Safety management method, device and system of distributed power supply and medium

Abstract

The invention discloses a safety management method, device and system for distributed power supplies and a medium, and the method comprises the steps: collecting the energy production and consumption data of each distributed power supply, an energy storage system, a key load node and a connection point between the microgrid and a main power grid in a microgrid in real time; the energy demand of each distributed power supply of the micro-grid is monitored in real time through the collected energy production and consumption data; according to the business requirements and the data characteristics, determining characteristics having significant influences on the energy requirements in each distributed power supply, and according to the characteristics having the significant influences, determining an optimal characteristic set and a prediction model of each distributed power supply; predicting future energy demands in the microgrid of each distributed power supply through the optimal feature set and the prediction model; and the multi-agent system is used for adjusting the operation state and output of each distributed power supply, so that the coordinated operation of the whole micro-grid is realized. According to the invention, by monitoring and predicting the energy demand of the micro-grid in real time, the output of each distributed power supply is adjusted, and efficient utilization of energy is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed power supply, and in particular to a method, device, system and medium for safe management of distributed power supply. Background Art

[0002] With the rapid development of renewable energy and the widespread application of distributed power sources, microgrids, as a new type of power system structure, are gaining more and more attention. Microgrids are composed of distributed power sources, energy storage devices, energy conversion devices, loads, monitoring and protection devices, etc. They can realize the integrated grid connection of renewable distributed energy such as wind and light, and uniformly control the distributed power sources that are connected in an unordered manner.

[0003] However, distributed power sources are random and volatile, and their large-scale application and access have brought challenges to traditional power grids. Therefore, how to ensure the coordinated operation of multiple distributed power sources in microgrids to achieve stable operation of the system and meet user needs has become the focus of current research. Summary of the invention

[0004] In view of this, the present invention provides a safe management method, device, system and medium for distributed power sources, which can achieve efficient use of energy by real-time monitoring and predicting the energy demand of the microgrid and adjusting the output of each distributed power source.

[0005] A method for safe management of distributed power sources includes the following steps: The energy production and consumption data of each distributed power source, energy storage system, key load node, and connection point between the microgrid and the main grid are collected in real time through smart sensors. The energy demand of each distributed power source in the microgrid is monitored in real time through the collected energy production and consumption data. According to business needs and data characteristics, determine the features that have a significant impact on the energy demand of each distributed power source, and determine the optimal feature set and prediction model for each distributed power source based on the features that have a significant impact; Predicting the future energy demand in the microgrid of each distributed power source by determining the optimal feature set and prediction model for each distributed power source; After predicting the energy demand of multiple distributed power sources, the multi-agent system is used to adjust the operating status and output of each distributed power source to achieve coordinated operation of the entire microgrid.

[0006] Furthermore, the characteristics that have a significant impact on the energy demand of each distributed power source are determined according to the business needs and data characteristics, and the optimal feature set and prediction model of each distributed power source are determined according to the characteristics that have a significant impact, specifically including: Conduct correlation analysis on the energy production and consumption data of each distributed power source in the microgrid to identify candidate features that have a significant impact on the energy demand of each distributed power source; The obtained candidate features are evaluated for feature importance, and the candidate features whose importance scores exceed the importance threshold are obtained as the important feature set; The important feature sets obtained for each distributed power source are cross-validated to determine the optimal feature set and prediction model in the important feature sets of each distributed power source.

[0007] Furthermore, the correlation analysis of the energy production and consumption data of each distributed power source in the microgrid is performed to determine the candidate features that have a significant impact on the energy demand of each distributed power source, specifically including: The collected energy production and consumption data are initially screened according to the purpose of the analysis to remove features that are obviously irrelevant to the energy demand in the distributed generation; Using statistical methods to calculate the energy correlations between various residual characteristics and energy demand, and obtaining a set of energy correlation coefficients; A feature whose energy correlation coefficient is higher than a correlation threshold is selected from the energy correlation coefficient set as a candidate feature.

[0008] Furthermore, the obtained candidate features are evaluated for feature importance, and candidate features whose importance scores exceed an importance threshold are obtained as an important feature set, specifically including: Use the machine learning model for preliminary training to evaluate the importance of each candidate feature and obtain the importance score of each candidate feature; Sort the candidate features according to their importance scores given by the machine learning model; Among the ranked candidate features, features whose importance scores exceed the importance threshold are selected as the important feature set.

[0009] Furthermore, the cross-validation of the important feature set obtained for each distributed power source to determine the optimal feature set and prediction model in the important feature set of each distributed power source specifically includes: Generate multiple candidate feature sets using different feature combinations; Select multiple machine learning models, and for each candidate evaluation feature set, train on each machine learning model; Use K-fold cross validation to evaluate the performance of each candidate evaluation feature set on each machine learning model. Divide the data set into K mutually exclusive subsets of similar size, train the model with K-1 subsets each time, and test the model on the remaining subset. Repeat this process K times, selecting a different subset as the test set each time, and calculate the average of all test results as the performance indicator of the candidate evaluation feature set and machine learning model. Use performance indicators to evaluate each feature set and model combination, compare the performance of different feature set and model combinations, and select the best performing feature set and model; The performance of the selected feature set and model is verified on an independent validation set or test set, and the optimal feature set for each distributed power source and the subsequent prediction model to be used are finally determined.

[0010] Furthermore, after predicting the energy demand of multiple distributed power sources, the multi-agent system is used to adjust the operating status and output of each distributed power source to achieve coordinated operation of the entire microgrid, specifically including: Each distributed power source and key equipment in the microgrid is regarded as an agent. Each agent has autonomous decision-making and communication capabilities. Agents exchange information through communication protocols and share energy supply and demand and operation data. Based on the local information of each distributed power source and the global optimization goal, each agent autonomously adjusts its operating status and output to achieve coordinated operation of all distributed power sources. A central controller is established to collect the local optimal operating status and output of each distributed power source and the optimal operating mode of each distributed power source and perform centralized processing. According to the predicted energy demand, the central controller formulates a global optimization strategy and sends it to each distributed power source for execution. The distributed power source adjusts its operating status and output according to the received instructions to meet the system requirements and optimization goals.

[0011] A safety management device for distributed power supply, comprising: The energy demand monitoring module is used to collect the energy production and consumption data of each distributed power source, energy storage system, key load node, and connection point between the microgrid and the main grid in real time through intelligent sensors, and monitor the energy demand of each distributed power source in the microgrid in real time through the collected energy production and consumption data; The module for determining the optimal feature set and prediction model is used to determine the features that have a significant impact on the energy demand of each distributed power source according to business requirements and data characteristics, and to determine the optimal feature set and prediction model for each distributed power source according to the features that have a significant impact; An energy demand prediction module is used to predict the future energy demand in the microgrid of each distributed power source by determining the optimal feature set and prediction model for each distributed power source; The operation adjustment module is used to adjust the operation status and output of each distributed power source by using a multi-agent system after predicting the energy demand of multiple distributed power sources, so as to achieve coordinated operation of the entire microgrid.

[0012] Furthermore, the optimal feature set and prediction model determination module is specifically used to: Conduct correlation analysis on the energy production and consumption data of each distributed power source in the microgrid to identify candidate features that have a significant impact on the energy demand of each distributed power source; The obtained candidate features are evaluated for feature importance, and the candidate features whose importance scores exceed the importance threshold are obtained as the important feature set; The important feature sets obtained for each distributed power source are cross-validated to determine the optimal feature set and prediction model in the important feature sets of each distributed power source.

[0013] A safety management system for distributed power sources, comprising: a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the distributed power supply safety management method.

[0014] A non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements the safe management method of distributed power sources.

[0015] The present invention has the following advantages and beneficial effects: 1. Through coordinated control, the distributed power sources within the microgrid can effectively cooperate and coordinate their output, thereby improving the grid's ability to accept intermittent distributed power sources, such as solar energy and wind energy. This helps solve the problem of large-scale photovoltaic output power fluctuations and improves the dynamic balance of power within the grid.

[0016] 2. Treat each distributed power source and key equipment in the microgrid as an agent. Each agent has autonomous decision-making and communication capabilities. Agents exchange information through communication protocols and share energy supply and demand and operating data. Based on the local information and global optimization goals of each distributed power source, each agent autonomously adjusts its operating status and output to achieve coordinated operation of the entire distributed power source.

[0017] 3. Each distributed power source and key equipment in the microgrid can act as an intelligent agent system to coordinate and operate autonomously to achieve optimized operation of the entire microgrid.

[0018] 4. Microgrids use electricity as a carrier to form a miniature integrated energy system and achieve the joint optimization operation of multiple energy sources, which will help promote the transformation and upgrading of the energy structure and improve energy utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a distributed power supply safety management method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1 The embodiment of the present invention provides a distributed power supply safety management method, including the following steps: Step 1: Collect the energy production and consumption data of each distributed power source in the microgrid in real time through smart sensors, and monitor the energy demand of each distributed power source in the microgrid in real time through the collected energy production and consumption data.

[0022] Among them, the intelligent sensors can be set at important nodes of distributed power sources: for example, sensors can be set at the access points of distributed power sources such as wind power, solar power, and small hydropower to monitor the real-time output of each power source. These sensors can collect key data such as voltage, current, and power factor in real time.

[0023] Setting up power intelligent sensors at the access points of the wind power generation system can monitor the output voltage, current, power factor and other parameters of the wind turbine generator set in real time, helping to evaluate the utilization efficiency of wind energy resources and the operating status of the generator set. The specific power intelligent sensors set up are as follows: Voltage sensor, used to monitor the output voltage of wind turbines in real time to ensure that the voltage is stable within a reasonable range; Current sensors are used to measure the output current of wind turbines to assess their power generation capacity and efficiency; Power factor sensor, used to measure power factor, which is the ratio of active power to apparent power, to assess the power quality of wind turbines; In addition, wind speed sensors, wind direction sensors, etc. may also be used to monitor the impact of natural environmental conditions on wind power generation).

[0024] At the solar access point, the main concern is the output power and efficiency of the photovoltaic panels. Power smart sensors can monitor various parameters of distributed power sources in real time, including voltage, current, power factor, etc., to ensure the accuracy and real-time nature of the data. Therefore, the power smart sensors set up at the solar access point mainly include one or more of the following functions: Voltage sensor, used to monitor the output voltage of the photovoltaic panel to ensure that it is within the optimal operating voltage range; Current sensors, used to measure the output current of photovoltaic panels to assess their power generation capabilities; Power sensors are used to measure the output power of photovoltaic panels, which is a key parameter for evaluating their power generation efficiency; Light sensors and temperature sensors are used to monitor the impact of light intensity and ambient temperature on the power generation efficiency of photovoltaic panels. Power smart sensors and hydropower smart sensors are installed at the access points of small hydropower stations to monitor parameters such as voltage, current and power factor of turbine generator sets, as well as environmental parameters such as water flow and water level in real time, to help evaluate the power generation capacity and operating status of hydropower stations.

[0025] At the small hydropower access point, power smart sensors mainly focus on parameters related to power production, transmission and distribution, such as voltage, current, power factor, frequency, etc. These sensors can measure these parameters with high accuracy and transmit the data digitally to the monitoring system or data center for real-time analysis and processing.

[0026] In addition to basic power parameters, the hydropower smart sensors installed at the small hydropower access points may also have some functions related to the specific environment of the hydropower station, such as: Water level monitoring: Monitoring water levels in upstream reservoirs to assess water inflow and water reserves (level sensors).

[0027] Flow monitoring: Measuring the flow of water entering a turbine to calculate the power generation capacity and efficiency of a hydroelectric power plant (flow sensor).

[0028] Pressure monitoring: monitoring water pressure in pipelines and turbines to ensure system stability and safety (pressure sensors).

[0029] Depending on the specific application scenario and requirements, small hydropower access points may also use other types of sensors, such as: Ultrasonic flow meter: When ultrasonic waves propagate in a flowing fluid, the flow of the fluid will cause a slight change in the propagation time, thereby measuring the flow rate of the fluid, and then calculating the flow rate based on the pipe diameter.

[0030] Radar water level meter: It adopts the transmission-reflection-reception working mode and calculates the distance to the water surface by measuring the time from the transmission to the reception of electromagnetic waves.

[0031] Magnetic float water level gauge: uses the buoyancy principle and magnetic coupling to measure the water level. When the measured water level rises or falls, the magnetic float of the water level gauge also rises or falls, thereby indicating the height of the water level.

[0032] Energy storage equipment: Sensors are installed at the input and output ends of battery energy storage systems, flywheel energy storage and other equipment to monitor their charging and discharging status, power, temperature and other parameters.

[0033] For battery energy storage systems, the sensors set at the input and output ends are mainly the following: Voltage sensor: used to measure the voltage of a battery pack or a single battery to understand the battery power and usage in real time. Voltage is an important indicator of the battery energy storage status. By monitoring the voltage, battery failures can be discovered in time and the battery life status can be diagnosed, so as to better maintain the battery health status.

[0034] Current sensor: used to measure the current changes during the battery charging and discharging process. By monitoring the current, we can understand information such as charging and discharging efficiency and battery life, which is crucial to the safety and performance evaluation of the battery.

[0035] Temperature sensor: used to detect the temperature of the battery and understand the working status of the battery. Temperature is one of the key factors affecting battery performance and life. By monitoring the temperature, overheating and overcooling can be prevented, and heat dissipation or heating control can be performed as needed to protect battery safety.

[0036] Capacity sensor (indirectly calculated from voltage and current): Although the capacity is not a directly measured parameter, it can be estimated from the changes in voltage and current and the capacity characteristics of the battery. Knowing the remaining capacity of the battery is important for properly arranging charging and discharging plans and preventing overcharging or over-discharging of the battery.

[0037] For the flywheel energy storage system, the sensors set at the input and output ends are mainly the following: Speed ​​sensor (indirectly reflects the charge and discharge status): used to measure the speed of the flywheel. The change in speed can indirectly reflect the charge and discharge status of the flywheel energy storage system. The flywheel energy storage system stores and releases energy through the rotation of the flywheel. The speed is an important indicator for evaluating its energy storage and release capabilities.

[0038] Temperature sensor: used to monitor the temperature of various components in the flywheel energy storage system, including flywheels, bearings, motors, etc. Temperature is one of the key factors affecting the performance and life of the flywheel energy storage system. By monitoring the temperature, overheating can be detected in time and corresponding heat dissipation measures can be taken to protect the safety of the system.

[0039] Vibration sensor: used to monitor the vibration of the flywheel energy storage system to evaluate the stability and operation of the system. Abnormal vibration may be a harbinger of system failure or imbalance. By monitoring vibration, potential problems can be discovered and dealt with in a timely manner.

[0040] It should be noted that different types of energy storage systems and equipment may require different types of sensor combinations to fully monitor their operating status and performance parameters. In addition, with the continuous development of technology, new sensor types and monitoring technologies are constantly emerging, providing more possibilities for the intelligent and efficient operation of energy storage systems.

[0041] Critical load nodes: Sensors are installed at critical load nodes (such as hospitals, data centers, etc.) within the microgrid to monitor their power demand and power quality in real time.

[0042] Sensors installed at key load nodes (such as hospitals, data centers, etc.) in the microgrid include the following types: (1) Power demand monitoring sensors Current sensor: used to monitor the current changes in the circuit in real time, so as to understand the power consumption of the load. Current is an important indicator reflecting power demand. By monitoring the current, the real-time power and power consumption of the load can be evaluated, providing a basis for power dispatch and supply-demand balance.

[0043] Voltage sensor: used to measure the voltage value in the circuit to ensure that the voltage is stable within a reasonable range and meet the power demand of the load. Voltage stability is crucial to the normal operation of the load. Too high or too low voltage may damage the equipment or affect the operating efficiency.

[0044] Power sensor: directly measures the power consumption of the load, including active power and reactive power. Power data is important for evaluating the energy efficiency and power demand of the load, helping to identify potential energy-saving opportunities and optimize power distribution.

[0045] (2) Power quality monitoring sensor Power quality sensor: Comprehensively monitor power quality parameters in the power grid, such as harmonics, voltage fluctuations, frequency deviation, etc. Power quality directly affects the operating stability and life of load equipment. By monitoring power quality, potential problems can be discovered and handled in a timely manner to ensure the normal operation of load equipment.

[0046] Harmonic sensor: specially used to measure the harmonic content in the power grid, including the amplitude and phase of each harmonic. Harmonic is one of the common power quality problems in the power grid, which has adverse effects on the normal operation of load equipment and the stability of the power grid. By monitoring harmonics, the degree of harmonic pollution in the power grid can be evaluated and corresponding treatment measures can be taken.

[0047] Frequency sensor: used to monitor the frequency change of the power grid and ensure that the frequency is stable within the rated range. Frequency is one of the important parameters of power grid operation. Frequency fluctuation will directly affect the operating performance and stability of load equipment. By monitoring the frequency, frequency deviation problems can be discovered and handled in time to ensure the stable operation of the power grid.

[0048] (3) Auxiliary sensor Temperature sensor: used to monitor the temperature of load equipment or the surrounding environment to prevent safety hazards such as equipment damage or fire caused by overheating. Temperature is one of the key factors affecting equipment performance and life. By monitoring the temperature, overheating problems can be discovered and handled in a timely manner to ensure the safe operation of the equipment.

[0049] Humidity sensor (for specific environments such as data centers): used to monitor ambient humidity to ensure that the equipment operates under appropriate humidity conditions. Humidity that is too high or too low may have adverse effects on the equipment, such as causing equipment corrosion, short circuits, or performance degradation. By monitoring humidity, humidity problems can be discovered and handled in a timely manner to ensure the normal operation of the equipment.

[0050] Grid connection point: Sensors are installed at the connection point between the microgrid and the main grid to monitor the grid-connected or islanded operation status of the microgrid and the power exchange situation.

[0051] The following types of sensors can be set up at the connection point between the microgrid and the main grid: (1) Grid-connected / Island status monitoring sensor State switching monitoring sensor: used to monitor the connection status between the microgrid and the main grid in real time, and determine whether the microgrid is in a grid-connected state or an islanded operation state. The state can be identified by monitoring the changes in parameters such as voltage, current, and frequency at the connection point, combined with preset logical judgments.

[0052] Synchronous monitoring sensors (if necessary): During the grid connection process, synchronous monitoring sensors may be required to ensure the synchronization of the microgrid and the main grid in parameters such as voltage and frequency to ensure a smooth and safe grid connection process.

[0053] (2) Power exchange monitoring sensor Power sensor: directly measures the active power and reactive power exchanged between the microgrid and the main grid. Power data is an important indicator for evaluating the operating efficiency and stability of the microgrid, and is also an important basis for power dispatch and supply and demand balance. The power sensor can be an analog output power sensor or a digital output power sensor, the latter of which can simultaneously output multiple power parameters, such as voltage, current, frequency, etc.

[0054] Current and voltage sensors: Although power sensors can directly measure power, current and voltage sensors are also indispensable. They measure the current and voltage values ​​at the connection points respectively, providing basic data for power calculation. Current sensors can be divided into non-contact and contact types, while voltage sensors include capacitive, resistive, and magnetic sensitive types.

[0055] (3) Auxiliary sensor Frequency sensor: monitors the frequency change at the connection point to ensure the frequency consistency between the microgrid and the main grid. Frequency stability is crucial to the stable operation of the power system. Power quality sensor (optional): comprehensively monitors the power quality parameters at the connection point, such as harmonics, voltage fluctuations, etc. Power quality directly affects the stable operation of the microgrid and the main grid and the life of the equipment.

[0056] High-precision smart sensors are used to sense various parameters in the power grid in real time and convert them into electrical or digital signals. These smart sensors are connected to the data acquisition system via wired or wireless means. The data acquisition system is responsible for receiving data from the sensors and performing preliminary processing and storage. It can filter, clean and standardize data to improve data quality; through communication networks (such as wired networks, wireless networks, the Internet of Things, etc.), the data acquisition system is connected to other systems (such as energy management systems, central controllers, etc.) to achieve real-time data transmission and sharing.

[0057] The data collected by the above sensors are the energy production and consumption data within the microgrid.

[0058] Taking the solar photovoltaic power generation system as an example, its data collection and monitoring process is as follows: Voltage sensors and current sensors are set at the input and output ends of the photovoltaic array to monitor the voltage, current and power output of the photovoltaic array in real time.

[0059] Sensors are set at the input and output ends of the photovoltaic inverter to monitor parameters such as the inverter's conversion efficiency, working status and temperature.

[0060] (1) Input sensor 1. Current sensor, used to measure the DC current at the input of the photovoltaic inverter, helps monitor the output current of the photovoltaic panel or battery in real time, ensures that the current is within the rated range of the inverter, and prevents problems such as overload and short circuit. Common current sensors include Hall current sensors (open loop and closed loop), fluxgate current sensors, etc. These sensors can measure current with high precision and convert current signals into electrical signals or other required forms of information output.

[0061] 2. Voltage sensor, which measures the DC voltage at the input of the photovoltaic inverter to ensure that the voltage is within the rated voltage range of the inverter and to ensure the normal operation of the inverter. The voltage sensor can directly measure the voltage value and convert it into a processable electrical signal for use by the control system.

[0062] (2) Output sensor 1. Power sensors directly measure the AC power at the output of the PV inverter to evaluate whether the inverter's conversion efficiency and output power are as expected. Power sensors can measure voltage and current and calculate power values. They are usually highly accurate and stable, and can provide real-time feedback on the inverter's output power.

[0063] 2. Current and voltage sensors At the output end, current and voltage sensors are equally important. They measure the AC current and voltage output of the inverter respectively to ensure the stability and accuracy of the output current and voltage. Similar to the input end, the current and voltage sensors at the output end also include Hall sensors, fluxgate sensors and other types.

[0064] (3) Temperature sensor: used to measure the internal temperature of the photovoltaic inverter to ensure that the inverter operates within the appropriate temperature range and prevent equipment damage or performance degradation due to excessive temperature.

[0065] Imported temperature sensor: installed at the air inlet of the photovoltaic inverter converter, measures the air temperature entering the inverter in real time, and feeds back to the controller to adjust the working state of the inverter.

[0066] SMD temperature sensor: It is close to the inverter chip, has high temperature resistance, and can work stably in the high temperature environment inside the inverter. When the inverter has abnormal temperature, the SMD temperature sensor will feedback the abnormal signal to the controller, triggering the protection mechanism.

[0067] The data from these sensors are collected by the data acquisition system and initially processed and stored. For example, interpolation, smoothing and correction techniques can be used to fill in missing data and repair erroneous data.

[0068] The processed data is transmitted to the energy management system or central controller through the communication network, and is comprehensively analyzed and optimized together with other data.

[0069] Step 2: According to business needs and data characteristics, determine the characteristics that have a significant impact on the energy demand of each distributed power source, such as historical demand, weather conditions, electricity prices, etc., and determine the optimal feature set and prediction model for each distributed power source based on the characteristics that have a significant impact. Step 2 specifically includes: Step 2.1: Perform correlation analysis on the energy production and consumption data of each distributed power source in the microgrid to determine the candidate features that have a significant impact on the energy demand of each distributed power source.

[0070] First, a preliminary screening is performed, that is, all the collected features (energy production and consumption data collected in step 1) are screened to remove those features that are obviously irrelevant to the energy demand in distributed power sources.

[0071] For example, the collected features include: key data such as voltage, current, power factor of distributed power sources such as wind power, solar power, and small hydropower, the charging and discharging status, power, and temperature of the battery energy storage system, the grid-connected or islanded operation status of the microgrid, and the power exchange status, voltage, current, and power output of the photovoltaic array in the distributed solar power source, the conversion efficiency of the photovoltaic inverter, the working status of the inverter, and the temperature of the inverter.

[0072] Initial screening to remove features that are clearly irrelevant to the energy demand in distributed generation can be determined based on the purpose of the analysis.

[0073] First, the purpose of the analysis needs to be clearly defined. For example, determine whether the purpose of the subsequent analysis is to evaluate the efficiency of the solar power generation system, to predict the operating requirements of the energy storage system, or to optimize the load balance of the power grid. According to the purpose of the analysis, identify the key features that are directly related to it. For example, when analyzing the efficiency of the solar power generation system, the daily power generation and the power generation efficiency curve are the key features; when predicting the operating requirements of the energy storage system, the maximum load and the energy storage system capacity are the key features. In some cases, feature selection techniques (such as variance selection, correlation coefficient analysis, model scoring, etc.) can also be used to further screen and extract key features.

[0074] For example, the efficiency of a solar power generation system in a specific period of time is to be evaluated. The relevant data is collected through system monitoring equipment or data loggers, and then the key features are determined: daily power generation, power generation efficiency curve, and power generation curve. Finally, methods such as variance selection or correlation coefficient analysis are used to screen out the features that are directly related to efficiency.

[0075] Secondly, correlation calculation is performed, and statistical methods (such as Pearson correlation coefficient, Spearman rank correlation coefficient, etc.) are used to calculate the energy correlation between various residual characteristics and energy demand respectively, and a set of energy correlation coefficients is obtained.

[0076] Taking the calculation of Pearson correlation coefficient as an example, the Pearson correlation coefficient is used to measure the linear correlation between two variables. The formula is: Axy = {{cov}(X,Y)} / {\sigma_X \sigma_Y}) Where Axy is the calculated energy correlation coefficient, ({cov}(X,Y)) is the covariance of (X) and (Y), (\sigma_X) and (\sigma_Y) are the standard deviations of (X) and (Y), respectively, X can be the collected features, and Y can be the energy demand.

[0077] Finally, a correlation threshold (such as 0.1) is set, and features whose energy correlation coefficients are higher than the correlation threshold are selected from the energy correlation coefficient set as candidate features.

[0078] Step 2.2: Evaluate the feature importance of the candidate features obtained in step 2.1, and obtain the candidate features with higher importance scores as the important feature set. Step 2.2 specifically includes: Step 2.21: Use a machine learning model (such as random forest, gradient boosting machine, etc.) for preliminary training to evaluate the importance of each candidate feature obtained in step 2.1 and obtain the importance score of each candidate feature.

[0079] Choose a machine learning model suitable for evaluating feature importance, such as Random Forest (RF) or Gradient Boosting Machine (GBM). Train the model in advance (collect a training set to divide the dataset into a training set and a test set). Here, we use a pre-trained model.

[0080] Perform preliminary training using a machine learning model (such as random forest, gradient boosting machine, etc.) to evaluate the importance of each candidate feature and output the importance score of each candidate feature.

[0081] Suppose there is a distributed photovoltaic power station and the following features are collected: light intensity, temperature, humidity, wind speed, voltage, current and power factor. We use the random forest model for training and obtain the importance score of each feature. Temperature (0.3), power factor (0.01), wind speed (0.15), humidity (0.08), voltage (0.05), light intensity (0.4), current (0.01), etc.

[0082] Step 2.22: Sort the candidate features according to their importance scores given by the machine learning model.

[0083] Taking the example in step 2.21 as an example, the feature importance ranking is as follows: the calculated feature importance scores from high to low are: light intensity (0.4), temperature (0.3), wind speed (0.15), humidity (0.08), voltage (0.05), current (0.01), and power factor (0.01).

[0084] Step 2.23: Select features with higher importance scores from the ranked candidate features as the important feature set.

[0085] First, determine the main influencing factors of the current business, and select highly representative features and auxiliary features from the sorted candidate features based on the main influencing factors.

[0086] The main factors affecting the current business are mainly determined based on the system type. The system type can be confirmed in advance. For example, if the system type is a photovoltaic power generation system, light intensity, temperature, humidity, etc. may be key factors that directly affect power generation. Considering the characteristics of the photovoltaic power generation system, light intensity and temperature are determined to be the main factors affecting power generation. Therefore, these two characteristics are very important in business.

[0087] Selecting highly representative features and auxiliary features from the sorted candidate features based on the main influencing factors can be done by selecting features that are highly correlated with the main influencing factors from the candidate features. If the system type is a photovoltaic power generation system, checking the data, it is found that wind speed, humidity, light intensity, and the main influencing factor (temperature) have a certain correlation, but not as direct as them. Voltage and current may be related to the state of the grid, but their direct contribution to the power generation forecast may be small. The power factor is usually close to 1 in photovoltaic power generation systems, so its changes may not have much impact on the forecast. Finally, it is determined that wind speed, humidity, light intensity, and the main influencing factor (temperature) are highly correlated.

[0088] If you need to further improve the performance of the model, you can consider adding wind speed as an auxiliary feature, because although it has a lower importance score, it has a certain correlation with light intensity and temperature and may provide additional information. Finally, light intensity, temperature and wind speed are selected as the important feature set.

[0089] Pay special attention to analyzing the distribution, correlation, missing values ​​and other characteristics of the data when selecting features to ensure that the selected features are representative and stable in the data set, and exclude those with a large number of missing values ​​or highly correlated with other features to avoid redundancy and noise.

[0090] In this way, features with higher importance scores (e.g., exceeding a preset importance threshold) can be selected as the important feature set, thereby improving the accuracy and stability of the prediction model.

[0091] Step 2.3: Cross-validate the important feature set obtained for each distributed power source to determine the optimal feature set and prediction model in the important feature set of each distributed power source.

[0092] Use different machine learning models (such as linear models, neural networks, etc.) for cross-validation to ensure that the selected important feature sets have good predictive performance under different models, and try different feature combinations to find the optimal feature set.

[0093] The above important feature set can be used as a training set, and the training set can be further divided into multiple subsets (such as K subsets in K-fold cross validation) for multiple training and validation.

[0094] Observe the stability of feature importance in different models and cross-validation iterations, compare the performance indicators of different models (such as linear models (logistic regression, linear regression), neural networks, decision trees, random forests, gradient boosting trees, etc.) in cross-validation, understand which model is more suitable for the current data set and feature set, analyze the performance fluctuations of the model in different iterations, understand the stability and robustness of the model, and then determine the optimal feature set and machine learning model, and prepare to apply it to subsequent actual distributed power prediction tasks.

[0095] Specifically: Step 2.3 includes: Step 2.3.1: Generate candidate evaluation feature set Generate multiple candidate feature sets using different feature combinations (such as single features, feature pairs, feature subsets, etc.).

[0096] In the present invention, the original features (important feature set) can also be converted and combined to generate new features and added to the candidate evaluation feature set to improve the performance of the prediction model. For example, the moving average or seasonal index of historical demand can be calculated.

[0097] Step 2.3.2, model selection and training: Select multiple machine learning models, such as linear models (logistic regression, linear regression), neural networks, decision trees, random forests, gradient boosted trees, etc. For each candidate evaluation feature set, train on each machine learning model.

[0098] Step 2.3.3, cross validation: K-fold cross-validation is used to evaluate the performance of each feature set on each model. The dataset is divided into K mutually exclusive subsets of similar size, and the model is trained with K-1 subsets each time, and the model is tested on the remaining subset. This process is repeated K times, and a different subset is selected as the test set each time. The average of all test results is calculated as the performance indicator of the candidate evaluation feature set and model.

[0099] Step 2.3.4: Performance evaluation and comparison Use performance metrics such as accuracy, recall, F1 score, mean square error, etc. to evaluate each feature set and model combination. Compare the performance of different feature set and model combinations and select the best performing feature set and model.

[0100] Step 2.3.5: Final verification The performance of the selected feature set and model is verified on an independent validation set or test set, and the optimal feature set for each distributed power source and the subsequent prediction model to be used are finally determined.

[0101] Step 3: Predict the future energy demand in the microgrid of each distributed power source by using the optimal feature set and prediction model selected for each distributed power source. Step 3 specifically includes: Step 3.1: Model selection and training Data collection: Collect historical data sets with the optimal feature set and divide the historical data sets into training data sets and test data sets.

[0102] Model selection: Select the prediction model finally determined in step 2.

[0103] Model training: Perform model training, such as using the training data set corresponding to the optimal feature set to train the model and adjust the model parameters so that the model can accurately fit the historical data.

[0104] Step 3.2: Model evaluation and optimization Model evaluation: Use the test data set to evaluate the performance of the model and calculate the model's prediction accuracy, recall rate, F1 score and other indicators.

[0105] Model optimization: Adjust and optimize the model based on the evaluation results, such as adjusting model parameters, adding new features, trying different model structures, etc.

[0106] Step 3.3: Prediction and result output Forecasting: Use the trained model to predict future energy demand.

[0107] Result output: Output the forecast results in the form of charts, reports, etc. for reference by decision makers.

[0108] Suppose we have a network with multiple distributed power sources, and we need to predict the power demand for the next week. After determining the features that have a significant impact on the energy demand in the distributed power sources in step 2, assuming that they are data such as light intensity, temperature, and electricity price data, we can collect power demand data, weather data (such as light intensity, temperature, etc.), and electricity price data for multiple distributed power sources in the past year. Then, we use this data to train a deep learning model (such as LSTM). During the training process, we continuously adjust the parameters of the model until the model can accurately fit the historical data. Finally, we use the trained model to predict the power demand of multiple distributed power sources in the next week, and output the prediction results in the form of charts. Decision makers can formulate corresponding energy management strategies based on the prediction results to ensure the stable operation of the microgrid and the efficient use of distributed power sources.

[0109] Step 4: After predicting the energy demand of multiple distributed power sources in step 3, the multi-agent system is used to adjust the operating status and output of each distributed power source to achieve coordinated operation of the entire microgrid. Step 4 specifically includes: Step 4.1: Treat each distributed power source and key equipment in the microgrid as an agent. Each agent has autonomous decision-making and communication capabilities. Agents exchange information through communication protocols and share energy supply and demand and operating data. Based on the local information of each distributed power source and the global optimization goal, each agent autonomously adjusts its operating status and output to achieve coordinated operation of all distributed power sources.

[0110] Each agent independently calculates the optimal operation mode and output based on its own power demand forecast, energy type (such as solar energy, wind energy, etc.) and operating status. Agents exchange information through standardized communication protocols (such as IEC 61850, MQTT, etc.) to share energy supply and demand, operating data and optimization goals. Step 4.1 specifically includes: Step 4.11, each agent collects the local power demand forecast (predicted in step 3), energy status (such as battery charging status, wind speed of wind turbine, etc.) and operation data, uploads the information to the central controller or central controller through the communication protocol (such as IEC 61850, MQTT, etc.), and receives information from other agents.

[0111] Step 4.12: Each agent calculates the optimal local operating state and output of each distributed power source based on the received global optimization objectives and local information, and selects the optimal operating mode of each distributed power source by considering factors such as energy complementarity, operating efficiency and cost.

[0112] Among them, each agent first collects local real-time data, including energy generation (such as solar photovoltaic power generation, wind power generation), energy storage status (such as battery charging status), energy demand forecast, equipment status (such as equipment failure, maintenance needs, etc.) and cost information (such as operating cost, maintenance cost, etc.).

[0113] Based on the collected local data, each agent performs local optimization calculations, such as optimization algorithms such as linear programming, dynamic programming, genetic algorithms, or machine learning algorithms, to calculate the optimal local operating status and output of each distributed power source and select the optimal operating mode for each distributed power source.

[0114] Assume that the microgrid contains a solar photovoltaic power generation system (agent A), a wind power generation system (agent B) and a battery energy storage system (agent C). Agent A (solar photovoltaic power generation system) predicts the solar power generation in the next few hours based on the weather forecast and calculates the optimal local operation mode and output. Agent B (wind power generation system) calculates the optimal local operation mode and output based on the wind speed forecast and local equipment status. Agent C (battery energy storage system) calculates the amount of electricity and time required for charging and discharging based on the power demand and the output forecast of the two power generation systems.

[0115] For example, Agent A calculates the amount of electricity that should be generated in a given time period and the optimal operation mode (such as whether to start the tracking system, whether to perform cleaning, etc.) based on solar radiation forecasts and cost information. Agent B calculates the amount of electricity that should be generated in a given time period and the optimal operation mode (such as whether to adjust the blade angle, whether to perform maintenance, etc.) based on wind speed forecasts and cost information. Agent C calculates the optimal charging and discharging strategy based on energy demand forecasts, battery charging status, and discharge costs.

[0116] In a specific example: (1) First, each agent needs to clearly define its optimization goal. This may include maximizing energy efficiency, minimizing costs, meeting energy demand, etc. Based on these goals, the agent can define a cost function (or loss function) that measures the gap between the current operating state and output and the optimal state. Each agent collects local real-time data, such as energy generation, energy storage status, energy demand forecast, equipment status, and cost information.

[0117] (2) Based on the current operating state and output, as well as the collected local data, the agent calculates the gradient of the cost function with respect to its parameters. The gradient is a vector that points in the direction in which the function value increases the fastest. In optimization problems, we want to find the parameter value that minimizes the cost function, so we need to search in the opposite direction of the gradient (i.e., the negative gradient direction).

[0118] (3) The agent updates its operating state and output based on the calculated gradient. This can be achieved by subtracting a step size proportional to the gradient from the current parameter value. The step size (also known as the learning rate) is an important hyperparameter that determines the magnitude of the parameter update. A step size that is too large may cause the algorithm to be unstable, while a step size that is too small may cause the algorithm to converge too slowly.

[0119] (4) The agent repeats steps 4 and 5 until a certain stopping condition is met. The stopping condition can be reaching the maximum number of iterations, the cost function value is less than a certain threshold, the gradient value is less than a certain threshold, etc. When the stopping condition is met, the agent outputs its optimized operating status and output.

[0120] Assume that agent A is a solar photovoltaic power generation system, and its optimization goal is to maximize energy efficiency while meeting energy demand. Agent A defines a cost function that measures the gap between current photovoltaic power generation and energy demand, as well as the efficiency of the photovoltaic power generation system. Agent A collects information such as the solar radiation forecast for the day, equipment status (such as inverter efficiency, component cleanliness, etc.), and energy demand forecast. Agent A initializes parameters such as the angle of its photovoltaic panels and the operating mode of the inverter based on historical data and expert knowledge.

[0121] Based on the current parameters and the local data collected, agent A calculates the gradient of the cost function with respect to its parameters. For example, it may find that the cost function value (i.e., energy efficiency) improves when the angle of the photovoltaic panel increases by a certain angle.

[0122] Agent A adjusts the angle of its PV panels and the operating mode of its inverter based on the calculated gradient. It may try multiple different angle and mode combinations to find the optimal solution that minimizes the cost function. Agent A repeats steps 4 and 5 until the cost function value reaches a stable low value or other stopping conditions are met. When the optimization process is completed, Agent A outputs its optimized PV panel angle, inverter operating mode, and estimated PV power generation. These optimization results will be used to guide the actual operation of the PV power generation system.

[0123] Among them, based on the current parameters and the collected local data, agent A can calculate the gradient of the cost function with respect to its parameters as follows: Assume that the cost function of agent A is J(θ), where θ is the parameter vector to be optimized. Then, the gradient of the cost function with respect to the parameter θ can be expressed as , which is a vector of the same dimension as θ.

[0124] Each component of the gradient can be obtained by taking the partial derivative of the cost function with respect to the corresponding parameter. That is:

[0125] In practice, these partial derivatives can be approximated or calculated exactly using numerical methods (such as the finite difference method) or symbolic computation.

[0126] For example, agent A is an agent of a photovoltaic power generation system, and its goal is to maximize the energy efficiency under the premise of meeting the energy demand. We can define a simple cost function J(θ) = -Efficiency(θ) + Penalty(θ), where Efficiency(θ) is the energy efficiency function (θ is a parameter such as the angle of the photovoltaic panel), and Penalty(θ) is a penalty term used to consider the situation where the energy demand is not met.

[0127] Determine the optimization parameters: Assume that θ is the angle of the PV panel in degrees.

[0128] Define the cost function: J(θ) = -Efficiency(θ) + Penalty(θ), where Efficiency(θ) is the energy efficiency calculated based on the angle of the photovoltaic panels and the current weather conditions, and Penalty(θ) is the penalty cost incurred when the energy demand is not met.

[0129] Collecting local data: Agent A collects current weather conditions (such as solar radiation intensity, temperature, etc.) and energy demand forecast data.

[0130] Calculate the cost function value: Use the current PV panel angle θ and the collected local data to calculate Efficiency(θ) and Penalty(θ) to obtain the cost function value J(θ).

[0131] Compute the gradient: Calculate the partial derivative of J(θ) with respect to θ by numerical methods (such as finite difference method) or symbolic calculation , which is the component of the gradient.

[0132] Apply gradient descent: Based on the calculated gradient And the pre-set learning rate α, update the angle of the photovoltaic panel .

[0133] By continuously iterating this process, agent A can gradually find the optimal photovoltaic panel angle that minimizes the cost function value, thereby maximizing energy efficiency.

[0134] Step 4.2: Establish a central controller to collect all information of each distributed power source (the local optimal operating status and output of each distributed power source in step 4.1, and the optimal operating mode of each distributed power source) and process it centrally. According to the predicted energy demand, the central controller formulates a global optimization strategy and sends it to each distributed power source for execution. The distributed power source adjusts its operating status and output according to the received instructions to meet the system requirements and optimization goals. Step 4.2 specifically includes: Step 4.21, the central controller performs global energy supply and demand balance calculation based on the local optimal operating status and output of each distributed power source fed back by each agent, and the optimal operating mode of each distributed power source. If it is found that the output of some agents does not meet the global optimization goal, the central controller sends adjustment instructions to the relevant agents through the communication protocol.

[0135] First, the central controller needs to collect feedback information from various distributed power sources (such as solar photovoltaic panels, wind turbines, energy storage systems, etc.), including: Local optimal operating status (such as maximum efficiency point, current health status) Current output (i.e. actual power generation or discharge) Forecasted output (power generation forecast for a period of time in the future based on weather forecasts or historical data) Optimal operating mode (based on comprehensive considerations of cost, efficiency, environmental protection, etc.) The central processing unit formulates a global optimization strategy through a global optimization processing model to perform global energy supply and demand balance calculations.

[0136] The global optimization model is as follows: Objective function: The goal of global optimization is usually to maximize energy efficiency, minimize costs, or meet specific supply and demand balance conditions. Here we take minimizing the total system cost as an example, including power generation cost, transmission loss, maintenance cost, etc.

[0137] Where N is the number of distributed generation, C gen,i is the generation cost of the ith power source (which may be proportional to its output), C trans,i is the transmission loss cost, C maint,i It is the maintenance cost.

[0138] The constraints are as follows: Supply and demand balance: total power generation should be equal to or slightly greater than total demand (taking into account a certain reserve capacity).

[0139] Among them, P gen,iis the output of the ith power supply, P demand is the total system demand.

[0140] Power output limitations: The output of each power supply must be within its physical and technical limitations.

[0141] Grid stability constraints: voltage, frequency, etc. must be kept within a safe range. (Relevant parameters exist in grid specifications).

[0142] The central processor formulates a global optimization strategy through a global optimization processing model, that is, the central processor determines the optimal output of each distributed power source through a global optimization processing model. The specific method is as follows: Initialization: Randomly generate a set of solutions (population), each of which represents a configuration of distributed generation output.

[0143] Evaluation: Calculate the objective function value (total cost) of each solution using the objective function above.

[0144] Selection: Select excellent solutions for reproduction based on the objective function value.

[0145] Crossover: Generate new solutions through crossover operations (such as single-point crossover and multi-point crossover).

[0146] Mutation: Randomly mutate the solution with a certain probability to increase population diversity.

[0147] Iteration: Repeat the selection, crossover, and mutation process until the stopping condition is met (such as reaching the maximum number of iterations, the objective function value converges, etc.).

[0148] The above method is used to determine the optimal output of each distributed power source. By comparing the local optimal operating output fed back by each distributed power source (each distributed power source has the optimal output, which is the global optimization goal), it is determined whether each distributed power source needs to adjust its output. If the output needs to be adjusted, it is sent to each distributed power source for execution. The distributed power source adjusts its operating status and output according to the received instructions to meet system requirements and optimization goals.

[0149] For example, agents A, B, and C send their respective optimization results to the central controller. If it is found that the output of some agents does not meet the global optimization goal, adjustment instructions are sent to the relevant agents through the communication protocol. For example, the central controller performs a global energy supply and demand balance calculation based on the information uploaded by the three agents, and finds that the output of agents A and B is insufficient to meet the power demand. Therefore, an instruction is sent to agent C to increase the discharge to supplement the power. For example, the central controller finds that the sum of the power generation of agents A and B may not meet the energy demand in the afternoon, so it sends an instruction to agent C to increase the discharge in the afternoon.

[0150] At this point, energy complementarity can also be considered between multiple agents. For example, if agent A is a solar photovoltaic power generation system and agent B is a wind power generation system, the two may be complementary in energy generation (for example, at night or when the wind is insufficient, the solar photovoltaic power generation is reduced, but there may be sufficient wind). Therefore, agents A and B need to consider each other's energy generation during the optimization process.

[0151] Step 4.22: The agent that receives the adjustment instruction adjusts its operating status and output autonomously according to the instruction content. After the adjustment is completed, it reports the new operating status and output to the central controller or central controller again through the communication protocol.

[0152] For example, after receiving the instruction, agent C autonomously adjusts its discharge amount and discharge time to meet the power demand, and reports the new operating status and output to the central controller through the communication protocol. Agent C sends the adjusted optimization results to the central controller for the next round of global optimization.

[0153] In this way, each agent can autonomously calculate the optimal local operating state and output based on the global optimization goal and local information, and work in collaboration with other agents and the central controller to achieve optimized operation of the entire microgrid.

[0154] Step 4.23: The central controller monitors the operation status of the microgrid in real time, evaluates the response speed and adjustment effect of each agent, obtains the evaluation results, and continuously optimizes and adjusts the logic strategy based on the evaluation results. Step 4.23 specifically includes: Step 4.231. Data collection and preprocessing Real-time data collection: The central controller collects the operating data of each agent (such as distributed power sources, energy storage systems, load controllers, etc.) in real time through the communication network, including output, efficiency, status code, response timestamp, etc.

[0155] Step 4.232, response speed evaluation Defining response time: Response time refers to the time difference from when the central controller sends an adjustment instruction to when each agent starts to execute the instruction.

[0156] Calculate response time: Use timestamps to record the time points when instructions are sent and received, and calculate the actual response time.

[0157] Response time calculation formula: {Response time} = {Receive timestamp} - {Send timestamp} Evaluation criteria: Set a reasonable response time threshold (such as seconds or milliseconds), compare the actual response time with the threshold, and evaluate whether the response speed of each agent meets the standard.

[0158] Step 4.233: Evaluate the effect of adjustments Set evaluation indicators: According to the operation objectives and optimization objectives of the microgrid, set appropriate evaluation indicators, such as output stability, energy utilization efficiency, supply and demand balance, cost savings rate, etc.

[0159] For example, output stability evaluation indicators (example): {Output stability} = 1 - {Output fluctuation} / {Average output} The output fluctuation amount may be the difference between the maximum and minimum output values ​​within a certain period of time.

[0160] Energy efficiency evaluation indicators (example): {Energy utilization efficiency} = {actual power generation} / {theoretical power generation}*100% Data collection and calculation: Within a period of time (such as a few minutes or hours) after the adjustment instruction is executed, relevant data is continuously collected and the actual value of the evaluation indicator is calculated.

[0161] Comparative analysis: Compare and analyze the actual value of the evaluation indicator with the expected target value, historical data or industry standards to evaluate whether the adjustment effect meets expectations.

[0162] Step 4.234. Comprehensive evaluation and feedback Comprehensive evaluation: Combined with the evaluation results of response speed and adjustment effect, each agent is comprehensively evaluated to determine its performance in microgrid operation.

[0163] Feedback mechanism: The evaluation results will be fed back to each agent and relevant management personnel as a basis for subsequent optimization and adjustment.

[0164] Step 4.3: Use energy storage devices (such as batteries, flywheels, etc.) to balance fluctuations in energy supply and demand and improve system stability and reliability.

[0165] When the predicted power generation exceeds the power demand, the energy storage system is used to charge and store the excess power. When the power demand exceeds the power generation, the energy storage system is discharged to supplement the power supply. Reasonable charge and discharge thresholds are formulated to avoid excessive charge and discharge of the energy storage system, which affects the system life and stability.

[0166] Assume that we have a distributed power system, including a solar photovoltaic power generation system and a wind power generation system, as well as a battery energy storage system. According to the forecast, in the next 24 hours, the power generation of the photovoltaic power generation system is expected to be 500kWh, the power generation of the wind power generation system is expected to be 300kWh, and the power demand is expected to be 100kWh. According to the forecast results, an energy storage system capable of storing 100kWh of electricity is required. Considering the charging and discharging efficiency and life, lithium-ion batteries are selected as energy storage devices.

[0167] During the day when there is sufficient sunlight and high wind power generation, the power generation of the photovoltaic power generation system and the wind power generation system exceeds the power demand. At this time, we charge the excess power into the energy storage system until the energy storage system reaches 80% of its capacity (to avoid overcharging).

[0168] At night or during periods of poor wind and light conditions, electricity demand exceeds power generation. At this time, discharge from the energy storage system to supplement the power supply. During the discharge process, we keep the SOC (state of charge) of the energy storage system not less than 20% (to avoid over-discharge).

[0169] Through the monitoring system, we monitor the operating status of the distributed power system, energy storage system and power grid in real time. If we find that the SOC of the energy storage system is close to the set threshold, we will make adjustments based on real-time data to ensure the stable operation of the system. We regularly maintain and inspect the energy storage system to ensure its normal operation. At the same time, through the optimization algorithm, we improve the operating efficiency and reliability of the energy storage system and reduce operating costs.

[0170] Through the above strategies, each distributed power source and key equipment in the microgrid can act as an intelligent agent system, autonomously coordinate operations, and achieve optimized operation of the entire microgrid.

[0171] The present invention can realize the coordinated operation of multiple distributed power sources in a microgrid, ensure the safe and stable operation of the system and meet user needs, and realize energy management and optimization of the microgrid.

[0172] Another embodiment of the present invention provides a distributed power supply safety management device, which is used to execute the above method, and the device includes: The energy demand monitoring module is used to collect the energy production and consumption data of each distributed power source, energy storage system, key load node, and connection point between the microgrid and the main grid in real time through intelligent sensors, and monitor the energy demand of each distributed power source in the microgrid in real time through the collected energy production and consumption data; The module for determining the optimal feature set and prediction model is used to determine the features that have a significant impact on the energy demand of each distributed power source according to business requirements and data characteristics, and to determine the optimal feature set and prediction model for each distributed power source according to the features that have a significant impact; An energy demand prediction module is used to predict the future energy demand in the microgrid of each distributed power source by determining the optimal feature set and prediction model for each distributed power source; The operation adjustment module is used to adjust the operation status and output of each distributed power source by using a multi-agent system after predicting the energy demand of multiple distributed power sources, so as to achieve coordinated operation of the entire microgrid. Another embodiment of the present invention provides a distributed power supply safety management system, including: a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the distributed power supply safety management method.

[0173] Another embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which implements the safe management method of distributed power sources when executed by a processor.

[0174] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0175] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0176] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0177] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for safe management of distributed power sources, characterized in that: The steps include: The energy production and consumption data of each distributed power source, energy storage system, key load node, and connection point between the microgrid and the main grid are collected in real time through smart sensors. The energy demand of each distributed power source in the microgrid is monitored in real time through the collected energy production and consumption data. According to business needs and data characteristics, determine the features that have a significant impact on the energy demand of each distributed power source, and determine the optimal feature set and prediction model for each distributed power source based on the features that have a significant impact; Predicting the future energy demand in the microgrid of each distributed power source by determining the optimal feature set and prediction model for each distributed power source; After predicting the energy demand of multiple distributed power sources, the multi-agent system is used to adjust the operating status and output of each distributed power source to achieve coordinated operation of the entire microgrid.

2. The safety management method for distributed power sources according to claim 1, characterized in that: The method of determining the features that have a significant impact on the energy demand of each distributed power source according to the business requirements and data characteristics, and determining the optimal feature set and prediction model for each distributed power source according to the features that have a significant impact, specifically includes: Conduct correlation analysis on the energy production and consumption data of each distributed power source in the microgrid to identify candidate features that have a significant impact on the energy demand of each distributed power source; The obtained candidate features are evaluated for feature importance, and the candidate features whose importance scores exceed the importance threshold are obtained as the important feature set; The important feature sets obtained for each distributed power source are cross-validated to determine the optimal feature set and prediction model in the important feature sets of each distributed power source.

3. The safety management method for distributed power sources according to claim 2, characterized in that: The correlation analysis of the energy production and consumption data of each distributed power source in the microgrid is performed to determine the candidate features that have a significant impact on the energy demand of each distributed power source, specifically including: The collected energy production and consumption data are initially screened according to the purpose of the analysis to remove features that are obviously irrelevant to the energy demand in the distributed generation; Using statistical methods to calculate the energy correlations between various residual characteristics and energy demand, and obtaining a set of energy correlation coefficients; A feature whose energy correlation coefficient is higher than a correlation threshold is selected from the energy correlation coefficient set as a candidate feature.

4. The safety management method for distributed power sources according to claim 2, characterized in that: The step of evaluating the importance of the obtained candidate features and obtaining the candidate features whose importance scores exceed the importance threshold as the important feature set specifically includes: Use the machine learning model for preliminary training to evaluate the importance of each candidate feature and obtain the importance score of each candidate feature; Sort the candidate features according to their importance scores given by the machine learning model; Among the ranked candidate features, features whose importance scores exceed the importance threshold are selected as the important feature set.

5. The safety management method for distributed power sources according to claim 2, characterized in that: The cross-validation of the important feature set obtained for each distributed power source to determine the optimal feature set and prediction model in the important feature set of each distributed power source specifically includes: Generate multiple candidate feature sets using different feature combinations; Select multiple machine learning models, and for each candidate evaluation feature set, train on each machine learning model; Use K-fold cross validation to evaluate the performance of each candidate evaluation feature set on each machine learning model. Divide the data set into K mutually exclusive subsets of similar size, train the model with K-1 subsets each time, and test the model on the remaining subset. Repeat this process K times, selecting a different subset as the test set each time, and calculate the average of all test results as the performance indicator of the candidate evaluation feature set and machine learning model. Use performance indicators to evaluate each feature set and model combination, compare the performance of different feature set and model combinations, and select the best performing feature set and model; The performance of the selected feature set and model is verified on an independent validation set or test set, and the optimal feature set for each distributed power source and the subsequent prediction model to be used are finally determined.

6. The safety management method for distributed power sources according to claim 1, characterized in that: After predicting the energy demand of multiple distributed power sources, the multi-agent system is used to adjust the operating status and output of each distributed power source to achieve coordinated operation of the entire microgrid, specifically including: Each distributed power source and key equipment in the microgrid is regarded as an agent. Each agent has autonomous decision-making and communication capabilities. Agents exchange information through communication protocols and share energy supply and demand and operation data. Based on the local information of each distributed power source and the global optimization goal, each agent autonomously adjusts its operating status and output to achieve coordinated operation of all distributed power sources. A central controller is established to collect the local optimal operating status and output of each distributed power source and the optimal operating mode of each distributed power source and perform centralized processing. According to the predicted energy demand, the central controller formulates a global optimization strategy and sends it to each distributed power source for execution. The distributed power source adjusts its operating status and output according to the received instructions to meet the system requirements and optimization goals.

7. A safety management device for distributed power supply, characterized in that: include: The energy demand monitoring module is used to collect the energy production and consumption data of each distributed power source, energy storage system, key load node, and connection point between the microgrid and the main grid in real time through intelligent sensors, and monitor the energy demand of each distributed power source in the microgrid in real time through the collected energy production and consumption data; The module for determining the optimal feature set and prediction model is used to determine the features that have a significant impact on the energy demand of each distributed power source according to business requirements and data characteristics, and to determine the optimal feature set and prediction model for each distributed power source according to the features that have a significant impact; An energy demand prediction module is used to predict the future energy demand in the microgrid of each distributed power source by determining the optimal feature set and prediction model for each distributed power source; The operation adjustment module is used to adjust the operation status and output of each distributed power source by using a multi-agent system after predicting the energy demand of multiple distributed power sources, so as to achieve coordinated operation of the entire microgrid.

8. The safety management device for distributed power source according to claim 7, characterized in that: The optimal feature set and prediction model determination module is specifically used for: Conduct correlation analysis on the energy production and consumption data of each distributed power source in the microgrid to identify candidate features that have a significant impact on the energy demand of each distributed power source; The obtained candidate features are evaluated for feature importance, and the candidate features whose importance scores exceed the importance threshold are obtained as the important feature set; The important feature sets obtained for each distributed power source are cross-validated to determine the optimal feature set and prediction model in the important feature sets of each distributed power source.

9. A safety management system for distributed power sources, characterized in that: include: A computer readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the distributed power supply safety management method according to any one of claims 1 to 6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the distributed power supply safety management method according to any one of claims 1 to 6.

Images