Search

CN-122000867-A - Micro-grid energy management optimization prediction method

CN122000867ACN 122000867 ACN122000867 ACN 122000867ACN-122000867-A

Abstract

The invention relates to the technical field of micro-grid management optimization prediction and discloses a micro-grid energy management optimization prediction method, which comprises the following steps of constructing a data acquisition network, synchronously acquiring basic meteorological data, enhanced meteorological data and micro-grid equipment operation data, and generating a multi-source feature vector through edge computing gateway fusion; and (3) constructing a physical enhancement type prediction model, taking a transducer as a core, embedding a dust accumulation influence index and a temperature coefficient correction factor, performing interval prediction on the photovoltaic output and load demand of the micro-grid, and outputting a confidence interval result. By introducing basic weather, enhanced weather and multisource fusion of micro-grid equipment operation data into the prediction model and combining the capability of transducer time sequence feature extraction, high-precision prediction of photovoltaic output and load demand is realized.

Inventors

  • YUAN NAN
  • LI YUE
  • MA YIFAN
  • TAN MINGHUI

Assignees

  • 苏州易助能源管理有限公司

Dates

Publication Date
20260508
Application Date
20251231

Claims (10)

  1. 1. The micro-grid energy management optimization prediction method is characterized by comprising the following steps of: Constructing a data acquisition network, synchronously acquiring basic meteorological data, enhanced meteorological data and micro-grid equipment operation data, and generating a multi-source feature vector through edge computing gateway fusion; Building a physical enhancement type prediction model, taking a transducer as a core, embedding a dust accumulation influence index and a temperature coefficient correction factor, performing interval prediction on photovoltaic output and load demand of a micro-grid, and outputting a confidence interval result; based on the multi-source feature vector and the prediction result, a digital twin extreme scene previewing unit is called, and energy storage charge and discharge thresholds and load adjustment priorities under different working conditions are simulated to generate an initial scheduling strategy; correcting the initial scheduling strategy, generating a final micro-grid energy scheduling instruction and transmitting the final micro-grid energy scheduling instruction to edge execution equipment; And collecting feedback data after the dispatching instruction is executed, inputting the feedback data into a parameter updating unit, automatically calibrating the characteristic weight of the prediction model, and iteratively optimizing the parameters of the digital twin scene library.
  2. 2. The method for optimizing and predicting energy management of a micro-grid according to claim 1, wherein the construction flow of the data acquisition network comprises the following steps: The method comprises the steps that a sensing terminal is distributed in a micro-grid area, basic meteorological data are collected through temperature and humidity sensors, wind speed sensors and precipitation sensors respectively, enhanced meteorological data are collected through solar radiation sensors and cloud cover collectors, and running data of micro-grid equipment are collected through current sensors, voltage sensors and power sensors; and providing partition edge acquisition nodes, receiving first-level data through wired and wireless communication, and performing duplication removal, outlier rejection and format standardization pretreatment to form structured data.
  3. 3. The method for optimizing and predicting energy management of a micro-grid according to claim 1, wherein the fusion generation flow of the multi-source eigenvectors comprises: the edge computing gateway receives basic meteorological data, enhanced meteorological data and micro-grid equipment operation data from an edge convergence layer; Performing time stamp calibration on data from different sources by taking a gateway clock as a reference, so as to ensure the consistency of time sequences; abnormal values are removed through a 3 sigma principle, and the missing data is supplemented by linear interpolation to complete format standardization; Extracting average value and change rate statistical characteristics from meteorological data, and extracting power fluctuation and state code operation characteristics from equipment data; And performing dimension matching and normalization processing on the extracted multi-dimensional features, and combining to form a multi-source feature vector containing meteorological and equipment operation features.
  4. 4. The microgrid energy management optimization prediction method according to claim 1, wherein the building process of the physical enhancement prediction model comprises the following steps: processing the time sequence characteristics of the photovoltaic output and load demands by taking a transducer model as a core, and optimizing model prediction deviation; Calculating an ash accumulation influence index and a temperature coefficient correction factor, and embedding the ash accumulation influence index and the temperature coefficient correction factor as additional features into a feature input layer of a transducer for fusion of a physical mechanism and a data model; Introducing a Bayesian probability optimization method in physical enhancement type prediction model training, fitting prediction error distribution based on historical data, and setting a confidence level; And predicting the photovoltaic output and the load demand of the micro-grid, and outputting a corresponding prediction interval and a confidence interval result.
  5. 5. The method for optimizing and predicting energy management of a micro-grid according to claim 4, wherein the calculation flow of the soot impact index and the temperature coefficient correction factor comprises: Collecting real-time monitoring data of the ash thickness on the surface area of the photovoltaic panel, and determining the current light transmittance attenuation coefficient by combining a corresponding relation table of the ash thickness and the light transmittance attenuation coefficient calibrated in a laboratory; Obtaining an accumulated ash influence index based on the weighted calculation of the attenuation coefficient and the accumulated ash duration; Obtaining an actual measurement value of the ambient temperature and rated working temperatures of the photovoltaic module and the load equipment, and calculating a temperature deviation value; According to the temperature-efficiency correction curve of the equipment delivery, the efficiency correction coefficient under the corresponding deviation value is obtained; And normalizing the correction coefficient to obtain the temperature coefficient correction factor.
  6. 6. The method of claim 1, wherein the workflow of the digital twin extreme scene pre-modeling unit comprises: Importing real-time working condition data in the physical topology, equipment parameters and multisource feature vectors of the micro-grid, and constructing a digital twin body which is 1:1 mapped with the physical micro-grid; recording typical extreme working conditions of a micro-grid, including extreme meteorological scenes, equipment fault scenes and load abrupt change scenes, and defining triggering conditions and parameter boundaries of the scenes; Taking a prediction result as input, calling an energy flow calculation model built in the unit, and simulating energy storage charging and discharging thresholds and load adjustment priorities under different terminal scenes; and through simulation result analysis, determining an energy storage operation threshold and a load regulation sequence which meet the supply and demand balance of the micro-grid in each extreme scene, and providing data support for generating an initial energy scheduling strategy.
  7. 7. The method for optimizing and predicting energy management of a micro-grid according to claim 1, wherein the generating process of the initial energy scheduling strategy comprises: importing the multisource feature vector and the photovoltaic output and load demand prediction result into a digital twin extreme scene previewing unit, and screening out an extreme scene matched with the current working condition from a unit scene library; the digital twin extreme scene previewing unit simulates the lowest discharge threshold and the highest charge threshold of energy storage under different scenes according to the current SOC of the energy storage in the multi-source feature vector and the rated charge-discharge multiplying power of the equipment by combining the surplus and gap of the predicted photovoltaic output; According to the load type in the multisource feature vector and the predicted load demand change trend, according to an important load priority supply maintaining principle, determining a load adjustment priority through a built-in grading rule of a digital twin extreme scene previewing unit; Invoking an energy flow model built in a digital twin extreme scene previewing unit, combining an energy storage charge-discharge threshold, a load adjustment priority and predicted photovoltaic output and load demands, calculating the supply-demand difference of the micro-grid in different scenes, and judging whether the balance condition of output more than or equal to the demand and equipment loss is satisfied; aiming at a supply-demand balance result, determining an energy storage operation instruction and a load adjustment scheme, and combining to form a preliminary scheduling strategy; and previewing an execution result of the preliminary strategy under the extreme scene by a digital twin extreme scene previewing unit, and if equipment overrun or supply and demand unbalance exists, finely adjusting a threshold value and a priority, so as to finally generate an initial energy scheduling strategy.
  8. 8. The method for optimizing and predicting energy management of a micro-grid according to claim 1, wherein the modification procedure of the initial scheduling strategy comprises: Invoking the multisource feature vector and the photovoltaic output and load demand prediction result to serve as a basis for strategy correction; determining a safe operation boundary of core equipment based on equipment delivery standards and micro-grid operation specifications, and extracting an equipment safety constraint threshold; comparing an energy storage charging and discharging instruction, a photovoltaic output distribution scheme and a load regulation plan in an initial scheduling strategy with the safety constraint threshold value to judge whether collision exists; Calculating the supply and demand difference of the micro-grid after the initial strategy is executed by combining the predicted photovoltaic output interval and the load demand interval, judging whether the balance condition that the photovoltaic output, the energy storage and the discharge are more than or equal to the load demand and the equipment operation loss is met or not, and marking the balance condition as a correction item if a supply and demand gap or surplus excess exists; aiming at a safety conflict item, correcting according to a safety threshold of equipment preferentially, and aiming at a supply-demand unbalanced item, if a gap exists, preferentially reducing non-core load, and if surplus is excessive, adjusting an energy storage charging instruction to an SOC safety upper limit; And integrating the corrected energy storage operation instruction, the photovoltaic output control scheme and the load regulation plan to form a final micro-grid energy dispatching instruction, and transmitting the final micro-grid energy dispatching instruction to the edge execution equipment through the industrial Ethernet.
  9. 9. The method of claim 1, wherein the workflow of the parameter updating unit comprises: Receiving micro-grid operation data fed back after the execution of the scheduling instruction, and extracting core related parameters from the micro-grid operation data, wherein the core related parameters comprise an actual photovoltaic output value, an actual load demand value, equipment real-time operation parameters and environment actual measurement data, so as to form a data set for unit processing; Comparing the actual value in the data set with the interval prediction result output by the prediction model, calculating a prediction deviation value, analyzing the contribution degree of the input characteristic of the prediction model to the deviation through a SHAP value characteristic importance algorithm, and marking high-influence characteristics with the contribution degree exceeding a preset threshold value; Based on the feature contribution degree calculation result, reversely adjusting the feature weight of the transform prediction model by adopting a gradient descent method; Comparing the equipment operation parameters and the environment actual measurement data in the data set with preset parameters of corresponding scenes in the digital twin scene library, calculating parameter deviation rate, and screening abnormal scene parameters with deviation rate exceeding a set threshold; For abnormal scene parameters, a least square method is adopted to fit the mapping relation between feedback data and preset parameters of a scene library, a parameter correction coefficient is generated, the corresponding parameters in the scene library are updated by the correction coefficient, new working conditions which are not covered in the feedback data are used as new scene entry scene libraries, and scene dimensions are supplemented; And outputting the characteristic weight of the calibrated prediction model and the optimized digital twin scene library parameters to the next round of micro-grid prediction and scheduling process, recalculating the prediction deviation rate and the scene parameter deviation rate after new round of scheduling feedback data is acquired, completing the round of work if both the deviation rates fall within a set threshold value, and repeating the steps until the running accuracy requirement of the micro-grid is met if the deviation rates do not reach the standard.
  10. 10. A microgrid energy management optimization prediction system for use in a microgrid energy management optimization prediction method according to any one of claims 1-9, comprising: The data acquisition and fusion module is used for constructing a data acquisition network, synchronously acquiring basic meteorological data, enhanced meteorological data and micro-grid equipment operation data, and generating a multi-source feature vector through edge computing gateway fusion; The physical enhancement type prediction module is used for constructing a physical enhancement type prediction model, taking a transducer as a core, embedding an accumulated ash influence index and a temperature coefficient correction factor, carrying out interval prediction on the photovoltaic output and the load demand of the micro-grid, and outputting a confidence interval result; The digital twin-type predicting and initial strategy generating module is used for calling a digital twin-type extreme scene predicting unit based on the multi-source feature vector and a predicting result, simulating energy storage charging and discharging thresholds and load adjusting priorities under different working conditions and generating an initial scheduling strategy; The scheduling strategy correction and instruction issuing module corrects the initial scheduling strategy, generates a final micro-grid energy scheduling instruction and issues the final micro-grid energy scheduling instruction to the edge execution equipment; And the parameter closed loop updating module is used for collecting feedback data after the scheduling instruction is executed, inputting the feedback data into the parameter updating unit, automatically calibrating the characteristic weight of the prediction model and iteratively optimizing the digital twin scene library parameters.

Description

Micro-grid energy management optimization prediction method Technical Field The invention relates to the technical field of micro-grid management optimization prediction, in particular to a micro-grid energy management optimization prediction method. Background The micro-grid energy management optimization technology belongs to the field of intelligent power grids and distributed energy management, and aims to realize reasonable configuration of unit start-stop, energy storage charge-discharge and demand response strategies by optimizing and dispatching operation states, load demands and energy storage systems of various energy sources in the micro-grid under the conditions of meeting electric energy balance, safety constraint, electric energy quality, standby capacity and the like. According to the technology, by constructing an optimization model and introducing constraint conditions and optimization targets such as operation cost, carbon emission and energy storage life, the micro-grid is dynamically controlled by combining a prediction result, so that the energy utilization efficiency is improved, the operation cost is reduced, and the safe and reliable operation of the micro-grid is ensured. With the wide application of distributed renewable energy sources in micro-grids, how to realize high-precision prediction and high-reliability scheduling under complex meteorological conditions, equipment operation fluctuation and load demand change has become a key problem of micro-grid energy management. In the prior art, prediction is carried out only by relying on a single data source, multi-source information of weather, environment and equipment operation is not fully fused, so that the prediction characteristics are incomplete, meanwhile, a conventional prediction model is mostly only driven by data, physical working condition factors such as dust accumulation, temperature and the like of a photovoltaic module are ignored, the real operation state is difficult to reflect, and systematic deviation exists in a prediction result. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a micro-grid energy management optimization prediction method, which solves the problem that the prediction characteristic is incomplete due to the fact that the prior art relies on a single data source for prediction. In order to achieve the purpose, the invention is realized by the following technical scheme that the micro-grid energy management optimizing and predicting method comprises the following steps: Constructing a data acquisition network, synchronously acquiring basic meteorological data, enhanced meteorological data and micro-grid equipment operation data, and generating a multi-source feature vector through edge computing gateway fusion; Building a physical enhancement type prediction model, taking a transducer as a core, embedding a dust accumulation influence index and a temperature coefficient correction factor, performing interval prediction on photovoltaic output and load demand of a micro-grid, and outputting a confidence interval result; based on the multi-source feature vector and the prediction result, a digital twin extreme scene previewing unit is called, and energy storage charge and discharge thresholds and load adjustment priorities under different working conditions are simulated to generate an initial scheduling strategy; correcting the initial scheduling strategy, generating a final micro-grid energy scheduling instruction and transmitting the final micro-grid energy scheduling instruction to edge execution equipment; And collecting feedback data after the dispatching instruction is executed, inputting the feedback data into a parameter updating unit, automatically calibrating the characteristic weight of the prediction model, and iteratively optimizing the parameters of the digital twin scene library. Preferably, the construction process of the data acquisition network includes: The method comprises the steps that a sensing terminal is distributed in a micro-grid area, basic meteorological data are collected through temperature and humidity sensors, wind speed sensors and precipitation sensors respectively, enhanced meteorological data are collected through solar radiation sensors and cloud cover collectors, and running data of micro-grid equipment are collected through current sensors, voltage sensors and power sensors; and providing partition edge acquisition nodes, receiving first-level data through wired and wireless communication, and performing duplication removal, outlier rejection and format standardization pretreatment to form structured data. Preferably, the fusion generation flow of the multi-source feature vector includes: the edge computing gateway receives basic meteorological data, enhanced meteorological data and micro-grid equipment operation data from an edge convergence layer; Performing time stamp calibration on data from different sources by takin