Manufacturing enterprise cluster energy consumption situation prediction method based on heterogeneous energy consumption characteristic analysis

Abstract

The invention discloses a manufacturing enterprise cluster energy consumption situation prediction method based on heterogeneous energy consumption characteristic analysis, and the method comprises the steps: collecting the power consumption data of a manufacturing enterprise, and constructing a power consumption data set Ei of the manufacturing enterprise and a whole power consumption data set E of the manufacturing enterprise cluster; constructing an energy consumption situation prediction network F of the manufacturing enterprise cluster, training by using Ei, storing a final parameter theta'during model fitting, and realizing analysis of heterogeneous energy consumption characteristics of different manufacturing enterprises in the manufacturing enterprise cluster; using E to train F, migrating a final parameter theta'as an initial parameter of training, and storing a final parameter theta during model fitting; and inputting manufacturing enterprise cluster overall power consumption data collected in real time into the trained F to realize manufacturing enterprise cluster overall energy consumption situation prediction based on heterogeneous energy consumption characteristic analysis. According to the method, on the basis of historical energy consumption data of the manufacturing enterprise cluster, heterogeneous energy consumption characteristics of the manufacturing enterprise cluster are analyzed, and overall energy consumption situation prediction of the manufacturing enterprise cluster is realized.

Description

Technical Field

[0001] The present invention relates to cross-technical fields such as machine learning, industrial manufacturing, and demand-side management, and in particular to a method for predicting energy consumption status of a manufacturing enterprise cluster based on analysis of heterogeneous energy consumption characteristics. Background Art

[0002] Energy consumption situation prediction refers to the pre-analysis of the total energy consumption and intensity of the energy consumption side in each future period by the energy supply side, such as the power company. It is the core of ensuring the stability of the energy system and the key technology for coordinating the balance of energy supply and consumption. A typical manufacturing enterprise cluster consists of multiple manufacturing enterprises that perform different types of manufacturing tasks, and their overall energy consumption is large and the intensity is high. Therefore, predicting the overall energy consumption situation can better promote peak shaving and valley filling and maintain the stability of the energy system. However, due to the diversity of product production and manufacturing modes of different manufacturing enterprises within the manufacturing cluster, there are significant heterogeneity differences in their energy consumption characteristics, which leads to the overall energy consumption situation of the cluster showing a complex periodic law. Therefore, how to analyze the heterogeneous energy consumption characteristics of multiple enterprises in the manufacturing enterprise cluster, analyze the complex energy consumption laws of the manufacturing enterprise cluster under its influence, and thus realize the accurate prediction of the overall energy consumption situation of the manufacturing enterprise cluster is a key problem that needs to be solved urgently.

[0003] The research on energy consumption situation prediction has always been a hot area in the industry, and it is mainly used for advance energy planning and management on the energy supply side. The current situation prediction methods in the manufacturing field are mainly applied to manufacturing processes, manufacturing equipment and manufacturing workshops. Energy consumption prediction is achieved by mining the data association between the energy consumption characteristics of specific processes, equipment and workshops and covariates such as process parameters and manufacturing tasks. Traditional prediction methods only realize energy consumption prediction of single processes, equipment and workshops, and cannot provide theoretical and data support for the global energy consumption situation prediction of manufacturing enterprise clusters on the energy supply side.

[0004] At present, the global and systematic energy consumption situation prediction methods are mainly concentrated in the fields related to people's livelihood, such as analyzing the energy consumption patterns and trends of buildings, urban areas or national regions. The existing methods also realize situation prediction by modeling the energy consumption situation in time, thereby mining the data association between energy consumption and multiple covariates. However, it mainly considers the impact of periodic spatiotemporal characteristics such as seasons and locations on the overall energy consumption situation. The difficulty of predicting the energy consumption situation of manufacturing enterprise clusters is mainly due to the complex energy consumption patterns caused by manufacturing heterogeneity caused by fluctuations such as production orders and tasks. The periodic patterns that can be mined are sparse, so it cannot be applied to the energy consumption situation prediction problem of manufacturing enterprise clusters. At the same time, the existing methods all rely on the association of covariates to realize situation prediction, but due to the complex types and structures of the internal energy consumption units of manufacturing enterprise clusters, it is difficult to screen covariates and collect data, and ensure the timeliness and accuracy required by demand response. Summary of the invention

[0005] The present invention provides a method for predicting the energy consumption status of a manufacturing enterprise cluster based on analysis of heterogeneous energy consumption characteristics. Based on the historical energy consumption data of the manufacturing enterprise cluster, the heterogeneous energy consumption characteristics of the manufacturing enterprise cluster are analyzed to realize the overall energy consumption status prediction of the manufacturing enterprise cluster.

[0006] The present invention provides a method for predicting energy consumption status of a manufacturing enterprise cluster based on analysis of heterogeneous energy consumption characteristics, comprising:

[0007] Collect the electricity consumption data of the manufacturing enterprise, and construct the electricity consumption data set E of the manufacturing enterprise based on the electricity consumption data i and a manufacturing enterprise cluster overall electricity consumption data set E; wherein the manufacturing enterprise cluster includes a plurality of manufacturing enterprises;

[0008] Construct a manufacturing enterprise cluster energy consumption situation prediction network F, and use the electricity consumption dataset E of the manufacturing enterprises i Conduct training, optimize network parameters by minimizing the loss function, save the final parameters θ′ during model fitting, and analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster;

[0009] The manufacturing enterprise cluster energy consumption situation prediction network is trained using the manufacturing enterprise cluster overall electricity consumption data set E, the final parameter θ′ is transferred as the initial parameter of the training, the network parameter is optimized by minimizing the loss function, and the final parameter θ during model fitting is saved;

[0010] The overall electricity consumption data of the manufacturing enterprise cluster collected in real time is input into the trained manufacturing enterprise cluster energy consumption situation prediction network F to realize the overall energy consumption situation prediction of the manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics.

[0011] Furthermore, the electricity consumption data of the manufacturing enterprise is collected, and the electricity consumption data set E of the manufacturing enterprise is constructed based on the electricity consumption data. i and the manufacturing enterprise cluster overall electricity consumption dataset E,

[0012] The granularity of the electricity consumption data is 15 minutes and the unit is kW·h;

[0013] The electricity consumption dataset E of the manufacturing enterprise i The overall electricity consumption dataset E of the manufacturing enterprise cluster is:

[0014]

[0015] E = {x1, x2, ..., x n}

[0016] in, represents the electricity consumption data of the nth manufacturing enterprise at the nth time step, x tn Represents the electricity consumption data of the manufacturing enterprise cluster at the nth time step.

[0017] Furthermore, the manufacturing enterprise cluster energy consumption situation prediction network includes a characteristic analysis module M a And situation prediction module M p .

[0018] Furthermore, the characteristic analysis module M a The steps include:

[0019] Energy consumption data frequency selection:

[0020]

[0021] Where x[n] represents the nth input time series energy consumption data, X[k] represents the kth frequency information after transformation, N represents the sample length of the input energy consumption data, and Complex roots of unity;

[0022] Two-dimensional reorganization of energy consumption data:

[0023]

[0024] Among them, A[k] represents the amplitude corresponding to the kth frequency, x X[k] represents the power consumption data of the kth input, f avg (·) represents the mean function, g amp (·) represents the amplitude acquisition function;

[0025]

[0026] Wherein, Lx represents the input length of energy consumption data, and P[k] represents the kth frequency;

[0027] Heterogeneous Characterization Extraction:

[0028]

[0029] in, represents the j-th two-dimensional convolution operation, x i Represents the energy consumption data of the first segment of the input, x′ i Represents the output data after the convolution operation;

[0030] One-dimensional reorganization of energy consumption data and importance weighting:

[0031] S i =Softmax(A i )

[0032] Among them, Softmax(·) represents the Softmax function, A i represents the i-th amplitude, S i Indicates the importance score corresponding to the i-th amplitude;

[0033]

[0034] in, and They represent the energy consumption data after one-dimensional reorganization and the energy consumption data after weighting respectively;

[0035]

[0036] In the formula, x f Represents the final output.

[0037] Furthermore, the situation prediction module M p It contains a unidirectional LSTM layer and a linear classification layer, which are used to perform temporal modeling and prediction on the energy consumption data that have completed the heterogeneous energy consumption characteristic analysis.

[0038] Furthermore, the manufacturing enterprise cluster energy consumption situation prediction network F is constructed, and the power consumption data set E of the manufacturing enterprise is used i Training is performed to optimize network parameters by minimizing the loss function, saving the final parameters θ′ during model fitting, and implementing the steps of analyzing the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster, including:

[0039] Set the number of pre-training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F;

[0040] For any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value:

[0041]

[0042] in, represents the predicted consumption value of ti time step, x ti Represents the actual consumption value of ti time step;

[0043] After each round of training, the parameters θ′ of the convolutional neural network are optimized:

[0044]

[0045] Among them, α represents the learning rate, (θ')' represents the parameters of the optimized convolutional neural network F';

[0046] Save the final parameters θ' of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of pre-training.

[0047] Furthermore, the step of using the manufacturing enterprise cluster overall electricity consumption data set E to train the manufacturing enterprise cluster energy consumption situation prediction network F, migrating the final parameter θ' as the initial parameter of the training, optimizing the network parameters by minimizing the loss function, and saving the final parameter θ during model fitting specifically includes:

[0048] Set the number of training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F;

[0049] For any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value:

[0050]

[0051] After each round of training, the parameters θ of the manufacturing enterprise cluster energy consumption situation prediction network F are optimized:

[0052]

[0053] Among them, (θ)' represents the parameters of the optimized manufacturing enterprise cluster energy consumption situation prediction network F;

[0054] The final parameters θ of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of training are saved, and the training is completed.

[0055] The present invention also provides a manufacturing enterprise cluster energy consumption situation prediction device based on heterogeneous energy consumption characteristic analysis, comprising:

[0056] The collection module is used to collect the electricity consumption data of the manufacturing enterprise and construct the electricity consumption data set E of the manufacturing enterprise based on the electricity consumption data. i and a manufacturing enterprise cluster overall electricity consumption data set E; wherein the manufacturing enterprise cluster includes a plurality of manufacturing enterprises;

[0057] A construction module is used to construct a manufacturing enterprise cluster energy consumption situation prediction network F and use the electricity consumption dataset E of the manufacturing enterprise i Conduct training, optimize network parameters by minimizing the loss function, save the final parameter θ' during model fitting, and analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster;

[0058] A training module is used to train the manufacturing enterprise cluster energy consumption situation prediction network F using the manufacturing enterprise cluster overall electricity consumption data set E, migrate the final parameter θ' as the initial parameter of the training, optimize the network parameters by minimizing the loss function, and save the final parameter θ during model fitting;

[0059] The prediction module is used to input the real-time collected overall electricity consumption data of the manufacturing enterprise cluster into the trained manufacturing enterprise cluster energy consumption situation prediction network F, so as to realize the overall energy consumption situation prediction of the manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics.

[0060] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0061] The present invention also provides a computer-readable storage medium on which a computer program is stored, and the computer program implements the steps of the above method when executed by a processor.

[0062] The beneficial effects of the present invention are:

[0063] 1. From the perspective of data science, the present invention transforms the problem of predicting the energy consumption status of manufacturing enterprises into an autoregressive problem that only requires historical energy consumption data, greatly reducing the difficulty and requirements of data collection.

[0064] 2. The present invention adopts the idea of ​​two-dimensional reorganization of one-dimensional time series data and extracts the cross-period variation representation of energy consumption time series data through convolution operation, thereby realizing the analysis of the heterogeneous energy consumption characteristics of different enterprises in the cluster.

[0065] 3. Through the transfer learning mechanism, the present invention assists the data-driven prediction network to build the overall energy consumption time series correlation of the cluster based on the reference prior energy consumption characteristic knowledge, thereby realizing accurate energy consumption situation prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 The present invention is a flow chart of a method for predicting energy consumption status of a manufacturing enterprise cluster based on analysis of heterogeneous energy consumption characteristics.

[0067] Figure 2 It is a structural schematic diagram of a manufacturing enterprise cluster energy consumption situation prediction device based on heterogeneous energy consumption characteristic analysis of the present invention.

[0068] Figure 3 It is a schematic diagram of the internal structure of the computer device in the present invention.

[0069] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0070] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0071] The energy consumption situation prediction of manufacturing enterprise clusters can be further understood as a time series autoregression problem that depends on the relationship between time steps, that is, through the historical time step T n The total energy consumption within the future time step T is predicted n+i The total energy consumption of a manufacturing enterprise cluster. Traditional time series prediction methods build deep neural networks, such as recurrent neural networks and convolutional neural networks, to mine data dependency associations between time steps, but these methods cannot effectively analyze and predict sequences with time changes. Therefore, the main technical problem to be solved by the present invention is to analyze the heterogeneous energy consumption characteristics based on the historical energy consumption data of the manufacturing enterprise cluster, so as to realize the prediction of the overall energy consumption situation of the manufacturing enterprise cluster.

[0072] like Figure 1 As shown, the present invention provides a method for predicting energy consumption status of a manufacturing enterprise cluster based on heterogeneous energy consumption characteristics analysis, comprising:

[0073] S1. Collect the electricity consumption data of the manufacturing enterprise and construct the electricity consumption data set E of the manufacturing enterprise based on the electricity consumption data i and a manufacturing enterprise cluster overall electricity consumption data set E; wherein the manufacturing enterprise cluster includes a plurality of manufacturing enterprises;

[0074] By installing electric meters in each manufacturing enterprise within the manufacturing enterprise cluster, electricity consumption data with a granularity of 15 minutes and a unit of kilowatt-hour (kW·h) is collected. Therefore, the granularity of the electricity consumption data is 15 minutes and a unit of kW·h.

[0075] The electricity consumption dataset E of a single manufacturing enterprise i The overall electricity consumption dataset E of the manufacturing enterprise cluster is:

[0076]

[0077] E={x1,x2,....,x n}

[0078] in, represents the electricity consumption data of the i-th manufacturing enterprise at the n-th time step, x tn Represents the electricity consumption data of the manufacturing enterprise cluster at the nth time step.

[0079] S2. Construct a manufacturing enterprise cluster energy consumption situation prediction network F, and use the electricity consumption dataset E of the manufacturing enterprises i Conduct training, optimize network parameters by minimizing the loss function, save the final parameter θ' during model fitting, and analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster;

[0080] The manufacturing enterprise cluster energy consumption situation prediction network F includes a characteristic analysis module M a And situation prediction module M p .

[0081] Step S2 includes:

[0082] S201, in order to analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster, a manufacturing enterprise cluster energy consumption situation prediction network F is constructed, wherein the characteristic analysis module M a It includes the following 5 operations:

[0083] 1. Energy consumption data frequency selection:

[0084]

[0085] Where x[n] represents the nth input time series energy consumption data, X[k] represents the kth frequency information after transformation, N represents the sample length of the input energy consumption data, and Complex roots of unity;

[0086] 2. Two-dimensional reorganization of energy consumption data:

[0087]

[0088] Among them, A[k] represents the amplitude corresponding to the kth frequency, x X[k] represents the power consumption data of the kth input, f avg (·) represents the mean function, g amp (·) represents the amplitude acquisition function;

[0089]

[0090] Wherein, Lx represents the input length of energy consumption data, and P[k] represents the kth frequency;

[0091] 3. Heterogeneous Characterization and Extraction:

[0092]

[0093] in, represents the j-th two-dimensional convolution operation, x i represents the energy consumption data of the i-th segment of the input, x' i Represents the output data after the convolution operation;

[0094] 4 and 5. One-dimensional reorganization of energy consumption data and importance weighting:

[0095] S i =Softmax(A i )

[0096] Among them, Softmax(·) represents the Softmax function, A i represents the i-th amplitude, S i Indicates the importance score corresponding to the i-th amplitude;

[0097]

[0098] in, and They represent the energy consumption data after one-dimensional reorganization and the energy consumption data after weighting respectively;

[0099]

[0100] In the formula, x f Represents the final output.

[0101] The situation prediction module M p It contains a unidirectional LSTM layer and a linear classification layer, which are used to perform temporal modeling and prediction on the energy consumption data that have completed the heterogeneous energy consumption characteristic analysis.

[0102] S202: constructing a manufacturing enterprise cluster energy consumption situation prediction network F and using the manufacturing enterprise power consumption data set E i Training is performed, that is, the number of pre-training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F are set; for any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value:

[0103]

[0104] in, represents the predicted consumption value of ti time step, x ti Represents the actual consumption value of ti time step;

[0105] S203. After each round of training, the parameters θ' of the convolutional neural network F are optimized:

[0106]

[0107] Among them, α represents the learning rate, (θ')' represents the parameters of the optimized convolutional neural network F';

[0108] S204, saving the final parameters θ' of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of pre-training.

[0109] S3, using the manufacturing enterprise cluster overall electricity consumption data set E to train the manufacturing enterprise cluster energy consumption situation prediction network F, migrating the final parameter θ' as the initial parameter of the training, optimizing the network parameters by minimizing the loss function, and saving the final parameter θ during model fitting;

[0110] S301, using the manufacturing enterprise cluster overall electricity consumption data set E to train the manufacturing enterprise cluster energy consumption situation prediction network F, setting the number of training rounds e of the manufacturing enterprise cluster energy consumption situation prediction network F and the input data batch size B;

[0111] S302. For any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value:

[0112]

[0113] S303. After each round of training, the parameter θ of the manufacturing enterprise cluster energy consumption situation prediction network F is optimized:

[0114]

[0115] Among them, (θ)' represents the parameters of the optimized manufacturing enterprise cluster energy consumption situation prediction network F;

[0116] S304. Save the final parameter θ of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of training, and the training is completed.

[0117] S4. Input the overall electricity consumption data of the manufacturing enterprise cluster collected in real time into the trained manufacturing enterprise cluster energy consumption situation prediction network F to realize the overall energy consumption situation prediction of the manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics.

[0118] The key to accurately predicting the energy consumption status of manufacturing enterprise clusters is to enhance the model's analysis of the energy consumption characteristics of different manufacturing enterprises in the cluster, so that the prediction model can accurately mine the dependency associations between time steps in historical energy consumption data. The present invention makes full use of the convolutional neural network's ability to extract two-dimensional data representations and the recurrent neural network's ability to model implicit time associations, and correspondingly solves the problem of the difficulty in predicting the overall energy consumption status of the cluster caused by the large differences in energy consumption characteristics caused by the manufacturing heterogeneity of different enterprises in the manufacturing enterprise cluster.

[0119] The present invention discloses a method for predicting the energy consumption situation of a manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics. By collecting the energy consumption data of a manufacturing enterprise cluster, an enterprise energy consumption data set and a cluster energy consumption data set are constructed; a manufacturing enterprise energy consumption situation prediction network including a characteristic analysis module and a situation prediction module is constructed; the enterprise energy consumption data set is used to pre-train the manufacturing enterprise cluster energy consumption situation prediction network, the mean square error function is used to optimize the model parameters, and the final parameters of the model are saved to realize the analysis of the heterogeneous energy consumption characteristics of the enterprises in the cluster; the cluster energy consumption data set is used to train the manufacturing enterprise energy consumption situation prediction network, the pre-training parameters are migrated as the initial parameters of the training, the mean square error function is used to optimize the model parameters, and the final parameters of the model are saved to realize the accurate prediction of the energy consumption situation of the manufacturing enterprise cluster. The present invention can realize the prediction of the energy consumption situation of a manufacturing enterprise cluster including multiple manufacturing enterprises.

[0120] like Figure 2 As shown, the present invention also provides a manufacturing enterprise cluster energy consumption situation prediction device based on heterogeneous energy consumption characteristic analysis, comprising:

[0121] The collection module 1 is used to collect the electricity consumption data of the manufacturing enterprise and construct the electricity consumption data set E of the manufacturing enterprise based on the electricity consumption data. i and a manufacturing enterprise cluster overall electricity consumption data set E; wherein the manufacturing enterprise cluster includes a plurality of manufacturing enterprises;

[0122] Construction module 2 is used to construct a manufacturing enterprise cluster energy consumption situation prediction network F and use the power consumption dataset E of the manufacturing enterprise i Conduct training, optimize network parameters by minimizing the loss function, save the final parameter θ' during model fitting, and analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster;

[0123] The training module 3 is used to train the manufacturing enterprise cluster energy consumption situation prediction network F using the manufacturing enterprise cluster overall electricity consumption data set E, migrate the final parameter θ' as the initial parameter of the training, optimize the network parameters by minimizing the loss function, and save the final parameter θ during model fitting;

[0124] The prediction module 4 is used to input the real-time collected overall electricity consumption data of the manufacturing enterprise cluster into the trained manufacturing enterprise cluster energy consumption situation prediction network F, so as to realize the overall energy consumption situation prediction of the manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics.

[0125] In one embodiment, in the acquisition module 1,

[0126] The granularity of the electricity consumption data is 15 minutes and the unit is kW·h;

[0127] The electricity consumption dataset E of the manufacturing enterprisei The overall electricity consumption dataset E of the manufacturing enterprise cluster is:

[0128]

[0129] E={x1,x2,....,x n}

[0130] in, represents the electricity consumption data of the i-th manufacturing enterprise at the n-th time step, x tn Represents the electricity consumption data of the manufacturing enterprise cluster at the nth time step.

[0131] In one embodiment, the manufacturing enterprise cluster energy consumption situation prediction network F includes a characteristic analysis module M a And situation prediction module M p .

[0132] In one embodiment, in the manufacturing enterprise cluster energy consumption situation prediction network F of the construction module 2, the characteristic analysis module M a The steps include:

[0133] Energy consumption data frequency selection:

[0134]

[0135] Where x[n] represents the nth input time series energy consumption data, X[k] represents the kth frequency information after transformation, N represents the sample length of the input energy consumption data, and Complex roots of unity;

[0136] Two-dimensional reorganization of energy consumption data:

[0137]

[0138] Among them, A[k] represents the amplitude corresponding to the kth frequency, x X[k] represents the power consumption data of the kth input, f avg (·) represents the mean function, g amp (·) represents the amplitude acquisition function;

[0139]

[0140] Wherein, Lx represents the input length of energy consumption data, and P[k] represents the kth frequency;

[0141] Heterogeneous Characterization Extraction:

[0142]

[0143] in, represents the j-th two-dimensional convolution operation, x i represents the energy consumption data of the i-th segment of the input, x' i Represents the output data after the convolution operation;

[0144] One-dimensional reorganization of energy consumption data and importance weighting:

[0145] S i =Softmax(A i )

[0146] Among them, Softmax(·) represents the Softmax function, A i represents the i-th amplitude, S i Indicates the importance score corresponding to the i-th amplitude;

[0147]

[0148] in, and They represent the energy consumption data after one-dimensional reorganization and the energy consumption data after weighting respectively;

[0149]

[0150] In the formula, x f Represents the final output.

[0151] In one embodiment, in the manufacturing enterprise cluster energy consumption situation prediction network F of the construction module 2, the situation prediction module M p It contains a unidirectional LSTM layer and a linear classification layer, which are used to perform temporal modeling and prediction on the energy consumption data that have completed the heterogeneous energy consumption characteristic analysis.

[0152] In one embodiment, the construction module 2 specifically includes:

[0153] A setting unit, used to set the number of pre-training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F;

[0154] For any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value:

[0155]

[0156] in, represents the predicted consumption value of ti time step, x ti Represents the actual consumption value of ti time step;

[0157] The first optimization unit is used to optimize the parameters θ' of the convolutional neural network F after each round of training:

[0158]

[0159] Among them, α represents the learning rate, (θ')' represents the parameters of the optimized convolutional neural network F';

[0160] The first storage unit is used to store the final parameter θ' of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of pre-training.

[0161] In one embodiment, the steps of using the manufacturing enterprise cluster overall electricity consumption data set E to train the manufacturing enterprise cluster energy consumption situation prediction network F, migrating the final parameter θ' as the initial parameter of the training, optimizing the network parameters by minimizing the loss function, and saving the final parameter θ during model fitting specifically include:

[0162] A setting unit, used to set the number of training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F;

[0163] The calculation unit is used to calculate the loss between the predicted value and the true value using the loss function L for any batch of data in any round of training:

[0164]

[0165] The second optimization unit is used to optimize the parameter θ of the manufacturing enterprise cluster energy consumption situation prediction network F after each round of training:

[0166]

[0167] Among them, (θ)' represents the parameters of the optimized manufacturing enterprise cluster energy consumption situation prediction network F;

[0168] The second saving unit is used to save the final parameters θ of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of training, and the training is completed.

[0169] The above modules and units are used to execute the corresponding steps in the above method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis. The specific implementation method is described in the above method embodiment and will not be repeated here.

[0170] like Figure 3 As shown, the present invention also provides a computer device, which can be a server, and its internal structure can be as shown in Figure 3As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store all data required for the process of the method for predicting the energy consumption status of a cluster of manufacturing enterprises based on the analysis of heterogeneous energy consumption characteristics. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the method for predicting the energy consumption status of a cluster of manufacturing enterprises based on the analysis of heterogeneous energy consumption characteristics is implemented.

[0171] Those skilled in the art will understand that Figure 3 The structure shown in is merely a block diagram of a portion of the structure related to the present application solution and does not constitute a limitation on the computer device to which the present application solution is applied.

[0172] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, any one of the above-mentioned methods for predicting energy consumption status of a manufacturing enterprise cluster based on analysis of heterogeneous energy consumption characteristics is implemented.

[0173] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0174] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0175] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting energy consumption status of manufacturing enterprise clusters based on analysis of heterogeneous energy consumption characteristics, characterized in that: include: Collect the electricity consumption data of the manufacturing enterprise, and construct the electricity consumption data set E of the manufacturing enterprise based on the electricity consumption data i and a manufacturing enterprise cluster overall electricity consumption data set E; wherein the manufacturing enterprise cluster includes a plurality of manufacturing enterprises; Construct a manufacturing enterprise cluster energy consumption situation prediction network F, and use the electricity consumption dataset E of the manufacturing enterprises i Conduct training, optimize network parameters by minimizing the loss function, save the final parameters θ′ during model fitting, and analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster; The manufacturing enterprise cluster energy consumption situation prediction network is trained using the manufacturing enterprise cluster overall electricity consumption data set E, the final parameter θ′ is transferred as the initial parameter of the training, the network parameter is optimized by minimizing the loss function, and the final parameter θ during model fitting is saved; The overall electricity consumption data of the manufacturing enterprise cluster collected in real time is input into the trained manufacturing enterprise cluster energy consumption situation prediction network F to realize the overall energy consumption situation prediction of the manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics.

2. The method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis according to claim 1 is characterized in that: The electricity consumption data of the manufacturing enterprise is collected, and the electricity consumption data set E of the manufacturing enterprise is constructed based on the electricity consumption data. i and the manufacturing enterprise cluster overall electricity consumption dataset E, The granularity of the electricity consumption data is 15 minutes and the unit is kW·h; The electricity consumption dataset E of the manufacturing enterprise i The overall electricity consumption dataset E of the manufacturing enterprise cluster is: And={x1,x2,...,x n } in, represents the electricity consumption data of the nth manufacturing enterprise at the nth time step, x tn Represents the electricity consumption data of the manufacturing enterprise cluster at the nth time step.

3. The method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis according to claim 1 is characterized in that: The manufacturing enterprise cluster energy consumption situation prediction network includes a characteristic analysis module M a And situation prediction module M p .

4. The method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis according to claim 3 is characterized in that: The characteristic analysis module M a The steps include: Energy consumption data frequency selection: Where x[n] represents the nth input time series energy consumption data, X[k] represents the kth frequency information after transformation, N represents the sample length of the input energy consumption data, and Complex roots of unity; Two-dimensional reorganization of energy consumption data: Among them, A[k] represents the amplitude corresponding to the kth frequency, x X[k] represents the power consumption data of the kth input, f avg (·) represents the mean function, g amp (·) represents the amplitude acquisition function; Wherein, Lx represents the input length of energy consumption data, and P[k] represents the kth frequency; Heterogeneous Characterization Extraction: in, represents the j-th two-dimensional convolution operation, x i represents the energy consumption data of the i-th segment of the input, x′ i Represents the output data after the convolution operation; One-dimensional reorganization of energy consumption data and importance weighting: S i =Softmax(A i ) Among them, Softmax(·) represents the Softmax function, A i represents the i-th amplitude, S i Indicates the importance score corresponding to the i-th amplitude; in, and They represent the energy consumption data after one-dimensional reorganization and the energy consumption data after weighting respectively; In the formula, x f Represents the final output.

5. The method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis according to claim 3 is characterized in that: The situation prediction module M p It contains a unidirectional LSTM layer and a linear classification layer, which are used to perform temporal modeling and prediction on the energy consumption data that have completed the heterogeneous energy consumption characteristic analysis.

6. The method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis according to claim 3 is characterized in that: The manufacturing enterprise cluster energy consumption situation prediction network F is constructed, and the power consumption data set E of the manufacturing enterprise is used i Training is performed to optimize network parameters by minimizing the loss function, saving the final parameter θ' during model fitting, and implementing the steps of analyzing the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster, including: Set the number of pre-training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F; For any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value: in, represents the predicted consumption value of ti time step, x ti Represents the actual consumption value of ti time step; After each round of training, the parameters θ' of the convolutional neural network F are optimized: Among them, α represents the learning rate, (θ')' represents the parameters of the optimized convolutional neural network F'; Save the final parameters θ' of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of pre-training.

7. The method for predicting energy consumption status of manufacturing enterprise clusters based on heterogeneous energy consumption characteristics analysis according to claim 6 is characterized in that: The step of using the manufacturing enterprise cluster overall electricity consumption data set E to train the manufacturing enterprise cluster energy consumption situation prediction network F, migrating the final parameter θ' as the initial parameter of the training, optimizing the network parameters by minimizing the loss function, and saving the final parameter θ during model fitting specifically includes: Set the number of training rounds e and the input data batch size B of the manufacturing enterprise cluster energy consumption situation prediction network F; For any batch of data in any round of training, the loss function L is used to calculate the loss between the predicted value and the true value: After each round of training, the parameters θ of the manufacturing enterprise cluster energy consumption situation prediction network F are optimized: Among them, (θ)' represents the parameters of the optimized manufacturing enterprise cluster energy consumption situation prediction network F; The final parameters θ of the manufacturing enterprise cluster energy consumption situation prediction network F after e rounds of training are saved, and the training is completed.

8. A manufacturing enterprise cluster energy consumption situation prediction device based on heterogeneous energy consumption characteristics analysis, characterized in that: include: The collection module is used to collect the electricity consumption data of the manufacturing enterprise and construct the electricity consumption data set E of the manufacturing enterprise based on the electricity consumption data. i and a manufacturing enterprise cluster overall electricity consumption data set E; wherein the manufacturing enterprise cluster includes a plurality of manufacturing enterprises; A construction module is used to construct a manufacturing enterprise cluster energy consumption situation prediction network F and use the electricity consumption dataset E of the manufacturing enterprise i Conduct training, optimize network parameters by minimizing the loss function, save the final parameter θ' during model fitting, and analyze the heterogeneous energy consumption characteristics of different manufacturing enterprises within the manufacturing enterprise cluster; A training module is used to train the manufacturing enterprise cluster energy consumption situation prediction network F using the manufacturing enterprise cluster overall electricity consumption data set E, migrate the final parameter θ' as the initial parameter of the training, optimize the network parameters by minimizing the loss function, and save the final parameter θ during model fitting; The prediction module is used to input the real-time collected overall electricity consumption data of the manufacturing enterprise cluster into the trained manufacturing enterprise cluster energy consumption situation prediction network F, so as to realize the overall energy consumption situation prediction of the manufacturing enterprise cluster based on the analysis of heterogeneous energy consumption characteristics.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images