Power data energy consumption prediction control method and system based on preview machine learning
Through the method based on rehearsal machine learning, a dynamic energy consumption prediction model is constructed, which solves the shortcomings of energy consumption prediction control in the existing technology, and realizes accurate prediction and effective control of power data energy consumption.
Patent Information
- Application Number
- CN202411820601.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-11
- Publication Date
- 2025-04-01
AI Technical Summary
The prior art only stays in the prediction process of static models in the energy consumption prediction of power data, and fails to propose a comprehensive model that can further control the trend of energy consumption through the estimated energy consumption data in combination with the prediction model.
Using a method based on rehearsal machine learning, all factors affecting energy consumption are classified into initial variables, process variables and result variables, and an initial rehearsal graph is constructed. The graph structure is dynamically updated through structural equation parameters and Bayesian method, and two independent and same-distributed parameter training sets and verification sets are constructed, indicating that the decision process is a constrained optimization problem, the candidate change selection set is output, and the change that meets the maximum mutual information optimization conditions is selected as the decision result of the model output.
Accurate prediction of energy consumption is achieved, and corresponding adjustment decisions are proposed based on the prediction results to effectively control the change trend of energy consumption and avoid the loss of control of the result data.
Smart Images
Figure CN119740701A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of machine learning, and in particular to a method and system for predicting and controlling power data energy consumption based on preview machine learning. Background Art
[0002] In the power data, analyzing the energy consumption and energy consumption of the whole province, monitoring the total energy consumption and intensity changes of each energy-consuming entity, is a long-term key task to assist the government in making scientific decisions, formulating energy consumption dual control measures, improving energy management level and optimizing energy-saving standards. At present, more and more work has been applied to the energy consumption prediction task of power data using machine learning methods. However, these methods only stay at the stage of studying the prediction process of system energy consumption, and have not combined the prediction model to propose a comprehensive model that can further control the trend of energy consumption changes through estimated energy consumption data.
[0003] The prior art proposes a method for displaying power energy consumption data, which mainly includes: first, obtaining real-time power energy consumption data of different areas based on pre-built intelligent sensors to obtain current standard energy consumption data; then, detecting the current standard energy consumption data according to the power energy consumption anomaly detection model to obtain abnormal data; finally, using the pre-built BP neural network to predict the change trend of the current standard energy consumption data to obtain the power energy consumption prediction value, and displaying the current standard energy consumption data, abnormal data and power energy consumption prediction value in a graphical form in a visualization interface.
[0004] The prior art proposes to provide a method and device for predicting power consumption, which relates to the field of power consumption prediction and analysis. The power consumption prediction method includes: first, obtaining a first historical energy consumption sample data set of a thermal power unit in accordance with each preset time category and each preset time interval within a historical preset time period and preprocessing the first historical energy consumption sample data set to obtain a processed second historical energy consumption sample data set; then, obtaining a first proportion and an average value of each historical energy consumption data interval occupying the second historical energy consumption sample data set, and performing a first neuron training on the second historical energy consumption sample data set, the first proportion and the average value to obtain a first neuron model; finally, obtaining the planned energy consumption data of the thermal power unit, and obtaining the proportion of each planned energy consumption data interval occupying the planned energy consumption data and the predicted value according to the planned energy consumption data and the first neuron model. Using this energy consumption prediction method, the planned power generation amount can be predicted more accurately and conveniently.
[0005] The above methods are all based on static models to predict the trend of energy consumption, and do not consider using dynamic models with two-way influence relationships to describe more realistic problem scenarios. At the same time, the above methods only consider the accuracy of power consumption forecast data, and cannot further provide solutions to unexpected results to avoid the loss of control of the result data. Summary of the invention
[0006] The purpose of the present invention is to provide a power data energy consumption prediction control method and system based on preview machine learning to solve the above problems.
[0007] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for predicting and controlling power data energy consumption based on preview machine learning, comprising: Classify all factors affecting energy consumption into initial variables, process variables and result variables, and construct an initial preview diagram; For the initial preview graph, the overall posterior is estimated, and the structural equation parameters are obtained based on maximizing the parameter posterior; According to the structural equation parameters, two sets of independent and identically distributed parameter training sets and validation sets are constructed, the decision-making process is expressed as a constrained optimization problem about the change, all changes are trained and validated, and a candidate change selection set is output; Based on the obtained candidate change selection set, the change that meets the maximum mutual information optimization condition is selected as the decision result output by the model.
[0008] Furthermore, all factors affecting the energy consumption of enterprises are classified as initial variables , process variables and the outcome variable , construct the initial preview graph In the preview diagram representation, the correlation between categorical variables is shown through the preview diagram. For each preview diagram, it is defined as: , where V represents a vertex and E represents an edge; Is Induced complete subgraph, any edge They are both bidirectional. middle, for All maximal bi-cliques in .
[0009] Furthermore, is a preview graph if and only if: [1] For any ; [2] , if there exists any point to , then for any Midpoint , both ; [3] Direction side A topological sort was performed; The structural equation is expressed as follows:
[0010] in, is the largest bi-cluster The parent node set of So is a multi-valued function of parameters, represents Gaussian noise.
[0011] Furthermore, for the initial preview graph, the overall posterior is estimated, and the structural equation parameters are obtained based on maximizing the parameter posterior, including: The Bayesian method is used to dynamically update the graph structure. Through the Bayesian formula, the entire preview graph The posterior estimate of the parameters is expressed as follows:
[0012] in, express All data collected before the decision round begins, which consists of the initial observation data and related variables Composition; for graph posterior Adopt the approximation of bootstrapping, that is, learn multiple preview graphs on Bootstrap data to build a set of graphs And each graph element is weighted equally, then its approximate posterior expression is:
[0013] For the parameter posterior Representation:
[0014] in, is included The set of graphs with their transformation graphs after a given change, express One of the largest bi-cluster in , is given middle and A set of regression parameters and the posterior of the noise variance of the parent node in ; the Bayesian ridge regression method is used to learn the parameter distribution that better fits the real data.
[0015] Furthermore, based on the structural equation parameters, two sets of independent and identically distributed parameter training sets and validation sets are constructed. The decision-making process is expressed as a constrained optimization problem about the change. All changes are trained and validated, and a candidate change selection set is output, including: The optimization objective of the entire energy consumption prediction problem is expressed as the following constrained parameter optimization problem:
[0016] in, Indicates all result variables set The expected interval of Preview The overall parameters and the observed values revealed after the change By maximizing the mutual information To avoid the graph structure after intervention changes from deviating from the real data; define is the intervention decision for each round of intervention execution, where Indicates that the setting sequence number is The intermediate variables operation; further, For preview The last time A preview of the round, and represented as a decision in the decision-making process; By sampling a set of independently distributed preview graph samples To estimate the expectation under the constraints:
[0017] in, is specified The results produced.
[0018] Further, the calculation of constraints: (1) From what I have learned Posterior and set noise distribution The sampling results are both of size The independent and identically distributed parameter training set and validation set ; (2) By Preview and find satisfaction Change collection , integration Build candidate changesets ; (3) By Rehearsal to verify Change collection , and from Delete Change collection .
[0019] The present invention is based on variables The expression can further find candidate changes that satisfy the constraint interval .
[0020] Furthermore, based on the obtained candidate change selection set, the change that satisfies the maximum mutual information optimization condition is selected as the decision result output by the model, including: Power data , by giving all candidate changes , and for the main optimization goals
[0021] Solve, where
[0022] Use the Entropyestimate method to estimate the information entropy value and calculate the mutual information of each candidate change to solve the optimization problem, and finally select the change that meets the maximum mutual information optimization condition. The decision result as the output of the model.
[0023] In a second aspect, the present invention provides a power data energy consumption prediction control system based on preview machine learning, comprising: A preview diagram construction module is used to classify all factors affecting energy consumption into initial variables, process variables and result variables, and construct an initial preview diagram; The structural equation parameter acquisition module is used to estimate the overall posterior for the initial preview diagram and obtain the structural equation parameters based on the maximized parameter posterior; The data processing module is used to construct two sets of independent and identically distributed parameter training sets and validation sets based on the structural equation parameters, express the decision-making process as a constrained optimization problem about the changes, train and validate all the changes, and output a candidate change selection set; The result output module is used to select the change that meets the maximum mutual information optimization condition as the decision result output by the model according to the obtained candidate change selection set.
[0024] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the power data energy consumption prediction and control method based on preview machine learning when executing the computer program.
[0025] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the power data energy consumption prediction and control method based on preview machine learning are implemented.
[0026] Compared with the prior art, the present invention has the following technical effects: The present invention realizes accurate prediction of energy consumption by dynamically modeling the relationship between various energy data, and proposes corresponding adjustment decisions based on the prediction results to achieve effective control of the trend of energy consumption changes. The present invention is divided into two modules, a graph learning module and an optimization control module. First, the method models the energy consumption prediction and control problem as a multi-round dynamic decision-making problem, and classifies each influencing factor. Then, the initial preview graph of the problem is generated by constructing a preview structure model (Structural Rehearsal Model, SRM). Secondly, for the Bayesian graph learning module, the present invention adopts the IC algorithm and the Bayesian Ridge regression method to learn the preview graph topology and the structural equation parameters of the preview structure model in the problem, respectively, so as to find the most realistic graph structure to ensure the effectiveness of subsequent optimization solutions. Then, the present invention transforms the energy consumption prediction and control problem into a parameter optimization problem with constraints, and introduces the learned parameters to solve the constraint interval of the optimization problem. At the same time, the entropy estimation method is used to estimate the mutual information of each change selection within the constraint interval. Finally, the change that satisfies the maximum mutual information is found as the decision result. The present invention adopts a preview structure model to express the mutual influence relationship between the variables, and captures the two-way effect between factors by defining two-way subgroups. At the same time, the present invention continuously learns based on the historical data of the variables, and dynamically models the graphical representation that best conforms to the actual data, and fully expresses the influence relationship between the variables and the energy consumption through the preview graph; at the same time, the present invention models the enterprise energy consumption prediction and control problem as a multi-round dynamic decision-making process, and further introduces a graph learning module and an optimization control module to realize the dynamic update of the model and the optimization of the real-time control strategy. In the prediction process, it is not limited to a fixed graph structure, but a graph learning process is set up. This process continuously learns based on the historical data of the variables and gives the graphical representation that best conforms to the actual data. The preview structure model is used to express the mutual influence relationship between the variables, and the two-way effect between the factors is captured by defining two-way subgroups, which can better model the influence relationship between the variables and improve the prediction effect. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 This is a block diagram of the enterprise energy consumption prediction and control method based on power data of preview machine learning in the present invention.
[0027] Figure 2 It is an overall flow chart represented by a preview diagram of the energy consumption prediction control problem.
[0028] Figure 3 It is a flowchart for dynamically updating the preview diagram parameters.
[0029] Figure 4 It is a flowchart for establishing a multi-round dynamic decision-making model.
[0030] Figure 5It is a flow chart of decision making for energy consumption predictive control. DETAILED DESCRIPTION
[0031] The present invention is further described below in conjunction with the accompanying drawings: Example 1, please refer to Figure 1 , a power data energy consumption prediction and control method based on preview machine learning, including: Classify all factors affecting energy consumption into initial variables, process variables and result variables, and construct an initial preview diagram; For the initial preview graph, the overall posterior is estimated, and the structural equation parameters are obtained based on maximizing the parameter posterior; According to the structural equation parameters, two sets of independent and identically distributed parameter training sets and validation sets are constructed, the decision-making process is expressed as a constrained optimization problem about the change, all changes are trained and validated, and a candidate change selection set is output; Based on the obtained candidate change selection set, the change that meets the maximum mutual information optimization condition is selected as the decision result output by the model.
[0032] Previous energy consumption prediction and control methods are based on fixed models, and dynamic modeling is often not considered. However, this method is not limited to a fixed graph structure in the prediction process, but sets up a graph learning process, which continuously learns based on the historical data of the variables and gives the graph representation that best matches the actual data.
[0033] Compared with other prediction methods based on machine learning, the present invention adopts a preview structural model to describe the mutual influence relationship between variables, and captures the two-way effects between factors by defining two-way subgroups, which can better model the influence relationship between variables and improve the prediction effect.
[0034] Embodiment 2, the present invention provides a power data energy consumption prediction and control method based on preview machine learning, specifically comprising: The specific implementation process of the present invention includes a preview diagram representation of energy consumption prediction control problems, a dynamic update process of preview diagram parameters based on Bayesian learning, and a decision model for model optimization control. Figure 1 This is a block diagram of the enterprise energy consumption prediction and control method based on power data of preview machine learning in the present invention.
[0035] Preview diagram representation of energy consumption prediction control problem The present invention uses an example to illustrate the specific implementation method. The example data set is a nine-dimensional data set, which contains data on comprehensive energy consumption, local GDP, clean energy ratio, total energy consumption of various raw materials, and corresponding enterprise energy loss. In this example, in order to reduce the energy loss of the enterprise as much as possible to improve its operating efficiency, it is necessary to adjust The six influencing factors of energy consumption prediction and control can be used to control the energy consumption within a reasonable range. Figure 2 The preview diagram of the initial structure of the example data set. Among them, the total energy consumption (Total), the local gross domestic product (GDP) and the proportion of clean energy (PCE, proportion of cleanenergy) are variables that cannot be changed, and it is only positively correlated to the size of the energy loss (Loss). The only variables that can be changed are the various energy consumptions: raw coal consumption / ton (Coal), crude oil consumption / ton (Oil), electricity consumption / 10,000 kilowatts (Electric), coke consumption / ton (Coke) and heat consumption / 10,000 cubic meters (Heat).
[0036] In the actual energy consumption process, there is also a mutual influence between various energy sources. Specifically, for the same comprehensive energy consumption, the total energy consumption includes the consumption and loss of various energy sources. The total energy loss (Loss) refers to the part of the energy that cannot be effectively utilized in the whole process, including losses due to equipment efficiency, transmission process and conversion process. Different types of energy (such as raw coal, crude oil, electricity, coke and heat) may show certain correlations in the energy conversion process, especially when they are all involved in the same industrial process or system. First of all, the comprehensive energy consumption (Total) can be regarded as the sum of all energy consumption, that is:
[0037] In this model, energy loss (Loss) is mainly reflected in the ineffective consumption of various energy sources in conversion and transmission. For example, the heat loss caused by resistance during the transmission of electricity can be included in the total loss. When raw coal and coke are burned, part of the heat energy may not be fully utilized due to low combustion efficiency. This part of the energy that is not effectively utilized will eventually be reflected in the total loss. In addition, there may also be heat loss during the transmission of heat. In the consumption process of each energy source, the effectively used energy (such as the actual use of electricity) will directly provide energy for various equipment, while the part that is not effectively used will be shown as loss. At the same time, the consumption of raw coal and crude oil may affect each other in different systems-if the use of coal increases, the use of oil or coke may be reduced, and vice versa; at the same time, the proportion of clean energy (PCE) and local gross domestic product (GDP) have a significant impact on energy loss (Loss). Clean energy such as wind energy, solar energy and hydropower can reduce total energy loss due to their high energy efficiency and low loss. When a region increases the proportion of clean energy in its energy structure, energy loss usually decreases. The impact of the increase in local GDP on energy loss depends on the economic structure and development mode. If high-energy-consuming industries dominate economic growth, energy consumption and loss will increase; but through technological progress and industrial upgrading, energy utilization efficiency can be improved, and GDP growth will not necessarily lead to more energy loss, and may even reduce energy waste. Therefore, the relationship between the consumption of various types of energy and the total energy consumption can be explained by energy conservation and energy conversion efficiency, and the mutual influence of various energy sources will also be reflected in the consumption ratio and total loss between different energy categories. Figure 2 One-way edges are used to represent the unilateral impact of various types of energy consumption, gross domestic product (GDP) and the proportion of clean energy (PCE) on energy loss (Loss), and two-way edges are used to connect the various types of energy consumption that influence each other, intuitively showing the interaction between them.
[0038] The structural equation of the entire preview diagram can be expressed by the following formula:
[0039] in, is the largest bi-cluster The parent node set of So is a multi-valued function of parameters, represents Gaussian noise.
[0040] Dynamically update preview image parameters The entire preview graph can be obtained through the Bayesian formula A posteriori estimate of the parameters:
[0041] in, express All node data collected before the round of decision making begins. At the same time, for the graph posterior have:
[0042] This example uses the IC algorithm to derive the Markov equivalence class of the Bayesian network from the sample data and constructs an initial structure graph that is consistent with the sample data. On this basis, by introducing bidirectional edges on the edges with uncertain directions, a set of candidate preview graphs that are indistinguishable from the sample data are generated as the basis for subsequent graph parameter estimation.
[0043] For the parameter posterior The representation is as follows:
[0044] This example uses the Bayesian Ridge regression method to learn a parameter distribution that fits the real data.
[0045] Decision Model for Model Optimization Control Preview diagram through intervention operation To modify all variables classified as modifiable variables , in order to achieve the control output result But in reality, not all variables are all modifiable, in the instance Figure 2 In the decision model, the variables that can be changed are set The collection is various energy consumption: coal consumption / ton (Coal), crude oil consumption / ton (Oil), electricity consumption / 10,000 kilowatts (Electric) and coke consumption / ton (Coke) and heat consumption / 10,000 cubic meters (Heat). The changeable value interval of is set to the set of all real numbers In this example, if the intervention decision is executed for a total of t rounds, only one intervention decision is made in each round. ,in Indicates that the setting sequence number is The intermediate variables operation, then all Defined as a graph A preview of round t on the previous one, writing on G, and call it a decision in the decision-making process. , the optimization goal of the entire energy consumption prediction problem can be expressed as the following constrained parameter optimization problem:
[0046] in, , that is, it is expected that the predicted energy consumption (Loss) remains within the 30th percentile of the overall energy consumption data.
[0047] This example maximizes the overall parameters of the SRM and the observed values revealed after the preview Mutual information between To prevent the SRM after the pre-evolution from deviating from the real data. At the same time, this method also sets artificial parameters To limit the lower bound of the optimization process, which means that the expected energy consumption after intervention is in the interval The probability is 50%.
[0048] To solve the optimization problem, this method samples a set of 500 independently distributed preview graphs. To estimate the expectation under the constraints:
[0049] in, is specified As a result, this step is divided into three stages to calculate the constraints.
[0050] (1) From what I have learned Posterior and set noise distribution The sampling results are both of size The independent and identically distributed parameter training set and validation set ; (2) By Preview and find satisfaction Change collection , integration Build candidate changesets ; (3) By Rehearsal to verify Change collection , and from Delete Change collection .
[0051] Furthermore, according to the linear structure expression between variables: (in , , , Reason and parameters The present invention finds all candidate changes that satisfy the constraint interval. , and make the best decision based on this.
[0052] Decision making for predictive energy control For the main optimization goals:
[0053] This method reduces the mutual information expression and converts it into the difference of conditional information entropy.
[0054] Based on the learned preview image parameters and the resulting candidate changeset In this step, the information entropy value is estimated by the Entropyestimate method And calculate each candidate change The mutual information is used to solve the optimization problem, and finally the change that satisfies the maximum mutual information optimization condition is selected. The decision result is the output of the model. The whole decision process flow chart is as follows: Figure 5 shown.
[0055] In yet another embodiment of the present invention, a power data energy consumption prediction control system based on preview machine learning is provided, which can be used to implement the above-mentioned power data energy consumption prediction control method based on preview machine learning. Specifically, the system includes: A preview diagram construction module is used to classify all factors affecting energy consumption into initial variables, process variables and result variables, and construct an initial preview diagram; The structural equation parameter acquisition module is used to estimate the overall posterior for the initial preview diagram and obtain the structural equation parameters based on the maximized parameter posterior; The data processing module is used to construct two sets of independent and identically distributed parameter training sets and validation sets based on the structural equation parameters, express the decision-making process as a constrained optimization problem about the changes, train and validate all the changes, and output a candidate change selection set; The result output module is used to select the change that meets the maximum mutual information optimization condition as the decision result output by the model according to the obtained candidate change selection set.
[0056] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.
[0057] In another embodiment of the present invention, a computer device is provided, the computer device including a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the power data energy consumption prediction control method based on pre-performance machine learning.
[0058] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the power data energy consumption prediction control method based on pre-performance machine learning in the above embodiment.
[0059] It will be appreciated by those skilled in the art 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. Furthermore, 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.
[0060] 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.
[0061] 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 including 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.
[0062] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A power data energy consumption prediction and control method based on preview machine learning, characterized in that: Includes: Classify all factors affecting energy consumption into initial variables, process variables and result variables, and construct an initial preview diagram; For the initial preview graph, the overall posterior is estimated, and the structural equation parameters are obtained based on maximizing the parameter posterior; According to the structural equation parameters, two sets of independent and identically distributed parameter training sets and validation sets are constructed, the decision-making process is expressed as a constrained optimization problem about the change, all changes are trained and validated, and a candidate change selection set is output; Based on the obtained candidate change selection set, the change that meets the maximum mutual information optimization condition is selected as the decision result output by the model.
2. The power data energy consumption prediction and control method based on preview machine learning according to claim 1 is characterized in that: Classify all factors that affect the energy consumption of enterprises into initial variables , process variables and the outcome variable , construct the initial preview graph In the preview diagram representation, the correlation between categorical variables is shown through the preview diagram. For each preview diagram, it is defined as: , where V represents a vertex and E represents an edge; Is Induced complete subgraph, any edge They are both bidirectional. middle, for All maximal bi-cliques in .
3. The power data energy consumption prediction and control method based on preview machine learning according to claim 2 is characterized in that: is a preview graph if and only if: [1] For any ; [2] , if there exists any point to , then for any Midpoint , both ; [3] Direction side A topological sort was performed; The structural equation is expressed as follows: in, is the largest bi-cluster The parent node set of So is a multi-valued function of parameters, represents Gaussian noise.
4. The power data energy consumption prediction and control method based on preview machine learning according to claim 1 is characterized in that: For the initial preview graph, estimate the overall posterior and obtain the structural equation parameters based on maximizing the parameter posterior, including: The Bayesian method is used to dynamically update the graph structure. Through the Bayesian formula, the entire preview graph The posterior estimate of the parameters is expressed as follows: in, express All data collected before the decision round begins, which consists of the initial observation data and related variables Composition; for graph posterior Adopt the approximation of bootstrapping, that is, learn multiple preview graphs on Bootstrap data to build a set of graphs And each graph element is weighted equally, then its approximate posterior expression is: For the parameter posterior Representation: in, is included The set of graphs with their transformation graphs after a given change, express One of the largest bi-cluster in , is given middle and A set of regression parameters and the posterior of the noise variance of the parent node in ; the Bayesian ridge regression method is used to learn the parameter distribution that better fits the real data.
5. The power data energy consumption prediction and control method based on preview machine learning according to claim 1 is characterized in that: According to the structural equation parameters, two sets of independent and identically distributed parameter training sets and validation sets are constructed. The decision-making process is expressed as a constrained optimization problem about the change. All changes are trained and validated, and a candidate change selection set is output, including: The optimization objective of the entire energy consumption prediction problem is expressed as the following constrained parameter optimization problem: in, Indicates all result variables set The expected interval of Preview The overall parameters and the observed values revealed after the change By maximizing the mutual information To avoid the graph structure after intervention changes from deviating from the real data; define is the intervention decision for each round of intervention execution, where Indicates that the setting sequence number is The intermediate variables operation; further, For preview The last time A preview of the round, and represented as a decision in the decision-making process; By sampling a set of independently distributed preview graph samples To estimate the expectation under the constraints: in, is specified The results produced.
6. The power data energy consumption prediction and control method based on preview machine learning according to claim 5 is characterized in that: Calculation of constraints: (1) From what I have learned Posterior and set noise distribution The sampling results are both of size The independent and identically distributed parameter training set and validation set ; (2) By Preview and find satisfaction Change collection , integration Build candidate changesets ; (3) By Rehearsal to verify Change collection , and from Delete Change collection ; Based on variables The expression finds candidate changes that satisfy the constraint interval .
7. The power data energy consumption prediction and control method based on preview machine learning according to claim 6 is characterized in that: Based on the obtained candidate change selection set, the changes that meet the maximum mutual information optimization condition are selected as the decision results output by the model, including: Power data , by giving all candidate changes , and for the main optimization goals Solve, where Use the Entropyestimate method to estimate the information entropy value and calculate the mutual information of each candidate change to solve the optimization problem, and finally select the change that meets the maximum mutual information optimization condition. The decision result as the output of the model.
8. The power data energy consumption prediction control system based on pre-performance machine learning is characterized by: include: A preview diagram construction module is used to classify all factors affecting energy consumption into initial variables, process variables and result variables, and construct an initial preview diagram; The structural equation parameter acquisition module is used to estimate the overall posterior for the initial preview diagram and obtain the structural equation parameters based on the maximized parameter posterior; The data processing module is used to construct two sets of independent and identically distributed parameter training sets and validation sets based on the structural equation parameters, express the decision-making process as a constrained optimization problem about the changes, train and validate all the changes, and output a candidate change selection set; The result output module is used to select the change that meets the maximum mutual information optimization condition as the decision result output by the model according to the obtained candidate change selection set.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the power data energy consumption prediction and control method based on preview machine learning as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the power data energy consumption prediction and control method based on preview machine learning as described in any one of claims 1 to 7 are implemented.