CN114777192B

CN114777192B - Secondary network heat supply autonomous optimization regulation and control method based on data association and deep learning

Info

Publication number: CN114777192B
Application number: CN202210429423.XA
Authority: CN
Inventors: 穆佩红; 谢金芳; 赵琼; 金鹤峰
Original assignee: Zhejiang Yingji Power Technology Co ltd
Current assignee: Zhejiang Yingji Power Technology Co ltd
Priority date: 2022-04-22
Filing date: 2022-04-22
Publication date: 2024-01-19
Anticipated expiration: 2042-04-22
Also published as: CN114777192A

Abstract

The invention discloses a secondary network heat supply autonomous optimization regulation method based on data association and deep learning, which comprises the following steps: establishing a secondary network digital twin model of the heating system; an electric regulating valve is arranged at the entrance of each unit building, a heat meter is arranged on a water supply main pipe at the entrance of each unit building, and a data concentrator is arranged in each building; constructing a multi-variable data sequence by using heat supply operation data and multi-variable data related to the building room temperature, and obtaining unit building room temperature characterization data after establishing association analysis of the multi-variable data sequence and the room temperature; building a load prediction model of the unit building by adopting a first mixed deep learning method; based on the demand load predicted value and the historical regulation target parameter, a second mixed deep learning method is adopted to establish a unit building flow prediction model, and then the opening of the electric regulating valve is regulated; based on a digital twin model of the secondary network of the heating system, deducing the operation pressure difference of the circulating pump under the condition of meeting the distribution of the demand flow of each unit building, and adjusting the operation frequency of the circulating pump of the secondary network by combining the total demand flow of the secondary network.

Description

Secondary network heat supply autonomous optimization regulation and control method based on data association and deep learning

Technical Field

The invention belongs to the technical field of intelligent heat supply, and particularly relates to a secondary network heat supply autonomous optimization regulation method based on data association and deep learning.

Background

The urban central heating is an important civil engineering which is always concerned by various levels of government and society, is an important subject for the research of the heat supply industry, and is mainly supported in the field of infrastructure of China, the heat supply quality is improved, the heat supply cost is reduced, and the pollution emission is reduced. For a long time, since hydraulic balance of a primary heat supply network relates to safe operation of the whole heat supply network, most heat supply enterprises pay great attention, and a great deal of funds and energy are invested for research and modification. The remarkable achievement is achieved, and the heat loss rate and the water loss rate of the pipe network are obviously reduced. The existing management means of the secondary network are mostly remained in the manual regulation stage, and the regulation fineness and the flexibility degree can not meet the requirements.

In the urban heat supply secondary network system at present, intelligent regulation and balance control of a pipe network are core parts of the whole system design, and the control quality can greatly influence the power consumption and the heat consumption performance of the whole heat supply system. At present, the problem faced by central heating is still that under the condition that the tail end lacks an effective adjusting means, indoor temperature is uneven due to hydraulic imbalance of different users, and meanwhile, in order to maintain the heating quality of users with lower indoor temperature, the problem of total overheat loss of excessive heating caused by improving the heat output of a heat source is solved. Therefore, the operation frequency of the diode network circulating pump and the flow rate of the entrance of the unit building must be controlled within a reasonable range, so that the requirements of heat users can be met, the whole heating system can save more energy and reduce consumption, room temperature data is used as a key index for load prediction and flow rate prediction, the method has an important effect on the opening adjustment of an electric regulating valve and the frequency adjustment of the circulating pump, the room temperature data is difficult to accurately obtain at low cost at present, the measurement of the room temperature is often not advisable, and the cost for installing a temperature measuring device for most users is high.

Based on the technical problems, a new secondary network heat supply autonomous optimization regulation method based on data association and deep learning needs to be designed.

Disclosure of Invention

The invention aims to solve the technical problem of overcoming the defects of the prior art and providing a secondary network heat supply autonomous optimization regulation method based on data association and deep learning.

In order to solve the technical problems, the technical scheme of the invention is as follows:

the invention provides a secondary network heat supply autonomous optimization regulation method based on data association and deep learning, which comprises the following steps:

s1, establishing a secondary network digital twin model of a heating system by adopting a mechanism modeling and data identification method;

step S2, reforming the building Internet of things hardware equipment at least comprises the following steps: an electric regulating valve is arranged at the entrance of each unit building, a heat meter is arranged on a water supply main pipe at the entrance of each unit building, and a data concentrator is arranged in each building;

s3, constructing a multi-variable data sequence by using heat supply operation data and multi-variable data related to the building room temperature, and acquiring unit building room temperature characterization data after establishing association analysis of the multi-variable data sequence and the room temperature;

s4, based on a heating system secondary network digital twin model, building a unit building load prediction model for historical weather data, unit building room temperature characterization data and heat metering data by adopting a first mixed deep learning method to obtain a predicted value of the demand load of each unit building;

S5, building a unit building flow prediction model by adopting a second mixed deep learning method based on the unit building demand load predicted value and the historical regulation target parameter, and regulating the opening of an electric regulating valve at the entrance of the unit building after obtaining the unit building demand flow predicted value;

and S6, based on a digital twin model of the secondary network of the heating system, deducing the operation pressure difference of the circulating pump under the condition of meeting the distribution of the demand flow of each unit building, and adjusting the operation frequency of the circulating pump of the secondary network of the heating system by combining the total demand flow of the secondary network of the heating system to realize the on-demand automatic regulation and control of the heat supply of the secondary network of the heating system.

Further, in the step S1, a mechanism modeling and data identification method is adopted to build a digital twin model of a secondary network of the heating system, which specifically includes:

step S101, constructing a secondary network virtual entity of a heating system, and establishing a secondary network digital twin model after virtual and real data connection, wherein the step comprises the following steps:

constructing a heating system secondary network structure model, a physical equipment entity model, a behavior model and a rule model; the heat supply system secondary network structure model at least comprises a heat exchanger, a diode network and a unit building heat user; the physical equipment entity model is obtained by adding equipment physical attributes; constructing a behavior model based on a thermodynamic basic theory of a secondary network of the heating system, and establishing a virtual simulation system of the secondary network of the heating system with an interactive function and simulating a real operation environment; finally, establishing a rule model of the virtual entity to formulate a control strategy of the virtual entity;

Driving corresponding virtual equipment by collecting actual operation data of the secondary network physical equipment of the heating system, and establishing a mapping relation of virtual and real data to form a secondary network operation strategy of the heating system; the connection and dynamic interaction of real-time data of a physical entity and a virtual space are realized through continuous iteration and optimization of a data acquisition control process, and a two-level network digital twin model is established;

step S102, identifying the digital twin model, which comprises the following steps:

and accessing the multi-working-condition real-time operation data of the secondary network of the heating system into the established digital twin model, and carrying out self-adaptive identification correction on the simulation result of the digital twin model by adopting a reverse identification method to obtain the digital twin model of the secondary network of the heating system after identification correction.

Further, in the step S2, the modifying the building internet of things hardware device at least includes: the electric regulating valve is installed at the entrance of the unit building, and the heat meter is installed on the water supply main pipe of each unit building, which comprises the following steps:

installing an electric regulating valve at the entrance of the unit building; a heat meter is arranged on a water supply main pipe of each unit building mouth; a data concentrator is arranged in each building and is connected with the heat meter; the electric regulating valve and the data concentrator transmit electric regulating valve data and heat metering data to an autonomous optimizing operation terminal arranged in a building through a communication module for data processing;

Wherein, every unit building thing networking hardware equipment still includes: the system comprises a pressure transmitter, a differential pressure transmitter, a calorimeter and a circulating pump frequency converter; the pressure transmitter is arranged on a straight pipe section of the main water supply and return pipe of each unit building mouth and is used for measuring the water supply pressure and return pressure of the unit building mouth; the differential pressure transmitter is sequentially arranged on the water supply main pipe of each unit building mouth and in front of and behind the circulating pump and is used for measuring the differential pressure of the water supply and return of the unit building and the differential pressure of the circulating pump in front of and behind; the heat meter is used for measuring the water supply temperature, the backwater temperature and the water supply flow of the building mouth of the unit; the circulating pump frequency converter is arranged in the heat exchange station and is used for measuring the frequency of the circulating pump.

Further, in step S3, the multivariate data of the heat supply operation data and the building room temperature are formed into a multivariate data sequence, and after the correlation analysis of the multivariate data sequence and the room temperature is established, the unit building room temperature characterization data is obtained, which specifically includes:

acquiring multi-element data related to room temperature, wherein the multi-element data at least comprises: regular household temperature measurement data, complaint data, house orientation type, building type, weather data and heat metering data;

forming a multi-element data sequence sample from the heat metering data, the electric valve regulating data and the multi-element data related to the room temperature;

Taking data in a multi-element data sequence sample as independent variables, taking unit building room temperature characterization data as dependent variables, and firstly, taking the relation between each independent variable and the dependent variable into consideration to obtain mathematical models of the independent variables and the dependent variables; and then stacking all mathematical models one by one, and if the stacked models do not meet the requirements, considering the interaction among respective variables to obtain corresponding multi-element nonlinear mathematical models, wherein the multi-element nonlinear mathematical models are expressed as follows:

wherein Y is a dependent variable; x is x _i 、x _j Is an independent variable; a, a _i 、b _m K is a regression coefficient; f (f) _i (x _i ) Is a functional relationship between a certain independent variable and a dependent variable; the correlation degree between each independent variable and the dependent variable adopts partial correlation analysis, the independent variable which is obviously correlated and generally correlated is screened out, and uncorrelated independent variables are proposed; if the independent variables have high collinearity, the mutual relation coefficient is taken as the collinearity judgment basis, and the factor relation coefficient is removed when the factor relation coefficient is larger than the threshold value.

Further, in the step S4, a unit building load prediction model is built for historical weather data, room temperature data and heat metering data by using a first hybrid deep learning method based on a two-level network digital twin model of the heating system, so as to obtain a predicted value of a demand load of each unit building, which specifically comprises:

S401, based on a heating system secondary network digital twin model, carrying out heat load related factor analysis on the obtained historical weather data, unit building room temperature characterization data and heat metering data by adopting an APRIORI method;

s402, performing Bayesian optimization on the CNN prediction model, taking influence factors screened after probability of association rules as an input sample set of the CNN prediction model after Bayesian optimization, applying the Bayesian optimization to super-parameter optimization to obtain an optimal CNN prediction model, and establishing a unit building thermal load prediction model to obtain a thermal load prediction result.

Further, the S401 includes:

preprocessing an original dataset of historical weather data, unit building room temperature characterization data and heat metering data, comprising: performing redundancy removal, denoising and standardization treatment on the data, filtering records with severely missing numerical values, and temporarily retaining only partially missing data;

thermal load correlation factor analysis, comprising: determining the relevance of the heat load influence factors by utilizing relevance rule probability, converting the APRIORI relevance factor analysis result to be used as a sample set of a heat load prediction model, and using R _i The higher the value is, the stronger the association degree of the influence factor is represented, and the probability calculation of the association degree is represented as:

Wherein,for influencing factor X in the current calculation term _i The sum of the support and the ratio of the number of factors contained in the item set; />For j-item set L _j For X _i Confidence of (2); s (L) _j ) For j-item set L _j Is a support degree of (2); s (X) _i ) To influence factor X _i Is a support of (1).

Further, the S402 includes:

adopting Bayes to optimize super parameters:

selecting one of the super parameters in each iteration process of the CNN prediction model training after selecting the probability agent model and the acquisition function, carrying out evaluation and optimization on the super parameters by using the acquisition function, and adding the obtained most potential evaluation points into historical data until the termination condition is met, thereby obtaining the CNN prediction model; the input of the Bayesian optimization algorithm is a parameter set X to be optimized, an objective function f of a Bayesian optimizer, an acquisition function S and a Gaussian process model M, and the output is an optimal CNN prediction model;

training a CNN prediction model:

the influence factors screened after the probability of the association rule are used as an input sample set and divided into a training set and a testing set; for a training set, respectively extracting heat load characteristics through a convolution layer, and then obtaining load characteristic mapping through batch normalization processing and input of a ReLU function; then compressing the output of the convolution layer through the pooling layer to obtain a compressed heat load characteristic map; finally, the heat load characteristics are further extracted through a newly added convolution layer, and a final characteristic time sequence is obtained through batch normalization processing and ReLU activation functions;

The feature time sequence is subjected to information fusion through the full connection layer, so that mapping from the features to a sample mark space is realized; assigning inputs to one of the mutually exclusive classes and calculating losses using a softmax layer for the probability returned by each input, the objective function employing a cross-loss function;

judging whether the termination condition of the CNN prediction model training is reached, if so, inputting a test set into the CNN prediction model and calculating the model accuracy; otherwise, training is carried out again until the termination condition of the CNN prediction model training is reached;

judging whether the termination condition of the Bayesian optimizer is met, if so, outputting an optimal CNN prediction model and a demand load prediction value of each unit building; otherwise, the next super parameter is reselected until the Bayesian optimizer termination condition is reached.

Further, in step S5, a second hybrid deep learning method is adopted to build a unit building flow prediction model based on the predicted value of the demand load of each unit building and the historical regulation target parameter, so as to obtain the predicted value of the demand flow of each unit building, which specifically includes:

s501, decomposing original data into K IMF subsequences with single frequency characteristics by adopting a fusion variation modal decomposition VMD technology based on the acquired demand load predicted value and the history regulation target parameter of each unit building, and determining a data sample set for model establishment according to the characteristics of the IMF subsequences;

S502, training the K IMF subsequences by using a DBN deep learning algorithm, and obtaining a final predicted value of the unit building demand flow by superposing predicted flow values of the subsequences on the same predicted sample point.

Further, the step S501 includes:

calculating analytic signals of K modal functions by utilizing Hilbert transformation to obtain a single-side frequency spectrum;

aliasing is carried out on each mode function and the index term of the corresponding center frequency, so that the frequency spectrum of each mode is converted into a baseband;

estimating the bandwidth of each mode signal by a Gaussian smoothing method of the demodulation signal, and solving the variation problem with constraint conditions;

a secondary punishment factor and a Lagrange multiplier are adopted to change the constraint variation problem into an unconstrained variation problem;

solving the variation problem by adopting an alternate direction multiplier method, updating each mode function and the center frequency thereof, and demodulating each mode to a corresponding baseband so as to minimize the sum of the estimated bandwidths of each mode;

the K value is determined by a Pelson correlation coefficient method, and the number of the finally decomposed IMF modal components is represented;

in S502, training the DBN deep learning algorithm includes:

unsupervised pre-training based on RBM: determining the number of input neurons through the dimension of the original sample data, and independently training RBM of each layer by adopting an unsupervised greedy algorithm;

Fine tuning: and fine tuning the weight and the threshold value of the DBN network through the BP neural network back propagation algorithm, so as to realize training of the DBN deep learning algorithm.

Further, optimizing the penalty factor alpha and the modal decomposition number K in the VMD algorithm by adopting a PSO particle swarm optimization algorithm, wherein the method comprises the following steps:

taking the minimum value of the envelope entropy as the fitness function of the particle swarm optimization algorithm, and initializing each parameter in the optimization algorithm;

initializing a population in a particle swarm optimization algorithm, randomly generating a plurality of groups [ alpha, K ] at the same time, and taking the groups as information positions of particles, wherein the initial particle speed of the particle swarm is also randomly generated;

and carrying the initial solution into the VMD, calculating the corresponding IMF envelope entropy value, and finding out the obtained minimum value which is taken as the local minimum value. Determining fitness values of different positions according to the position transformation;

comparing the fitness values of different positions, and comparing the fitness values with the extremum of the local part of the individual and the global extremum of the population so as to further update and iterate through the comparison result;

updating the particle speed and the position of the particle swarm, if the particle speed and the position do not meet the requirements, returning to recalculate the local minimum value until iteration is completed, and finally outputting the optimal fitness function value and combining the optimal fitness function value into an optimal parameter set [ alpha, k ];

The GA genetic algorithm is adopted to optimize the learning parameter theta in the DBN algorithm, and the method comprises the following steps:

initializing parameters in the DBN structure while the reconstructed data is treated as a corresponding chromosome;

randomly generating an initial population by a genetic algorithm, wherein the population comprises M chromosomes;

randomly selecting a chromosome from the group, taking the chromosome as theta in a DBN model, training the DBN, and calculating a corresponding fitness function;

judging the condition, namely judging whether the termination condition is met or not by calculating the fitness value of the population; if the termination condition is met, outputting DBN optimization parameters, and if the termination condition is not met, selecting a genetic algorithm, performing cross operation, mutation and other operations, and performing a new round of calculation and training.

The beneficial effects of the invention are as follows:

the invention reforms the hardware equipment of the building internet of things, at least comprising: an electric regulating valve is arranged at the entrance of a unit building, and a heat meter is arranged on a water supply main pipe at each unit building entrance; constructing a multi-variable data sequence by using heat supply operation data and multi-variable data related to the building room temperature, and obtaining unit building room temperature characterization data after establishing association analysis of the multi-variable data sequence and the room temperature; establishing a unit building load prediction model by adopting a first mixed deep learning method to obtain a predicted value of the demand load of each unit building; establishing a unit building flow prediction model by adopting a second mixed deep learning method, and adjusting the opening of an electric regulating valve at the entrance of the unit building after obtaining the predicted value of the required flow of each unit building; based on a digital twin model of the secondary network of the heating system, deducing the operation pressure difference of the circulating pump under the condition of meeting the distribution of the demand flow of each unit building, and adjusting the operation frequency of the circulating pump of the secondary network by combining the total demand flow of the secondary network to realize the independent regulation and control of the heat supply of the secondary network according to the demand; after the transformation of hardware equipment and the association analysis of a multi-element data sequence and the room temperature are established, the room temperature characterization data of the unit building can be obtained, the problem that the room temperature data is difficult to obtain is avoided, and key data can be provided for heat load prediction through the effective acquisition of the room temperature; and building a unit building thermal load and flow prediction model by mining key factors of model prediction and adopting a mixed deep learning method, and determining an intelligent regulation and control strategy of regulation and control target parameters and adjustable equipment so as to realize intelligent on-demand heating regulation and control of the heating network.

Additional features and advantages will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of a secondary network heat supply autonomous optimization regulation method based on data association and deep learning;

FIG. 2 is a schematic diagram of a building Internet of things hardware device according to the present invention;

FIG. 3 is a flow chart of a method for optimizing the heat load of each unit building of the CNN by APRIORI-Bayesian method of the invention;

FIG. 4 is a flow chart of a method for predicting the flow of each unit building of the VMD-DBN of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Example 1

Fig. 1 is a flow chart of a secondary network heating autonomous optimization regulation method based on data association and deep learning.

Fig. 2 is a schematic diagram of a building internet of things hardware device according to the present invention.

FIG. 3 is a flow chart of an APRIORI-Bayesian optimization CNN unit building thermal load method related to the invention.

Fig. 4 is a flowchart of a flow prediction method of each unit building of the VMD-DBN according to the present invention.

As shown in fig. 1-4, embodiment 1 provides a method for automatically optimizing and controlling heat supply of a secondary network based on data association and deep learning, which includes:

It should be noted that, the precise regulation and control system of the secondary network of the heating system needs to obtain the pipe network operation data of the inlets of each building, including operation flow, operation heat, water supply and return pressure, pressure difference, water supply and return temperature, etc., to support the functions of system load prediction and pipe network resistance identification, consider that if hydraulic unbalance occurs between the unit buildings, the hydraulic unbalance is regulated by the time period of "full open" or "full closed" of the shutoff valve before the building, and it is difficult to realize balanced regulation and control. Therefore, an electrically-controlled valve needs to be arranged in front of the building.

In this embodiment, in the step S1, a mechanism modeling and data identification method is adopted to build a digital twin model of a secondary network of a heating system, which specifically includes:

In this embodiment, in the step S2, the modifying the building internet of things hardware device at least includes: the electric regulating valve is installed at the entrance of the unit building, the heat meter is installed on the water supply main pipe of each unit building, and the data concentrator is arranged in each building, and specifically comprises:

In this embodiment, in step S3, the multivariate data related to the building room temperature and the heating operation data are combined into a multivariate data sequence, and after the correlation analysis of the multivariate data sequence and the room temperature is established, the unit building room temperature characterization data are obtained, which specifically includes:

It should be noted that, through partial correlation analysis and factor collinearity judgment, the factors affecting the room temperature are screened, and the change rule of the independent variables is determined; and constructing a model between the room temperature data and related influencing variables by using a multi-element nonlinear regression method.

In this embodiment, in step S4, a first hybrid deep learning method is used to build a unit building load prediction model for historical weather data, unit building room temperature characterization data and heat metering data based on a two-level network digital twin model of a heating system, so as to obtain a predicted value of a demand load of each unit building, which specifically includes:

It should be noted that, for a large amount of original alarm information, the heat load association rule is mined by adopting an APRIORI algorithm, the key influencing factors are extracted by utilizing probability function conversion, the redundancy of input data in a Bayesian optimization CNN prediction model is reduced, the value density of the input data is improved, and the disturbance of invalid data to a prediction result is avoided; 2) Establishing an APRIORI-Bayesian optimized CNN prediction model, searching a parameter optimal solution of the CNN prediction model by using Bayesian, simplifying the model optimizing process, and further improving the model efficiency and the thermal load prediction precision.

In this embodiment, the step S401 includes:

In this embodiment, the step S402 includes:

adopting Bayes to optimize super parameters:

Training a CNN prediction model:

dividing the screened influence factors into a training set and a testing set by taking the influence factors as an input sample set; for a training set, respectively extracting heat load characteristics through a convolution layer, and then obtaining load characteristic mapping through batch normalization processing and input of a ReLU function; then compressing the output of the convolution layer through the pooling layer to obtain a compressed heat load characteristic map; finally, the heat load characteristics are further extracted through a newly added convolution layer, and a final characteristic time sequence is obtained through batch normalization processing and ReLU activation functions;

In this embodiment, in step S5, a second hybrid deep learning method is used to build a unit building flow prediction model based on the predicted value of the demand load of each unit building and the historical regulation target parameter, so as to obtain the predicted value of the demand flow of each unit building, which specifically includes:

In this embodiment, the step S501 includes:

in S502, training the DBN deep learning algorithm includes:

In the DBN structure model, the input of the original data is mainly performed by the lower layer, and the output of the original data is performed by the feature extraction of the lower layer. Each layer of RBM can train data, the feature data obtained by the previous layer of training can be used as the data of the next layer of RBM to train, the weight and the threshold value in the RBM can be continuously updated through continuous iteration, the final stopping condition is that the maximum iteration times are reached, the iteration is stopped when the requirement is met, and the updating is completed. The deep belief network performs multiple layer-by-layer data feature extraction so as to show some finer features of the data, provides a certain foundation for the flow prediction of the subsequent unit building, and improves the accuracy of the flow prediction.

In this embodiment, optimizing the penalty factor α and the modal decomposition number K in the VMD algorithm by using the PSO particle swarm optimization algorithm includes:

In the DBN algorithm, the influence of the selection of the structural parameters on the algorithm model is large, and in particular, in the RBM structure, θ= { ω, a, b } plays a very large role in the accuracy of DBN flow model prediction. Because the maximum likelihood estimation method has great limitation in application, θ is easy to fall into a local optimal value in calculation, and finally convergence is impossible, and finally a global optimal solution is not obtained. Therefore, the adaptive function GA is used for optimizing the model, and meanwhile, the genetic algorithm has the characteristics of simple principle, strong global searching capability and the like, so that the method has wide application in optimization and model prediction methods.

In the several embodiments provided in this application, it should be understood that the disclosed systems and methods may be implemented in other ways as well. The system embodiments described above are merely illustrative, for example, of the flowcharts and block diagrams in the figures that illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, functional modules in the embodiments of the present invention may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored on a computer readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method of the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

With the above-described preferred embodiments according to the present invention as an illustration, the above-described descriptions can be used by persons skilled in the relevant art to make various changes and modifications without departing from the scope of the technical idea of the present invention. The technical scope of the present invention is not limited to the description, but must be determined according to the scope of claims.

Claims

1. A secondary network heat supply autonomous optimization regulation and control method based on data association and deep learning is characterized by comprising the following steps:

2. The method for autonomous optimization and control of heat supply of a secondary network according to claim 1, wherein in step S1, a mechanism modeling and data identification method is adopted to build a digital twin model of the secondary network of the heat supply system, and the method specifically comprises the following steps:

constructing a heating system secondary network structure model, a physical equipment entity model, a behavior model and a rule model;

the heat supply system secondary network structure model at least comprises a heat exchanger, a diode network and a unit building heat user;

the physical equipment entity model is obtained by adding equipment physical attributes;

constructing a behavior model based on a thermodynamic basic theory of a secondary network of the heating system, and establishing a virtual simulation system of the secondary network of the heating system with an interactive function and simulating a real operation environment;

Finally, establishing a rule model of the virtual entity to formulate a control strategy of the virtual entity;

3. The method for autonomous optimization and control of heat supply of a secondary network according to claim 1, wherein in the step S2, the hardware equipment of the building internet of things is modified, and the method at least comprises: the electric regulating valve is installed at the entrance of the unit building, the heat meter is installed on the water supply main pipe of each unit building, and the data concentrator is arranged in each building, and specifically comprises:

Installing an electric regulating valve at the entrance of the unit building; a heat meter is arranged on a water supply main pipe of each unit building mouth; a data concentrator is arranged in each building and is connected with a heat meter; the electric regulating valve and the data concentrator transmit electric regulating valve data and heat metering data to an autonomous optimized operation terminal arranged on a building through a communication module for data processing;

4. The method for autonomous optimizing and controlling heat supply of a secondary network according to claim 1, wherein in step S3, multivariate data of heat supply operation data and building room temperature are formed into a multivariate data sequence, and after correlation analysis of the multivariate data sequence and the room temperature is established, unit building room temperature characterization data are obtained, and specifically comprising:

taking data in the multi-element data sequence sample as independent variables, taking unit building room temperature characterization data as the dependent variables, obtaining mathematical models of respective variables and the dependent variables, then superposing all the mathematical models one by one, and if the superposed models do not meet the requirements, taking interaction among the respective variables into consideration, so as to obtain corresponding multi-element nonlinear mathematical models, wherein the mathematical models are expressed as follows:

5. The method according to claim 1, wherein in step S4, based on the heating system two-level network digital twin model, the unit building load prediction model is built by using a first hybrid deep learning method on historical weather data, unit building room temperature characterization data and heat metering data, so as to obtain the predicted value of the demand load of each unit building, and the method specifically comprises:

6. The method for autonomous optimization regulation of secondary network heating as recited in claim 5, wherein S401 includes:

preprocessing the original data set of the historical weather data, the unit building room temperature characterization data and the heat metering data, wherein the preprocessing comprises the following steps: performing redundancy removal, denoising and standardization treatment on the data, filtering records with severely missing numerical values, and temporarily retaining only partially missing data;

The thermal load correlation factor analysis includes: determining the relevance of the heat load influence factors by using relevance rule probability, and converting the APRIORI relevance factor analysis resultAs a sample set of the thermal load prediction model, R is used _i Representing the degree of association of the influencing factors, R _i The larger the value is, the stronger the association degree is, and the probability calculation of the association degree is expressed as:

wherein,for influencing factor X in the current calculation term _i The sum of the support and the ratio of the number of factors contained in the item set;

for j-item set L _j For X _i Confidence of (2);

S(L _j ) For j-item set L _j Is a support degree of (2); s (X) _i ) To influence factor X _i Is a support of (1).

7. The method for autonomous optimization regulation of secondary network heating as recited in claim 5, wherein S402 includes:

adopting Bayes to optimize super parameters:

Training a CNN prediction model:

judging whether the termination condition of the CNN prediction model is met, if so, inputting a test set into the CNN prediction model and calculating the model accuracy; otherwise, training is carried out again until the termination condition of the CNN prediction model training is reached;

8. The method for autonomous optimization and control of heat supply in a secondary network according to claim 6, wherein in step S5, a second hybrid deep learning method is used to build a predicted unit building flow model based on the predicted unit building demand load value and the historical control target parameter, so as to obtain the predicted unit building demand flow value, and the method specifically comprises:

s502, training the K IMF subsequences by using a DBN deep learning algorithm, and obtaining a final predicted value of the unit building demand flow by superposing predicted flow values of the IMF subsequences on the same predicted sample point.

9. The method for autonomous optimization regulation of secondary network heating according to claim 8, wherein S501 comprises:

in S502, training the DBN deep learning algorithm includes:

10. The method for automatically optimizing and regulating the heat supply of the secondary network according to claim 9, wherein optimizing the penalty factor alpha and the modal decomposition number K in the VMD by using a PSO particle swarm optimization algorithm comprises:

the initial solution is brought into the VMD, the corresponding IMF envelope entropy value is calculated, the obtained minimum value is found out and is used as a local minimum value, and the fitness values of different positions are determined according to the position transformation;