A photovoltaic microgrid energy management and optimization scheduling system

Abstract

The present invention discloses a photovoltaic microgrid energy management and optimization scheduling system, relates to the technical field of smart grids, and solves the problems of unstable output power, complex control strategy and poor real-time response capability of the existing system; the present invention predicts future photovoltaic power generation and load curves through an intelligent prediction and analysis module, provides a reliable data basis for the optimal scheduling of the system, adjusts the energy distribution between the photovoltaic system, energy storage device and controllable load in real time through a dynamic scheduling optimization module, realizes efficient energy utilization and stable operation of the system, optimizes the output of distributed energy resources and the charging and discharging strategy of the energy storage device through a distributed energy resource control module and an energy storage management and optimization module, and further improves the real-time response capability and operation efficiency of the system; the present invention greatly improves the energy management, prediction accuracy, real-time response capability and safety control capability of the existing photovoltaic microgrid.

Description

Technical Field

[0001] The present invention relates to the field of smart grid technology, and more specifically to a photovoltaic microgrid energy management and optimization scheduling system. Background Art

[0002] In the tide of energy transformation in the 21st century, photovoltaic microgrid systems came into being, symbolizing a leap forward in distributed energy technology. As the global consensus on sustainable development and energy conservation and emission reduction deepens, photovoltaic microgrids have become a key link in building future smart grids with their flexibility and environmental friendliness. These systems can not only provide reliable power supply in remote areas, but also serve as backup power sources to ensure continuous power supply to important loads when the main power grid fails, demonstrating strong self-healing capabilities and energy self-sufficiency potential.

[0003] The background technology of the existing photovoltaic microgrid energy management system relies on the deep integration of advanced information technology and traditional power engineering. Real-time energy dispatching technology can dynamically adjust power distribution according to photovoltaic power generation and load demand to ensure the stable operation of the microgrid. Demand-side management reduces energy demand and improves energy efficiency through intelligent load control. At the same time, the energy management system integrates functions such as power generation optimization, load forecasting, energy storage management and real-time monitoring, and can automatically synchronize the operation of the microgrid and the main grid to achieve intelligent energy management.

[0004] However, the photovoltaic microgrid energy management system has exposed a series of challenges in practice. The primary problem is the instability of photovoltaic output power, which is caused by the natural fluctuation of light intensity. This intermittent nature brings great difficulties to energy scheduling. Secondly, the complexity of the control strategy increases the cost of system design and maintenance. Especially for microgrids that require fine adjustment, how to achieve efficient coordination among many distributed power sources is a difficult task. In addition, the lack of prediction accuracy and real-time response capabilities further exacerbates the uncertainty of microgrid operation. Due to the unpredictability of meteorological conditions and the randomness of electricity demand, it is often difficult to plan the system accurately in advance. In the face of emergencies, delays in data processing and decision-making limit the system's immediate response, which may lead to an imbalance in energy supply and demand, affecting the stability and economic benefits of the microgrid.

[0005] Therefore, the present invention discloses a photovoltaic microgrid energy management and optimization scheduling system. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention discloses a photovoltaic microgrid energy management and optimization scheduling system, which greatly improves the energy management, prediction accuracy, real-time response capability and safety control capability of the existing photovoltaic microgrid.

[0007] In order to achieve the above technical effects, the present invention adopts the following technical solutions:

[0008] A photovoltaic microgrid energy management and optimization scheduling system, comprising:

[0009] Intelligent prediction and analysis module, which integrates real-time meteorological data, historical photovoltaic output records and user load data, and predicts future photovoltaic power generation and load curves through deep learning prediction algorithms;

[0010] A dynamic scheduling optimization module is used to receive the output data of the intelligent prediction and analysis module, adjust the energy distribution between the photovoltaic system, the energy storage device and the controllable load in real time, and send scheduling commands to the distributed energy resource control module and the energy storage management and optimization module;

[0011] A distributed energy resource control module, which is used to dynamically adjust the output of each distributed energy resource according to the scheduling instructions sent by the dynamic scheduling optimization module, and report the execution results to the real-time monitoring and data feedback module; the distributed energy resource control module is based on edge computing and microservice architecture, and realizes intelligent local control of distributed energy resources through a protocol adapter;

[0012] An energy storage management and optimization module is used to dynamically adjust the charge and discharge status of the energy storage device based on the scheduling instructions sent by the dynamic scheduling optimization module, and to feed back the execution status to the real-time monitoring and data feedback module; the energy storage management and optimization module optimizes the energy storage charge and discharge strategy through a deep health status assessment algorithm;

[0013] The real-time monitoring and data feedback module is used to monitor the status of the microgrid system in real time, collect the execution results of each module, and feed back to the intelligent prediction and analysis module and the dynamic scheduling optimization module; the real-time monitoring and data feedback module realizes real-time monitoring and secure data transmission of the system status through the Internet of Things sensor network and blockchain.

[0014] As a further technical solution of the present invention,

[0015] The working steps of the intelligent prediction and analysis module are as follows:

[0016] Step 201, data collection and preprocessing, integrating historical photovoltaic power generation, meteorological conditions and power load data from different sources through a heterogeneous data fusion method, and ensuring data integrity and accuracy through data cleaning and quality control; the heterogeneous data fusion method converts data from different data sources and formats into the same framework through data standardization and data mapping, and ensures data consistency and integrity through data warehouse ETL;

[0017] Step 202, feature engineering, extracting the periodic characteristics of photovoltaic power generation and load changes through wavelet transform and Fourier transform, and identifying the feature set that has the greatest impact on the prediction result through the mutual information method; the mutual information method quantifies the dependency between the feature and the target variable through the entropy difference between the joint distribution of random variables and the marginal distribution, and selects the most valuable features;

[0018] Step 203, model training and verification, capturing the time dependency of data through a long short-term memory network LSTM and a gated recurrent unit GRU, and constructing a deep learning prediction model; the deep learning prediction model is adjusted by a grid search and a random search method to include at least a learning rate and the number of hidden layers of the model;

[0019] Step 204: real-time prediction and adjustment, using the trained model to predict future photovoltaic power generation and load, and performing real-time data fusion and model online learning through a data stream processing framework and an online gradient descent method; the data stream processing framework integrates photovoltaic sensor and smart meter data in real time through a message queue and a streaming data processing method; the online gradient descent method enables the model to dynamically adjust the prediction results according to the newly arrived data through differential geometry and numerical optimization methods;

[0020] Step 205: Result feedback and model iteration. The prediction effect is regularly checked through cross-validation and leave-one-out method, and a prediction accuracy threshold is set. If the prediction accuracy is lower than the set threshold, the model is automatically retrained through the model update mechanism to maintain the prediction accuracy. The model update mechanism automatically loads new data, retrains the model, and deploys the updated model through version control and automated deployment methods.

[0021] Step 206, anomaly detection and correction, identifying abnormal patterns in the prediction through an anomaly detection algorithm, and adjusting the prediction output through the expert system and rule engine according to the abnormal situation; the anomaly detection algorithm establishes a baseline model of normal behavior through box plots and identifies observations that are out of the normal range;

[0022] Step 207, performance monitoring and reporting, build a visual dashboard to display the model operation status in real time, and the visual dashboard provides a periodic forecast performance overview through automated report generation; the model operation status includes but is not limited to forecast accuracy, response time, and computing resource utilization indicators.

[0023] As a further technical solution of the present invention,

[0024] The working steps of the deep learning prediction algorithm in the photovoltaic microgrid energy management and optimization scheduling system include:

[0025] Step 301, obtaining a photovoltaic IoT sensor data set;

[0026] The photovoltaic IoT sensor data set is composed of multiple data clusters, including but not limited to historical photovoltaic power generation, power load and meteorological condition information; and the photovoltaic IoT sensor data set includes m historical monitoring events, where m is a natural number greater than 1;

[0027] Step 302: Obtain control data features of the monitoring event;

[0028] For each of the m historical monitoring events, a control data feature is obtained; each of the monitoring events is associated with one or more control data features; the control data features include but are not limited to time series data and meteorological parameters;

[0029] Step 303: focus on feature analysis and knowledge collision processing;

[0030] 1) Select monitoring event n as the analysis object, and obtain the similarity score between the control data feature of monitoring event n and all other features through knowledge collision processing;

[0031] 2) Obtaining sparse coefficients, and performing sparse transformation on each score in the consistency score set to obtain a sparse transformation result;

[0032] 3) Numerical normalization of the sparse transformation results to obtain a set of importance indicators for each control data feature;

[0033] 4) Based on the importance index set, all control data features of the nth monitoring event are adjusted in importance and added to obtain the kth control data feature after iteration;

[0034] Step 304: Determine the characteristics of the monitoring event data;

[0035] Determine the monitoring event data characteristics according to the control data characteristics after each monitoring event iteration;

[0036] Step 305: Determine the association between the classification mapping and the event;

[0037] Classify and map the monitoring event data features of the m monitoring events to determine the event association of the monitoring events in the photovoltaic microgrid IoT sensor data set;

[0038] Step 306: Deep learning prediction and optimization scheduling;

[0039] The monitoring event data features after classification and mapping in step 305 are used as input, and the prediction model is trained through the deep learning model LSTM to predict and generate future photovoltaic power generation and power load predictions.

[0040] As a further technical solution of the present invention,

[0041] The dynamic scheduling optimization module includes a prediction data integration unit, a real-time monitoring information processing unit, a mixed integer linear programming scheduling unit, an instruction generation and transmission unit, and an adaptive adjustment and learning unit; the prediction data integration unit integrates the photovoltaic power generation, load demand and meteorological condition prediction data output by the intelligent prediction analysis module through a data fusion method, and the data fusion method extracts information from multiple data sources through Kalman filter and Bayesian estimation, and forms a prediction result through weight distribution and error correction; the real-time monitoring information processing unit is used to receive the real-time system status information of the real-time monitoring and data feedback module, and ensures that the scheduling decision is based on the latest system status through real-time data analysis and anomaly detection algorithm; the real-time system status information includes the current status of the photovoltaic system, energy storage device and controllable load; the real-time data analysis realizes the rapid processing of real-time data through a stream processing framework; the anomaly detection algorithm identifies the anomalies of the system status through the standard deviation threshold method and the machine learning model Chang; the mixed integer linear programming scheduling unit generates an optimal scheduling scheme based on the predicted data and real-time monitoring information by taking into account the system constraints through the mixed integer linear programming algorithm; the mixed integer linear programming scheduling unit also performs complex scheduling decisions for multiple optimization objectives through particle swarm optimization; the system constraints include but are not limited to power balance and equipment operation restrictions; based on the output result of the mixed integer linear programming scheduling unit, the instruction generation and transmission unit generates operation instructions through instruction encoding, and sends the operation instructions to the distributed energy resource control module and the energy storage management and optimization module through a secure communication protocol; the operation instructions include power adjustment commands for photovoltaic systems, energy storage devices and controllable loads; the adaptive adjustment and learning unit learns based on historical scheduling results and system feedback through deep reinforcement learning and model predictive control methods, and continuously adjusts the parameters of the optimization model; the model predictive control method achieves optimal scheduling under closed-loop control through dynamic programming and rolling optimization.

[0042] As a further technical solution of the present invention,

[0043] The algorithm working steps of the mixed integer linear programming scheduling unit include:

[0044] Step 501: define a total cost objective function to minimize the total operating cost while satisfying all system constraints; the total operating cost includes power purchase cost, power generation cost, energy storage operation cost and penalty cost; the total cost objective function is expressed as:

[0045]

[0046] In formula (1), T represents the total number of time steps, which is used to represent the number of time periods in the entire scheduling cycle; c i 、r i ,yi and u i They represent the amount of electricity purchased, generated, charged, discharged and penalized at time step i, respectively. i is the electricity purchase cost coefficient at time step i, which is used to calculate the cost of purchasing electricity from the grid; R i represents the power generation cost coefficient at time step i, which is used to calculate the cost of using the generator to generate electricity; Y i represents the energy storage operation cost coefficient at time step i, including the cost of charging and discharging; U i represents the penalty cost coefficient, which is used to calculate the penalty cost for not meeting the load demand;

[0047] Step 502: Constrain the system through an energy balance function to ensure that the generated energy matches the consumed energy; the system constraints include power balance constraints, energy storage constraints, equipment operation constraints, and grid interaction constraints; the formula expression of the energy balance function is:

[0048]

[0049] In formula (2), P L represents the total load demand at time step i; E i represents the energy storage state at time step i; P max Indicates the maximum charging and discharging power of the energy storage system; P V represents the power generated by the photovoltaic panel at time step i; P S represents the electricity sold to the grid at time step i; μ c and μ d They represent the charging and discharging efficiencies, respectively, and are used to measure the efficiency of energy storage devices; Δi represents the time step, which is used to calculate the change in energy storage per unit time.

[0050] Step 503: Process the start and stop status of the device through the device status formula. The device status formula is based on the binary variable Y h (i) Controlling the start-up decision of device h at time i: The formula of the device state formula is:

[0051] P h (i) = Y h (i)*P MAX_h +1 (3)

[0052] In formula (3), P h (i) represents the operating power of device h at time i, P MAX_h represents the maximum operating power of device i;

[0053] Step 504: When there are multiple optimization objectives, multiple objective functions are optimized by particle swarm optimization (PSO) update function to find the Pareto optimal solution set; the formula of the PSO update function is:

[0054]

[0055] In formula (4), θ1, θ2, θ3 and g represent the inertia weight, cognitive acceleration constant, social acceleration constant and random number, respectively, which control the degree of dependence of particles on past states, individual experience and social experience; v e (i) represents the velocity of particle e at time step i, which is used to indicate the speed at which the particle moves toward the optimal solution; x e (i) represents the position of particle e at time step i, which is used to represent the position of the particle in the search space; pbest e represents the best individual position of particle e; gbest represents the best position of the group.

[0056] As a further technical solution of the present invention,

[0057] The energy resources controlled by the distributed energy resource control module include photovoltaic systems, energy storage systems, wind power converters and cogeneration systems.

[0058] As a further technical solution of the present invention,

[0059] The distributed energy resource control module includes an edge processing submodule, a microservice control logic submodule, a protocol adapter submodule and a status monitoring feedback submodule; the edge processing submodule collects the status information of various distributed energy resources in real time through IoT sensors and edge gateways, and ensures data quality and availability through data standardization and anomaly detection; the microservice control logic submodule includes a microservice management and orchestration unit and a dynamic scheduling instruction execution unit; the microservice management and orchestration unit realizes the modular deployment and management of distributed energy resource control logic through containerization and microservice orchestration platform; the dynamic scheduling instruction execution unit is used to receive instructions from the dynamic scheduling optimization module, and perform instruction parsing and execution sequence planning based on event-driven architecture and state machine model; the protocol adapter submodule uses protocol The stack and middleware services realize the standardization and conversion of different energy resource communication protocols, and ensure the security of control instructions and status feedback data during transmission through encryption algorithms and digital signatures; the status monitoring feedback submodule includes a distributed energy resource status monitoring unit and a feedback information aggregation and reporting unit; the distributed energy resource status monitoring unit continuously monitors the operating status and health status of distributed energy resources through a real-time monitoring system and a remote sensing method, and the real-time monitoring system stores historical data through a time series database to provide data support for predictive analysis and fault diagnosis; the feedback information aggregation and reporting unit aggregates and optimizes the execution results and status information of distributed energy resources through data aggregation and compression algorithms, and reports them to the real-time monitoring and data feedback module through a secure communication channel.

[0060] As a further technical solution of the present invention,

[0061] The deep health status assessment algorithm includes an input layer, a convolutional layer, a recurrent layer, a fully connected layer, a policy network layer, a value network layer and an output layer; the operation process steps of the deep health status assessment algorithm in the system are:

[0062] Step 801: Receive real-time data provided by the system through the input layer, including but not limited to the current charging and discharging state of the energy storage device, battery voltage, current and temperature data, and clean and standardize the real-time data provided by the system to ensure data quality;

[0063] Step 802: extracting battery status features through the convolution layer and the circulation layer, and integrating and abstracting the extracted battery status features through the fully connected layer; the battery status features include but are not limited to the temperature distribution on the battery surface and the performance changes of historical charge and discharge cycles;

[0064] Step 803: outputting the best charging and discharging action or the probability distribution of the action of the energy storage device in the current state through the strategy network layer according to the extracted features and historical data; the strategy network layer trains the charging and discharging strategy of the energy storage device through a reinforcement learning algorithm;

[0065] Step 804: Evaluate the long-term returns of different charging and discharging strategies through the value network layer to assist the strategy network layer in selecting the optimal strategy;

[0066] Step 805: According to the outputs of the strategy network layer and the value network layer, the energy storage management and optimization module dynamically adjusts the charging and discharging state of the energy storage device, and feeds back the execution state to the real-time monitoring and data feedback module;

[0067] Step 806: Evaluate the battery health status by integrating the historical and real-time data of the battery through the fully connected layer and the output layer. The battery health status includes but is not limited to the remaining life and performance degradation.

[0068] Step 807: According to the health status assessment results and real-time monitoring data, the weights of the strategy network and the value network are updated through the loss function and the optimizer to optimize the charging and discharging strategy;

[0069] Step 808, looping through steps 801 to 807, to achieve dynamic management and optimal scheduling of energy storage equipment.

[0070] As a further technical solution of the present invention,

[0071] The operation process of the real-time monitoring and data feedback module includes the following steps:

[0072] Step 901: Collect system status data in real time through the IoT sensor network, where the system status data includes but is not limited to photovoltaic power generation, battery energy storage status, load demand, and grid interface status;

[0073] Step 902: The collected data is cleaned, outlier detected and format converted at the sensor node through a stream processing engine;

[0074] Step 903: The data is packaged into blocks through the blockchain, each block contains the hash value of the previous block, forming a chain structure to record and verify the data. When the data is transmitted, the information is encrypted during the data transmission process through the encrypted communication protocol SSL; the blockchain ensures the immutability of the data through a distributed ledger;

[0075] Step 904: The processed and verified data is fed back to the intelligent prediction analysis module and the dynamic scheduling optimization module in real time through a two-way communication mechanism, and the data is stored in a real-time database for historical data comparison and trend analysis; the two-way communication mechanism realizes two-way communication of data from the sensor to the central server and vice versa through the lightweight message transmission protocol MQTT;

[0076] Step 905: According to the decisions of the intelligent prediction and analysis module and the dynamic scheduling optimization module, the real-time monitoring and data feedback module guides various parts of the system to perform optimization operations through the real-time data analysis platform; the optimization operations include but are not limited to photovoltaic array angle adjustment and energy storage device charge and discharge control;

[0077] Step 906: Identify abnormal conditions or faults in the system through the statistical method IQR. When an abnormality is detected, execute predefined response measures through the automatic response mechanism. The predefined response measures include switching the backup power supply, adjusting the load, and notifying the maintenance personnel.

[0078] As a further technical solution of the present invention,

[0079] The system also includes a security and privacy protection module for ensuring the security of user data and system operations; the security and privacy protection module includes an access control submodule, a transmission security submodule, a logging submodule, a data anonymization submodule and a compliance check submodule; the access control submodule prevents unauthorized access and operation through attribute-based access control; the transmission security submodule ensures the security of data during transmission through a symmetric encryption algorithm and a secure socket layer; the logging submodule records all user operations and system activities through a log file system and timestamp; the data anonymization submodule adds random noise to the data set through a differential privacy method to ensure the accuracy of statistical results and the non-traceability of personal data; the compliance check submodule regularly checks whether the system complies with security policies and laws and regulations through automatic compliance scanning tools and expert reviews, and generates compliance reports.

[0080] Positive beneficial effects:

[0081] In the present invention, the intelligent prediction and analysis module uses a deep learning prediction algorithm to integrate real-time meteorological data, historical photovoltaic output records and user load data, effectively solving the problem of poor prediction accuracy in traditional systems, and can more accurately predict future photovoltaic power generation and load curves, providing a reliable data basis for the system's optimal scheduling. The dynamic scheduling optimization module adjusts the energy distribution between the photovoltaic system, energy storage device and controllable load in real time, significantly improving the problems of unstable output power and complex control strategy of the traditional system, and realizing efficient energy utilization and stable operation of the system. The distributed energy resource control module and the energy storage management and optimization module are based on advanced edge computing and microservice architectures. Through intelligent local control and deep health status assessment algorithms, they optimize the output of distributed energy resources and the charging and discharging strategies of energy storage devices, further improving the real-time response capability and operating efficiency of the system. The real-time monitoring and data feedback module uses the Internet of Things sensor network and blockchain technology to realize real-time monitoring and secure data transmission of the system status, effectively ensuring the safe and reliable operation of the system. In addition, the addition of the security and privacy protection module further enhances the data security and user privacy protection of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0083] Figure 1 A framework diagram of a photovoltaic microgrid energy management and optimization scheduling system of the present invention;

[0084] Figure 2 A diagram showing the working principle of the deep learning prediction algorithm of the present invention;

[0085] Figure 3 It is a step diagram of the algorithm working principle of the mixed integer linear programming scheduling unit of the present invention;

[0086] Figure 4 It is a principle framework diagram of the dynamic scheduling optimization module of the present invention;

[0087] Figure 5 This is a framework diagram of the security and privacy protection module of the present invention. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] like Figure 1-Figure 5 As shown, a photovoltaic microgrid energy management and optimization scheduling system includes:

[0090] Intelligent prediction and analysis module, which integrates real-time meteorological data, historical photovoltaic output records and user load data, and predicts future photovoltaic power generation and load curves through deep learning prediction algorithms;

[0091] A dynamic scheduling optimization module is used to receive the output data of the intelligent prediction and analysis module, adjust the energy distribution between the photovoltaic system, the energy storage device and the controllable load in real time, and send scheduling commands to the distributed energy resource control module and the energy storage management and optimization module;

[0092] A distributed energy resource control module, which is used to dynamically adjust the output of each distributed energy resource according to the scheduling instructions sent by the dynamic scheduling optimization module, and report the execution results to the real-time monitoring and data feedback module; the distributed energy resource control module is based on edge computing and microservice architecture, and realizes intelligent local control of distributed energy resources through a protocol adapter;

[0093] An energy storage management and optimization module is used to dynamically adjust the charge and discharge status of the energy storage device based on the scheduling instructions sent by the dynamic scheduling optimization module, and to feed back the execution status to the real-time monitoring and data feedback module; the energy storage management and optimization module optimizes the energy storage charge and discharge strategy through a deep health status assessment algorithm;

[0094] The real-time monitoring and data feedback module is used to monitor the status of the microgrid system in real time, collect the execution results of each module, and feed back to the intelligent prediction and analysis module and the dynamic scheduling optimization module; the real-time monitoring and data feedback module realizes real-time monitoring and secure data transmission of the system status through the Internet of Things sensor network and blockchain.

[0095] In a specific embodiment, the intelligent prediction and analysis module integrates real-time meteorological data, historical photovoltaic output records and user load data, and uses deep learning algorithms to model complex time series data to achieve accurate prediction of future photovoltaic power generation and load curves. The dynamic scheduling optimization module is based on the prediction results and combined with the current operating status of the system. It uses complex optimization algorithms to adjust the energy distribution between the photovoltaic system, energy storage device and controllable load in real time to maximize the economic benefits and energy utilization of the system. The distributed energy resource control module and the energy storage management and optimization module use edge computing and microservice architecture to achieve intelligent local control of distributed energy resources and in-depth health status assessment of energy storage equipment to ensure efficient and stable operation of the system. The real-time monitoring and data feedback module comprehensively collects system status data through the Internet of Things sensor network, and uses blockchain technology to ensure the security and immutability of data transmission, providing real-time and accurate data support for the system's intelligent decision-making.

[0096] In the specific implementation, first, the system collects real-time meteorological data, photovoltaic panel output data and user load data of the photovoltaic power station through the IoT sensor network, and transmits them to the intelligent prediction and analysis module. The module uses deep learning algorithms to process and analyze these data to generate predictions of photovoltaic power generation and load curves in the future. Then, the dynamic scheduling optimization module calculates the optimal energy allocation plan based on the prediction results and the current system status through complex optimization algorithms, and generates corresponding scheduling instructions. These instructions are then sent to the distributed energy resource control module and the energy storage management and optimization module. The control module adjusts the output power of the photovoltaic system, the charging and discharging status of the energy storage equipment, and the power consumption plan of the controllable load according to the instructions. At the same time, the real-time monitoring and data feedback module continuously monitors the system status, collects the execution results of each module, and feeds back the data to the intelligent prediction and analysis module and the dynamic scheduling optimization module for further optimization and adjustment. In addition, the security and privacy protection module ensures the security of system data and the protection of user privacy throughout the process.

[0097] In the specific implementation, through accurate prediction and intelligent optimization and scheduling, the system can significantly improve the stability and reliability of photovoltaic power generation and reduce the impact of unstable output power on the power grid. Secondly, the system's intelligent local control and deep health status assessment functions extend the service life of distributed energy resources and energy storage equipment and reduce operation and maintenance costs. Thirdly, the application of real-time monitoring and data feedback modules ensures the comprehensive perception of system status and real-time transmission of data, providing a strong guarantee for the safe and stable operation of the system. Finally, the implementation of the system promotes the large-scale access and efficient utilization of renewable energy, and helps to promote the optimization and upgrading of my country's energy structure and the realization of green and low-carbon development goals. In short, the implementation of this system has greatly improved the energy management, prediction accuracy, real-time response capability and safety control capability of photovoltaic microgrids, and provided strong support for the utilization of renewable energy and the intelligent management of energy systems.

[0098] In the above embodiment, the working steps of the intelligent prediction and analysis module are:

[0099] Step 201, data collection and preprocessing, integrating historical photovoltaic power generation, meteorological conditions and power load data from different sources through a heterogeneous data fusion method, and ensuring data integrity and accuracy through data cleaning and quality control; the heterogeneous data fusion method converts data from different data sources and formats into the same framework through data standardization and data mapping, and ensures data consistency and integrity through data warehouse ETL;

[0100] Step 202, feature engineering, extracting the periodic characteristics of photovoltaic power generation and load changes through wavelet transform and Fourier transform, and identifying the feature set that has the greatest impact on the prediction result through the mutual information method; the mutual information method quantifies the dependency between the feature and the target variable through the entropy difference between the joint distribution of random variables and the marginal distribution, and selects the most valuable features;

[0101] Step 203, model training and verification, capturing the time dependency of data through a long short-term memory network LSTM and a gated recurrent unit GRU, and constructing a deep learning prediction model; the deep learning prediction model is adjusted by a grid search and a random search method to include at least a learning rate and the number of hidden layers of the model;

[0102] Step 204: real-time prediction and adjustment, using the trained model to predict future photovoltaic power generation and load, and performing real-time data fusion and model online learning through a data stream processing framework and an online gradient descent method; the data stream processing framework integrates photovoltaic sensor and smart meter data in real time through a message queue and a streaming data processing method; the online gradient descent method enables the model to dynamically adjust the prediction results according to the newly arrived data through differential geometry and numerical optimization methods;

[0103] Step 205: Result feedback and model iteration. The prediction effect is regularly checked through cross-validation and leave-one-out method, and a prediction accuracy threshold is set. If the prediction accuracy is lower than the set threshold, the model is automatically retrained through the model update mechanism to maintain the prediction accuracy. The model update mechanism automatically loads new data, retrains the model, and deploys the updated model through version control and automated deployment methods.

[0104] Step 206, anomaly detection and correction, identifying abnormal patterns in the prediction through an anomaly detection algorithm, and adjusting the prediction output through the expert system and rule engine according to the abnormal situation; the anomaly detection algorithm establishes a baseline model of normal behavior through box plots and identifies observations that are out of the normal range;

[0105] Step 207, performance monitoring and reporting, build a visual dashboard to display the model operation status in real time, and the visual dashboard provides a periodic forecast performance overview through automated report generation; the model operation status includes but is not limited to forecast accuracy, response time, and computing resource utilization indicators.

[0106] In the above embodiment, the working steps of the deep learning prediction algorithm in the photovoltaic microgrid energy management and optimization scheduling system include:

[0107] Step 301, obtaining a photovoltaic IoT sensor data set;

[0108] The photovoltaic IoT sensor data set is composed of multiple data clusters, including but not limited to historical photovoltaic power generation, power load and meteorological condition information; and the photovoltaic IoT sensor data set includes m historical monitoring events, where m is a natural number greater than 1;

[0109] Step 302: Obtain control data features of the monitoring event;

[0110] For each of the m historical monitoring events, a control data feature is obtained; each of the monitoring events is associated with one or more control data features; the control data features include but are not limited to time series data and meteorological parameters;

[0111] Step 303: focus on feature analysis and knowledge collision processing;

[0112] 1) Select monitoring event n as the analysis object, and obtain the similarity score between the control data feature of monitoring event n and all other features through knowledge collision processing;

[0113] 2) Obtaining sparse coefficients, and performing sparse transformation on each score in the consistency score set to obtain a sparse transformation result;

[0114] 3) Numerical normalization of the sparse transformation results to obtain a set of importance indicators for each control data feature;

[0115] 4) Based on the importance index set, all control data features of the nth monitoring event are adjusted in importance and added to obtain the kth control data feature after iteration;

[0116] Step 304: Determine the characteristics of the monitoring event data;

[0117] Determine the monitoring event data characteristics according to the control data characteristics after each monitoring event iteration;

[0118] Step 305: Determine the association between the classification mapping and the event;

[0119] Classify and map the monitoring event data features of the m monitoring events to determine the event association of the monitoring events in the photovoltaic microgrid IoT sensor data set;

[0120] Step 306: Deep learning prediction and optimization scheduling;

[0121] The monitoring event data features after classification and mapping in step 305 are used as input, and the prediction model is trained through the deep learning model LSTM to predict and generate future photovoltaic power generation and power load predictions.

[0122] In a specific embodiment, the intelligent prediction and analysis module constructs a multi-dimensional data set by integrating real-time meteorological data (such as light intensity, temperature, humidity, etc.), historical photovoltaic output records, and user load data. Deep learning algorithms, such as long short-term memory networks (LSTM) or gated recurrent units (GRU), are used to process these time series data to capture the dynamic changes in photovoltaic power generation and load curves. The algorithm automatically learns and extracts complex patterns and laws from the data through multiple layers of nonlinear transformations, thereby achieving accurate predictions of future photovoltaic power generation and load curves.

[0123] LSTM and GRU introduce gating mechanisms (such as input gate, forget gate, and output gate) to allow the model to selectively remember or forget information during training, thereby effectively handling long-term dependency problems in time series. In addition, spatially related data, such as meteorological images, are processed through convolutional neural networks (CNNs) to enhance the input diversity of the prediction model.

[0124] In the specific implementation, the intelligent prediction and analysis module first receives real-time data from meteorological stations, photovoltaic power plants and load monitoring systems. These data undergo a preprocessing phase, including data cleaning, missing value filling, outlier detection and normalization, to ensure the quality and consistency of the data. Subsequently, the processed data is fed into the deep learning model for training. During the training process, the model calculates the prediction results through forward propagation and adjusts the model parameters through the backpropagation algorithm to minimize the prediction error. This process usually requires a lot of computing resources and time, but once the model training is completed, it can quickly and accurately predict future photovoltaic power generation and load curves. In the prediction stage, the module receives new input data in real time, uses the trained model for prediction, and outputs the prediction results to the dynamic scheduling optimization module for further decision-making.

[0125] Specifically, Experiment 1 is designed to evaluate the performance of deep learning prediction algorithms in photovoltaic output prediction; the deep learning prediction algorithm uses long short-term memory network (LSTM) and traditional ARIMA model as comparison benchmarks. The evaluation indicators use mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) to measure the prediction accuracy.

[0126] The data comes from the real-time monitoring system of the photovoltaic microgrid, including photovoltaic output power, light intensity, temperature, humidity and user load data every 15 minutes. In the preprocessing step, the model architecture is an LSTM network, which contains two LSTM layers, each with 128 hidden units, a Dropout layer to reduce overfitting, and a fully connected layer to output the predicted value.

[0127] The training process is as follows: 1. Data partitioning: Divide the dataset into training set (80%), validation set (10%) and test set (10%). 2. Hyperparameter setting: Learning rate is 0.001, batch size is 32, and training cycle is 100 rounds. 3. Loss function: Use mean square error as loss function. 4. Optimizer: Adam optimizer. 5. Prediction stage: Use the trained model to predict the photovoltaic output value in the test set.

[0128] For comparison, the ARIMA (2,1,2) model is used to predict the photovoltaic output. The model parameters are obtained through grid search optimization. The training set is also used for model training, and prediction is performed on the test set. The final data record is shown in Table 1:

[0129] Table 1 Comparison of deep learning prediction algorithm and traditional method for predicting photovoltaic output

[0130]

[0131] In Data Table 1, Actual PV Output (kW) represents the actual output power of the PV microgrid at a specific time point. Deep Learning Prediction (kW) represents the PV output power predicted by the deep learning algorithm. Error (Deep Learning) (kW) The difference between the actual PV output and the deep learning prediction value. Traditional Method Prediction (kW) The PV output power predicted using traditional prediction methods (such as ARIMA, linear regression, etc.). Error (Traditional Method) (kW) The difference between the actual PV output and the traditional method prediction value.

[0132] Next, for the deep learning model and the ARIMA model, the mean absolute error (MAE), mean square error (MSE), and mean absolute percentage error (MAPE) on the test set are calculated respectively; as shown in Data Table 2:

[0133] Table 2 Comparison of evaluation indicators

[0134] <![CDATA[MSE / kW 2 ]]> MAE / kW MAPE Deep Learning Models 3.5 1.5 2.0％ ARIMA Model 15.5 4.5 5.0％

[0135] It can be seen from the data table 2 that the deep learning prediction algorithm shows a significant prediction accuracy advantage in the photovoltaic microgrid energy management and optimal scheduling system. The lower MSE, MAE and MAPE values ​​prove its superior performance in processing photovoltaic output prediction tasks. Compared with the traditional ARIMA model, the deep learning algorithm can more accurately predict photovoltaic output and provide more reliable data support for the energy management and optimal scheduling of photovoltaic microgrids, thereby improving the operating efficiency and economic benefits of the system.

[0136] In the above embodiment, the dynamic scheduling optimization module includes a prediction data integration unit, a real-time monitoring information processing unit, a mixed integer linear programming scheduling unit, an instruction generation and transmission unit, and an adaptive adjustment and learning unit; the prediction data integration unit integrates the photovoltaic power generation, load demand and meteorological condition prediction data output by the intelligent prediction analysis module through a data fusion method, and the data fusion method extracts information from multiple data sources through Kalman filters and Bayesian estimation, and forms prediction results through weight allocation and error correction; the real-time monitoring information processing unit is used to receive the real-time system status information of the real-time monitoring and data feedback module, and ensure that the scheduling decision is based on the latest system status through real-time data analysis and anomaly detection algorithms; the real-time system status information includes the current status of the photovoltaic system, energy storage device and controllable load; the real-time data analysis realizes rapid processing of real-time data through a stream processing framework; the anomaly detection algorithm recognizes the system through a standard deviation threshold method and a machine learning model state abnormality; the mixed integer linear programming scheduling unit generates an optimal scheduling scheme based on the predicted data and real-time monitoring information by taking into account the system constraints through the mixed integer linear programming algorithm; the mixed integer linear programming scheduling unit also performs complex scheduling decisions for multiple optimization objectives through particle swarm optimization; the system constraints include but are not limited to power balance and equipment operation restrictions; based on the output result of the mixed integer linear programming scheduling unit, the instruction generation and transmission unit generates operation instructions through instruction encoding, and sends the operation instructions to the distributed energy resource control module and the energy storage management and optimization module through a secure communication protocol; the operation instructions include power adjustment commands for photovoltaic systems, energy storage devices and controllable loads; the adaptive adjustment and learning unit learns based on historical scheduling results and system feedback through deep reinforcement learning and model predictive control methods, and continuously adjusts the parameters of the optimization model; the model predictive control method achieves optimal scheduling under closed-loop control through dynamic programming and rolling optimization.

[0137] In the above embodiment, the algorithm working steps of the mixed integer linear programming scheduling unit include:

[0138] Step 501: define a total cost objective function to minimize the total operating cost while satisfying all system constraints; the total operating cost includes power purchase cost, power generation cost, energy storage operation cost and penalty cost; the total cost objective function is expressed as:

[0139]

[0140] In formula (1), T represents the total number of time steps, which is used to represent the number of time periods in the entire scheduling cycle; c i 、r i ,y i and u iThey represent the amount of electricity purchased, generated, charged, discharged and penalized at time step i, respectively. i is the electricity purchase cost coefficient at time step i, which is used to calculate the cost of purchasing electricity from the grid; R i represents the power generation cost coefficient at time step i, which is used to calculate the cost of using the generator to generate electricity; Y i represents the energy storage operation cost coefficient at time step i, including the cost of charging and discharging; U i represents the penalty cost coefficient, which is used to calculate the penalty cost for not meeting the load demand;

[0141] Step 502: Constrain the system through an energy balance function to ensure that the generated energy matches the consumed energy; the system constraints include power balance constraints, energy storage constraints, equipment operation constraints, and grid interaction constraints; the formula expression of the energy balance function is:

[0142]

[0143] In formula (2), P L represents the total load demand at time step i; E i represents the energy storage state at time step i; P max Indicates the maximum charging and discharging power of the energy storage system; P V represents the power generated by the photovoltaic panel at time step i; P S represents the electricity sold to the grid at time step i; μ c and μ d They represent the charging and discharging efficiencies, respectively, and are used to measure the efficiency of energy storage devices; Δi represents the time step, which is used to calculate the change in energy storage per unit time.

[0144] Step 503: Process the start and stop status of the device through the device status formula. The device status formula is based on the binary variable Y h (i) Controlling the start-up decision of device h at time i: The formula of the device state formula is:

[0145] P h (i) = Y h (i)*P MAX_h +1 (3)

[0146] In formula (3), P h (i) represents the operating power of device h at time i, P MAX_h represents the maximum operating power of device i;

[0147] Step 504: When there are multiple optimization objectives, multiple objective functions are optimized by particle swarm optimization (PSO) update function to find the Pareto optimal solution set; the formula of the PSO update function is:

[0148]

[0149] In formula (4), θ1, θ2, θ3 and g represent the inertia weight, cognitive acceleration constant, social acceleration constant and random number, respectively, which control the degree of dependence of particles on past states, individual experience and social experience; v e (i) represents the velocity of particle e at time step i, which is used to indicate the speed at which the particle moves toward the optimal solution; x e (i) represents the position of particle e at time step i, which is used to represent the position of the particle in the search space; pbest e represents the best individual position of particle e; gbest represents the best position of the group.

[0150] In a specific embodiment, the specific implementation process of the dynamic scheduling optimization module is as follows: First, the prediction data integration unit receives the photovoltaic power generation forecast, load demand forecast and energy storage device status information from the intelligent prediction analysis module, and performs preprocessing to ensure the accuracy and consistency of the data. At the same time, the real-time monitoring information processing unit collects real-time data such as the output power of the photovoltaic system, the charge state of the energy storage device, and the power demand of the controllable load in real time through means such as sensor networks, data acquisition and monitoring systems. Subsequently, the mixed integer linear programming scheduling unit constructs an optimization scheduling model based on the prediction data and real-time data with the goal of minimizing operating costs, maximizing energy utilization efficiency and ensuring system stability. The model takes into account factors such as the uncertainty of photovoltaic power generation, the charging and discharging characteristics of the energy storage device, and the dynamic changes in load demand, and obtains the optimal energy allocation plan through a solution algorithm. The instruction generation and transmission unit generates a specific scheduling command based on the optimization results, and transmits it to the distributed energy resource control module and the energy storage management and optimization module through the communication network to achieve real-time control of the photovoltaic system, energy storage device and controllable load. Finally, the adaptive adjustment and learning unit iteratively optimizes the scheduling strategy based on the actual feedback of the system operation to improve the response speed and scheduling accuracy of the system.

[0151] like Figure 4As shown in the figure, in the actual operation process of the photovoltaic microgrid energy management and optimization scheduling system, the dynamic scheduling optimization module realizes the precise scheduling of the system energy through real-time monitoring and predictive analysis. The specific operation and analysis process is as follows: First, after the system is started, the prediction data integration unit and the real-time monitoring information processing unit start working to collect prediction data and real-time data respectively. The mixed integer linear programming scheduling unit constructs an optimization scheduling model based on these data and solves the optimal energy allocation scheme. Subsequently, the instruction generation and transmission unit transmits the scheduling command to the distributed energy resource control module and the energy storage management and optimization module to perform the corresponding energy adjustment operation. In this process, the system continuously receives new real-time data, compares and analyzes it with the prediction data, and dynamically adjusts the scheduling strategy to adapt to the real-time changes of the system. The adaptive adjustment and learning unit continuously optimizes the scheduling strategy based on the system operation feedback to improve the intelligence level and operation efficiency of the system. Through this series of operation analysis processes, the dynamic scheduling optimization module can ensure that the photovoltaic microgrid maintains an efficient and stable operation state under a complex and changeable operating environment.

[0152] Among them, the mixed integer linear programming MILP is responsible for finding the optimal energy allocation scheme in the photovoltaic microgrid energy management and optimization scheduling system while considering the system constraints. These system constraints may include power balance, equipment operation restrictions, battery charge and discharge status, electricity price fluctuations, etc. The MILP model usually consists of a set of linear equations and inequalities that define the system's operating rules and constraints, while the objective function expresses the system goal to be minimized or maximized, such as cost, emissions, or instability.

[0153] In the photovoltaic microgrid environment, the working principle of the MILP dispatch unit is to establish a mathematical model that includes the status and operating costs of all dispatchable resources (photovoltaic panels, energy storage devices, controllable loads). The decision variables in the model include both continuous variables (such as power generation and power consumption) and discrete variables (such as the on / off status of the device). The MILP solver will try to find a set of decision variable values ​​so that the objective function is optimal while satisfying all constraints. Its specific implementation is as follows:

[0154] Data preparation: First, collect and integrate forecast data from the intelligent forecast analysis module, including photovoltaic output forecast, load demand forecast and market electricity price forecast. At the same time, the real-time monitoring information processing unit provides the current system status, such as energy storage level, equipment status, etc.

[0155] Model construction: Build a MILP model based on forecast data and real-time information. The model needs to include the following parts:

[0156] Objective function: For example, minimizing total operating costs or maximizing system benefits.

[0157] Constraints: ensuring power balance, equipment operation within a safe range, energy storage status meeting requirements, etc.

[0158] Decision variables: photovoltaic output, energy storage charge and discharge rate, controllable load adjustment, etc.

[0159] Optimization solution: Use MILP solver to solve the above model and obtain the optimal scheduling strategy. During the solution process, the particle swarm optimization algorithm (PSO) can be used to assist in processing multi-objective optimization problems by simulating the foraging behavior of bird flocks and iteratively updating decision variables until the global optimal solution is found or the predetermined convergence criteria are reached.

[0160] Result analysis and feedback: Analyze the optimal scheduling strategy obtained, confirm that it meets all constraints, and evaluate its economy and reliability. Send the optimal scheduling plan to the instruction generation and transmission unit, and further convert it into specific scheduling commands.

[0161] In the actual operation process, the mixed integer linear programming scheduling unit dynamically adjusts the energy scheduling scheme according to the output data of the intelligent prediction and analysis module and the real-time operation data collected by the real-time monitoring information processing unit. First, a set of feasible solution sets that meet the system constraints is quickly generated through the MILP solver. Then, the PSO algorithm is started on the basis of the feasible solution set, and the scheduling scheme is gradually optimized through multiple iterations, considering the trade-offs between multiple optimization objectives (such as cost, efficiency, reliability, etc.). During the iteration process, the PSO algorithm updates the speed and position of the particles according to the fitness value of the particles (i.e., the evaluation index of the scheduling scheme), so that the particles gradually approach the global optimal solution. At the same time, the real-time monitoring information processing unit continuously receives new real-time operation data and feeds these data back to the MILP model and the PSO algorithm, so as to adjust the scheduling scheme in time to adapt to the real-time changes of the system. Finally, the mixed integer linear programming scheduling unit outputs the optimal energy scheduling scheme and sends it to the distributed energy resource control module and the energy storage management and optimization module for execution through the instruction generation and transmission unit.

[0162] Specifically, Experiment 2 is designed to compare the control effects of the MILP dispatch unit and the traditional rule-based dispatch algorithm; the MILP dispatch unit is based on a mixed integer linear programming algorithm, taking into account all system constraints to generate the optimal dispatch plan. The rule-based dispatch algorithm uses fixed rules, such as starting charging when the energy storage SOC is below a certain threshold and starting discharging when it is above another threshold; or storing excess energy when the photovoltaic output is higher than the load, and using energy storage to supplement when the photovoltaic output is insufficient.

[0163] Evaluation indicators include: Total operating cost: including electricity purchase cost, equipment operating cost and potential penalty fees. Total revenue: including electricity sales revenue and other possible revenue. Supply and demand balance: evaluate whether power balance is met at all time points, that is, the power generation is equal to the consumption plus the change in energy storage. Equipment utilization: ensure that the equipment is within the safe operating range to avoid overcharging, over-discharging and overloading. Response speed: evaluate the reaction speed of the algorithm when dealing with sudden changes, such as a drop in photovoltaic output caused by a sudden change in weather. The data recording table is shown in Table 3:

[0164] Table 3 MILP scheduling unit performance comparison table

[0165] Timestamp MILP Scheduling Unit Rule-based Scheduling Algorithm Total operating cost $2,350.00 $2,800.00 Total revenue $1,200.00 $1,000.00 Supply and demand balance 100％ 95％ Equipment utilization 90％ 85％ Response speed 1 second 10 seconds

[0166] From the experimental data in Table 3, it can be seen that the MILP dispatch unit has significant cost savings, increased revenue, better supply and demand balance and equipment utilization in the photovoltaic microgrid energy management and optimization dispatch system compared with the traditional rule-based dispatch algorithm, which proves the advantages of MILP in dealing with complex dispatch decisions of photovoltaic microgrids. Although the MILP dispatch unit may need more consideration in computing resources and response speed, the economic benefits and system stability improvement it brings are worthwhile.

[0167] In general, the dynamic scheduling optimization module realizes the precise scheduling and optimal allocation of system energy through real-time monitoring and predictive analysis, effectively improving energy utilization efficiency and power supply reliability. Secondly, through the application of advanced optimization algorithms such as mixed integer linear programming, the module can obtain the optimal solution under the consideration of multiple constraints, ensuring the scientificity and rationality of the scheduling strategy. In addition, the application of adaptive adjustment and learning units enables the system to continuously optimize the scheduling strategy according to operation feedback, improving the system's adaptive ability and intelligence level. These beneficial effects not only help reduce system operating costs and improve economic benefits, but also help promote the widespread application and sustainable development of photovoltaic microgrid technology. It is of great significance to promote energy structure optimization, reduce environmental pollution, and achieve green and low-carbon development.

[0168] In the above embodiment, the energy resources controlled by the distributed energy resource control module include a photovoltaic system, an energy storage system, a wind power converter and a cogeneration system.

[0169] In the above embodiment, the distributed energy resource control module includes an edge processing submodule, a microservice control logic submodule, a protocol adapter submodule and a status monitoring feedback submodule; the edge processing submodule collects the status information of various distributed energy resources in real time through Internet of Things sensors and edge gateways, and ensures data quality and availability through data standardization and anomaly detection; the microservice control logic submodule includes a microservice management and orchestration unit and a dynamic scheduling instruction execution unit; the microservice management and orchestration unit realizes the modular deployment and management of distributed energy resource control logic through containerization and microservice orchestration platform; the dynamic scheduling instruction execution unit is used to receive instructions from the dynamic scheduling optimization module, and perform instruction parsing and execution sequence planning based on event-driven architecture and state machine model; the protocol adapter submodule Standardization and conversion of different energy resource communication protocols are achieved through protocol stacks and middleware services, and the security of control instructions and status feedback data during transmission is guaranteed by encryption algorithms and digital signatures; the status monitoring feedback submodule includes a distributed energy resource status monitoring unit and a feedback information aggregation and reporting unit; the distributed energy resource status monitoring unit continuously monitors the operating status and health status of distributed energy resources through a real-time monitoring system and a remote sensing method, and the real-time monitoring system stores historical data through a time series database to provide data support for predictive analysis and fault diagnosis; the feedback information aggregation and reporting unit aggregates and optimizes the execution results and status information of distributed energy resources through data aggregation and compression algorithms, and reports them to the real-time monitoring and data feedback module through a secure communication channel.

[0170] In a specific embodiment, the distributed energy resource control module uses the edge processing submodule to perform real-time data processing and analysis, the microservice control logic submodule executes specific control logic, the protocol adapter submodule implements communication protocol adaptation with different distributed energy resources, and the status monitoring feedback submodule is responsible for monitoring the operating status of distributed energy resources and reporting the execution results to the real-time monitoring and data feedback module. Through the collaborative work of this series of components, the distributed energy resource control module can realize intelligent and efficient control of various distributed energy resources such as photovoltaics, energy storage, and controllable loads.

[0171] In the photovoltaic microgrid energy management and optimization scheduling system, the dynamic scheduling optimization module generates the optimal energy scheduling plan based on the output data of the intelligent prediction and analysis module and the information of the real-time monitoring information processing unit, and sends the scheduling instructions to the distributed energy resource control module through the instruction generation and transmission unit. After the distributed energy resource control module receives the scheduling instruction, the edge processing submodule immediately parses and processes the instruction, identifies the distributed energy resources that need to be adjusted and their adjustment amount.

[0172] Next, the microservice control logic submodule calls the corresponding microservice to execute specific control logic according to the processing results of the edge processing submodule. Each microservice is independently responsible for the control strategy of one or more distributed energy resources. Through the collaborative work between microservices, the overall control of the complex system is achieved. At the same time, the protocol adapter submodule converts the control instructions into a format that can be recognized by the device according to the actual communication protocol of the distributed energy resources, and sends them to the corresponding distributed energy resources through the network.

[0173] In the process of executing control instructions, the state monitoring feedback submodule continuously monitors the operating status of distributed energy resources, including output power, operating efficiency, fault information, etc., and reports the monitoring results to the real-time monitoring and data feedback module in real time. The real-time monitoring and data feedback module further analyzes these monitoring data and provides feedback to the dynamic scheduling optimization module so as to adjust the energy scheduling plan in time.

[0174] During the operation analysis process, the distributed energy resource control module ensures that each distributed energy resource operates according to the optimal scheduling plan through real-time monitoring and data analysis. When the power generation of the photovoltaic system is higher than the load demand, the control module will instruct the energy storage device to charge to store excess electricity; when the load demand increases, the control module will prioritize the power generation of the photovoltaic system and start the energy storage device to discharge in time to meet the load demand. At the same time, for controllable loads, the control module will dynamically adjust its operating power according to factors such as real-time electricity prices and system operating status to achieve efficient energy utilization and economical operation.

[0175] During operation, the distributed energy resource control module also needs to consider the handling of various abnormal situations. For example, when a distributed energy resource fails, the control module will immediately identify and take corresponding measures, such as switching to backup power supplies, adjusting the output of other energy resources, etc., to ensure the stable operation of the system. In addition, the control module will continuously optimize the control strategy based on real-time monitoring data to improve the system's adaptability and robustness.

[0176] The application of distributed energy resource control module in photovoltaic microgrid energy management and optimization scheduling system has important significance and beneficial effects. First, through intelligent local control strategy, the module can respond to scheduling instructions in real time and quickly adjust the output of distributed energy resources, which improves the response speed and operation efficiency of the system. Secondly, based on the design of edge computing and microservice architecture, the system has better scalability and maintainability, which is convenient for subsequent functional upgrades and expansions.

[0177] In addition, through real-time monitoring and data analysis, the module can accurately grasp the operating status and performance of distributed energy resources, providing strong data support for the dynamic scheduling optimization module. This helps to optimize the energy scheduling plan and improve energy utilization efficiency and economy. At the same time, timely handling of abnormal situations and strategy optimization also enhance the stability and reliability of the system.

[0178] In the above embodiment, the deep health status assessment algorithm includes an input layer, a convolutional layer, a recurrent layer, a fully connected layer, a policy network layer, a value network layer and an output layer; the operation process steps of the deep health status assessment algorithm in the system are:

[0179] Step 801: Receive real-time data provided by the system through the input layer, including but not limited to the current charging and discharging state of the energy storage device, battery voltage, current and temperature data, and clean and standardize the real-time data provided by the system to ensure data quality;

[0180] Step 802: extracting battery status features through the convolution layer and the circulation layer, and integrating and abstracting the extracted battery status features through the fully connected layer; the battery status features include but are not limited to the temperature distribution on the battery surface and the performance changes of historical charge and discharge cycles;

[0181] Step 803: outputting the best charging and discharging action or the probability distribution of the action of the energy storage device in the current state through the strategy network layer according to the extracted features and historical data; the strategy network layer trains the charging and discharging strategy of the energy storage device through a reinforcement learning algorithm;

[0182] Step 804: Evaluate the long-term returns of different charging and discharging strategies through the value network layer to assist the strategy network layer in selecting the optimal strategy;

[0183] Step 805: According to the outputs of the strategy network layer and the value network layer, the energy storage management and optimization module dynamically adjusts the charging and discharging state of the energy storage device, and feeds back the execution state to the real-time monitoring and data feedback module;

[0184] Step 806: Evaluate the battery health status by integrating the historical and real-time data of the battery through the fully connected layer and the output layer. The battery health status includes but is not limited to the remaining life and performance degradation.

[0185] Step 807: According to the health status assessment results and real-time monitoring data, the weights of the strategy network and the value network are updated through the loss function and the optimizer to optimize the charging and discharging strategy;

[0186] Step 808, looping through steps 801 to 807, to achieve dynamic management and optimal scheduling of energy storage equipment.

[0187] In a specific embodiment, the deep health status assessment algorithm combines data-driven and model-driven methods. By collecting the operating data of the energy storage device (such as charging and discharging current, voltage, temperature, etc.), the algorithm uses machine learning or deep learning models to extract features, classify and predict the data. The algorithm first constructs an assessment model for the health status of the energy storage device, which can learn and capture the trends and patterns of equipment performance degradation; then, during real-time operation, the algorithm continuously receives new data, updates and calibrates the model to reflect the latest health status of the energy storage device. Through deep health status assessment, the algorithm can accurately predict the remaining life and capacity attenuation of the energy storage device, thereby providing a scientific basis for the optimization of the charging and discharging strategy.

[0188] In the photovoltaic microgrid energy management and optimization scheduling system, the specific implementation process of the deep health status assessment algorithm includes the following key steps: First, collect comprehensive operating data of energy storage equipment, including but not limited to the number of charge and discharge cycles, charge and discharge depth of each cycle, operating temperature, voltage fluctuation, etc. These data are the basis for the algorithm to evaluate the health status of energy storage equipment. Secondly, preprocess the collected data, including data cleaning (removing noise and outliers), feature extraction (selecting features useful for health status assessment), etc. Then, use machine learning or deep learning algorithms to build an energy storage equipment health status assessment model. In the model training stage, historical data is used to train the model, and the model parameters are adjusted to minimize the prediction error. After the model training is completed, it is deployed to the energy storage management and optimization module to achieve real-time health status assessment. Finally, according to the evaluation results, the charging and discharging strategy of the energy storage equipment is dynamically adjusted. For example, when it is predicted that the capacity of the energy storage equipment is significantly attenuated, its discharge depth is reduced or the charging frequency is increased to protect the equipment; during the peak load period of the power grid, energy storage equipment with good health status is preferentially dispatched to participate in the discharge to maximize the system benefits.

[0189] During the operation analysis process, the deep health status assessment algorithm continuously monitors the operating status of the energy storage equipment and updates the health status assessment results in real time. By comparing historical assessment data with current assessment data, the algorithm can identify the trend and abnormal changes in the performance degradation of energy storage equipment. Once accelerated performance degradation or abnormal status is found, the algorithm will immediately trigger the early warning mechanism and notify the operation and maintenance personnel to conduct inspections and maintenance. At the same time, based on the real-time assessment results, the energy storage management and optimization module dynamically adjusts the charging and discharging strategy to ensure that the energy storage equipment operates in the optimal state. For example, during periods of sufficient photovoltaic power generation, priority is given to charging energy storage equipment in good health; during peak load periods, discharge tasks are reasonably allocated according to the remaining capacity and health status of the energy storage equipment. In addition, the algorithm also improves the accuracy of health status assessment and the effectiveness of charging and discharging strategies by continuously learning and optimizing its own parameters.

[0190] Specifically, Experiment 3 is designed to apply the deep health status assessment algorithm to evaluate the health status of energy storage equipment in the photovoltaic microgrid energy management and optimization scheduling system; the data record table is shown in Table 4:

[0191] Table 4 Data Record Table

[0192]

[0193] According to the experimental data shown in Table 4, the deep health status assessment algorithm successfully conducted a comprehensive health status assessment of the energy storage equipment and obtained various health indicators. By comparing the initial value, minimum value, maximum value, average value and standard deviation, it can be found that the energy storage equipment has shown good stability in terms of capacity, number of charge and discharge cycles, and average charge and discharge depth. At the same time, the algorithm also successfully predicted the remaining life of the energy storage equipment, providing a scientific basis for subsequent maintenance plans.

[0194] In terms of strategy optimization, based on the health status assessment results, the energy storage management and optimization module successfully adjusted the charging and discharging strategy of the energy storage equipment, improving the equipment utilization and system stability. Experimental data showed that both the system stability improvement index and the economic benefit improvement index reached a high level, proving the positive and beneficial effects of the deep health status assessment algorithm in the photovoltaic microgrid energy management and optimization scheduling system.

[0195] In general, the deep health status assessment algorithm can detect and solve potential problems in a timely manner by accurately assessing the health status of energy storage equipment, thereby extending the service life of energy storage equipment and reducing operation and maintenance costs. Secondly, the optimization of charging and discharging strategies based on health status can maximize the utilization rate and economic benefits of energy storage equipment and improve the stability and reliability of the entire microgrid system. In addition, the deep health status assessment algorithm also provides strong support for the intelligent management of energy storage equipment and promotes the development of photovoltaic microgrids in a more intelligent, efficient and sustainable direction. In short, the application of this algorithm is of great significance to improving the energy management and optimal scheduling level of photovoltaic microgrids.

[0196] In the above embodiment, the operation process of the real-time monitoring and data feedback module includes the following steps:

[0197] Step 901: Collect system status data in real time through the IoT sensor network, where the system status data includes but is not limited to photovoltaic power generation, battery energy storage status, load demand, and grid interface status;

[0198] Step 902: The collected data is cleaned, outlier detected and format converted at the sensor node through a stream processing engine;

[0199] Step 903: The data is packaged into blocks through the blockchain, each block contains the hash value of the previous block, forming a chain structure to record and verify the data. When the data is transmitted, the information is encrypted during the data transmission process through the encrypted communication protocol SSL; the blockchain ensures the immutability of the data through a distributed ledger;

[0200] Step 904: The processed and verified data is fed back to the intelligent prediction analysis module and the dynamic scheduling optimization module in real time through a two-way communication mechanism, and the data is stored in a real-time database for historical data comparison and trend analysis; the two-way communication mechanism realizes two-way communication of data from the sensor to the central server and vice versa through the lightweight message transmission protocol MQTT;

[0201] Step 905: According to the decisions of the intelligent prediction and analysis module and the dynamic scheduling optimization module, the real-time monitoring and data feedback module guides various parts of the system to perform optimization operations through the real-time data analysis platform; the optimization operations include but are not limited to photovoltaic array angle adjustment and energy storage device charge and discharge control;

[0202] Step 906: Identify abnormal conditions or faults in the system through the statistical method IQR. When an abnormality is detected, execute predefined response measures through the automatic response mechanism. The predefined response measures include switching the backup power supply, adjusting the load, and notifying the maintenance personnel.

[0203] In a specific embodiment, the real-time monitoring and data feedback module is based on the Internet of Things (IoT) sensor network and blockchain technology. The IoT sensor network is responsible for collecting the operating status data of each distributed energy resource and energy storage device in the microgrid in real time, such as the output power of photovoltaic panels, the charging and discharging status of energy storage devices, the real-time demand of loads, etc. These data are transmitted to the central processing unit in real time through sensor nodes. Blockchain technology is used to ensure the security and immutability of data transmission, and record the detailed information of all data transmission and processing through a distributed ledger, thereby enhancing the transparency and reliability of the system.

[0204] During the operation of the system, the real-time monitoring and data feedback module first collects real-time data from various parts of the microgrid through IoT sensors. These data include but are not limited to the output power of photovoltaic panels, the charging and discharging status of energy storage equipment, and the real-time demand of loads. After receiving these data, the central server performs preliminary data cleaning and preprocessing to ensure the accuracy and availability of the data. Subsequently, data fusion technology is used to integrate data from different sensors to form a comprehensive understanding of the overall operating status of the microgrid. The anomaly detection algorithm further analyzes the data, identifies possible equipment failures or operating anomalies, and promptly feeds back to the operation and maintenance team. In addition, the central server also records and verifies the processed data through the blockchain network to ensure the immutability and traceability of the data.

[0205] In the specific implementation, the real-time monitoring and data feedback module collects real-time data through the IoT sensor network, so that the system can respond to the operation changes of the microgrid in a timely manner and improve the flexibility and adaptability of the system. Secondly, the introduction of blockchain technology greatly enhances the security and reliability of data, reduces the risk of data tampering and loss, and improves the credibility of the system. In addition, the implementation of this module also helps to optimize the operation strategy of the microgrid, dynamically adjust energy distribution and scheduling through real-time monitoring and feedback, improve energy utilization efficiency, and reduce operating costs. More importantly, the transparency and traceability of the system have been significantly improved, providing strong data support for the operation and maintenance management of the microgrid. In short, the implementation of the real-time monitoring and data feedback module not only improves the operation efficiency and safety of the photovoltaic microgrid, but also provides a solid foundation for the realization of intelligent, efficient and safe microgrid management.

[0206] In the above embodiment, the system also includes a security and privacy protection module for ensuring the security of user data and system operations; the security and privacy protection module includes an access control submodule, a transmission security submodule, a logging submodule, a data anonymization submodule and a compliance check submodule; the access control submodule prevents unauthorized access and operation through attribute-based access control; the transmission security submodule ensures the security of data during transmission through symmetric encryption algorithms and secure socket layers; the logging submodule records all user operations and system activities through a log file system and timestamp; the data anonymization submodule adds random noise to the data set through a differential privacy method to ensure the accuracy of statistical results and the non-traceability of personal data; the compliance check submodule regularly checks whether the system complies with security policies and laws and regulations through automatic compliance scanning tools and expert reviews, and generates compliance reports.

[0207] In the specific embodiment, the security of the system is significantly improved through access control and data transmission encryption, preventing unauthorized access and data leakage. In addition, the application of data anonymization technology ensures that the privacy of user data is effectively protected when conducting data analysis and publishing statistical results. Secondly, the logging system provides detailed records for security audits and problem tracking, enhancing the transparency and traceability of the system. At the same time, the compliance check mechanism ensures that the system always complies with relevant security policies and laws and regulations, reducing compliance risks.

[0208] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A photovoltaic microgrid energy management and optimization scheduling system, characterized by: The system comprises: Intelligent prediction and analysis module, which integrates real-time meteorological data, historical photovoltaic output records and user load data, and predicts future photovoltaic power generation and load curves through deep learning prediction algorithms; A dynamic scheduling optimization module is used to receive the output data of the intelligent prediction and analysis module, adjust the energy distribution between the photovoltaic system, the energy storage device and the controllable load in real time, and send scheduling commands to the distributed energy resource control module and the energy storage management and optimization module; A distributed energy resource control module is used to dynamically adjust the output of each distributed energy resource according to the scheduling instructions sent by the dynamic scheduling optimization module, and report the execution results to the real-time monitoring and data feedback module; the distributed energy resource control module is based on edge computing and microservice architecture, and realizes intelligent local control of distributed energy resources through a protocol adapter; the distributed energy resource control module includes an edge processing submodule, a microservice control logic submodule, a protocol adaptation submodule and a status monitoring feedback submodule; the edge processing submodule collects the status information of various distributed energy resources in real time through Internet of Things sensors and edge gateways, and ensures data quality and availability through data standardization and anomaly detection; the microservice control logic submodule includes a microservice management and orchestration unit and a dynamic scheduling instruction execution unit; the microservice management and orchestration unit realizes modular deployment and management of distributed energy resource control logic through containerization and microservice orchestration platform; the dynamic scheduling instruction execution unit The element is used to receive instructions from the dynamic scheduling optimization module, and parses instructions and executes sequence planning based on event-driven architecture and state machine model; the protocol adapter submodule realizes the standardization and conversion of different energy resource communication protocols through protocol stack and middleware services, and ensures the security of control instructions and status feedback data during transmission through encryption algorithm and digital signature; the status monitoring feedback submodule includes a distributed energy resource status monitoring unit and a feedback information aggregation and reporting unit; the distributed energy resource status monitoring unit continuously monitors the operating status and health status of distributed energy resources through real-time monitoring system and remote sensing method, and the real-time monitoring system stores historical data through time series database to provide data support for predictive analysis and fault diagnosis; the feedback information aggregation and reporting unit aggregates and optimizes the execution results and status information of distributed energy resources through data aggregation and compression algorithm, and reports to the real-time monitoring and data feedback module through a secure communication channel; The energy storage management and optimization module is used to dynamically adjust the charging and discharging status of the energy storage equipment based on the scheduling instructions sent by the dynamic scheduling optimization module, and feed back the execution status to the real-time monitoring and data feedback module; the energy storage management and optimization module optimizes the energy storage charging and discharging strategy through a deep health status assessment algorithm; the deep health status assessment algorithm includes an input layer, a convolutional layer, a recurrent layer, a fully connected layer, a policy network layer, a value network layer and an output layer; the operation process steps of the deep health status assessment algorithm in the system are: Step 801: Receive real-time data provided by the system through the input layer, including but not limited to the current charging and discharging state of the energy storage device, battery voltage, current and temperature data, and clean and standardize the real-time data provided by the system to ensure data quality; Step 802: extracting battery status features through the convolution layer and the circulation layer, and integrating and abstracting the extracted battery status features through the fully connected layer; the battery status features include but are not limited to the temperature distribution on the battery surface and the performance changes of historical charge and discharge cycles; Step 803: outputting the best charging and discharging action or the probability distribution of the action of the energy storage device in the current state through the strategy network layer according to the extracted features and historical data; the strategy network layer trains the charging and discharging strategy of the energy storage device through a reinforcement learning algorithm; Step 804: Evaluate the long-term returns of different charging and discharging strategies through the value network layer to assist the strategy network layer in selecting the optimal strategy; Step 805: According to the outputs of the strategy network layer and the value network layer, the energy storage management and optimization module dynamically adjusts the charging and discharging state of the energy storage device, and feeds back the execution state to the real-time monitoring and data feedback module; Step 806: Evaluate the battery health status by integrating the historical and real-time data of the battery through the fully connected layer and the output layer. The battery health status includes but is not limited to the remaining life and performance degradation. Step 807: According to the health status assessment results and real-time monitoring data, the weights of the strategy network and the value network are updated through the loss function and the optimizer to optimize the charging and discharging strategy; Step 808, looping through steps 801 to 807 to achieve dynamic management and optimal scheduling of energy storage equipment; The real-time monitoring and data feedback module is used to monitor the state of the microgrid system in real time, collect the execution results of each module, and feed back to the intelligent prediction and analysis module and the dynamic scheduling optimization module; the real-time monitoring and data feedback module realizes real-time monitoring and secure data transmission of the system state through the Internet of Things sensor network and blockchain; the operation process of the real-time monitoring and data feedback module includes the following steps: Step 901: Collect system status data in real time through the IoT sensor network, where the system status data includes but is not limited to photovoltaic power generation, battery energy storage status, load demand, and grid interface status; Step 902: The collected data is cleaned, outlier detected and format converted at the sensor node through a stream processing engine; Step 903: The data is packaged into blocks through the blockchain, each block contains the hash value of the previous block, forming a chain structure to record and verify the data. When the data is transmitted, the information is encrypted during the data transmission process through the encrypted communication protocol SSL; the blockchain ensures the immutability of the data through a distributed ledger; Step 904: The processed and verified data is fed back to the intelligent prediction analysis module and the dynamic scheduling optimization module in real time through a two-way communication mechanism, and the data is stored in a real-time database for historical data comparison and trend analysis; the two-way communication mechanism realizes two-way communication of data from the sensor to the central server and vice versa through the lightweight message transmission protocol MQTT; Step 905: According to the decisions of the intelligent prediction and analysis module and the dynamic scheduling optimization module, the real-time monitoring and data feedback module guides various parts of the system to perform optimization operations through the real-time data analysis platform; the optimization operations include but are not limited to photovoltaic array angle adjustment and energy storage device charge and discharge control; Step 906: Identify abnormal conditions or faults in the system through the statistical method IQR. When an abnormality is detected, execute predefined response measures through the automatic response mechanism. The predefined response measures include switching the backup power supply, adjusting the load, and notifying the maintenance personnel.

2. A photovoltaic microgrid energy management and optimization scheduling system according to claim 1, characterized in that: The working steps of the intelligent prediction and analysis module are as follows: Step 201, data collection and preprocessing, integrating historical photovoltaic power generation, meteorological conditions and power load data from different sources through a heterogeneous data fusion method, and ensuring data integrity and accuracy through data cleaning and quality control; the heterogeneous data fusion method converts data from different data sources and formats into the same framework through data standardization and data mapping, and ensures data consistency and integrity through data warehouse ETL; Step 202, feature engineering, extracting the periodic characteristics of photovoltaic power generation and load changes through wavelet transform and Fourier transform, and identifying the feature set that has the greatest impact on the prediction result through the mutual information method; the mutual information method quantifies the dependency between the feature and the target variable through the entropy difference between the joint distribution of random variables and the marginal distribution, and selects the most valuable features; Step 203, model training and verification, capturing the time dependency of data through a long short-term memory network LSTM and a gated recurrent unit GRU, and constructing a deep learning prediction model; the deep learning prediction model is adjusted by a grid search and a random search method to include at least a learning rate and the number of hidden layers of the model; Step 204: real-time prediction and adjustment, using the trained model to predict future photovoltaic power generation and load, and performing real-time data fusion and model online learning through a data stream processing framework and an online gradient descent method; the data stream processing framework integrates photovoltaic sensor and smart meter data in real time through a message queue and a streaming data processing method; the online gradient descent method enables the model to dynamically adjust the prediction results according to the newly arrived data through differential geometry and numerical optimization methods; Step 205: Result feedback and model iteration. The prediction effect is regularly checked through cross-validation and leave-one-out method, and a prediction accuracy threshold is set. If the prediction accuracy is lower than the set threshold, the model is automatically retrained through the model update mechanism to maintain the prediction accuracy. The model update mechanism automatically loads new data, retrains the model, and deploys the updated model through version control and automated deployment methods. Step 206, anomaly detection and correction, identifying abnormal patterns in the prediction through an anomaly detection algorithm, and adjusting the prediction output through the expert system and rule engine according to the abnormal situation; the anomaly detection algorithm establishes a baseline model of normal behavior through box plots and identifies observations that are out of the normal range; Step 207, performance monitoring and reporting, build a visual dashboard to display the model operation status in real time, and the visual dashboard provides a periodic forecast performance overview through automated report generation; the model operation status includes but is not limited to forecast accuracy, response time, and computing resource utilization indicators.

3. A photovoltaic microgrid energy management and optimization scheduling system according to claim 1, characterized in that: The working steps of the deep learning prediction algorithm in the photovoltaic microgrid energy management and optimization scheduling system include: Step 301, obtaining a photovoltaic IoT sensor data set; The photovoltaic IoT sensor data set is composed of multiple data clusters, including but not limited to historical photovoltaic power generation, power load and meteorological condition information; and the photovoltaic IoT sensor data set includes m historical monitoring events, where m is a natural number greater than 1; Step 302: Obtaining control data features of the monitoring event; For each of the m historical monitoring events, a control data feature is obtained; each of the monitoring events is associated with one or more control data features; the control data features include but are not limited to time series data and meteorological parameters; Step 303: focus on feature analysis and knowledge collision processing; 1) Select monitoring event n as the analysis object, and obtain the similarity score between the control data feature of monitoring event n and all other features through knowledge collision processing; 2) Obtaining sparse coefficients, and performing sparse transformation on each score in the consistency score set to obtain a sparse transformation result; 3) Numerical normalization of the sparse transformation results to obtain a set of importance indicators for each control data feature; 4) Based on the importance index set, all control data features of the nth monitoring event are adjusted in importance and added to obtain the kth control data feature after iteration; Step 304: determine the monitoring event data characteristics; Determine the monitoring event data characteristics according to the control data characteristics after each monitoring event iteration; Step 305: Determine the association between the classification mapping and the event; Classify and map the monitoring event data features of the m monitoring events to determine the event association of the monitoring events in the photovoltaic microgrid IoT sensor data set; Step 306: Deep learning prediction and optimization scheduling; The monitoring event data features after classification and mapping in step 305 are used as input, and the prediction model is trained through the deep learning model LSTM to predict and generate future photovoltaic power generation and power load predictions.

4. A photovoltaic microgrid energy management and optimization scheduling system according to claim 1, characterized in that: The dynamic scheduling optimization module includes a prediction data integration unit, a real-time monitoring information processing unit, a mixed integer linear programming scheduling unit, an instruction generation and transmission unit, and an adaptive adjustment and learning unit; the prediction data integration unit integrates the photovoltaic power generation, load demand and meteorological condition prediction data output by the intelligent prediction analysis module through a data fusion method, and the data fusion method extracts information from multiple data sources through a Kalman filter and Bayesian estimation, and forms a prediction result through weight allocation and error correction; The real-time monitoring information processing unit is used to receive the real-time system status information of the real-time monitoring and data feedback module, and ensure that the scheduling decision is based on the latest system status through real-time data analysis and anomaly detection algorithm; the real-time system status information includes the current status of the photovoltaic system, energy storage device and controllable load; the real-time data analysis realizes rapid processing of real-time data through the stream processing framework; the anomaly detection algorithm identifies anomalies in the system status through the standard deviation threshold method and machine learning model; the mixed integer linear programming scheduling unit generates the optimal scheduling plan based on the predicted data and real-time monitoring information through the mixed integer linear programming algorithm considering the system constraints; the mixed integer linear programming scheduling unit also performs multiple optimizations through particle swarm optimization The system constraints include but are not limited to power balance and equipment operation restrictions; based on the output results of the mixed integer linear programming scheduling unit, the instruction generation and transmission unit generates operation instructions through instruction encoding, and sends the operation instructions to the distributed energy resource control module and the energy storage management and optimization module through a secure communication protocol; the operation instructions include power adjustment commands for photovoltaic systems, energy storage devices and controllable loads; the adaptive adjustment and learning unit learns based on historical scheduling results and system feedback through deep reinforcement learning and model predictive control methods, and continuously adjusts the parameters of the optimization model; the model predictive control method achieves optimal scheduling under closed-loop control through dynamic programming and rolling optimization.

5. A photovoltaic microgrid energy management and optimization scheduling system according to claim 4, characterized in that: The algorithm working steps of the mixed integer linear programming scheduling unit include: Step 501: define a total cost objective function to minimize the total operating cost while satisfying all system constraints; the total operating cost includes power purchase cost, power generation cost, energy storage operation cost and penalty cost; the total cost objective function is expressed as: In formula (1), T represents the total number of time steps, which is used to represent the number of time periods in the entire scheduling cycle; c i 、r i ,y i and u i They represent the amount of electricity purchased, generated, charged, discharged and penalized at time step i, respectively. i is the electricity purchase cost coefficient at time step i, which is used to calculate the cost of purchasing electricity from the grid; R i represents the power generation cost coefficient at time step i, which is used to calculate the cost of using the generator to generate electricity; Y i represents the energy storage operation cost coefficient at time step i, including the cost of charging and discharging; U i represents the penalty cost coefficient, which is used to calculate the penalty cost for not meeting the load demand; Step 502: Constrain the system through an energy balance function to ensure that the generated energy matches the consumed energy; the system constraints include power balance constraints, energy storage constraints, equipment operation constraints, and grid interaction constraints; the formula expression of the energy balance function is: In formula (2), P L represents the total load demand at time step i; E i represents the energy storage state at time step i; P max Indicates the maximum charging and discharging power of the energy storage system; P V represents the power generated by the photovoltaic panel at time step i; P S represents the electricity sold to the grid at time step i; μ c and μ d They represent the charging and discharging efficiencies, respectively, and are used to measure the efficiency of energy storage devices; Δi represents the time step, which is used to calculate the change in energy storage per unit time. Step 503: Process the start and stop status of the device through the device status formula. The device status formula is based on the binary variable Y h (i) Controlling the start-up decision of device h at time i: The formula of the device state formula is: P h (u)=Y h (i)*P MAX_h +1 (3) In formula (3), P h (i) represents the operating power of device h at time i, P MAX_h represents the maximum operating power of device i; Step 504: When there are multiple optimization objectives, multiple objective functions are optimized by particle swarm optimization (PSO) update function to find the Pareto optimal solution set; the formula of the PSO update function is: In formula (4), θ1, θ2, θ3 and g represent the inertia weight, cognitive acceleration constant, social acceleration constant and random number, respectively, which control the degree of dependence of particles on past states, individual experience and social experience; v e (i) represents the velocity of particle e at time step i, which is used to indicate the speed at which the particle moves toward the optimal solution; x e (i) represents the position of particle e at time step i, which is used to represent the position of the particle in the search space; pbest e represents the best individual position of particle e; gbest represents the best position of the group.

6. A photovoltaic microgrid energy management and optimization scheduling system according to claim 1, characterized in that: The energy resources controlled by the distributed energy resource control module include photovoltaic systems, energy storage systems, wind power converters and cogeneration systems.

7. A photovoltaic microgrid energy management and optimization scheduling system according to claim 1, characterized in that: The system also includes a security and privacy protection module for ensuring the security of user data and system operations; the security and privacy protection module includes an access control submodule, a transmission security submodule, a logging submodule, a data anonymization submodule and a compliance check submodule; the access control submodule prevents unauthorized access and operation through attribute-based access control; the transmission security submodule ensures the security of data during transmission through a symmetric encryption algorithm and a secure socket layer; the logging submodule records all user operations and system activities through a log file system and timestamp; the data anonymization submodule adds random noise to the data set through a differential privacy method to ensure the accuracy of statistical results and the non-traceability of personal data; The compliance check submodule regularly checks whether the system complies with security policies and laws and regulations through automatic compliance scanning tools and expert reviews, and generates a compliance report.

Images