Solar power generation power peak regulation control method and system based on multi-source data analysis

Abstract

The invention provides a solar power generation power peak regulation control method and system based on multi-source data analysis, and the method comprises the steps: obtaining multi-source data from a plurality of data sources in real time through a data collection interface, and enabling the plurality of data sources to comprise a meteorological data source, a power grid operation data source and a solar power station monitoring data source; wherein the meteorological data source provides temperature, humidity, wind speed and cloud cover data, the power grid operation data source provides historical load curves and real-time load demand data, and the solar power station monitoring data source provides solar irradiance, photovoltaic panel temperature, inverter operation parameters and the like. According to the invention, multi-source real-time data of meteorology, a power grid and a power station are fused, and an advanced machine learning algorithm is adopted to carry out high-precision power prediction, so that peak regulation control has perspectiveness and predictability, the fluctuation trend of solar power generation can be sensed in advance, and passive response is converted into active management.

Description

Technical Field

[0001] This invention relates to the field of power management technology, and in particular to a method and system for peak shaving control of solar power generation based on multi-source data analysis. Background Technology

[0002] With the acceleration of the global energy transition, the installed capacity of solar power generation continues to grow rapidly. Its inherent intermittent and volatile characteristics pose a severe challenge to the stable operation of the power grid. In particular, when the solar intensity changes drastically under sunny weather or when the afternoon load peak overlaps with the power generation peak and then drops rapidly, it can lead to significant power fluctuations and an aggravated peak-valley difference in the power grid. In severe cases, it may cause frequency overruns or require the use of expensive peak-shaving units, increasing the risk and cost of power grid operation. Traditional solar power generation control methods are mostly focused on the localized control of a single power plant, such as simple power limiting based on inverter characteristics or rough power generation forecasting based on single meteorological data. These methods lack the ability to perceive and proactively regulate the overall operating status of the power grid. They are unable to achieve precise coordination with grid demand and cannot effectively cope with the complex and ever-changing actual operating environment. This may lead to wasted solar power and may also fail to provide effective power support when the grid needs it.

[0003] Therefore, in order to address the above problems, a solar power generation peak-shaving control method and system based on multi-source data analysis is now being developed. Summary of the Invention

[0004] In order to overcome the shortcomings of existing technologies in practical use, this invention provides a solar power generation peak shaving control method and system based on multi-source data analysis.

[0005] The technical solution of this invention is: a solar power generation peak-shaving control method based on multi-source data analysis, comprising the following steps: Step 1: Real-time acquisition of multi-source data from multiple data sources through the data acquisition interface. The multiple data sources include meteorological data sources, power grid operation data sources, and solar power plant monitoring data sources. The meteorological data sources provide temperature, humidity, wind speed, and cloud cover data. The power grid operation data sources provide historical load curves and real-time load demand data. The solar power plant monitoring data sources provide solar irradiance, photovoltaic panel temperature, and inverter operating parameters. Step 2: Perform data preprocessing on the acquired multi-source data, including data cleaning to remove outliers and fill in missing values, data normalization using the min-max normalization method to unify the data scale, and time alignment using interpolation algorithms to synchronize data with different sampling frequencies to the same timestamp, thereby generating a standardized dataset; Step 3: Input the standardized dataset into the pre-trained machine learning prediction model, which is a long short-term memory network model. Its training process uses historical multi-source data and corresponding historical solar power generation data. The model parameters are optimized through backpropagation algorithm and cross-validation to predict the fluctuation of solar power generation in the near future. Step 4: Based on the predicted solar power generation and real-time grid load data, an optimization algorithm is used to calculate the peak-shaving control command. The optimization algorithm is based on the particle swarm optimization algorithm, and the objective function is to minimize the peak load of the grid and maximize the utilization rate of solar power generation, while also considering grid security constraints. Step 5: Generate specific control strategies based on peak shaving control instructions, including adjusting the output power of the solar inverter, controlling the charging and discharging operation of the energy storage system to smooth power output, and coordinating distributed energy resources such as backup generators or load-side management. Finally, execute the control strategies through the actuator module to achieve peak shaving control of solar power generation and reduce peak load on the power grid.

[0006] Preferably, in the data preprocessing step, data cleaning uses statistical methods to identify outliers, including using the Z-score method to detect data points that deviate from the mean by more than three times the standard deviation, and using linear interpolation to fill in missing values; data normalization scales each data feature to the [0,1] interval, and time alignment uses linear interpolation or spline interpolation algorithms to ensure that all data are a uniform time series with a one-minute interval.

[0007] Preferably, the training process of the machine learning prediction model further includes a data augmentation step, which expands the training dataset by adding Gaussian noise or time series shifting to improve the generalization ability of the model. The model uses a rolling prediction method, updating the input data at preset intervals for real-time prediction, with the prediction time range being 1 to 4 hours in the future.

[0008] Preferably, when the optimization algorithm calculates the peak-shaving control command, it also integrates the grid topology and real-time fault information, and ensures voltage and frequency stability through constraints. The control strategy includes priority settings, wherein the charging and discharging of the energy storage system takes precedence over the inverter power adjustment, and the backup power supply is automatically started when the predicted power generation is insufficient.

[0009] Preferably, the system also includes a solar power generation peak-shaving control system based on multi-source data analysis, comprising: a data acquisition module for acquiring multi-source data in real time via wired or wireless communication interfaces, the multi-source data including meteorological data, grid load data, and solar power plant operation data; a data preprocessing module connected to the data acquisition module for cleaning, normalizing, and time-aligning the multi-source data to output a standardized dataset; a power prediction module connected to the data preprocessing module, integrating a long short-term memory network machine learning model for predicting solar power generation based on the standardized dataset; a peak-shaving strategy generation module connected to the power prediction module for generating a peak-shaving control strategy based on the predicted power and grid load using an optimization algorithm; an execution control module connected to the peak-shaving strategy generation module for converting the control strategy into control signals and adjusting the power of the solar inverter, the charging and discharging of the energy storage system, and distributed energy resources through actuators; and a central processing unit for coordinating the operation of each module and providing a human-machine interface for monitoring and parameter setting.

[0010] Preferably, the data acquisition module includes multiple sub-interfaces: a meteorological data interface connected to the meteorological bureau API or a local meteorological station, a power grid data interface connected to the power grid dispatch center SCADA system, and a power plant monitoring data interface connected to the monitoring and data acquisition system of the photovoltaic power plant or an IoT sensor; the data preprocessing module is deployed on an edge computing device to achieve low-latency processing.

[0011] Preferably, the power prediction module is deployed on a cloud server or a local server and uses a distributed computing framework to process large-scale data; the peak shaving strategy generation module integrates a real-time optimization engine, which supports multi-objective optimization and dynamic constraint adjustment.

[0012] Preferably, the execution control module includes a communication unit that is connected to the solar inverter, energy storage system controller, and load switch via a 4G / 5G network or optical fiber; the system also includes a security protection module for data encryption and access control to ensure system network security.

[0013] By adopting the above technical solution, the present invention has the following advantages: 1. This invention integrates real-time data from multiple sources, including meteorology, power grid, and power plants, and employs advanced machine learning algorithms for high-precision power prediction. This enables peak-shaving control to be forward-looking and predictive, allowing for early detection of fluctuations in solar power generation trends. This transforms passive response into proactive management. By optimizing algorithms to coordinate various resources such as inverters and energy storage systems, it achieves precise smoothing and proactive shaping of power generation, effectively reducing peak loads on the power grid, alleviating peak-shaving pressure on traditional units, significantly improving the grid's capacity to accommodate high proportions of solar power generation, reducing curtailment, and enhancing the utilization rate of clean energy.

[0014] 2. By closely integrating power plant control with the needs of power grid operation, this invention not only improves the operational efficiency of a single power plant, but also enhances the overall stability and intelligence level of the power grid. It provides reliable technical support for building a new type of power system, has broad application prospects and significant socio-economic value, and makes a positive contribution to promoting the green and low-carbon transformation of the energy structure. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of the solar power generation peak-shaving control method based on multi-source data analysis of the present invention.

[0016] Figure 2 This is a schematic diagram of the structure of the solar power generation peak-shaving control system based on multi-source data analysis of the present invention. Detailed Implementation

[0017] References to embodiments herein mean that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0018] A method for peak shaving control of solar power generation based on multi-source data analysis, such as... Figure 1 As shown, it includes the following steps: Step 1: Real-time acquisition of multi-source data from multiple data sources through the data acquisition interface. These data sources include meteorological data sources, power grid operation data sources, and solar power plant monitoring data sources. The meteorological data sources provide temperature, humidity, wind speed, and cloud cover data. The power grid operation data sources provide historical load curves and real-time load demand data. The solar power plant monitoring data sources provide solar irradiance, photovoltaic panel temperature, and inverter operating parameters. Step 2: Perform data preprocessing on the acquired multi-source data, including data cleaning to remove outliers and fill in missing values, data normalization using the min-max normalization method to unify the data scale, and time alignment using interpolation algorithms to synchronize data with different sampling frequencies to the same timestamp, thereby generating a standardized dataset; Step 3: Input the standardized dataset into the pre-trained machine learning prediction model. The machine learning prediction model is a long short-term memory network model. Its training process uses historical multi-source data and corresponding historical solar power generation data. The model parameters are optimized through backpropagation algorithm and cross-validation to predict the fluctuation of solar power generation in the near future. Step 4: Based on the predicted solar power generation and real-time grid load data, the peak-shaving control command is calculated using an optimization algorithm. The optimization algorithm is based on the particle swarm optimization algorithm, and the objective function is to minimize the peak load of the grid and maximize the utilization rate of solar power generation, while also considering grid security constraints. Step 5: Generate specific control strategies based on peak shaving control instructions, including adjusting the output power of the solar inverter, controlling the charging and discharging operation of the energy storage system to smooth power output, and coordinating distributed energy resources such as backup generators or load-side management. Finally, execute the control strategies through the actuator module to achieve peak shaving control of solar power generation and reduce peak load on the power grid.

[0019] Specifically, in the data preprocessing steps, data cleaning uses statistical methods to identify outliers, including using the Z-score method to detect data points that deviate from the mean by more than three standard deviations, and using linear interpolation to fill in missing values; data normalization scales each data feature to the [0,1] interval, and time alignment uses linear interpolation or spline interpolation algorithms to ensure that all data are in a uniform time series with a one-minute interval. The training process of the machine learning prediction model also includes a data augmentation step, which expands the training dataset by adding Gaussian noise or time series shifting to improve the model's generalization ability. The model uses a rolling prediction method, updating the input data at preset intervals for real-time prediction, with a prediction time range of 1 to 4 hours in the future. When the optimization algorithm calculates peak-shaving control commands, it also integrates the grid topology and real-time fault information, and ensures voltage and frequency stability through constraints. The control strategy includes priority settings, in which the charging and discharging of the energy storage system takes precedence over the inverter power adjustment, and the backup power supply is automatically activated when the predicted power generation is insufficient.

[0020] It also includes a solar power generation peak-shaving control system based on multi-source data analysis, such as... Figure 2As shown, it includes: a data acquisition module for acquiring multi-source data in real time via wired or wireless communication interfaces, including meteorological data, grid load data, and solar power plant operation data; a data preprocessing module connected to the data acquisition module for cleaning, normalizing, and time-aligning the multi-source data to output a standardized dataset; a power prediction module connected to the data preprocessing module, integrating a long short-term memory network machine learning model for predicting solar power generation based on the standardized dataset; a peak-shaving strategy generation module connected to the power prediction module, using an optimization algorithm to generate peak-shaving control strategies based on predicted power and grid load; an execution control module connected to the peak-shaving strategy generation module for converting control strategies into control signals and adjusting the power of the solar inverter, the charging and discharging of the energy storage system, and distributed energy resources through actuators; and a central processing unit to coordinate the various modules. The system operates in blocks and provides a human-machine interface for monitoring and parameter setting. The data acquisition module includes multiple sub-interfaces: a meteorological data interface connects to the meteorological bureau's API or a local meteorological station; a power grid data interface connects to the power grid dispatch center's SCADA system; and a power plant monitoring data interface connects to the photovoltaic power plant's monitoring and data acquisition system or IoT sensors. The data preprocessing module is deployed on edge computing devices for low-latency processing, and the power prediction module is deployed on a cloud server or local server, using a distributed computing framework to process large-scale data. The peak-shaving strategy generation module integrates a real-time optimization engine, supporting multi-objective optimization and dynamic constraint adjustment. The execution control module includes a communication unit that connects to the solar inverter, energy storage system controller, and load switch via 4G / 5G networks or fiber optics. The system also includes a security protection module for data encryption and access control to ensure system network security.

[0021] It should be noted that through highly integrated data-driven closed-loop control, accurate prediction and proactive management of solar power generation fluctuations are achieved, thereby effectively participating in grid peak shaving and improving the grid's ability to absorb high proportions of renewable energy and its operational stability. The workflow begins with the collection and fusion of multi-source heterogeneous data, followed by precise data preprocessing, advanced machine learning power prediction, and multi-objective optimization decision-making, ultimately transforming it into specific control commands for execution, forming a real-time, adaptive, and intelligent control loop. Specifically, after system startup, the data acquisition module continuously and synchronously acquires real-time data streams from three key data sources. The first type of data source is meteorological data. The system acquires high-precision geographic location-related meteorological parameters at a frequency of once per minute by calling the application programming interface provided by the meteorological bureau or connecting to a micro-meteorological station deployed locally at the power station. These parameters include ambient temperature, relative humidity, wind speed and direction, and key illumination-related data such as total horizontal irradiance, direct normal irradiance, and diffuse irradiance. Particularly important is the cloud cover information and cloud movement trajectory prediction data obtained through satellite cloud image data analysis. The most direct and leading factor affecting photovoltaic power generation is the second type of data source: the power grid operation data source. This system communicates with the power grid dispatch center's data acquisition and monitoring system or energy management system through a security gateway to obtain real-time power grid operation status data provided by the wide-area measurement system. This includes real-time active load demand, system frequency, voltage level, and short-term load forecast curves for the next few hours to one day for the entire regional power grid or key nodes. In addition, it also includes power grid topology information, line transmission capacity limitations, and possible fault or maintenance status information. This data ensures that peak-shaving control strategies are always within the boundary constraints of safe power grid operation. The third type of data source is the solar power plant monitoring data source. This system connects to the monitoring and data acquisition system inside the power plant through industrial Ethernet or wireless sensor networks to read in real time the DC side voltage and current, AC side output power, operating efficiency, internal temperature, combiner box data, photovoltaic panel backsheet temperature sensor readings of each inverter, and, if the power plant is equipped with an energy storage system, it also reads key parameters such as the state of charge, available charge and discharge power, and health status of the energy storage battery packs provided by the battery management system.All data packets from different sources are timestamped and temporarily stored in the buffer of the data acquisition module. Next, the data preprocessing module begins its work, standardizing the raw multi-source data to eliminate noise and heterogeneity, providing high-quality input for subsequent prediction models. The data cleaning submodule first performs outlier detection on the incoming real-time data stream, using statistical methods, such as calculating the Z-score for each data stream. Data points that continuously exceed the historical mean by three standard deviations are marked as outliers and removed. For missing values ​​caused by temporary sensor malfunctions or communication interruptions, linear interpolation or smoother spline interpolation is used to fill in the missing values ​​based on the time-series characteristics of the data, ensuring the continuity of the data sequence. Subsequently, the data normalization submodule starts, using a minimum-maximum normalization algorithm to linearly transform the raw data with different units and numerical ranges (e.g., temperature in degrees Celsius, irradiance in watts per square meter, power in megawatts) to [0, 0, 1 ... Within the range of 1], this step effectively avoids certain features with large values ​​from dominating the machine learning model, accelerating model convergence and improving prediction accuracy. Finally, the time alignment submodule is responsible for solving the problem of inconsistent sampling frequencies of multi-source data (for example, meteorological data may be updated every 5 minutes, while power plant data is updated every second). It takes a unified clock reference (such as every minute on the hour) as its target, downsamples high-frequency data (such as power per second) and averages it, and upsamples low-frequency data (such as meteorological data every 5 minutes) using hold or interpolation algorithms (such as forward padding or linear interpolation). Finally, it generates a standardized dataset in which all feature parameters are at one-minute intervals and the timestamps are completely synchronized, and stores it in the real-time database for subsequent modules to call. Then, the power prediction module loads this standardized dataset and inputs it into the pre-trained long short-term memory network machine learning model. This model is a recurrent neural network specifically designed for processing time series data. It has a complex gating mechanism (input gate, forget gate, output gate) and can effectively capture long-term dependencies and dynamic change patterns in meteorological and power generation data.The model was trained offline, using several years of historical multi-source data (including historical meteorological, grid load, and power plant operation data) and corresponding actual solar power generation data as training samples. The backpropagation algorithm was used to backpropagate over time and gradient descent to minimize the error loss function between the predicted and actual power. K-fold cross-validation was used to adjust the network hyperparameters (such as the number of layers, neurons, and learning rate) to prevent overfitting and ensure the model's generalization ability. In the online prediction phase, the module used the standardized data sequence of the current moment and a past period (such as the past 2 hours) as the model input to predict the solar power generation value every minute within a specific future time window (such as the next 1 to 4 hours). To adapt to real-time changes, the module usually operates in a sliding window manner, for example, re-executing the prediction every 5 minutes to incorporate the latest data and dynamically update the prediction results, thereby achieving a rapid response to sudden changes in solar irradiance (such as rapid cloud movement).The forecast results include not only point predictions but also uncertainty interval estimates based on probabilistic models, providing risk references for subsequent optimization decisions. After obtaining accurate short-term power generation forecasts, the peak-shaving strategy generation module begins core optimization calculations. Its objective function is set as a multi-objective optimization problem, with the main objectives being: first, to minimize the peak load of the power grid during the forecast period, i.e., to smooth out peaks and valleys as much as possible by adjusting the output of solar power plants and the behavior of energy storage, making the total load curve of the power grid flatter; second, to maximize the utilization rate of solar power generation, maximizing the absorption of clean electricity and reducing curtailment. Simultaneously, multiple constraints must be considered, such as the upper and lower limits of safe grid operation (node ​​voltage must not exceed limits, line power must not be overloaded), and the energy storage system. Physical constraints (maintaining the state of charge between 20% and 90% to prevent overcharging and over-discharging, and ensuring that charging and discharging power does not exceed the rated value) and the inverter's maximum adjustable power range, along with the particle swarm optimization algorithm, simulate the foraging behavior of a flock of birds. It initializes a swarm of particles (each particle representing a set of possible control schemes, such as the inverter's set power and energy storage charging and discharging power at each moment in the future). The particles fly in the solution space, iteratively updating their position and velocity by tracking individual and group historical best solutions. After tens to hundreds of iterations, it converges to an optimal or near-optimal solution that satisfies all constraints. The optimization calculation process tightly integrates real-time grid topology and fault information; for example, when a critical transmission line is out of service due to a fault... The optimization algorithm immediately incorporates this new network constraint into its calculations, automatically adjusting the control strategy to ensure the feasibility of the solution. Based on the optimization results, this module generates a specific, time-seriesd set of peak-shaving control instructions, following a preset priority logic. For example, it prioritizes calling the energy storage system for charging and discharging to smooth power fluctuations and achieve peak-valley arbitrage, then adjusts the inverter's output power limit. In extreme cases (such as when a severe power generation shortage is predicted and the energy storage capacity is low), it initiates coordination instructions with the standby generator set or sends a load-cutting request to the load-side management system. Finally, the execution control module receives the instruction set from the peak-shaving strategy generation module. This module first converts the abstract instructions output by the optimization algorithm into standard control signals that can be recognized by specific devices. Then, through a highly reliable industrial communication network (such as a fiber optic private network or an encrypted 4G / 5G wireless network), these control signals are sent to the corresponding actuators. The photovoltaic inverter is sent with a setpoint command to adjust the active power output so that it operates at the specified power point. The energy storage converter is sent with commands for charging and discharging power and direction to control the energy throughput of the battery pack. If the strategy includes demand response, the switch-on and switch-off commands are sent to the smart switch of the interruptible load. All commands are issued with strict time synchronization tags to ensure that the control actions occur at the accurate time. After the actuators act, their status feedback signals (such as actual output power and switch status) are acquired again by the data acquisition module through the monitoring data source, forming a closed-loop feedback.Meanwhile, the central processing unit coordinates the entire process from data acquisition to execution control. It runs the task scheduling program, manages the start and stop of each module and resource allocation, and provides maintenance personnel with a panoramic visualization of the entire system's operating status through a human-machine interface. This includes real-time data curves, comparisons between predicted results and actual conditions, details of optimization strategies, alarm information, etc. Authorized personnel are also allowed to perform intervention operations such as adjusting model parameters and switching control modes (e.g., manual / automatic). The entire system executes the above workflow repeatedly. Through continuous data perception, intelligent analysis, optimization decision-making, and precise control, it achieves effective peak shaving of solar power generation, significantly improving the intelligence level of the power grid and the friendly access capability of renewable energy.

[0022] The above embodiments are provided for those skilled in the art to implement or use the present invention. Those skilled in the art can make various modifications or changes to the above embodiments without departing from the spirit of the present invention. Therefore, the scope of protection of the present invention is not limited to the above embodiments, but should be the maximum scope that conforms to the innovative features mentioned in the claims.

Claims

1. A solar power generation power peak shaving control method based on multi-source data analysis, characterized in that, The method comprises the following steps: Step 1: Real-time acquisition of multi-source data from multiple data sources through a data acquisition interface, including meteorological data sources, power grid operation data sources, and solar power station monitoring data sources, wherein the meteorological data sources provide temperature, humidity, wind speed, and cloud cover data, the power grid operation data sources provide historical load curves and real-time load demand data, and the solar power station monitoring data sources provide solar irradiance, photovoltaic panel temperature, and inverter operation parameters; Step 2: Data preprocessing of the acquired multi-source data, including data cleaning to remove outliers and fill in missing values, data normalization using the min-max normalization method to unify the data scale, and time alignment through an interpolation algorithm to synchronize data with different sampling frequencies to the same timestamp, thereby generating a standardized data set; Step 3: Inputting the standardized data set into a pre-trained machine learning prediction model, which is a long short-term memory network model, whose training process uses historical multi-source data and corresponding historical solar power generation data, and optimizes model parameters through a backpropagation algorithm and cross-validation to predict short-term solar power generation power fluctuations; Step 4: Based on the predicted solar power generation power and real-time power grid load data, an optimization algorithm is used to calculate peak shaving control instructions, which is based on a particle swarm optimization algorithm, with the objective function being to minimize the peak load of the power grid and maximize the utilization rate of solar power generation, while considering the safety constraints of the power grid; Step 5: Generating specific control strategies according to the peak shaving control instructions, including adjusting the output power of the solar inverter, controlling the charge and discharge operations of the energy storage system to smooth the power output, and coordinating distributed energy resources such as backup generators or load-side management, and finally executing the control strategies through an actuator module to achieve peak shaving control of solar power generation and reduce the peak load of the power grid.

2. The solar power generation power peak shaving control method based on multi-source data analysis according to claim 1, characterized in that, In the data preprocessing step, data cleaning uses a statistical-based method to identify outliers, including using the Z-score method to detect data points deviating from the mean by more than three standard deviations, and using linear interpolation to fill in missing values; data normalization scales each data feature to the [0, 1] interval, and time alignment uses linear interpolation or spline interpolation algorithms to ensure that all data are in a unified time series with one-minute intervals.

3. The solar power generation power peak shaving control method based on multi-source data analysis according to claim 1, characterized in that, The training process of the machine learning prediction model also includes a data augmentation step, which expands the training data set by adding Gaussian noise or time series shifting to improve the model's generalization ability, and the model uses a rolling prediction method during prediction, updating the input data every preset time to perform real-time prediction, with a prediction time range of 1 to 4 hours in the future.

4. The solar power generation power peak shaving control method based on multi-source data analysis according to claim 1, characterized in that, When calculating the peak shaving control instructions, the optimization algorithm also integrates the power grid topology and real-time fault information to ensure voltage stability and frequency stability through constraint conditions, and the control strategy includes priority settings, where the charge and discharge of the energy storage system takes priority over inverter power adjustment, and when the predicted power generation is insufficient, the backup power is automatically started.

5. The solar power generation power peak shaving control method based on multi-source data analysis according to claim 1, characterized in that, Also included is a solar power generation power peak shaving control system based on multi-source data analysis, characterized in that it comprises: a data acquisition module for acquiring multi-source data in real time through wired or wireless communication interfaces, the multi-source data including meteorological data, power grid load data and solar power station operation data; a data preprocessing module connected with the data acquisition module for cleaning, normalizing and time aligning the multi-source data, and outputting a standardized data set; a power prediction module connected with the data preprocessing module and integrated with a long short-term memory network machine learning model for predicting solar power generation power based on the standardized data set; a peak shaving strategy generation module connected with the power prediction module and using an optimization algorithm to generate a peak shaving control strategy according to the predicted power and the power grid load; an execution control module connected with the peak shaving strategy generation module for converting the control strategy into a control signal and adjusting the solar inverter power, the energy storage system charge and discharge and the distributed energy resources through an actuator; and a central processing unit for coordinating the operation of each module and providing a man-machine interface for monitoring and parameter setting.

6. The solar power generation power peak shaving control method based on multi-source data analysis according to claim 5, characterized in that, The data acquisition module includes multiple sub-interfaces: a meteorological data interface connected to a meteorological bureau API or a local meteorological station, a power grid data interface connected to a power grid dispatching center SCADA system, and a power station monitoring data interface connected to a photovoltaic power station monitoring and data acquisition system or an Internet of Things sensor; the data preprocessing module is deployed on an edge computing device to realize low-delay processing.

7. The solar power generation power peak shaving control method based on multi-source data analysis according to claim 5, characterized in that, The power prediction module is deployed on a cloud server or a local server and uses a distributed computing framework to process large-scale data; the peak shaving strategy generation module is integrated with a real-time optimization engine and supports multi-objective optimization and dynamic constraint adjustment. 8.The solar power generation power peak shaving control method based on multi-source data analysis of claim 5, wherein, The execution control module includes a communication unit connected to the solar inverter, the energy storage system controller and the load switch through a 4G / 5G network or an optical fiber; the system also includes a security protection module for data encryption and access control to ensure the network security of the system.

Images