CN-122022323-A - Distributed energy collaborative management system based on edge intelligence

CN122022323ACN 122022323 ACN122022323 ACN 122022323ACN-122022323-A

Abstract

The invention relates to the technical field of intelligent power grids and distributed energy management, in particular to an edge intelligent-based distributed energy collaborative management system. According to the scheme, a multi-scale weighted regression model is built, the multi-scale space-time characteristics of distributed energy sources are analyzed, the characteristics of an adaptive weight matrix accurate capturing area are built, and scale parameters are optimized by utilizing a corrected red pool information criterion, so that the wind abandoning and the light abandoning and the power supply shortage are obviously reduced; the method adopts an optimized self-adaptive differential evolution algorithm to carry out regional global optimization scheduling, dynamically adjusts scaling factors and cross probabilities based on a history memory bank, and realizes millisecond response by matching with a linear population reduction mechanism, thereby reducing the running cost of the system and improving the energy utilization efficiency.

Inventors

JIANG YAN
LIU PIN

Assignees

湖南信息学院

Dates

Publication Date: 20260512
Application Date: 20260128

Claims (4)

1. The distributed energy collaborative management system based on the edge intelligence is characterized by comprising an end side sensing layer, an edge intelligent node, a cloud collaborative layer, a secure communication layer and a visual interaction layer; the end-side sensing layer deploys a photovoltaic inverter, an energy storage BMS and a smart electric meter on a distributed power supply, an energy storage and load smart terminal, and is used for collecting running state data, preprocessing the collected space-time data, constructing an input feature matrix and accessing an edge smart node; The edge intelligent nodes are deployed at the network edge close to the end side sensing layer, an AI reasoning model is constructed to conduct multi-scale prediction on the input feature matrix, prediction error characteristics are obtained, and real-time power adjustment, fault isolation and anomaly detection are conducted according to the prediction error characteristics; The cloud collaborative layer performs regional global optimization scheduling by using an optimized self-adaptive differential evolution algorithm, updates an AI reasoning model, and transmits a global optimization scheduling result and the updated AI reasoning model to the edge intelligent node; the secure communication layer realizes bidirectional identity authentication and data encryption transmission between the end-side-cloud based on a zero trust architecture; The visual interaction layer is connected with a key interface of the system and a user, wherein the key interface comprises a local man-machine interaction interface, cloud Web, a mobile terminal platform, a digital twin visual module, a transaction scheduling signboard and audit.
2. The distributed energy collaborative management system based on edge intelligence of claim 1, wherein the edge intelligence node comprises an AI reasoning module, a local control module, and a federal learning client; The AI reasoning module uses a multi-scale weighted regression model to construct an AI reasoning model, analyzes multi-scale space-time characteristics of the distributed energy sources, extracts prediction error characteristics, and encrypts and uploads abstracts of the prediction error characteristics to a cloud coordination layer through a secure communication layer; The local control module generates a control instruction according to the results of the cloud cooperative layer and the AI reasoning model and transmits the control instruction to the end side equipment of the end side sensing layer; And the federal learning client uses a federal learning mechanism to encrypt and upload training gradients and feature summaries of the AI reasoning model to the cloud coordination layer to participate in the aggregation of the global AI reasoning model.
3. The distributed energy collaborative management system based on edge intelligence according to claim 2, wherein the AI reasoning module uses a multi-scale weighted regression model to construct an AI reasoning model, comprising the steps of: a1, constructing a multivariable response model, defining a multiscale geographic weighted regression model, and setting Interpretation variables, the total number of observation points is Response variable ; A2, initializing scale parameters of each variable, distributing independent scale parameters for each interpretation variable, and initializing each scale parameter by using a uniform sampling method; step A3, iterative fitting, wherein a component updating strategy is adopted to optimize an interpretation variable so as to obtain a predicted value; A4, optimizing all scale parameters in a cooperative manner, optimizing all scale parameters by using a genetic algorithm, and calculating to obtain optimal scale parameters by adopting a corrected red pool information criterion as fitness; A5, outputting a prediction result, and re-executing the steps A1 to A3 by using the optimal scale parameters to obtain a final prediction value and an optimized multi-scale weighted regression model; And step A6, edge side deployment and online updating, compressing the optimized multi-scale weighted regression model to obtain a local AI reasoning model, and directly calculating an input feature matrix constructed by the end side perception layer by using the local AI reasoning model to obtain the prediction error feature.
4. The distributed energy collaborative management system based on edge intelligence according to claim 2, wherein the cloud collaborative layer uses an optimized self-adaptive differential evolution algorithm to perform regional global optimization scheduling, updates an AI reasoning model, and specifically comprises the following steps: Step B1, initializing a population, wherein a cloud cooperative layer receives prediction error characteristics from edge intelligent nodes and a local AI reasoning model, and initializes a population based on a self-adaptive differential evolution algorithm; Step B2, evaluating the fitness, namely calculating a fitness value for each individual in the population to obtain the fitness value of each individual; step B3, a parameter updating mechanism is used for constructing a history memory, initializing the history memory, storing successfully used parameter combinations, and dynamically adjusting scaling factors and crossover probabilities in the self-adaptive differential evolution algorithm based on the parameter combinations in the history memory; step B4, generating a new solution, and performing mutation operation and crossover operation on each individual by using a scaling factor to obtain the new solution; Step B5, selecting operation, namely comparing the fitness of the individual and the new solution by using a greedy algorithm, if the fitness of the new solution is smaller than that of the compared individual, replacing the compared individual with the new solution, adding a parameter pair corresponding to the new solution into a history memory, and updating the history memory, otherwise, reserving the compared individual; and B6, performing iterative optimization, setting maximum iterative times and linear population reduction parameters, performing iterative steps B2 to B5, gradually reducing the population scale by using the linear population reduction parameters, performing iterative steps to the maximum iterative times to obtain an optimal solution, and issuing the optimal solution to each edge intelligent node.

Description

Distributed energy collaborative management system based on edge intelligence Technical Field The invention relates to the technical field of intelligent power grids and distributed energy management, in particular to a distributed energy collaborative management system based on edge intelligence. Background Along with the large-scale access of renewable energy sources such as photovoltaic, wind power and the like, the traditional centralized energy scheduling architecture faces challenges such as high response delay (> 200 ms), limited communication bandwidth, high fault diffusion risk and the like. Meanwhile, the energy system consists of a large number of scattered units, the management complexity is high, the response of the traditional centralized scheduling mode is slow, the cooperative efficiency is low, and the dynamic change of the energy system is difficult to adapt. Disclosure of Invention Aiming at the problems that the distributed energy is obviously affected by weather, the output volatility is high, and the precision of a traditional prediction method is insufficient, the scheme builds a multi-scale weighted regression model, and the characteristics of a region are accurately captured by analyzing the multi-scale space-time characteristics of the distributed energy, and the characteristics of a self-adaptive weight matrix are built, and the scale parameters are optimized by utilizing a corrected red pool information rule, so that the waste wind, the waste light and the power shortage are obviously reduced. The invention provides an edge intelligent-based distributed energy collaborative management system, which comprises an end side sensing layer, an edge intelligent node, a cloud collaborative layer, a secure communication layer and a visual interaction layer, wherein the end side sensing layer is connected with the edge intelligent node; the end-side sensing layer deploys a photovoltaic inverter, an energy storage BMS and a smart electric meter on a distributed power supply, an energy storage and load smart terminal, and is used for collecting running state data, preprocessing the collected space-time data, constructing an input feature matrix and accessing an edge smart node; The edge intelligent nodes are deployed at the network edge close to the end side sensing layer, an AI reasoning model is constructed to conduct multi-scale prediction on the input feature matrix, prediction error characteristics are obtained, and real-time power adjustment, fault isolation and anomaly detection are conducted according to the prediction error characteristics; The cloud collaborative layer performs regional global optimization scheduling by using an optimized self-adaptive differential evolution algorithm, updates an AI reasoning model, and transmits a global optimization scheduling result and the updated AI reasoning model to the edge intelligent node; the secure communication layer realizes bidirectional identity authentication and data encryption transmission between the end-side-cloud based on a zero trust architecture; The visual interaction layer is connected with a key interface of the system and a user, wherein the key interface comprises a local man-machine interaction interface, a cloud Web, a mobile terminal platform, a digital twin visual module, a transaction scheduling billboard and an audit center. Further, the edge intelligent node comprises an AI reasoning module, a local control module and a federal learning client; The AI reasoning module uses a multi-scale weighted regression model to construct an AI reasoning model, analyzes multi-scale space-time characteristics of the distributed energy sources, extracts prediction error characteristics, and encrypts and uploads abstracts of the prediction error characteristics to a cloud coordination layer through a secure communication layer; The local control module generates a control instruction according to the results of the cloud cooperative layer and the AI reasoning model and transmits the control instruction to the end side equipment of the end side sensing layer; And the federal learning client uses a federal learning mechanism to encrypt and upload training gradients and feature summaries of the AI reasoning model to the cloud coordination layer to participate in the aggregation of the global AI reasoning model. Further, the AI reasoning module uses a multi-scale weighted regression model to construct an AI reasoning model, which specifically comprises the following steps: a1, constructing a multivariable response model, defining a multiscale geographic weighted regression model, and setting Interpretation variables, the total number of observation points isResponse variableThe formula used is as follows: ; In the formula, In order to be an intercept term,Represent the firstThe local regression coefficients of the individual interpretation variables,Represent the firstThe spatial coordinates of the individual observation points,Is th