CN112053006A - Optimization time acceleration method and system for combined cooling, heating and power system based on transfer learning - Google Patents

Abstract

The invention discloses a method and system for accelerating the optimization time of a combined cooling, heating and power system based on migration learning. Denote the source domain and the target domain; cluster the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes; based on the maximum mean difference, the target domain sample class and the source domain The sample classes are compared to determine whether there is a matching source domain sample class data set; if so, based on the source domain sample class data set, the initial population of the genetic algorithm is determined by migration learning; if not, the initial population of the genetic algorithm is randomly generated; Optimization of combined cooling, heating and power system based on genetic algorithm. Based on the similar historical optimization problem, the invention determines the initial population of the genetic algorithm from the historical data through migration learning, and delimits the area where the optimal solution may exist, thereby realizing the acceleration of the optimization of the combined cooling, heating and power system.

Description

Optimization time acceleration method and system for combined cooling, heating and power system based on transfer learning

technical field

The invention belongs to the technical field of optimization of a combined cooling, heating and power supply system, and in particular relates to a method and a system for accelerating the optimization time of a combined cooling, heating and power supply system based on migration learning.

Background technique

The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

Combined Cooling Heating and Power System (CCHP) realizes the cascade utilization of energy while meeting the needs of users for various loads of cooling, heating and electricity through the optimal scheduling of energy. Energy solutions and economic and low-carbon benefits have become one of the important directions for the development of distributed energy. The well-developed CCHP system optimization method at this stage is day-ahead optimization, that is, based on the relevant data predicted the day before the work, such as user cooling load, heating load, electrical load and weather data to achieve the optimization goal. an optimization model. The model determines the optimal work plan of each main device in the new energy combined cooling, heating and power system in units of hours.

Most of the calculation methods used for optimization are genetic algorithms. There are three basic frameworks of genetic algorithms, namely coding, fitness function and initial population selection. However, since the genetic algorithm involves the computation of a large number of individuals, when the problem is complex, the computation time will increase exponentially. As far as the inventors know, when genetic algorithms are used to optimize the combined cooling, heating and power system in existing literature, the method of randomly generating initial populations is mostly used, resulting in a long system optimization time and a heavy computational burden on the system.

SUMMARY OF THE INVENTION

In order to overcome the above-mentioned shortcomings of the prior art, the present invention provides a method and system for accelerating the optimization time of a combined cooling, heating and power system based on migration learning. Based on similar historical optimization problems, the initial population of genetic algorithm is determined from historical data through migration learning. , to delineate the area where the optimal solution may exist, and to accelerate the optimization of the combined cooling, heating and power system.

To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:

A transfer learning-based optimization time acceleration method for a combined cooling, heating and power system, comprising the following steps:

Receive the cooling, heating and power load sample data corresponding to the optimization history of the combined cooling, heating and power system and the optimization target, which are recorded as the source domain and the target domain respectively;

Clustering the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes;

Compare the target domain sample class with the source domain sample class based on the maximum mean difference to determine whether there is a matching source domain sample class dataset;

If it exists, based on the source domain sample class data set, use transfer learning to determine the initial population of the genetic algorithm; if not, randomly generate the initial population of the genetic algorithm;

Optimization of combined cooling, heating and power system based on genetic algorithm.

One or more embodiments provide a transfer learning-based optimization time acceleration system for a combined cooling, heating and power system, including:

The data acquisition module is configured to: receive the cooling, heating and power load sample data corresponding to the optimization history of the combined cooling, heating and power system and the optimization target, which are respectively recorded as the source domain and the target domain;

The data clustering module is configured to: cluster the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes;

The initial population determination module is configured to: compare the target domain sample class with the source domain sample class based on the maximum mean difference, and determine whether there is a matching source domain sample class data set; if there is, based on the source domain sample class data set, Use transfer learning to determine the initial population of the genetic algorithm; if not, randomly generate the initial population of the genetic algorithm;

The system optimization module is configured to: optimize the combined cooling, heating and power system based on the genetic algorithm.

One or more embodiments provide an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the transfer-based learning when the processor executes the program A time-acceleration method for optimization of combined cooling, heating and power systems.

One or more embodiments provide a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the transfer learning-based method for optimizing time acceleration of a combined cooling, heating and power system.

One or more of the above technical solutions have the following beneficial effects:

Based on the sample data of cooling, heating and power loads corresponding to the optimization problem, the present invention uses the maximum mean difference to find similar historical cooling, heating and power load sample data, that is, finds a data set corresponding to a historical optimization problem similar to the current optimization problem, and determines through migration learning. The initial population of the genetic algorithm is used to delineate the area where the optimal solution of the genetic algorithm may exist, which realizes the acceleration of the optimization of the combined cooling, heating and power system, and reduces the computational burden of the system.

Description of drawings

The accompanying drawings forming a part of the present invention are used to provide further understanding of the present invention, and the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation of the present invention.

FIG. 1 is a flowchart of a method for accelerating time optimization of a combined cooling, heating and power system based on transfer learning in an embodiment of the present invention.

Detailed ways

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that when the terms "comprising" and/or "including" are used in this specification, it indicates that There are features, steps, operations, devices, components and/or combinations thereof.

Embodiments of the invention and features of the embodiments may be combined with each other without conflict.

Example 1

In order to solve the deficiencies of existing CCHP system optimization using genetic algorithm, the present embodiment discloses a transfer learning-based method for accelerating the optimization time of combined cooling, heating and power system, including the following steps:

Step 1: Receive the cooling load, heating load, and electric load prediction data of the corresponding target of the combined cooling, heating and power system optimization, and load it into the target domain; data, and load the model library, also known as the source domain.

Wherein, the target cooling load, heating load, and electric load prediction data are obtained by the combined cooling, heating and power system based on the problem to be optimized and according to the optimization goal.

Step 2: Cluster the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes.

For the source domain or target domain samples generated by the evolutionary population, since the optimization problem usually contains several peaks, all samples in the source domain or target domain are regarded as a whole subject to the same distribution, and MMD is used to judge the source domain and the target domain. The similarity of the target domain problems will be difficult to accurately evaluate their similarity. In view of this, this paper firstly uses the K-means algorithm to cluster the samples in the source domain and the target domain respectively, and assumes that the samples in each class after clustering obey the same distribution (set a total of j sample classes after clustering). Specifically, the cold, heat, and electric load data corresponding to each sample are used as multi-dimensional data, and the distance between samples is calculated by Euclidean distance.

This embodiment uses the K-means clustering algorithm to perform clustering. K-means algorithm originated from a vector quantization method in signal processing, and now it is more popular in the field of data mining as a cluster analysis method. The purpose of K-means clustering is to divide n points into k clusters, so that each point belongs to the cluster corresponding to the nearest mean (this is the cluster center), which is used as the clustering standard.

Step 3: Compare the target domain sample class with the source domain sample class based on the maximum mean difference, and determine whether there is a matching source domain sample class data set. If there is, go to Step 4, otherwise randomly generate the initial population of the genetic algorithm.

和一个满足Q₂分布的目标域域数据集

令H为再生希尔伯特空间(RKHS)，并且存在一个从原始空间到希尔伯特空间的映射函数

那么，当n和m趋于无穷时，X^s和X^t在RKHS上的最大均值差异可以表示为：Maximum mean difference (MMD) is an indicator used to judge whether two data distributions are the same, and was originally mainly used for two-sample detection problems. The basic principle of MMD is: Assume that there is _a source domain dataset that satisfies the Q1 distribution

and a target domain dataset that satisfies the _Q2 distribution

Let H be a regenerated Hilbert space (RKHS) and there exists a mapping function from the original space to the Hilbert space

Then, when n and m tend to infinity, the maximum mean difference of X ^s and X ^t on RKHS can be expressed as:

The step 3 specifically includes:

Step 3.1: For each target domain sample class, determine the source domain sample class with the smallest difference based on the maximum mean difference, and record it as the matching sample class;

Specifically, let the sample class of the model library be the source domain data set, and the sample class of the problem to be optimized be the target domain data set. Specifically, for each sample class in the target domain, calculate the difference between it and all the historical model sample classes in the model library Maximum mean difference f(X ^s , X ^t ). In this embodiment, for j sample classes in the target domain, first initialize k=1, calculate the average value of the maximum mean difference between the sample class and the kth historical model sample class in the model library, and denote it as Ave _k ; then, Let k=k+1, until the maximum mean difference Ave _k between the sample class and all historical models in the model library is obtained, k=1, 2, 3, ..., take the sample class of the historical model with the smallest Ave value as the The matching sample class for the sample class.

Based on the above method, the corresponding matching sample class of each target domain sample class is obtained.

Step 3.2: Calculate the average value of the maximum mean difference between the target domain sample class and its corresponding matching sample class. If the average value is less than the set threshold, it is considered that there is a matching source domain sample class data set, that is, the corresponding target domain sample class. Match the sample data set corresponding to the sample class.

Step 4: Determine the initial population of the genetic algorithm through transfer learning.

Perform transfer learning to determine the initial population of the problem to be optimized through the matched source domain sample class data set, that is, randomly extract some individuals from the matched source domain sample class data set to form the initial population of the genetic algorithm.

Step 5: Use the genetic algorithm to optimize the CCHP system a few days ago, and update the model library at the same time.

The daily cost saving rate, primary energy daily saving rate, and CO2 daily emission reduction rate of the new energy combined cooling, heating and power system are selected as the optimal scheduling goals, and the genetic algorithm is used to optimize the output plan of each equipment in the system.

At the same time, the sample data of cooling load, heating load and electric load corresponding to this optimization result are added to the model library, that is, the model library is updated.

Embodiment 2

The purpose of this embodiment is to provide a transfer learning-based optimization time acceleration system for a combined cooling, heating and power system, including:

The data acquisition module is configured to: receive the cooling, heating and power load sample data corresponding to the optimization history of the combined cooling, heating and power system and the optimization target, which are respectively recorded as the source domain and the target domain;

The data clustering module is configured to: cluster the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes;

The initial population determination module is configured to: compare the target domain sample class with the source domain sample class based on the maximum mean difference, and determine whether there is a matching source domain sample class data set; if there is, based on the source domain sample class data set, Use transfer learning to determine the initial population of the genetic algorithm; if not, randomly generate the initial population of the genetic algorithm;

The system optimization module is configured to: optimize the combined cooling, heating and power system based on the genetic algorithm.

Embodiment 3

The purpose of this embodiment is to provide an electronic device.

An electronic device, comprising a memory, a processor and a computer program stored on the memory and running on the processor, when the processor executes the program, the transfer learning-based cold-heat-electricity system as described in the first embodiment is realized Joint supply system optimization time acceleration method.

Embodiment 4

The purpose of this embodiment is to provide a computer-readable storage medium.

A computer-readable storage medium stores a computer program thereon, and when the program is executed by a processor, implements the method for accelerating the optimization time of a combined cooling, heating and power system based on migration learning as described in the first embodiment.

The steps involved in the apparatuses of the second, third, and fourth embodiments above correspond to the method embodiment 1, and the specific implementation can refer to the relevant description part of the embodiment 1. The term "computer-readable storage medium" should be understood to include a single medium or multiple media including one or more sets of instructions; it should also be understood to include any medium capable of storing, encoding or carrying for use by a processor The executed instruction set causes the processor to perform any of the methods of the present invention.

By adopting the technical solutions in one or more of the above embodiments, the problem of too long calculation time caused by using the genetic algorithm alone to optimize the new energy combined cooling, heating and power system can be improved, and the idea of transfer learning is integrated into the genetic algorithm. In the method, valuable historical information is extracted from the similar historical problems that have been solved to guide the optimal solution of new problems. This method can accelerate the evolution process of the population and improve the efficiency of system optimization.

Those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by the computing device, so that they can be stored in a storage device. The device is executed by a computing device, or they are separately fabricated into individual integrated circuit modules, or multiple modules or steps in them are fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they do not limit the scope of protection of the present invention. Those skilled in the art should understand that on the basis of the technical solutions of the present invention, those skilled in the art do not need to pay creative efforts. Various modifications or deformations that can be made are still within the protection scope of the present invention.

Claims (10)

1. a method for optimizing time acceleration of combined cooling, heating and power system based on migration learning, is characterized in that, comprises the following steps: Receive the cooling, heating and power load sample data corresponding to the optimization history of the combined cooling, heating and power system and the optimization target, which are recorded as the source domain and the target domain respectively; Clustering the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes; Compare the target domain sample class with the source domain sample class based on the maximum mean difference to determine whether there is a matching source domain sample class dataset; If it exists, based on the source domain sample class data set, use transfer learning to determine the initial population of the genetic algorithm; if not, randomly generate the initial population of the genetic algorithm; Optimization of combined cooling, heating and power system based on genetic algorithm. 2. The transfer learning-based optimization time acceleration method for a combined cooling, heating and power system according to claim 1, wherein the clustering adopts a K-Means clustering method. 3. The transfer learning-based optimization time acceleration method for a combined cooling, heating and power system according to claim 1, wherein judging whether there is a matching source domain sample class data set comprises: For each target domain sample class, determine the source domain sample class with the smallest difference based on the maximum mean difference, and record it as the matching sample class; According to the maximum mean difference value between the target domain sample class and the matching sample class, it is judged whether there is a matching source domain sample class dataset. 4. The transfer learning-based optimization time acceleration method for a combined cooling, heating and power system according to claim 3, wherein the average value of the maximum mean difference between the target domain sample class and its corresponding matching sample class is calculated, if the average value is If it is less than the set threshold, it is considered that there is a matching source domain sample class data set, that is, the sample data set corresponding to the matching sample class corresponding to each target domain sample class. 5. The migration learning-based optimization time acceleration method for a combined cooling, heating and power system according to claim 1, wherein determining the initial population of the genetic algorithm using migration learning comprises: randomly extracting from a matched source domain sample class data set Some individuals form the initial population of the genetic algorithm. 6. The method for accelerating the optimization time of a combined cooling, heating and power system based on migration learning as claimed in claim 1, wherein the optimization of the combined cooling, heating and power system based on a genetic algorithm comprises: The daily cost saving rate, primary energy daily saving rate, and CO2 daily emission reduction rate of the new energy combined cooling, heating and power system are selected as the optimal scheduling goals, and the genetic algorithm is used to optimize the output plan of each equipment in the system. 7. The transfer learning-based optimization time acceleration method for a combined cooling, heating and power system as claimed in claim 1, characterized in that, when optimizing the combined cooling, heating and power system based on a genetic algorithm, the obtained cooling load, heat load will also be obtained. , Electric load sample data is added to the source domain. 8. An optimization time acceleration system for a combined cooling, heating and power system based on migration learning, characterized in that it comprises: The data acquisition module is configured to: receive the cooling, heating and power load sample data corresponding to the optimization history of the combined cooling, heating and power system and the optimization target, which are respectively recorded as the source domain and the target domain; The data clustering module is configured to: cluster the sample data in the source domain and the target domain respectively to obtain multiple target domain sample classes and multiple source domain sample classes; The initial population determination module is configured to: compare the target domain sample class with the source domain sample class based on the maximum mean difference, and determine whether there is a matching source domain sample class data set; if there is, based on the source domain sample class data set, Use transfer learning to determine the initial population of the genetic algorithm; if not, randomly generate the initial population of the genetic algorithm; The system optimization module is configured to: optimize the combined cooling, heating and power system based on the genetic algorithm. 9. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements any one of claims 1-7 when the processor executes the program The transfer learning-based optimization time acceleration method for combined cooling, heating and power systems. 10. A computer-readable storage medium on which a computer program is stored, characterized in that, when the program is executed by a processor, the transfer learning-based combined cooling, heating and power supply as claimed in any one of claims 1-7 is realized System optimization time acceleration method.

Images