Flexible resource allocation and scheduling system based on edge intelligence

Abstract

The invention relates to a flexible resource allocation and scheduling system based on edge intelligence. The flexible resource allocation and scheduling system comprises an edge layer, a communication layer and a cloud layer, the edge layer comprises an intelligent electric meter, a sensor network, a local calculation unit and a micro-grid control center; the communication layer adopts gateway equipment as a data aggregation point and manages communication between the edge layer and the cloud layer; the cloud layer comprises a database, a big data analysis module, an intelligent algorithm module and a main control center; the system provided by the invention not only improves the resource utilization rate, but also enhances the ability to deal with dynamic requirements and uncertainty.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to a flexible resource allocation and scheduling system based on edge intelligence. Background Art

[0002] In modern power systems, efficient resource allocation and flexible scheduling are one of the key challenges. With the continuous growth of electricity demand and the widespread use of renewable energy, traditional power systems are facing significant pressure. This complexity requires the system to have higher flexibility and response speed to ensure the reliability and stability of power supply.

[0003] As an emerging technology, edge intelligence provides an innovative solution for power systems by sinking computing and data processing capabilities to locations close to data sources. It can achieve real-time data processing and rapid response, thereby improving the overall efficiency and reliability of the system. Combined with the powerful processing capabilities of cloud computing, this architecture can meet both real-time requirements and large-scale data analysis needs. Summary of the invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide a flexible resource allocation and scheduling system based on edge intelligence, which not only improves resource utilization, but also enhances the ability to cope with dynamic demands and uncertainties.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A flexible resource allocation and scheduling system based on edge intelligence, comprising an edge layer, a communication layer and a cloud layer; the edge layer comprises a smart meter and a sensor network, a local computing unit and a microgrid control center; the communication layer uses a gateway device as a data aggregation point to manage the communication between the edge layer and the cloud layer; the cloud layer comprises a database, a big data analysis module, an intelligent algorithm module and a main control center;

[0007] The smart meter and sensor network are connected to the local computing unit, and data is transmitted in real time through wired or wireless communication. The local computing unit pre-processes and preliminarily analyzes the acquired data; the local computing unit is integrated with the distributed energy resource controller to directly perform real-time control and data exchange; the microgrid control center collects data from the smart meter, sensor network and distributed energy resource controller, and uploads the processing results to the communication layer;

[0008] The database obtains data from the communication layer and uses Apache Hadoop for storage; the big data analysis module extracts data from the data warehouse for analysis; the intelligent algorithm module provides prediction and optimization services, and the results are fed back to the main control center; the main control center formulates global strategies based on the algorithm results and sends them to the edge devices through the communication layer.

[0009] Furthermore, the local computing unit preprocesses and preliminarily analyzes the acquired data, as follows:

[0010] Clean the acquired sensor data and process missing values ​​and outliers;

[0011] Use a sliding average or low-pass filter to smooth the data and remove short-term fluctuations and noise;

[0012] Based on the deployed lightweight decision tree model, the real-time preprocessed data is input into the model to obtain the prediction results, and the prediction results are used to predict short-term electricity demand.

[0013] Furthermore, the lightweight decision tree model is constructed as follows:

[0014] The key features selected from the historical sensor and meter data, each feature corresponding to an input variable, are standardized, normalized and missing value processed to construct the training data set

[0015] The lightweight decision tree model includes multiple independently trained decision trees, each tree uses a different feature subset and data subset, each non-leaf node of the decision tree stores a decision, that is, a feature and its threshold, which is used to divide the subset into two parts, and the selection criterion is based on information gain; the terminal node stores the predicted value of the target variable;

[0016] The information gain calculation formula is as follows:

[0017]

[0018] Among them, H(S) is entropy; IG(S, A) is information gain, which means the amount of information uncertainty reduction after selecting feature A to split the data set S; Values(A) is the set of possible values ​​of feature A; S v is a subset;

[0019]

[0020] Among them, c is the number of different categories in the data set, p i is the probability of belonging to the i-th category;

[0021] The maximum depth of each tree is limited to less than the preset number of levels; for further splitting of each node, the minimum number of samples required is set;

[0022] Train a lightweight decision tree model based on the training set data.

[0023] Furthermore, the local computing unit and the distributed energy resource controller adopt the standard communication protocol Modbus to establish a stable communication link and realize seamless data exchange; the local computing unit uses the optimized lightweight decision tree model for real-time inference and obtains the prediction results based on the current data input; based on the prediction results, the control parameters of the distributed energy resource controller DER are dynamically adjusted to meet the load demand and optimize energy utilization.

[0024] Furthermore, the big data analysis module uses the distributed computing capabilities provided by Hadoop / Spark to perform large-scale data storage and batch processing tasks; and uses the database to build multidimensional data sets and OLAP models to support fast query and analysis; and uses Apache Kafka for real-time data stream collection and processing to support dynamic analysis and fast response.

[0025] Furthermore, the intelligent algorithm module is equipped with a long-term electricity demand forecasting model built based on LSTM to predict load demand, and integrates genetic algorithms to provide a global optimization solution for energy distribution.

[0026] Furthermore, the genetic algorithm provides a global optimization solution for energy distribution, as follows:

[0027] Set the objective function f(x) to minimize operating costs, maximize user satisfaction, or optimize grid stability; constraints include generation capacity, policies and regulations, and user demand;

[0028] f(x)=w1·C(x)+w2·(1-S(x))+w3·(1-R(x));

[0029] Among them, w 1, w2, w3 are weights, C(x) is the operating cost function; S(x) is the user satisfaction function; R(x) is the grid stability function;

[0030] The representation of individuals is in the form of real number coding, and the initial solution group is randomly generated to ensure the diversity of solutions;

[0031] The fitness function is defined according to the objective function to evaluate the quality of each individual:

[0032]

[0033] Among them, ∈ is a constant;

[0034] using methods such as roulette wheel selection to select individuals from a population for reproduction;

[0035] The single-point crossover method is used to generate new individuals;

[0036] Introduce random mutations to increase population diversity and avoid local optimality;

[0037] The rate of change of the fitness function is used to determine whether convergence has been achieved. Convergence is achieved when the maximum number of generations is reached or the change in the objective function value is lower than the set threshold.

[0038] A computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the method steps described above.

[0039] The present invention has the following beneficial effects:

[0040] 1. The present invention not only improves resource utilization, but also enhances the ability to cope with dynamic demands and uncertainties. By deploying computing power at the edge layer, real-time analysis and processing of data can be achieved, reducing dependence on cloud processing, reducing latency, alleviating the pressure of cloud data processing, and improving the response speed and efficiency of the overall system;

[0041] 2. The system structure of the present invention is clear and highly modularized, and it is easy to expand and adapt to new technologies and changes in demand. For example, new sensor devices or algorithm modules can be easily integrated; through a flexible resource allocation mechanism, the system can dynamically adjust energy use strategies according to real-time demand and improve resource utilization;

[0042] 3. The cloud-based big data analysis module of the present invention can extract valuable information from historical and real-time data to provide support for optimized decision-making. Through the prediction and optimization functions of the intelligent algorithm module, the system can identify and solve potential problems in advance, effectively improving the decision-making level of energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0044] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0045] refer to Figure 1 The present invention provides a flexible resource allocation and scheduling system based on edge intelligence, including an edge layer, a communication layer and a cloud layer; the edge layer includes a smart meter and a sensor network, a local computing unit and a microgrid control center; the communication layer uses a gateway device as a data aggregation point to manage the communication between the edge layer and the cloud layer; the cloud layer includes a database, a big data analysis module, an intelligent algorithm module and a main control center;

[0046] The smart meter and sensor network are connected to the local computing unit, and data is transmitted in real time through wired or wireless communication (such as Zigbee, Wi-Fi). The local computing unit pre-processes and preliminarily analyzes the acquired data; the local computing unit is integrated with the distributed energy resource controller to directly perform real-time control and data exchange; the microgrid control center collects data from the smart meter, sensor network and distributed energy resource controller, and uploads the processing results to the communication layer;

[0047] The database obtains data from the communication layer and uses Apache Hadoop for storage; the big data analysis module extracts data from the data warehouse for analysis; the intelligent algorithm module provides prediction and optimization services, and the results are fed back to the main control center; the main control center formulates global strategies based on the algorithm results and sends them to the edge devices through the communication layer.

[0048] In this embodiment, the local computing unit performs preprocessing and preliminary analysis on the acquired data, specifically as follows:

[0049] Clean the acquired sensor data and process missing values ​​and outliers (e.g., fill in with interpolation and remove outliers);

[0050] Use a sliding average or low-pass filter to smooth the data and remove short-term fluctuations and noise.

[0051] Based on the deployed lightweight decision tree model, the real-time preprocessed data is input into the model to obtain the prediction results, and the prediction results are used to predict short-term electricity demand.

[0052] In this embodiment, the lightweight decision tree model is constructed as follows:

[0053] The key features selected from the historical sensor and meter data, each feature corresponding to an input variable, are standardized, normalized and missing value processed to construct the training data set

[0054] The lightweight decision tree model includes multiple independently trained decision trees, each tree uses a different feature subset and data subset, each non-leaf node of the decision tree stores a decision, that is, a feature and its threshold, which is used to divide the subset into two parts, and the selection criterion is based on information gain; the terminal node stores the predicted value of the target variable;

[0055] The information gain calculation formula is as follows:

[0056]

[0057] Among them, H(S) is entropy; IG(S, A) is information gain, which means the amount of information uncertainty reduction after selecting feature A to split the data set S; Values(A) is the set of possible values ​​of feature A; S v is a subset;

[0058]

[0059] Among them, c is the number of different categories in the data set, p i is the probability of belonging to the i-th category;

[0060] The maximum depth of each tree is limited to less than the preset number of levels; for further splitting of each node, the minimum number of samples required is set;

[0061] Train a lightweight decision tree model based on the training set data.

[0062] In this embodiment, the local computing unit and the distributed energy resource controller use the standard communication protocol Modbus to establish a stable communication link and realize seamless data exchange; the local computing unit uses the optimized lightweight decision tree model for real-time inference and obtains prediction results based on the current data input; based on the prediction results, the control parameters (such as output power and voltage level) of the distributed energy resource controller DER are dynamically adjusted to meet load requirements and optimize energy utilization.

[0063] In this embodiment, the big data analysis module uses the distributed computing capabilities provided by Hadoop / Spark to perform large-scale data storage and batch processing tasks; and uses the database to build multidimensional data sets and OLAP models to support fast query and analysis; and uses Apache Kafka for real-time data stream collection and processing to support dynamic analysis and fast response.

[0064] In this embodiment, the intelligent algorithm module is provided with a long-term electricity demand forecasting model based on LSTM to predict load demand, and integrates a genetic algorithm to provide a global optimization solution for energy distribution.

[0065] In this embodiment, the genetic algorithm provides a global optimization solution for energy allocation, as follows:

[0066] Set the objective function f(x) to minimize operating costs, maximize user satisfaction, or optimize grid stability; constraints include generation capacity, policies and regulations, and user demand;

[0067] f(x)=w1·C(x)+w2·(1-S(x))+w3·(1-R(x));

[0068] Among them, w 1,w2, w3 are weights, C(x) is the operating cost function; S(x) is the user satisfaction function; R(x) is the grid stability function;

[0069] The representation of individuals is in the form of real number coding, and the initial solution group is randomly generated to ensure the diversity of solutions;

[0070] The fitness function is defined according to the objective function to evaluate the quality of each individual:

[0071]

[0072] Among them, ∈ is a constant;

[0073] using methods such as roulette wheel selection to select individuals from a population for reproduction;

[0074] The single-point crossover method is used to generate new individuals;

[0075] Introduce random mutations to increase population diversity and avoid local optimality;

[0076] The rate of change of the fitness function is used to determine whether convergence has been achieved. Convergence is achieved when the maximum number of generations is reached or the change in the objective function value is lower than the set threshold.

[0077] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take 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 code.

[0078] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0079] 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.

[0080] 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.

[0081] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A flexible resource allocation and scheduling system based on edge intelligence, characterized in that: It includes edge layer, communication layer and cloud layer; the edge layer includes smart meters and sensor networks, local computing units and microgrid control center; the communication layer uses gateway devices as data aggregation points to manage the communication between the edge layer and the cloud layer; the cloud layer includes database, big data analysis module, intelligent algorithm module and main control center; The smart meter and sensor network are connected to the local computing unit, and data is transmitted in real time through wired or wireless communication. The local computing unit pre-processes and preliminarily analyzes the acquired data; the local computing unit is integrated with the distributed energy resource controller to directly perform real-time control and data exchange; the microgrid control center collects data from the smart meter, sensor network and distributed energy resource controller, and uploads the processing results to the communication layer; The database acquires data from the communication layer and uses Apache Hadoop for storage; the big data analysis module extracts data from the data warehouse for analysis; The intelligent algorithm module provides prediction and optimization services, and the results are fed back to the main control center; the main control center formulates global strategies based on the algorithm results and sends them to edge devices through the communication layer.

2. According to the flexible resource allocation and scheduling system based on edge intelligence according to claim 1, it is characterized in that: The local computing unit performs preprocessing and preliminary analysis on the acquired data, as follows: Clean the acquired sensor data and process missing values ​​and outliers; Use a sliding average or low-pass filter to smooth the data and remove short-term fluctuations and noise; Based on the deployed lightweight decision tree model, the real-time preprocessed data is input into the model to obtain the prediction results, and the prediction results are used to predict short-term electricity demand.

3. According to claim 2, a flexible resource allocation and scheduling system based on edge intelligence is characterized in that: The lightweight decision tree model is constructed as follows: The key features selected from the historical sensor and meter data, each feature corresponding to an input variable, are standardized, normalized and missing value processed to construct the training data set The lightweight decision tree model includes multiple independently trained decision trees, each tree uses a different feature subset and data subset, each non-leaf node of the decision tree stores a decision, that is, a feature and its threshold, which is used to divide the subset into two parts, and the selection criterion is based on information gain; the terminal node stores the predicted value of the target variable; The information gain calculation formula is as follows: Among them, H(S) is entropy; IG(S, A) is information gain, which means the amount of information uncertainty reduction after selecting feature A to split the data set S; Values(A) is the set of possible values ​​of feature A; S v is a subset; Among them, c is the number of different categories in the data set, p i is the probability of belonging to the i-th category; The maximum depth of each tree is limited to less than the preset number of levels; for further splitting of each node, the minimum number of samples required is set; Train a lightweight decision tree model based on the training set data.

4. According to claim 1, a flexible resource allocation and scheduling system based on edge intelligence is characterized in that: The local computing unit and the distributed energy resource controller use the standard communication protocol Modbus to establish a stable communication link and realize seamless data exchange; the local computing unit uses the optimized lightweight decision tree model for real-time inference and obtains prediction results based on the current data input; the control parameters of the distributed energy resource controller DER are dynamically adjusted based on the prediction results to meet load demand and optimize energy utilization.

5. The flexible resource allocation and scheduling system based on edge intelligence according to claim 1, characterized in that: The big data analysis module uses the distributed computing capabilities provided by Hadoop / Spark to perform large-scale data storage and batch processing tasks; The database is used to build multidimensional data sets and OLAP models to support fast query and analysis; Apache Kafka is used for real-time data stream collection and processing to support dynamic analysis and fast response.

6. The flexible resource allocation and scheduling system based on edge intelligence according to claim 1, characterized in that: The intelligent algorithm module is equipped with a long-term electricity demand forecasting model based on LSTM to predict load demand, and integrates genetic algorithm to provide a global optimization solution for energy distribution.

7. The flexible resource allocation and scheduling system based on edge intelligence according to claim 6, characterized in that: The genetic algorithm provides a global optimization solution for energy distribution as follows: Set the objective function f(x) to minimize operating costs, maximize user satisfaction, or optimize grid stability; constraints include generation capacity, policies and regulations, and user demand; f(x)=w1·C(x)+w2·(1-S(x))+w3·(1-R(x)); Among them, w1, w2, w3 are weights, C(x) is the operating cost function; S(x) is the user satisfaction function; R(x) is the grid stability function; The representation of individuals is in the form of real number coding, and the initial solution group is randomly generated to ensure the diversity of solutions; The fitness function is defined according to the objective function to evaluate the quality of each individual: Among them, ∈ is a constant; using methods such as roulette wheel selection to select individuals from a population for reproduction; The single-point crossover method is used to generate new individuals; Introduce random mutations to increase population diversity and avoid local optimality; The rate of change of the fitness function is used to determine whether convergence has been achieved. Convergence is achieved when the maximum number of generations is reached or the change in the objective function value is lower than the set threshold.

8. A computer storage medium, characterized in that: The computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method steps according to any one of claims 1 to 7.

Images