Distributed optical storage micro-grid control system based on large model and energy management method

Abstract

The invention discloses a distributed optical storage micro-grid control system based on a large model and an energy management method, and the system collects various data through a data collection module, captures a time sequence long-term dependence relation based on a self-attention mechanism through a large model prediction system, and predicts the photovoltaic power generation amount, the load demand and the energy storage charging and discharging demand. The network-forming inverter integration module dynamically adjusts the output power, the energy storage strategy and the interaction power of the power generation system according to a prediction result, the distributed control strategy module adopts a distributed consensus algorithm to realize information sharing and collaborative decision making, and the energy management module makes a multi-time scale plan and introduces an economic optimization model. The energy management method comprises the steps of data collection, real-time monitoring, prediction modeling, plan making, distributed control, economic optimization, system monitoring, fault processing and the like. The method can improve the new energy utilization rate, the electric energy quality and the system stability, adapts to the change of environmental factors, and maximizes the economic and environmental benefits of the micro-grid.

Description

Technical Field

[0001] This paper relates to the technical field of distributed energy management systems, specifically a distributed photovoltaic and energy storage microgrid control system and energy management method based on a large language model based on the Transformer architecture, which uses a self-attention mechanism to capture long-term dependencies in time series and combines it with grid-connected inverters for coordinated control. This system significantly improves renewable energy utilization, power quality, and system stability through multi-scale prediction and optimization. Background Art

[0002] In recent years, my country has increasingly prioritized the development of distributed energy and microgrids, particularly photovoltaic systems and energy storage technologies. These technologies have become key measures for ensuring power supply reliability, promoting the use of renewable energy, and reducing pollution. With the widespread adoption of the "photovoltaic + energy storage + charging" combination, integrated photovoltaic, storage, and charging has emerged as a highly efficient energy utilization model suitable for a variety of scenarios, including electric vehicle charging stations, industrial parks, commercial buildings, residential communities, remote areas, microgrids, grid ancillary services, and energy substitution.

[0003] Experts in my country's new energy sector have conducted in-depth research on photovoltaic energy storage control systems, emphasizing their application value and the practical effectiveness of their business models. This has strategic significance for project planning and economic development. Furthermore, to optimize the amount of renewable energy storage, it is necessary to implement smoothing measures based on power fluctuations and rationally configure energy storage control systems based on typical scenarios and probability distributions of renewable energy output. The configuration and planning of energy storage systems are closely linked to short-term operation, necessitating comprehensive modeling of energy storage and renewable energy to scientifically configure energy storage and address related issues.

[0004] Furthermore, energy storage management for photovoltaic power generation systems focuses on controlling the charge and discharge of energy storage devices and the grid-connected inverter power. Research has shown that energy storage systems can effectively address grid absorption issues and ensure stable operation of the power system through power balancing. The selection of energy storage devices should consider different needs and technologies, including high-energy-density energy storage (such as superconducting energy storage, flywheel energy storage, and supercapacitors) and high-power-density energy storage (such as lead-acid batteries and lithium batteries).

[0005] In terms of microgrid energy management systems, my country has developed systems such as the Acrel-2000MG, which aim to optimize economic operation while maintaining a safe and stable foundation, promote the use of renewable energy, enhance grid stability, balance load fluctuations, manage demand, and reduce power supply costs. These systems utilize a layered, distributed architecture and support multiple communication protocols and physical media. They are suitable for renewable energy monitoring and energy management in cities, highways, industrial parks, commercial areas, residential complexes, intelligent buildings, islands, and areas without electricity.

[0006] Although significant progress has been made in distributed photovoltaic and energy storage microgrid control systems and energy management methods, current technologies still have some drawbacks:

[0007] Low utilization of new energy: A large number of distributed energy systems are being abandoned after installation, or are directly connected to the grid, resulting in increased payback periods. This is mainly due to improper energy storage system configuration and suboptimal control system strategies.

[0008] Unstable power quality: Loads such as new energy vehicle charging can cause power surges in local power grids, leading to unstable power quality at charging stations and impacting equipment operation. This requires more refined control and management of energy storage equipment and charging facilities.

[0009] Incomplete data analysis: A lack of comprehensive analysis of key data such as renewable energy utilization, the operating status of photovoltaic and storage equipment, and electricity consumption and costs leads to a lack of basis for coordinated output. This limits the optimized operation and energy efficiency of microgrid systems.

[0010] System stability needs improvement: Grid-connected photovoltaic energy storage systems face technical challenges and a lack of experimental data, making their impact on the grid unclear. Furthermore, system stability, energy density, and regulation capabilities also need to be improved.

[0011] Environmental factors: Distributed solar-energy storage microgrids are susceptible to environmental factors such as weather and climate, which can reduce power generation efficiency and reliability. This requires more intelligent design and optimization of energy storage equipment and control systems. Summary of the Invention

[0012] The purpose of the present invention is to provide a distributed photovoltaic and energy storage microgrid control system and energy management method based on the Transformer large model, which captures the long-term dependencies of time series data through the self-attention mechanism, combines the dynamic adjustment capabilities of the grid-connected inverter, and realizes the intelligent scheduling and optimal configuration of photovoltaic and energy storage resources.

[0013] To achieve the above object, the present invention adopts the following technical solutions:

[0014] A distributed photovoltaic and energy storage microgrid control system based on a large model is characterized by including:

[0015] a) Data acquisition module, which collects historical meteorological data, historical load data, grid status information and real-time operation data of distributed energy;

[0016] b) A large-scale model prediction system, which uses a large language model based on the Transformer architecture to capture long-term dependencies in time series data through a self-attention mechanism. Based on the data collected by the data acquisition module, it predicts photovoltaic power generation, load demand, and charging and discharging requirements of the energy storage system over a period of time;

[0017] c) a grid-connected inverter integrated module, including an intelligent control algorithm, which dynamically adjusts the output power of the photovoltaic power generation system, the charging and discharging strategy of the energy storage system, and the interaction power between the microgrid and other power grids based on the prediction results of the large model prediction system;

[0018] d) Distributed control strategy module, which adopts a distributed control architecture and uses a high-speed communication network through a distributed consensus algorithm to achieve information sharing and collaborative decision-making among PCS nodes;

[0019] e) An energy management module, which formulates a multi-time-scale energy management plan based on the prediction results of the large-scale model prediction system, including a day-ahead scheduling plan and a real-time adjustment plan.

[0020] Furthermore, the prediction model of the large model prediction system is trained and optimized through the following steps:

[0021] Pre-training phase:

[0022] Use large-scale historical data for unsupervised pre-training, involving many years of historical meteorological data and load data, to learn universal feature representations of data;

[0023] Use masked language model or autoregressive model as pre-training target;

[0024] Fine-tuning phase:

[0025] Based on the pre-trained model, fine-tuning is performed using task-specific data, involving photovoltaic power generation and load demand;

[0026] The loss function uses a combination of mean square error and cross entropy loss to optimize the prediction performance of the model;

[0027] Optimization algorithm:

[0028] Use the AdamW optimizer combined with the learning rate scheduler to dynamically adjust the learning rate to accelerate convergence;

[0029] Distributed training and mixed precision training are used to improve training efficiency.

[0030] Furthermore, the large model prediction system also includes a model verification step, specifically including:

[0031] k-fold cross validation:

[0032] Divide the dataset into k subsets, k = 5 or 10, and use one of the subsets as the validation set in turn, and the remaining subsets as the training set for k training and validation;

[0033] Calculate the performance index of each validation and take the average of k validation results as the final performance index of the model;

[0034] Generalization ability assessment:

[0035] Evaluate the generalization ability of the model on an independent test set to ensure the stability and reliability of the model in different scenarios;

[0036] Use SHAP to analyze the feature importance of the model and optimize input feature selection.

[0037] Furthermore, the grid-connected inverter integrated module implements its control strategy based on the following formula:

[0038] P out (t) = P PV (t)+P bat (t)-P load (t)

[0039] Among them, P out (t) is the output power at time t, P PV (t) is the predicted output power of the photovoltaic power generation system at time t, P bat (t) is the charge and discharge power of the energy storage system at time t, a positive value indicates discharge, and a negative value indicates charge, P load (t) is the predicted power demand of the load at time t

[0040] Furthermore, the distributed control strategy module adopts a distributed consensus algorithm, i.e., a consistency algorithm, to achieve information synchronization and collaborative control among the PCS nodes, specifically:

[0041] Consensus algorithm:

[0042] Realize information synchronization and state estimation among PCS nodes;

[0043] Based on the consistency algorithm, an event trigger mechanism is introduced to reduce communication overhead and computing burden;

[0044] Fault tolerance mechanism:

[0045] Set up communication protocols to ensure the timeliness and accuracy of information;

[0046] When some nodes fail, the system can still maintain normal operation.

[0047] Furthermore, the energy management module also includes an economic optimization model, specifically including:

[0048] a. Considering electricity price fluctuations, establish an electricity price prediction model based on time series;

[0049] b. Develop charging and discharging strategies for the energy storage system based on electricity price forecasts to maximize economic benefits;

[0050] c. Introduce environmental benefit factors to evaluate the environmental impact of different energy use options;

[0051] d. Comprehensively consider electricity prices, energy costs, and environmental benefits, and use multi-objective optimization algorithms to develop the optimal energy management plan;

[0052] The economic optimization model is as follows:

[0053]

[0054] Among them, P sell (t) and P buy (t) are the power sold and purchased by the microgrid to the grid at time t, C sell (t) and C buy (t) are the electricity sales and purchase prices at time t, C operation (t) is the operating cost at time t, B environment (t) is the environmental benefit at time t.

[0055] An energy management method for a distributed photovoltaic and energy storage microgrid control system based on a large model includes the following steps:

[0056] S1: Data Collection and Analysis

[0057] Collect historical meteorological data, historical load data, grid status information and real-time operation data of distributed energy;

[0058] Perform data cleaning, denoising, and normalization preprocessing operations to ensure data accuracy and completeness;

[0059] S2: Real-time data monitoring

[0060] Real-time monitoring of key data on photovoltaic power generation, load demand, and energy storage system charging and discharging status;

[0061] Use sensors and smart devices to achieve real-time data collection and transmission;

[0062] S3: Prediction model building

[0063] A large language model based on the Transformer architecture is used to capture long-term dependencies in time series data through a self-attention mechanism. The model is trained and validated using both historical and real-time data to improve prediction accuracy.

[0064] S4: Energy Management Plan Development

[0065] Formulate day-ahead dispatch plans and real-time adjustment plans based on forecast results;

[0066] Formulate optimal energy utilization strategies by comprehensively considering electricity price fluctuations, energy costs, and environmental benefits;

[0067] S5: Distributed Control Strategy Implementation

[0068] Using a distributed control method, each distributed energy source is controlled by a local controller;

[0069] Local controllers exchange information and collaborate through the communication network to achieve optimized scheduling and coordinated operation of distributed energy;

[0070] S6: Application of Economic Optimization Model

[0071] Construct an economic optimization model that comprehensively considers electricity price fluctuations, energy costs, and environmental benefits;

[0072] Use multi-objective optimization algorithms to solve the optimal energy management plan and maximize the economic operation and environmental benefits of the microgrid;

[0073] S7: System Monitoring and Optimization

[0074] Real-time monitoring of the microgrid's operating status, including photovoltaic power generation, load demand, and energy storage system status;

[0075] Regularly evaluate the system's operating performance and economic benefits, and optimize and improve the system based on the evaluation results;

[0076] S8: Fault Detection and Recovery

[0077] Utilize advanced fault detection technology to monitor the system's operating status and parameter changes in real time;

[0078] Once a fault or abnormality is detected, an alarm is immediately issued and the protection mechanism is activated;

[0079] According to the fault type and severity, appropriate recovery measures are taken to ensure that the system can quickly resume normal operation after the fault occurs. Further, the fault detection and recovery steps include:

[0080] Fault Detection:

[0081] Utilize fault detection algorithms and smart devices to monitor faults and abnormal conditions in the microgrid in real time;

[0082] Detecting a fault or abnormality triggers the alarm mechanism;

[0083] Fault recovery:

[0084] Develop appropriate fault recovery strategies based on fault type and severity;

[0085] Utilize backup power sources and redundant equipment to ensure the reliability of the microgrid's power supply during faults.

[0086] The distributed photovoltaic and energy storage microgrid control system and energy management method based on the large model can improve and optimize the above-mentioned drawbacks and achieve the following beneficial effects:

[0087] Improve the utilization rate of new energy: By optimizing the energy storage system configuration and control system strategy, reduce the phenomenon of wind and solar power curtailment, and improve the grid connection rate and utilization rate of new energy.

[0088] Stable power quality: Through refined control and management of energy storage equipment and charging facilities, power shocks and power quality fluctuations are reduced to ensure stable and reliable equipment operation.

[0089] Comprehensive data analysis: Comprehensive analysis and visualization of key data such as new energy utilization rate, photovoltaic storage equipment operating status, electricity consumption and cost, providing strong support for system optimization.

[0090] Enhance system stability: Through in-depth research on grid-connected system models and optimized control system design, the stability and reliability of photovoltaic energy storage systems connected to the grid can be improved.

[0091] Adapt to environmental factors: Improve the system's adaptability and robustness to environmental factors such as weather and climate through intelligent energy storage equipment and control system design.

[0092] In summary, the distributed photovoltaic storage microgrid control system and energy management method based on large models can significantly improve the energy efficiency and stability of the system, and provide strong support for the development of distributed energy and microgrids. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a structural diagram of the distributed photovoltaic storage microgrid control system based on the large model of the present invention;

[0094] Figure 2 This is a flow chart of the energy management method of the distributed photovoltaic storage microgrid control system based on the large model of the present invention. DETAILED DESCRIPTION

[0095] Technical solution content

[0096] 1. System Overview

[0097] The present invention proposes a large-model-based distributed photovoltaic and energy storage microgrid control system. By integrating large-scale language model prediction based on Transformer architecture, grid-connected inverter intelligent control, distributed control strategy and energy management optimization, it realizes efficient, economical and reliable operation of the distributed photovoltaic and energy storage microgrid.

[0098] 2. Data acquisition module

[0099] The data acquisition module is responsible for collecting historical meteorological data, historical load data, grid status information, and real-time operating data of distributed energy resources. This data will serve as input for subsequent predictions in the large language model based on the Transformer architecture.

[0100] The specific implementation process is:

[0101] 2.1 Determine the data collection objectives:

[0102] Historical meteorological data: including light intensity, temperature, humidity, and wind speed, which are crucial for predicting photovoltaic power generation;

[0103] Historical load data: including load demand changes over a period of time, used for load demand forecasting;

[0104] Grid status information: including voltage, frequency, and phase, reflecting the stability and operating status of the grid;

[0105] Real-time operating data of distributed energy: including the output power of the photovoltaic power generation system, the charge and discharge status of the energy storage system, and the SOC (state of charge) of the battery pack.

[0106] 2.2 Select the data collection method:

[0107] For historical data, it can be obtained through database query, data warehouse, etc.

[0108] For real-time data, sensors, intelligent instruments and other equipment are needed to collect it and transmit it to the data acquisition module through communication protocols (such as RS485, RS232, Modbus, etc.).

[0109] 2.3 Configure data acquisition equipment:

[0110] Select and configure appropriate data collection equipment, such as weather stations, smart meters, sensors, etc., based on data collection objectives and methods;

[0111] Ensure the accuracy, reliability and stability of data acquisition equipment to meet the system's requirements for data quality.

[0112] 2.4 Establishing a data transmission channel:

[0113] Configure the communication protocol and transmission channel between the data acquisition module and the database and server to ensure that data can be transmitted to the designated location in real time and accurately;

[0114] You can use wired (such as Ethernet, optical fiber, etc.) or wireless (such as GPRS, WIFI, LoRa, etc.) communication methods, and choose the most appropriate transmission solution according to actual needs.

[0115] 2.5 Data Preprocessing:

[0116] Perform pre-processing operations such as cleaning, denoising, and normalization on the collected data to improve data quality and the accuracy of the prediction model;

[0117] For missing values, outliers, etc., appropriate methods need to be used to fill or correct them.

[0118] 2.6 Data Storage and Management:

[0119] Store the preprocessed data in the database and establish a historical database and a real-time database to facilitate subsequent data analysis and prediction model training;

[0120] Use technologies such as data warehouses and data lakes to efficiently store and manage data and improve data availability and scalability.

[0121] The specific data types are as follows:

[0122] Historical weather data:

[0123] Light intensity (unit: W / m 2 ): reflects the intensity of solar radiation and has an important impact on the prediction of photovoltaic power generation;

[0124] Temperature (unit: °C): affects the efficiency of photovoltaic cells and the performance of energy storage systems;

[0125] Humidity (unit: %RH): reflects the humidity of the air and has a certain impact on the working efficiency of photovoltaic cells and the heat dissipation performance of energy storage systems;

[0126] Wind speed (unit: m / s): reflects the intensity of wind energy. Although it has little direct impact on photovoltaic power generation, it can be used as a reference for system status monitoring.

[0127] Historical load data:

[0128] Load demand (unit: kW or MW): reflects the change in power demand over a period of time and is used for load demand forecasting;

[0129] Peak load (unit: kW or MW): reflects the maximum value of load demand and is of guiding significance for grid planning and scheduling;

[0130] Load fluctuation rate (dimensionless): reflects the variation in load demand and is used to evaluate the stability of the power grid and the effectiveness of load management.

[0131] Grid status information:

[0132] Voltage (unit: V): reflects the voltage level of the power grid and is an important indicator of power grid stability;

[0133] Frequency (unit: Hz): reflects the frequency stability of the power grid and is crucial to the normal operation of power equipment;

[0134] Phase (unit: degree or radian): reflects the phase relationship between each node in the power grid and has an important impact on the synchronous operation of the power system.

[0135] Distributed energy real-time operation data:

[0136] Photovoltaic power generation system output power (unit: kW or MW): reflects the real-time power generation capacity of the photovoltaic power generation system;

[0137] Energy storage system charge and discharge status (unit: Ah or kWh): reflects the real-time charge and discharge status of the energy storage system;

[0138] Battery pack SOC (dimensionless): reflects the remaining power and charging status of the battery pack, which is of great significance to the management and maintenance of the energy storage system.

[0139] In summary, the data acquisition module provides rich and accurate input information for the subsequent large language model based on the Transformer architecture by collecting and processing various historical and real-time data. This data not only helps to improve the accuracy of the prediction model, but also provides strong support for the system's optimized scheduling and energy management.

[0140] 3. Large model prediction system

[0141] 3.1 Model Structure

[0142] The large model prediction system uses a large language model (LLM) based on the Transformer architecture, which captures long-term dependencies in time series data through the self-attention mechanism. This model has the following features:

[0143] Large-scale parameters: Models contain billions to hundreds of billions of parameters and are capable of learning complex nonlinear relationships;

[0144] Pre-training and fine-tuning: The model learns general feature representations through large-scale unsupervised pre-training, and then optimizes prediction performance through task-specific fine-tuning;

[0145] Contextual understanding: The model can understand the context of the input data and dynamically adjust the prediction strategy to adapt to different scenarios.

[0146] 3.2 Input Features

[0147] Input features include historical meteorological data, historical load data, grid status information, and real-time operating data of distributed energy resources. These features are processed through the following steps and then input into the model:

[0148] Data preprocessing: clean, denoise, and normalize the raw data to ensure data quality;

[0149] Feature embedding: converts structured data (such as light intensity, temperature, load demand, etc.) into high-dimensional embedding vectors through the embedding layer, simulating word embedding in natural language.

[0150] Sequential input: The embedding vectors are arranged in time series to form an input sequence similar to natural language.

[0151] 3.3 Output Target

[0152] The output targets are the predicted values ​​of photovoltaic power generation, load demand, and charging and discharging demand of the energy storage system in the future. The model achieves output through the following methods:

[0153] Sequence Generation: The model maps the input sequence to the output sequence, generating predicted values ​​for future time points.

[0154] Multi-task learning: The model simultaneously predicts photovoltaic power generation, load demand, and energy storage demand, improving prediction accuracy through shared feature representation.

[0155] 3.4 Model Training

[0156] Pre-training phase:

[0157] Use large-scale historical data (such as years of historical meteorological data, load data, etc.) for unsupervised pre-training to learn universal feature representations of the data;

[0158] The Masked Language Model (MLM) or Autoregressive Model is used as the pre-training target.

[0159] Fine-tuning phase:

[0160] Based on the pre-trained model, fine-tune it using task-specific data (such as photovoltaic power generation, load demand, etc.);

[0161] The loss function uses a combination of mean square error (MSE) and cross entropy loss to optimize the prediction performance of the model.

[0162] Optimization algorithm:

[0163] Use the AdamW optimizer combined with a learning rate scheduler (such as Cosine Annealing) to dynamically adjust the learning rate to accelerate convergence;

[0164] Distributed training and mixed precision training are used to improve training efficiency.

[0165] 3.5 Model Validation

[0166] k-fold cross validation:

[0167] Divide the dataset into k subsets (usually k = 5 or 10), use one of the subsets as the validation set in turn, and the remaining subsets as the training set for k training and validation;

[0168] Calculate the performance indicators of each validation (such as MSE, MAE, R 2 The average of k validation results is taken as the final performance indicator of the model.

[0169] Generalization ability assessment:

[0170] Evaluate the generalization ability of the model on an independent test set to ensure the stability and reliability of the model in different scenarios;

[0171] Use SHAP (SHapley Additive exPlanations) to analyze the feature importance of the model and optimize the input feature selection.

[0172] 3.6 Model Deployment and Application

[0173] Real-time predictions:

[0174] The trained model is deployed to edge computing devices or cloud servers, and the predicted value for a period of time in the future is generated by collecting input feature data (such as light intensity, temperature, load demand, etc.) in real time.

[0175] Dynamic Adjustment:

[0176] Based on real-time data and prediction results, the output power of the photovoltaic power generation system, the charging and discharging strategy of the energy storage system, and the interaction power between the microgrid and other power grids are dynamically adjusted.

[0177] Feedback optimization:

[0178] Compare the predicted results with the actual observed values, and continuously optimize the model parameters through online learning to improve the prediction accuracy.

[0179] 4. Grid-connected inverter (PCS) integrated module

[0180] Intelligent control algorithm

[0181] The PCS integrated module uses an intelligent control algorithm to dynamically adjust the output power of the photovoltaic power generation system, the charging and discharging strategy of the energy storage system, and the interaction power between the microgrid and other power grids based on the prediction results of the large model prediction system. The control algorithm is as follows:

[0182] P out (t) = P PV (t)+P bat (t)-P load (t)

[0183] Among them, P out (t) is the output power at time t, P PV (t) is the predicted output power of the photovoltaic power generation system at time t, P bat (t) is the charge and discharge power of the energy storage system at time t, a positive value indicates discharge, and a negative value indicates charge, P load (t) is the predicted required power of the load at time t.

[0184] Fault detection and recovery mechanisms

[0185] The PCS integrated module also has a fault detection and recovery mechanism, which monitors the working status of the PCS in real time, including parameters such as current, voltage, and power. When the monitored parameters exceed the preset threshold, a fault alarm is triggered and the system automatically switches to the backup control strategy or starts the protection mechanism to ensure the stable operation of the microgrid.

[0186] The specific implementation process is:

[0187] 4.1 Intelligent control algorithm implementation process

[0188] 4.1.1 Data Reception and Preprocessing

[0189] The PCS integrated module receives the prediction results of the large model prediction system, including the predicted value of photovoltaic power generation, load demand and charging and discharging demand of the energy storage system in the future.

[0190] Preprocess the received prediction data to ensure the accuracy and completeness of the data.

[0191] 4.1.2 Dynamic Adjustment Strategy Formulation

[0192] Formulate an output power adjustment strategy for the photovoltaic power generation system based on the forecast results. For example, when the photovoltaic power generation is predicted to be high, the output power of the photovoltaic power generation system can be appropriately increased; when the load demand is predicted to be low, the output power of the photovoltaic power generation system can be reduced.

[0193] Formulate charging and discharging strategies for the energy storage system, and reasonably arrange the charging and discharging time and power of the energy storage system based on the predicted charging and discharging demand of the energy storage system to ensure the stable operation and efficient utilization of the energy storage system;

[0194] Formulate an interactive power adjustment strategy between the microgrid and other power grids, and adjust the interactive power between the microgrid and other power grids based on the prediction results and actual needs to achieve optimal scheduling and efficient utilization of energy.

[0195] 4.1.3 Control strategy execution and monitoring

[0196] Transmit the formulated adjustment strategy to the photovoltaic power generation system, energy storage system and microgrid control system to perform corresponding control operations;

[0197] Monitor the operating status and parameter changes of each system in real time to ensure the effective implementation of the control strategy;

[0198] Fine-tune the control strategy according to actual conditions to improve the stability and efficiency of the system.

[0199] 4.2 Fault Detection and Recovery Mechanism Implementation Process

[0200] 4.2.1 Real-time monitoring and parameter collection

[0201] The PCS integrated module monitors the working status of the PCS in real time, including key parameters such as current, voltage, and power.

[0202] These parameters are collected in real time through sensors and data acquisition devices and transmitted to the monitoring system for processing and analysis.

[0203] 4.2.2 Fault detection and alarm

[0204] The monitoring system monitors and analyzes the collected parameters in real time to determine whether each parameter exceeds the preset threshold;

[0205] When a parameter is detected to exceed the threshold, the fault alarm function is triggered and an alarm message is sent to the operator;

[0206] The alarm information includes detailed information such as fault type, fault location, fault time, etc., so that operators can quickly locate and handle the fault.

[0207] 4.2.3 Backup control strategy switching and protection mechanism activation

[0208] After a fault alarm, the PCS integrated module automatically switches to the backup control strategy to ensure stable operation of the system;

[0209] The backup control strategy is developed based on the normal control strategy to deal with possible fault conditions;

[0210] At the same time, the corresponding protection mechanisms are activated, such as cutting off the faulty power supply, isolating the faulty equipment, etc., to prevent the fault from spreading and causing greater losses.

[0211] 4.2.4 Fault recovery and system restart

[0212] After the fault is resolved, the operator performs fault recovery operations on the system;

[0213] Take appropriate recovery measures based on the fault type and severity, such as replacing faulty parts and repairing damaged equipment;

[0214] After the fault recovery is complete, restart and test the system to ensure that the system can operate normally and meet the design requirements.

[0215] Through the above implementation process, the PCS integrated module can realize intelligent control of photovoltaic power generation systems, energy storage systems and microgrids, and at the same time has fault detection and recovery mechanisms to ensure stable operation and efficient utilization of the system.

[0216] 5. Distributed control strategy module

[0217] The distributed control strategy module uses distributed consensus algorithms, such as the consistency algorithm, to achieve information sharing and collaborative control among PCS nodes. The consistency algorithm iteratively updates the state estimate of each node until the state estimates of all nodes are consistent, thereby achieving information synchronization and collaborative control. In addition, an event triggering mechanism is introduced to reduce communication overhead and computational burden.

[0218] The specific implementation process is:

[0219] 5.1 Distributed Consensus Algorithm (Consistency Algorithm) Implementation Process

[0220] 5.1.1 System Initialization

[0221] Determine the number of PCS nodes involved in distributed control and their initial state estimates;

[0222] Configure the communication network to ensure reliable information transmission between nodes.

[0223] 5.1.2 Algorithm startup and iterative update

[0224] Each PCS node calculates the initial state estimate based on the information it obtains (such as local load demand, photovoltaic power generation, etc.);

[0225] Nodes exchange information through the communication network and receive state estimates from other nodes;

[0226] According to the principle of the consensus algorithm, each node updates its own state estimate based on the state estimates received from other nodes through certain iterative rules;

[0227] The iteration process continues until the state estimates of all nodes reach a consensus or the preset number of iterations / time limit is reached.

[0228] 5.1.3 Consistency Judgment of State Estimation

[0229] After each iteration, determine whether the state estimates of all nodes meet the consistency conditions;

[0230] The consistency condition may be that the difference between the state estimation values ​​is less than a preset threshold, or the variance of the state estimation values ​​is less than a preset threshold, etc.

[0231] 5.1.4 Collaborative Control Strategy Development and Implementation

[0232] When the state estimates of all nodes reach a consensus, a collaborative control strategy is formulated based on these consistent state estimates;

[0233] The coordinated control strategy can include output power adjustment of each PCS node, charging and discharging strategy of the energy storage system, etc.

[0234] The collaborative control strategy is transmitted to each PCS node for execution to achieve collaborative control of the system.

[0235] 5.2 Implementation process of event trigger mechanism

[0236] 5.2.1 Event Definition and Threshold Setting

[0237] Define the types of events that trigger communication or calculation, such as changes in state estimates, changes in system load requirements, etc.

[0238] Set a corresponding threshold for each event type to determine whether the event occurs.

[0239] 5.2.2 Event Monitoring and Judgment

[0240] Real-time monitoring of changes in parameters such as the estimated state of each node and system load demand;

[0241] Compare the monitored parameter value with the preset threshold to determine whether an event has occurred.

[0242] 5.2.3 Event triggering and response

[0243] When an event is detected, the corresponding communication or computing operation is triggered;

[0244] For example, when the change in the state estimate exceeds a threshold, communication operations between nodes are triggered to update the state estimate; when the system load demand changes, the formulation and execution of the collaborative control strategy is triggered.

[0245] (1) Specific implementation methods and technical means for triggering communication operations based on changes in estimated state values

[0246] Implementation method:

[0247] In the distributed photovoltaic and energy storage microgrid control system, each node (photovoltaic power generation unit, energy storage unit, load) is equipped with a state estimation module to estimate and monitor the node status (electricity, power, voltage) in real time;

[0248] Set a threshold for the change in the state estimate value, which is set according to the actual situation and stability requirements of the system;

[0249] When the change in the estimated state of a node exceeds a set threshold, the node will trigger a communication operation and send state update information to adjacent nodes or the central controller;

[0250] After receiving the status update information, the adjacent nodes or the central controller update the status estimate and reconfigure the system according to the latest status estimate value.

[0251] Technical means:

[0252] Use advanced sensors and measurement technology to obtain node status information in real time;

[0253] Use distributed computing and data fusion algorithms to process and update state estimates in real time;

[0254] Use wireless or wired communication technology to achieve information transmission and communication operations between nodes.

[0255] (2) Specific implementation methods and technical means for triggering collaborative control strategies based on changes in system load demand

[0256] Implementation method:

[0257] In the distributed photovoltaic storage microgrid control system, a load demand monitoring module is set up to monitor the load demand changes of the system in real time;

[0258] When the load demand changes (e.g., a sudden increase or decrease in load), the load demand monitoring module will trigger the formulation and execution of the collaborative control strategy;

[0259] According to the changes in load demand, the system adopts corresponding coordinated control strategies, including adjusting the output of photovoltaic power generation units, regulating the charge and discharge status of energy storage units, and adjusting the power consumption plan of the load;

[0260] The execution results of the collaborative control strategy will be fed back to the load demand monitoring module to evaluate and adjust the effectiveness of the strategy.

[0261] Technical means:

[0262] Use load forecasting algorithms and data analysis techniques to predict and evaluate system load demand;

[0263] Use advanced control algorithms and optimization methods to develop and implement collaborative control strategies;

[0264] Utilize real-time monitoring and feedback mechanisms to evaluate and adjust the effectiveness of collaborative control strategies in real time.

[0265] Through the application of the above-mentioned implementation methods and technical means, state estimation and coordinated control in the distributed photovoltaic storage microgrid control system can be effectively realized, thereby improving the stability and economy of the system.

[0266] 5.2.4 Optimization of communication overhead and computational burden

[0267] By reasonably setting event trigger thresholds, unnecessary communication and computing operations can be reduced, thereby reducing communication overhead and computing burden;

[0268] Specifically,

[0269] Specific methods for reasonably setting event trigger thresholds to reduce communication overhead and computational burden:

[0270] (1) Clarify optimization goals

[0271] First, the optimization goal needs to be clearly defined, namely, reducing communication overhead and computational burden. This requires that when designing the system, the communication needs between nodes and the limitations of computing resources should be fully considered, and unnecessary communication and computing operations can be reduced by reasonably setting event trigger thresholds.

[0272] (2) Analyze system characteristics

[0273] Node characteristics: Understand the computing power, communication capabilities, and storage capabilities of each node so that appropriate event trigger thresholds can be set based on the node's characteristics;

[0274] Network characteristics: Analyze network characteristics such as bandwidth, latency, and reliability to determine reasonable communication frequency and data transmission volume;

[0275] System stability: Consider the stability and reliability requirements of the system to ensure that the normal operation of the system is not affected when setting event trigger thresholds.

[0276] (3) Set event trigger threshold

[0277] State estimation threshold:

[0278] For the state estimation value of a node, a change threshold can be set; when the change of the state estimation value exceeds the threshold, the communication operation between nodes is triggered to update the state estimation value;

[0279] The threshold setting should take into account the stability and accuracy requirements of the system to avoid frequent communication operations due to a too low threshold or inaccurate state estimation due to a too high threshold;

[0280] Load demand threshold:

[0281] For the system load demand, a change rate threshold can be set; when the load demand change rate exceeds the threshold, the formulation and execution of the collaborative control strategy is triggered;

[0282] The threshold setting should take into account the system's response speed and regulation capabilities to ensure timely adjustments when load demand changes, while avoiding frequent coordinated control operations due to too low a threshold;

[0283] Dynamically adjust the threshold:

[0284] Dynamically adjust the event trigger threshold based on the actual operation and stability requirements of the system. For example, in the early stages of system operation, a lower threshold can be set to quickly detect and handle problems. After the system is running stably, the threshold can be gradually increased to reduce unnecessary communication and computing operations.

[0285] (4) Evaluation and Optimization

[0286] Performance evaluation: Evaluate the impact of set event trigger thresholds on communication overhead and computational burden through simulation experiments or actual operation data;

[0287] Optimization and adjustment: Based on the evaluation results, the event trigger threshold is optimized and adjusted to achieve the best communication and computing efficiency;

[0288] Continuous monitoring: Continuously monitor changes in communication overhead and computing load during system operation to promptly identify and address potential problems.

[0289] (5) Notes

[0290] Avoid over-optimization: When setting event trigger thresholds, avoid over-optimization that may cause system instability or performance degradation.

[0291] Consider system scalability: When setting event trigger thresholds, consider system scalability and possible future changes so that thresholds can be easily adjusted when needed.

[0292] Comprehensive consideration of multiple factors: When setting event trigger thresholds, it is necessary to comprehensively consider multiple factors such as system stability, accuracy requirements, response speed, and adjustment capabilities to ensure that the set thresholds are both reasonable and effective.

[0293] Through the above method, the event trigger threshold can be reasonably set to reduce unnecessary communication and computing operations, thereby reducing communication overhead and computing burden, and improving the overall performance and efficiency of the distributed photovoltaic storage microgrid control system.

[0294] Optimize the communication and computing processes, such as using efficient data transmission protocols and optimization algorithms, to further improve the efficiency and performance of the system.

[0295] 5.3 System stability and security assurance

[0296] 5.3.1 Communication Fault Handling

[0297] When a communication network failure occurs, appropriate fault recovery measures are taken, such as retransmission mechanisms and backup communication paths, to ensure reliable information transmission between nodes.

[0298] 5.3.2 Data Security Protection

[0299] Encrypt communication data to prevent data from being stolen or tampered with during transmission;

[0300] Authentication and authorization management of nodes ensure that only legitimate nodes can participate in the distributed control process.

[0301] 5.3.3 Enhanced system robustness

[0302] Improve the robustness and reliability of the system by introducing redundant nodes, fault-tolerant mechanisms and other measures;

[0303] Regularly maintain and upgrade the system to ensure system stability and security;

[0304] Through the above implementation process, the distributed control strategy module can realize information sharing and collaborative control among PCS nodes, while reducing communication overhead and computing burden, and improving system efficiency and performance.

[0305] 6. Energy management module

[0306] The energy management module formulates multi-timescale energy management plans based on the prediction results of the large-scale model prediction system, including day-ahead scheduling plans and real-time adjustment plans;

[0307] The day-ahead dispatch plan formulates the charging and discharging strategies of the energy storage system and the interactive power plan between the microgrid and other power grids based on the forecast values ​​of photovoltaic power generation and load demand for the next day. The real-time adjustment plan dynamically adjusts the day-ahead dispatch plan based on real-time data and forecast errors to ensure the economic operation and power supply reliability of the microgrid.

[0308] The energy management module also introduces an economic optimization model, which comprehensively considers factors such as electricity price fluctuations, energy costs, and environmental benefits, and uses a multi-objective optimization algorithm to formulate the optimal energy management plan. The economic optimization model is as follows:

[0309]

[0310] Among them, P sell (t) and P buy (t) are the power sold and purchased by the microgrid to the grid at time t, C sell (t) and C buy (t) are the electricity sales and purchase prices at time t, C operation (t) is the operating cost at time t, B environment (t) is the environmental benefit at time t.

[0311] The specific implementation process is:

[0312] 6.1 Developing a Day-Ahead Scheduling Plan

[0313] 6.1.1 Data Collection and Analysis:

[0314] Collect the photovoltaic power generation forecast and load demand forecast for the next day;

[0315] Analyze historical data to understand the fluctuation patterns of photovoltaic power generation and load demand.

[0316] 6.1.2 Formulate energy storage system charging and discharging strategies:

[0317] Formulate charging and discharging strategies for the energy storage system based on the forecast results of photovoltaic power generation and load demand;

[0318] When the photovoltaic power generation is higher than the load demand, the excess electricity is stored in the energy storage system; when the load demand is higher than the photovoltaic power generation, the electricity in the energy storage system is released to meet the demand.

[0319] 6.1.3 Develop a power plan for interaction between microgrids and other power grids:

[0320] Develop interactive power plans between the microgrid and other power grids based on photovoltaic power generation, load demand, and the charging and discharging strategies of the energy storage system;

[0321] When the microgrid has excess electricity, it sells electricity to other power grids; when the microgrid has insufficient electricity, it purchases electricity from other power grids.

[0322] 6.2 Develop real-time adjustment plans

[0323] 6.2.1 Real-time data monitoring:

[0324] Real-time monitoring of key data such as photovoltaic power generation, load demand, energy storage system status, and interaction power between microgrids and other power grids.

[0325] 6.2.2 Forecast Error Analysis:

[0326] Compare real-time data with forecast data and analyze the causes and magnitude of forecast errors. Specific implementation method:

[0328] (1) Data collection and organization

[0329] Real-time data collection:

[0330] Real-time acquisition of key data such as photovoltaic power generation, load demand, and energy storage system status through sensors and measuring equipment;

[0331] Ensure the accuracy and completeness of data, and process and correct abnormal data.

[0332] Prediction data acquisition:

[0333] Use the prediction model to predict photovoltaic power generation, load demand, etc. in the future;

[0334] Record the generation time and predicted value of the forecast data.

[0335] (2) Prediction error calculation

[0336] Define the prediction error:

[0337] The prediction error is the difference between the predicted value and the actual observed value;

[0338] Indicators such as absolute error, relative error, mean square error (MSE), and normalized root mean square error (nRMSE) are used to measure the size of the prediction error.

[0339] Calculate the prediction error:

[0340] Calculate the forecast error at each time point based on real-time data and forecast data;

[0341] Summarize and calculate indicators such as the average prediction error and maximum prediction error during the entire prediction period.

[0342] (3) Analysis of causes of prediction errors

[0343] Model error:

[0344] Errors caused by imperfections in the prediction model itself or unreasonable parameter settings;

[0345] Model error can be reduced by optimizing model structure, adjusting model parameters, increasing training data, etc.

[0346] Data error:

[0347] Errors in the collection, transmission and processing of real-time data;

[0348] Data errors can be reduced by improving data collection accuracy, optimizing data transmission protocols, and strengthening data processing algorithms.

[0349] External environmental factors:

[0350] The impact of external environmental factors such as weather changes, light intensity, and temperature on photovoltaic power generation and load demand;

[0351] The prediction accuracy can be improved by introducing external environmental factors as input features of the prediction model or adopting methods such as multi-model fusion.

[0352] System uncertainty:

[0353] Uncertain factors in distributed photovoltaic and energy storage microgrid systems, such as equipment failures and load fluctuations;

[0354] The impact of system uncertainty on prediction accuracy can be reduced by strengthening system monitoring, improving fault warning and response capabilities, etc.

[0355] (4) Prediction error optimization strategy

[0356] Model optimization:

[0357] According to the prediction error analysis results, the prediction model is iteratively optimized to improve the prediction accuracy and generalization ability of the model.

[0358] Data preprocessing:

[0359] Strengthen data preprocessing, including data cleaning, outlier processing, data standardization, etc., to improve data quality and availability.

[0360] Introducing external environmental factors:

[0361] Introducing more external environmental factors as input features into the prediction model can improve the model's adaptability to changes in the external environment.

[0362] Strengthen system monitoring and early warning:

[0363] Strengthen the monitoring and early warning capabilities of distributed photovoltaic storage microgrid systems, promptly detect and handle abnormal situations in the system, and reduce the impact of system uncertainty on prediction accuracy.

[0364] (5) Implementation effect evaluation

[0365] Comparative evaluation:

[0366] The optimized prediction results are compared and evaluated with those before optimization to analyze the effectiveness of the optimization strategy.

[0367] Continuous monitoring:

[0368] Continuously monitor changes in prediction errors during system operation to promptly identify and address potential problems.

[0369] Feedback Adjustment:

[0370] Based on the implementation effect evaluation results and continuous monitoring results, feedback adjustments are made to the prediction model and optimization strategy to further improve prediction accuracy and system performance.

[0371] Through the above implementation, we can systematically analyze the causes and magnitude of prediction errors and adopt corresponding optimization strategies to improve prediction accuracy and system performance. This helps distributed photovoltaic and energy storage microgrid control systems better cope with external environmental changes and system uncertainties, achieving more efficient and stable operation.

[0372] 6.2.3 Dynamically adjust the day-ahead scheduling plan:

[0373] Dynamically adjust the day-ahead dispatch plan based on real-time data and forecast errors;

[0374] Specifically,

[0375] (1) Collect and analyze data

[0376] Real-time data collection:

[0377] Real-time acquisition of key data such as photovoltaic power generation, load demand, and energy storage system status through sensors and measuring equipment;

[0378] Ensure the accuracy and completeness of data, and process and correct abnormal data.

[0379] Forecast error analysis:

[0380] Compare real-time data with forecast data, calculate forecast errors, and analyze the causes and magnitude of the errors;

[0381] Forecast errors may come from a variety of factors, including inaccurate models, poor data quality, and changes in the external environment.

[0382] (2) Evaluate the effectiveness of the day-ahead scheduling plan

[0383] Comparing actual and planned:

[0384] Compare real-time data with the day-ahead scheduling plan to evaluate the actual execution of the plan;

[0385] Analyze the differences between plans and actuals, and identify existing problems and potential risks.

[0386] Performance evaluation indicators:

[0387] Appropriate performance evaluation indicators, such as power supply reliability, economic cost, energy utilization efficiency, etc., are used to quantitatively evaluate the effectiveness of the day-ahead dispatch plan.

[0388] (3) Develop dynamic adjustment strategies

[0389] Adjustment principles:

[0390] When formulating dynamic adjustment strategies based on real-time data and prediction errors, we should follow the principles of ensuring system safety and stability, optimizing energy utilization, and reducing economic costs.

[0391] Adjustment method:

[0392] Photovoltaic power generation adjustment: Adjust the output plan of the photovoltaic power generation system based on real-time light intensity and prediction error to ensure efficient use of photovoltaic power generation;

[0393] Load demand adjustment: Adjust the load distribution plan based on real-time load demand and forecast errors, give priority to meeting the needs of important loads, and ensure power supply reliability;

[0394] Energy storage system adjustment: Adjust the charging and discharging plan of the energy storage system based on the real-time status and prediction error of the energy storage system to optimize the utilization efficiency and life of the energy storage system.

[0395] (4) Implement dynamic adjustments

[0396] Adjustment execution:

[0397] Adjust the day-ahead scheduling plan accordingly based on the developed dynamic adjustment strategy;

[0398] During the adjustment process, the accuracy and timeliness of the adjustment instructions should be ensured to avoid unnecessary impact on the system.

[0399] Monitoring and Feedback:

[0400] During the adjustment process, continuously monitor the system's operating status to promptly identify and address potential problems;

[0401] Based on the actual response and effect of the system, feedback and adjustments are made to the adjustment strategy to further improve the accuracy and effect of the adjustment.

[0402] (5) Continuous optimization and improvement

[0403] Lessons learned:

[0404] Summarize and analyze the process and results of dynamic adjustment, and extract effective adjustment strategies and methods.

[0405] Model optimization:

[0406] Based on real-time data and adjustment experience, the prediction model is optimized and improved to improve the model's prediction accuracy and adaptability.

[0407] Strategy Improvement:

[0408] Based on the actual operation of the system and changes in demand, the dynamic adjustment strategy is continuously improved and adjusted to ensure continuous optimization and improvement of the system.

[0409] Through the above steps and methods, the day-ahead dispatch plan can be dynamically adjusted based on real-time data and forecast errors, ensuring the safe and stable operation and efficient energy utilization of the distributed photovoltaic and energy storage microgrid control system. This helps improve the system's flexibility and adaptability, reduces economic costs, and provides strong support for the widespread application of renewable energy.

[0410] Adjust the charging and discharging strategies of the energy storage system, the interactive power plan between the microgrid and other power grids, etc. to ensure the economic operation and power supply reliability of the microgrid.

[0411] 6.3 Introduction of Economic Optimization Model

[0412] 6.3.1 Establishing an economic optimization model:

[0413] Establish an economic optimization model by comprehensively considering factors such as electricity price fluctuations, energy costs and environmental benefits;

[0414] The objective function of the model is to maximize the economic and environmental benefits of the microgrid.

[0415] 6.3.2 Using multi-objective optimization algorithm:

[0416] Use multi-objective optimization algorithms such as NSGA-II to find the optimal solution of the economic optimization model;

[0417] The algorithm searches for the optimal solution by simulating the natural selection process, optimizing multiple conflicting objectives simultaneously.

[0418] 6.3.3 Develop an optimal energy management plan:

[0419] Formulate an optimal energy management plan based on the results of the multi-objective optimization algorithm;

[0420] The plan includes charging and discharging strategies for energy storage systems, interactive power plans between microgrids and other power grids, and optimization of energy costs and environmental benefits.

[0421] 6.4 Implementation and Monitoring

[0422] 6.4.1 Implement energy management plan:

[0423] Operate according to the established energy management plan, including charging and discharging of the energy storage system, and interaction between the microgrid and other power grids.

[0424] 6.4.2 Monitoring and Evaluation:

[0425] Real-time monitoring of the implementation of energy management plans, including the status of energy storage systems and the power supply of microgrids;

[0426] Regularly evaluate the effectiveness of energy management plans, including economic benefits, environmental benefits, and power supply reliability.

[0427] 6.4.3 Optimization and improvement:

[0428] Optimize and improve energy management plans based on monitoring and evaluation results;

[0429] Continuously optimize the charging and discharging strategies of energy storage systems, interactive power plans between microgrids and other power grids, etc., to improve energy utilization efficiency and economic benefits.

[0430] Through the above implementation process, the energy management module can achieve refined management and optimization of the microgrid, improve energy utilization efficiency and economic benefits, and reduce the impact on the environment.

[0431] 7. System implementation effect

[0432] 7.1 Efficient Operation

[0433] 7.1.1 Improved prediction accuracy:

[0434] The system uses large-scale model prediction technology, combining historical and real-time data, to provide highly accurate forecasts of photovoltaic power generation, load demand, and energy storage requirements. This significantly reduces energy waste and scheduling imbalances caused by inaccurate forecasts.

[0435] 7.1.2 Enhanced dynamic adjustment capabilities:

[0436] Based on the prediction results, the system can dynamically adjust the output power of distributed energy resources (such as photovoltaic and energy storage systems), as well as the interaction power between the microgrid and other power grids. This dynamic adjustment capability ensures local energy balance and efficient utilization, avoiding long-distance energy transmission and loss.

[0437] 7.1.3 Improved energy utilization:

[0438] Through precise prediction and dynamic adjustment, the system can maximize the use of renewable energy (such as photovoltaics) and reduce dependence on traditional energy. At the same time, the rational charging and discharging of the energy storage system further improves the overall energy utilization rate.

[0439] 7.2 Economic Operation

[0440] 7.2.1 Reduction of economic costs:

[0441] Through distributed control strategies and economic optimization models, the system can comprehensively consider factors such as electricity price fluctuations, energy costs, and environmental benefits to formulate an optimal energy management plan. This not only reduces the system's operating costs but also improves economic benefits.

[0442] 7.2.2 Maximizing Profits:

[0443] The system can intelligently adjust the microgrid's interactive power based on electricity price fluctuations and energy demand, achieving the optimal decision for selling or purchasing electricity from the grid. This helps maximize the microgrid's economic benefits.

[0444] 7.2.3 Significant environmental benefits:

[0445] By optimizing energy utilization and reducing dependence on traditional energy, the system significantly reduces carbon emissions and other pollutant emissions, making a positive contribution to environmental protection and sustainable development.

[0446] 7.3 Reliable Operation

[0447] 7.3.1 Fault detection and recovery mechanism:

[0448] The system is equipped with advanced fault detection and recovery mechanisms that monitor the operating status and parameter changes of each node in real time. Once a fault or abnormality is detected, the system can quickly switch to a backup control strategy or activate protection mechanisms to ensure the stable operation of the microgrid.

[0449] 7.3.2 Enhanced system robustness:

[0450] By introducing redundant nodes and fault-tolerant mechanisms, the system has improved its resistance to external interference and failures. This ensures that the system can maintain stable operation and power supply reliability even in severe weather or emergencies.

[0451] 7.3.3 Improved user satisfaction:

[0452] The system's efficient, economical, and reliable operation not only reduces energy costs and environmental impact, but also improves power supply quality and stability. This provides users with more reliable and high-quality power services, increasing user satisfaction and trust.

[0453] In summary, the distributed photovoltaic and energy storage microgrid control system based on the large-scale model of the present invention has achieved remarkable results in terms of efficient, economical, and reliable operation. These results not only improve the overall performance and economic benefits of the system, but also make important contributions to promoting the utilization and sustainable development of renewable energy.

[0454] 8. Energy management methods

[0455] The energy management method of a distributed photovoltaic storage microgrid control system based on a large model includes the following steps:

[0456] S1: Data Collection and Analysis

[0457] Collect historical meteorological data, historical load data, grid status information and real-time operation data of distributed energy;

[0458] Perform data cleaning, denoising, and normalization preprocessing operations to ensure data accuracy and completeness;

[0459] S2: Real-time data monitoring

[0460] Real-time monitoring of key data on photovoltaic power generation, load demand, and energy storage system charging and discharging status;

[0461] Use sensors and smart devices to achieve real-time data collection and transmission;

[0462] S3: Prediction model building

[0463] A large language model based on the Transformer architecture is used to capture long-term dependencies in time series data through the self-attention mechanism.

[0464] Combine historical data with real-time data to train and validate the model and improve the accuracy of predictions;

[0465] S4: Energy Management Plan Development

[0466] Formulate day-ahead dispatch plans and real-time adjustment plans based on forecast results;

[0467] Formulate optimal energy utilization strategies by comprehensively considering electricity price fluctuations, energy costs, and environmental benefits;

[0468] S5: Distributed Control Strategy Implementation

[0469] Using a distributed control method, each distributed energy source is controlled by a local controller;

[0470] Local controllers exchange information and collaborate through the communication network to achieve optimized scheduling and coordinated operation of distributed energy;

[0471] S6: Application of Economic Optimization Model

[0472] Construct an economic optimization model that comprehensively considers electricity price fluctuations, energy costs, and environmental benefits;

[0473] Use multi-objective optimization algorithms to solve the optimal energy management plan and maximize the economic operation and environmental benefits of the microgrid;

[0474] S7: System Monitoring and Optimization

[0475] Real-time monitoring of the microgrid's operating status, including photovoltaic power generation, load demand, and energy storage system status;

[0476] Regularly evaluate the system's operating performance and economic benefits, and optimize and improve the system based on the evaluation results;

[0477] S8: Fault Detection and Recovery

[0478] Utilize advanced fault detection technology to monitor the system's operating status and parameter changes in real time;

[0479] Once a fault or abnormality is detected, an alarm is immediately issued and the protection mechanism is activated;

[0480] Take appropriate recovery measures based on the fault type and severity to ensure that the system can quickly resume normal operation after a fault occurs.

[0481] The fault detection and recovery steps include:

[0482] Fault Detection:

[0483] Utilize fault detection algorithms and smart devices to monitor faults and abnormal conditions in the microgrid in real time;

[0484] Detecting a fault or abnormality triggers the alarm mechanism;

[0485] Fault recovery:

[0486] Develop appropriate fault recovery strategies based on fault type and severity;

[0487] Utilize backup power sources and redundant equipment to ensure the reliability of the microgrid's power supply during faults.

[0488] In summary, the energy management approach for a distributed photovoltaic and energy storage microgrid control system based on a large model includes multiple aspects, including data collection and forecasting, energy management plan formulation, distributed control and economic optimization, system monitoring and optimization, and fault detection and recovery. Together, these methods form a complete and efficient energy management system that enables the efficient, economical, and reliable operation of distributed photovoltaic and energy storage microgrids.

[0489] Specific application example 1:

[0490] 1. Application Background

[0491] A power engineering company specializes in various engineering construction activities and electrical installation services, with rapid growth in its new energy business. The company's integrated photovoltaic, storage, and charging energy system within its campus was no longer meeting its refined operational needs with its existing monitoring and management methods. To achieve scientific monitoring and optimized energy allocation, and improve operational and maintenance efficiency, the company decided to introduce a distributed photovoltaic, storage, and charging microgrid control system based on a large model.

[0492] 2. System Configuration

[0493] Distributed energy: 200kW distributed photovoltaic on the roof and 40kW photovoltaic on the carport;

[0494] Energy storage system: energy storage container 100kW / 200kWh;

[0495] Power equipment: 5 DC fast charging stations, as well as transformers, cables, and switchgear;

[0496] Control system: CET-7320 intelligent strategy control server, integrated energy IoT cloud platform.

[0497] 3. Energy Management Methods

[0498] Data Collection and Analysis: Historical data on photovoltaic power generation, load demand, and energy storage system status from the past year are collected. Using a large language model based on the Transformer architecture and a self-attention mechanism to capture long-term dependencies in time series data, this data is analyzed to build a predictive model. The collected photovoltaic power generation data covers various weather conditions and seasonal variations; the load demand data reflects seasonal variations in load demand, differences between weekdays and weekends, and the impact of special events; and the energy storage system status data covers its operating status under different load demand and photovoltaic power generation conditions. The collected data is then preprocessed with cleaning, denoising, and normalization to extract useful features, and then divided into training, validation, and test sets.

[0499] Real-time data monitoring: Real-time monitoring of key data such as photovoltaic power generation, load demand, and energy storage system charging and discharging status. Utilizing sensors and smart devices, real-time data collection and transmission is achieved.

[0500] Prediction Model Construction and Optimization: A large language model based on the Transformer architecture was used to build the prediction model. During the pre-training phase, unsupervised pre-training was performed using years of historical meteorological and load data to learn universal feature representations of the data. A masked language model or autoregressive model was used as the pre-training target. During the fine-tuning phase, the pre-trained model was fine-tuned using task-specific data such as photovoltaic power generation and load demand. The loss function used a combination of mean squared error and cross-entropy loss to optimize the model's predictive performance. The AdamW optimizer combined with a learning rate scheduler dynamically adjusted the learning rate. Distributed training and mixed-precision training were used to improve training efficiency. The model also optimized input feature selection through k-fold cross-validation, calculating the average performance metrics, and evaluating generalization on an independent test set. SHAP was used to analyze the model's feature importance.

[0501] Energy Management Plan Development: Based on the prediction results of the large-scale model prediction system, a multi-timescale energy management plan is developed, including day-ahead scheduling and real-time adjustment plans. Taking into account electricity price fluctuations, a time-series-based electricity price prediction model is established. Based on the forecast results, a charging and discharging strategy for the energy storage system is formulated. Environmental benefit factors are introduced, and a multi-objective optimization algorithm is used to comprehensively consider electricity price, energy cost, and environmental benefits to develop an optimal energy management plan.

[0502] Distributed Control Strategy Implementation: Utilizing a distributed control architecture, a distributed consensus algorithm leverages high-speed communication networks to enable information sharing and collaborative decision-making among PCS nodes. The grid-connected inverter integrated module dynamically adjusts the output power of the photovoltaic power generation system, the charging and discharging strategies of the energy storage system, and the interaction power between the microgrid and other power grids based on the prediction results of the large-scale model prediction system. Its control strategy follows a given formula, using an event-triggered mechanism to reduce communication overhead and computational burdens. Furthermore, a communication protocol is established to ensure real-time and accurate information. This ensures that the system can maintain normal operation even if some nodes fail.

[0503] Application of economic optimization model: Construct an economic optimization model, comprehensively consider electricity price fluctuations, energy costs and environmental benefits, and use a multi-objective optimization algorithm to solve the optimal energy management plan to maximize the economic operation and environmental benefits of the microgrid.

[0504] System monitoring and optimization: Use the integrated energy IoT cloud platform to monitor the operating status of the microgrid in real time, regularly evaluate the system's operating performance and economic benefits, and optimize and improve the system based on the evaluation results.

[0505] Fault detection and recovery: Utilize advanced fault detection technology to monitor the system's operating status and parameter changes in real time. Once a fault or abnormality is detected, an alarm is immediately issued and the protection mechanism is activated. Based on the type and severity of the fault, appropriate recovery measures are taken to ensure that the system can quickly resume normal operation after a fault occurs.

[0506] 4. Output results

[0507] After a week of trial operation, the distributed photovoltaic storage microgrid control system based on the large model achieved remarkable results. The accuracy of photovoltaic power generation prediction reached more than 95%, the charging and discharging strategy of the energy storage system was optimized, and the economic benefits of the microgrid increased by about 20% compared with traditional energy management methods. The energy utilization efficiency increased by about 15%, and the environmental benefits reduced carbon emissions and other pollutant emissions.

[0508] Specific application example 2:

[0509] I. Background Overview

[0510] Due to the closure and renovation of small thermal power plants in the Guanzhong region of Shaanxi Province, the Guanzhong power grid's load centers are lacking nearby conventional power sources, facing a power outage. To ensure stable and reliable power supply, the region decided to adopt a large-scale model-based distributed photovoltaic and energy storage microgrid control system and energy management approach, building a distributed microgrid system based on photovoltaic and energy storage equipment.

[0511] 2. System Configuration

[0512] Distributed power supply: Distributed photovoltaic power generation systems with a total installed capacity of 500kW will be installed in areas with good lighting conditions in the Guanzhong region.

[0513] Energy storage system: A 200kWh energy storage battery system is configured to store excess electricity generated by photovoltaic power generation and release it when the grid needs it.

[0514] Microgrid control system: Adopts advanced microgrid control system, integrates big data analysis and advanced control algorithms, and realizes real-time monitoring and intelligent scheduling of photovoltaic power generation, energy storage equipment and loads.

[0515] 3. Energy Management Methods

[0516] Data collection and analysis: Collect historical data such as photovoltaic power generation, load demand, and energy storage system status. Utilize a large language model based on the Transformer architecture and a self-attention mechanism to capture long-term dependencies in time series data. Analyze historical data and establish a prediction model to predict photovoltaic power generation and load demand for the next week.

[0517] Day-ahead dispatch plan formulation: Based on the results of the prediction model, a day-ahead dispatch plan is formulated. When photovoltaic power generation is sufficient, photovoltaic power generation is used first, and excess electricity is stored in the energy storage system. When photovoltaic power generation is insufficient, the energy storage system releases electricity to ensure that load demand is met.

[0518] Economic Optimization: Considering electricity price fluctuations and energy costs, a time-series-based electricity price forecasting model is established. Based on the forecasted electricity prices, a charging and discharging strategy for the energy storage system is formulated to maximize economic benefits. When electricity prices are low, electricity is purchased from the grid to charge the energy storage system. When electricity prices are high, photovoltaic power generation and energy storage systems are used to supply power, reducing energy costs. Environmental benefits are also factored in, integrating electricity prices, energy costs, and environmental benefits, using a multi-objective optimization algorithm to develop an optimal energy management plan.

[0519] Fault response: Integrate fault detection algorithms into the microgrid control system to monitor the operating status of the microgrid in real time. Once a fault or abnormality is detected, the alarm mechanism is immediately triggered and the corresponding fault recovery strategy is implemented. The backup power supply and redundant equipment are used to ensure the power supply reliability of the microgrid during the fault.

[0520] 4. Output results

[0521] After a week of trial operation, the distributed photovoltaic storage microgrid control system and energy management method based on the large model achieved remarkable results. The photovoltaic power generation reached 3500kWh within a week, meeting part of the load demand, and the energy storage system utilization rate reached 80%. Compared with purchasing electricity directly from the grid, it saved about 15% of energy costs. In extreme operating conditions such as faults, the microgrid system provided the necessary active and reactive power support to ensure voltage stability.

[0522] In summary, the application of distributed photovoltaic microgrid control system and energy management method based on large model in Guanzhong area of ​​Shaanxi Province has achieved remarkable results, providing strong guarantee for the power supply in the region.

[0523] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A distributed photovoltaic and storage microgrid control system based on a large model, characterized by: include: a) Data acquisition module, which collects historical meteorological data, historical load data, grid status information and real-time operation data of distributed energy; b) A large-scale model prediction system, which uses a large language model based on the Transformer architecture to capture long-term dependencies in time series data through a self-attention mechanism. Based on the data collected by the data acquisition module, it predicts photovoltaic power generation, load demand, and charging and discharging requirements of the energy storage system over a period of time; c) a grid-connected inverter integrated module, including an intelligent control algorithm, which dynamically adjusts the output power of the photovoltaic power generation system, the charging and discharging strategy of the energy storage system, and the interaction power between the microgrid and other power grids based on the prediction results of the large model prediction system; d) Distributed control strategy module, which adopts a distributed control architecture and uses a high-speed communication network through a distributed consensus algorithm to achieve information sharing and collaborative decision-making among PCS nodes; e) An energy management module, which formulates a multi-time-scale energy management plan based on the prediction results of the large-scale model prediction system, including a day-ahead scheduling plan and a real-time adjustment plan.

2. The distributed photovoltaic and energy storage microgrid control system based on a large model according to claim 1 is characterized in that: The prediction model of the large model prediction system is trained and optimized through the following steps: Pre-training phase: Use large-scale historical data for unsupervised pre-training, involving many years of historical meteorological data and load data, to learn universal feature representations of data; Use masked language model or autoregressive model as pre-training target; Fine-tuning phase: Based on the pre-trained model, fine-tuning is performed using task-specific data, involving photovoltaic power generation and load demand; The loss function uses a combination of mean square error and cross entropy loss to optimize the prediction performance of the model; Optimization algorithm: Use the AdamW optimizer combined with the learning rate scheduler to dynamically adjust the learning rate to accelerate convergence; Distributed training and mixed precision training are used to improve training efficiency.

3. The distributed photovoltaic and energy storage microgrid control system based on a large model according to claim 1 is characterized in that: The large model prediction system further includes a model verification step, specifically including: k-fold cross validation: Divide the dataset into k subsets, k = 5 or 10, and use one of the subsets as the validation set in turn, and the remaining subsets as the training set for k training and validation; Calculate the performance index of each validation and take the average of k validation results as the final performance index of the model; Generalization ability assessment: Evaluate the generalization ability of the model on an independent test set to ensure the stability and reliability of the model in different scenarios; Use SHAP to analyze the feature importance of the model and optimize input feature selection.

4. The distributed photovoltaic and energy storage microgrid control system based on a large model according to claim 1 is characterized in that: The grid-type inverter integrated module implements its control strategy based on the following formula: P out (t)=P PV (t)+P bat (t)-P load (t) Among them, P out (t) is the output power at time t, P PV (t) is the predicted output power of the photovoltaic power generation system at time t, P bat (t) is the charge and discharge power of the energy storage system at time t, a positive value indicates discharge, and a negative value indicates charge, P load (t) is the predicted required power of the load at time t.

5. The large-scale model-based distributed photovoltaic and energy storage microgrid control system according to claim 1 is characterized in that: The distributed control strategy module adopts a distributed consensus algorithm, i.e., a consistency algorithm, to achieve information synchronization and collaborative control among PCS nodes. Specifically: Consensus algorithm: Realize information synchronization and state estimation among PCS nodes; Based on the consistency algorithm, an event trigger mechanism is introduced to reduce communication overhead and computing burden; Fault tolerance mechanism: Set up communication protocols to ensure the timeliness and accuracy of information; When some nodes fail, the system can still maintain normal operation.

6. The large-scale model-based distributed photovoltaic and energy storage microgrid control system according to claim 1 is characterized in that: The energy management module also includes an economic optimization model, specifically including: a. Considering electricity price fluctuations, establish an electricity price prediction model based on time series; b. Develop charging and discharging strategies for the energy storage system based on electricity price forecasts to maximize economic benefits; c. Introduce environmental benefit factors to evaluate the environmental impact of different energy use options; d. Comprehensively consider electricity prices, energy costs, and environmental benefits, and use multi-objective optimization algorithms to develop the optimal energy management plan; The economic optimization model is as follows: Among them, P sell (t) and P buy (t) are the power sold and purchased by the microgrid to the grid at time t, C sell (t) and C buy (t) are the electricity sales and purchase prices at time t, C operation (t) is the operating cost at time t, B environment (t) is the environmental benefit at time t.

7. An energy management method for a distributed photovoltaic and energy storage microgrid control system based on a large model according to any one of claims 1 to 7, characterized in that: The following steps are involved: S1: Data Collection and Analysis Collect historical meteorological data, historical load data, grid status information and real-time operation data of distributed energy; Perform data cleaning, denoising, and normalization preprocessing operations to ensure data accuracy and completeness; S2: Real-time data monitoring Real-time monitoring of key data on photovoltaic power generation, load demand, and energy storage system charging and discharging status; Use sensors and smart devices to achieve real-time data collection and transmission; S3: Prediction model building A large language model based on the Transformer architecture is used to capture long-term dependencies in time series data through the self-attention mechanism. Combine historical data with real-time data to train and validate the model and improve the accuracy of predictions; S4: Energy Management Plan Development Formulate day-ahead dispatch plans and real-time adjustment plans based on forecast results; Formulate optimal energy utilization strategies by comprehensively considering electricity price fluctuations, energy costs, and environmental benefits; S5: Distributed Control Strategy Implementation Using a distributed control method, each distributed energy source is controlled by a local controller; Local controllers exchange information and collaborate through the communication network to achieve optimized scheduling and coordinated operation of distributed energy; S6: Application of Economic Optimization Model Construct an economic optimization model that comprehensively considers electricity price fluctuations, energy costs, and environmental benefits; Use multi-objective optimization algorithms to solve the optimal energy management plan and maximize the economic operation and environmental benefits of the microgrid; S7: System Monitoring and Optimization Real-time monitoring of the microgrid's operating status, including photovoltaic power generation, load demand, and energy storage system status; Regularly evaluate the system's operating performance and economic benefits, and optimize and improve the system based on the evaluation results; S8: Fault Detection and Recovery Utilize advanced fault detection technology to monitor the system's operating status and parameter changes in real time; Once a fault or abnormality is detected, an alarm is immediately issued and the protection mechanism is activated; Take appropriate recovery measures based on the fault type and severity to ensure that the system can quickly resume normal operation after a fault occurs.

8. The energy management method of a distributed photovoltaic and energy storage microgrid control system based on a large model according to claim 7 is characterized by: The fault detection and recovery steps include: Fault Detection: Utilize fault detection algorithms and smart devices to monitor faults and abnormal conditions in the microgrid in real time; Detecting a fault or abnormality triggers the alarm mechanism; Fault recovery: Develop appropriate fault recovery strategies based on fault type and severity; Utilize backup power sources and redundant equipment to ensure the reliability of the microgrid's power supply during faults.

Images