation system for a 5G integrated power supply, including:

[0150] The parameter acquisition module includes a voltage sensor, a current sensor, a light sensor, and a temperature sensor, used to acquire real-time photovoltaic parameters, energy storage status parameters, and load parameters.

[0151] The differential calculation module is used to construct a three-dimensional surface model, perform differential geometric curvature analysis, calculate Gaussian curvature, and obtain the rate of change of light intensity based on the Gaussian curvature.

[0152] The analysis and judgment module is used to analyze and determine the type of light change event, and to perform preset trigger condition judgment and preset time threshold calculation.

[0153] The power positioning module is used to locate the photovoltaic maximum power point by adopting the corresponding positioning strategy according to the type of light change event.

[0154] The coordinated adjustment module is used to perform photovoltaic output parameter adjustment and energy storage charging and discharging strategy generation, and output control signals to the photovoltaic inverter and energy storage battery management system.

[0155] The power positioning module and the collaborative adjustment module interact directly through an internal high-speed data bus that supports multi-priority scheduling. The output of the differential calculation module is fed into the analysis and judgment module and the collaborative adjustment module in parallel.

[0156] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention 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.

[0157] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as 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.

[0158] 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 function specified in one or more boxes.

[0159] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment 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.

[0160] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A photovoltaic control and regulation method applied to 5G integrated power supplies, characterized in that, The photovoltaic control and regulation method includes: Collect photovoltaic characteristic parameters and their corresponding time stamp data. The photovoltaic characteristic parameters include real-time photovoltaic parameters, energy storage status parameters, and load parameters. Based on the real-time photovoltaic parameters, a three-dimensional surface model is constructed for differential geometric curvature analysis to calculate the Gaussian curvature, and the rate of change of light intensity is obtained based on the Gaussian curvature. Based on the load parameters, a load demand matrix is ​​constructed, and the types of light change events are determined by combining the real-time photovoltaic parameters, Gaussian curvature, and rate of change of light intensity. A time series prediction model is constructed based on the Gaussian curvature to generate a future photovoltaic supply matrix. Virtual shading features are extracted based on the Gaussian curvature to correct the future photovoltaic supply matrix and obtain a photovoltaic output matrix. When the type of light change event meets the preset triggering conditions, a corresponding positioning strategy is adopted according to the type of light change event to locate the photovoltaic maximum power point within a preset time threshold. The preset time threshold is determined according to the load priority and light intensity change rate in the load parameters. Based on the photovoltaic maximum power point, energy storage state parameters, load parameters, and photovoltaic output matrix, the photovoltaic output parameters are dynamically adjusted and an energy storage charging and discharging strategy is generated. Virtual shading features are extracted based on Gaussian curvature, and the future photovoltaic supply matrix is ​​corrected. This correction includes power attenuation coefficient superposition correction and time phase calibration correction. The virtual shading features include virtual shading area and virtual shading duration, wherein: The power attenuation coefficient superposition correction is based on the virtual shading area to construct a piecewise power attenuation model, determine the attenuation function type through an adaptive function selection mechanism, and combine a multi-layer superposition algorithm to attenuate the photovoltaic power value of the future photovoltaic supply matrix. The virtual shading area is calculated based on the area of ​​abnormal regions exceeding a preset threshold in the spatial distribution of the Gaussian curvature. The time phase calibration correction is constructed based on the virtual shading duration. The time phase correction model rearranges the time series of the future photovoltaic supply matrix using a sliding window technique and combines an interpolation algorithm to ensure the data continuity before and after the time series rearrangement. The virtual shading duration is determined based on the continuous span of the abnormal region area on the time axis. The correction results of the power attenuation coefficient superposition correction and the time phase calibration correction are coupled and calculated to obtain the corrected future photovoltaic supply matrix, which is used as the photovoltaic output matrix.

2. The photovoltaic control and regulation method applied to a 5G integrated power supply according to claim 1, characterized in that, The collected photovoltaic characteristic parameters include: Collect the real-time photovoltaic parameters, including photovoltaic array voltage, current, and irradiance; Collect the energy storage state parameters, including the state of charge value, voltage, current and battery pack temperature of the energy storage battery; Collect the load parameters, including core load power, flexible load power, real-time power requirements, and load priority; The time stamp data corresponds one-to-one with the parameter acquisition time.

3. The photovoltaic control and regulation method applied to a 5G integrated power supply according to claim 2, characterized in that, Based on the real-time photovoltaic parameters, a three-dimensional surface model is constructed for differential geometric curvature analysis to calculate the Gaussian curvature, including: Based on the photovoltaic array voltage, current and time stamp data in the real-time photovoltaic parameters, the photovoltaic power at each sampling time is calculated and a photovoltaic power data sequence is formed within the sliding time window; By performing data filtering, outlier removal, and data standardization, the photovoltaic power data sequence, photovoltaic array voltage, and time stamp data are preprocessed to obtain standardized photovoltaic power data sequence, standardized photovoltaic array voltage, and standardized time stamp data. A three-dimensional surface model was constructed by fitting the standardized photovoltaic power data sequence, standardized photovoltaic array voltage, and standardized time stamp data using a multinomial regression method. Based on the three-dimensional surface model, the first-order partial derivatives and the second-order partial derivatives are calculated, and the first fundamental form coefficients and the second fundamental form coefficients are calculated according to the principles of differential geometry. Based on the first fundamental form coefficients, the second fundamental form coefficients, and the Gaussian curvature formula, the Gaussian curvature is calculated.

4. The photovoltaic control and regulation method applied to a 5G integrated power supply according to claim 3, characterized in that, The rate of change of light intensity is obtained based on the Gaussian curvature, including: Based on the Gaussian curvature, a mapping model between the Gaussian curvature and the rate of change of light intensity is constructed; Based on the real-time photovoltaic parameters, the light intensity variation characteristics are calculated. Based on the range of Gaussian curvature and the light intensity variation characteristics, the segmented intervals of the mapping relationship model are divided. Based on the Gaussian curvature within each segment interval and its corresponding range of light intensity variation, the correlation rules are determined, and the linear transformation equation of the mapping relationship model is constructed based on the correlation rules. The coefficients of the linear transformation equation are determined by theoretical derivation and verification. Based on the segmented intervals and linear transformation equations of the mapping relationship model, the initial rate of change of light intensity is obtained. The initial rate of change of light intensity is then dynamically verified using the light intensity in the real-time photovoltaic parameters to obtain the final rate of change of light intensity.

5. The photovoltaic control and regulation method for 5G integrated power supplies according to claim 4, characterized in that, Based on the aforementioned real-time photovoltaic parameters, Gaussian curvature, and rate of change of light intensity, the types of light change events are determined, including: Based on the real-time photovoltaic parameters, Gaussian curvature, rate of change of irradiance, and load demand matrix, the third derivative, spatial curvature characteristics, irradiance dynamic characteristics, and load power fluctuation coefficient of the photovoltaic power curve are calculated respectively. The third derivative, spatial curvature feature, illumination dynamic feature, and load power fluctuation coefficient of the photovoltaic power curve are fused using a weighted fusion algorithm to obtain a multidimensional feature vector. The weight of the spatial curvature feature is determined based on the area of ​​the abnormal region of the Gaussian curvature, the weight of the illumination dynamic feature is determined based on the amplitude of the change rate of illumination intensity, and the weight of the load power fluctuation coefficient is dynamically adjusted based on the load priority in the load parameters. Based on the numerical range and weight ratio of each feature component in the multidimensional feature vector, a discrimination criterion for illumination change events is constructed. The type of illumination change event is determined according to the discrimination criterion, and the type of illumination change event includes local shadow events, rapid cloud movement events, and steady-state events.

6. The photovoltaic control and regulation method applied to a 5G integrated power supply according to claim 1, characterized in that, The preset time threshold is determined based on the load priority and the rate of change of light intensity in the load parameters, including: A threshold reference range is determined based on the rate of change of light intensity, and the lower limit of the threshold reference range increases with the rate of change of light intensity. Based on the load priority in the load parameters, the threshold reference interval is dynamically corrected to obtain the corrected threshold reference interval. The upper limit of the corrected threshold reference interval decreases as the load priority increases, and the lower limit is greater than or equal to the minimum response time of the system hardware. The median of the corrected threshold reference interval is used as the preset time threshold.

7. The photovoltaic control and regulation method for 5G integrated power supplies according to claim 6, characterized in that, Based on the type of illumination change event, a corresponding positioning strategy is adopted to locate the new photovoltaic maximum power point. The positioning strategy includes a first positioning strategy, a second positioning strategy, and a third positioning strategy, wherein: The first positioning strategy corresponds to a local shadow event as the light change event type. The voltage sub-interval is divided according to the photovoltaic output matrix. By increasing the division density and scanning frequency of the voltage sub-interval, the new photovoltaic maximum power point is located. The division density and scanning frequency of the voltage sub-interval increase with the increase of the virtual shading area. The second positioning strategy corresponds to the light change event type of cloud rapid movement event. A photovoltaic power trend prediction curve is constructed based on the future photovoltaic supply matrix. Based on the photovoltaic power trend prediction curve, Gaussian curvature and virtual shading duration, the new photovoltaic maximum power point is located by dynamically adjusting the frequency of the scanning step and the response speed. The frequency of the scanning step increases as the virtual shading duration decreases, and the response speed increases as the absolute value of the Gaussian curvature increases. The third positioning strategy corresponds to a steady-state event type of light change event. The current photovoltaic maximum power point is taken as the new photovoltaic maximum power point and periodically corrected. The correction frequency of the periodic correction decreases as the rate of change of light intensity decreases.

8. The photovoltaic control and regulation method for 5G integrated power supplies according to claim 7, characterized in that, The dynamic adjustment of photovoltaic output parameters and generation of energy storage charging and discharging strategies include: The energy storage charging and discharging power is determined based on the state of charge value, battery temperature, and maximum allowable charging and discharging power in the energy storage state parameters. The maximum allowable charging and discharging power is calculated based on the voltage, current, and battery rated parameters in the energy storage state parameters. Based on the core load power in the load parameters and the state of charge value in the energy storage state parameters, the flexible load power in the load parameters is corrected to obtain the corrected flexible load power. The target output power of the photovoltaic system is calculated based on the core load power, energy storage charging and discharging power, corrected flexible load power, new photovoltaic maximum power point, and photovoltaic output matrix in the load parameters. The photovoltaic output parameters are adjusted according to the photovoltaic target output power. When there is a deviation between the actual photovoltaic output power and the photovoltaic target output power, the power prediction value of the sliding time window is obtained according to the photovoltaic output matrix. The energy storage charging and discharging power adjustment amount is calculated according to the battery pack temperature and power prediction value, and the energy storage charging and discharging power is dynamically adjusted according to the adjustment amount.

9. A photovoltaic control and regulation system for 5G integrated power supplies, used to implement the photovoltaic control and regulation method for 5G integrated power supplies as described in any one of claims 1-8, characterized in that, It includes a parameter acquisition module, a differential calculation module, an analysis and judgment module, a power positioning module, and a coordinated adjustment module, among which: The parameter acquisition module includes a voltage sensor, a current sensor, a light sensor, and a temperature sensor, and is used to acquire real-time photovoltaic parameters, energy storage status parameters, and load parameters. The differential calculation module is used to construct a three-dimensional surface model, perform differential geometric curvature analysis, calculate Gaussian curvature, and obtain the rate of change of light intensity based on the Gaussian curvature. The analysis and judgment module is used to analyze and determine the type of light change event, and to perform preset trigger condition judgment and preset time threshold calculation. The power positioning module is used to locate the photovoltaic maximum power point by adopting the corresponding positioning strategy according to the type of light change event. The coordinated adjustment module is used to perform photovoltaic output parameter adjustment and energy storage charging and discharging strategy generation, and output control signals to the photovoltaic inverter and energy storage battery management system.

Images