Integrated prediction method of energy consumption in industrial parks based on multi-source and multi-scale information fusion

Through multi-source and multi-scale information fusion and adversarial training strategies, a comprehensive integrated model is built, which solves the problem of limited energy consumption prediction accuracy in traditional methods, and achieves higher accuracy and stability, which is suitable for complex industrial systems.

CN118690921BActive Publication Date: 2024-11-05BEIJING INST OF TECH +3
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411162934.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-05
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

Traditional industrial energy consumption prediction methods are mainly based on a single data source or single-scale information, and cannot fully consider the comprehensive impact of multiple factors on energy consumption, resulting in limited prediction accuracy and difficult to meet the needs of complex and diverse industrial systems.

Method used

A method based on multi-source and multi-scale information fusion is adopted to collect and preprocess industrial energy consumption data, a comprehensive integrated model of hybrid deep neural network model and physical empirical model is constructed, and an adversarial training strategy is adopted to evaluate the prediction results in combination with composite evaluation factors.

Benefits of technology

It improves the accuracy and stability of industrial energy consumption prediction, enhances the robustness and generalization capabilities of the model, and provides more comprehensive and reliable energy consumption prediction support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118690921B_ABST
    Figure CN118690921B_ABST
Patent Text Reader

Abstract

The present application proposes an integrated prediction method for energy consumption in industrial parks based on multi-source and multi-scale information fusion, the method comprising: collecting multi-source data on industrial energy consumption, and performing data preprocessing based on periodic alignment on the multi-source data on industrial energy consumption; constructing a multi-source and multi-scale comprehensive integration model, wherein the multi-source and multi-scale comprehensive integration model is obtained by fusion of a hybrid deep neural network model and a physical empirical model; for a large amount of unlabeled data and a small amount of labeled data in the multi-source data on industrial energy consumption, an adversarial training strategy is used to train the multi-source and multi-scale comprehensive integration model; the preprocessed multi-source data on industrial energy consumption is input into the trained multi-source and multi-scale comprehensive integration model to obtain a predicted energy consumption result, and the predicted energy consumption result is evaluated according to a composite evaluation factor. By utilizing multi-source and multi-scale information, the present application can more comprehensively consider the impact of various factors on industrial energy consumption.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of industrial energy consumption forecasting, and in particular to an integrated forecasting method for industrial park energy consumption based on multi-source and multi-scale information fusion. Background Art

[0002] Traditional industrial energy consumption forecasting methods are mainly based on a single data source or single-scale information, such as time series analysis methods that only consider historical energy consumption data or degree-day models based on weather data.

[0003] However, these methods often fail to fully consider the combined impact of multiple factors on energy consumption, resulting in limited prediction accuracy. In addition, due to the complexity and diversity of industrial systems, a single method is often difficult to meet actual needs, and it is necessary to comprehensively consider multi-source and multi-scale information to improve prediction accuracy and reliability. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first purpose of this application is to propose an integrated prediction method for industrial park energy consumption based on multi-source and multi-scale information fusion to provide a comprehensive and reliable solution for industrial energy consumption prediction.

[0006] The second objective of this application is to propose an integrated prediction device for energy consumption of industrial parks based on multi-source and multi-scale information fusion.

[0007] The third objective of the present application is to provide an electronic device.

[0008] A fourth objective of the present application is to provide a computer-readable storage medium.

[0009] To achieve the above objectives, the first embodiment of the present application proposes an integrated prediction method for energy consumption of an industrial park based on multi-source and multi-scale information fusion, including:

[0010] Collecting multi-source data of industrial energy consumption, and performing data preprocessing based on periodic registration on the multi-source data of industrial energy consumption, wherein the multi-source data of industrial energy consumption includes weather forecast information, on-site temperature information of industrial parks, humidity information, power consumption data, production line production data, and physical parameter data of industrial parks;

[0011] Constructing a multi-source multi-scale comprehensive integration model, wherein the multi-source multi-scale comprehensive integration model is obtained by fusing a hybrid deep neural network model and a physical empirical model;

[0012] For a large amount of unlabeled data and a small amount of labeled data in the industrial energy consumption multi-source data, an adversarial training strategy is used to train the multi-source and multi-scale comprehensive integration model;

[0013] The preprocessed industrial energy consumption multi-source data is input into the trained multi-source and multi-scale comprehensive integration model to obtain the predicted energy consumption results, and the predicted energy consumption results are evaluated according to the composite evaluation factors.

[0014] Optionally, the collecting of multi-source data on industrial energy consumption includes:

[0015] Obtain weather data for the industrial park within a preset time in the future from the meteorological department or a third-party meteorological service provider;

[0016] Collect on-site temperature and humidity information of the industrial park through the environmental monitoring station in the industrial park;

[0017] Obtain power consumption data through electricity meters, smart grid management systems or energy management systems in industrial parks;

[0018] Provide production line production data based on the production management system or automated control system in the industrial park, including the output, start-up time, and downtime of each production line;

[0019] The physical parameter data of the industrial park is obtained through the equipment efficiency, thermal insulation performance, moisture retention performance, and production load fluctuations in the industrial park.

[0020] Optionally, the data preprocessing based on periodic registration for the industrial energy consumption multi-source data includes:

[0021] The industrial energy consumption multi-source data is time-series calibrated, and the timestamps of all the industrial energy consumption multi-source data are converted into a unified time format.

[0022] Optionally, the prediction process of the hybrid deep neural network model includes:

[0023] For the time series data consisting of the temperature information, humidity information, and power consumption data of the industrial park site, the Fourier decomposition method is used to decompose it into narrowband components of multiple channels. After Fourier decomposition, each time series data Decomposed into multiple narrowband components , the expression of Fourier decomposition process is:

[0024]

[0025] in, is time series data, are the Fourier coefficients, is the total number of data points in the time series data, It’s a point in time;

[0026] The convolutional neural network is used to perform convolution operations on each narrowband component after decomposition to extract local features. The expression is:

[0027]

[0028] in, is the convolution output, is the convolution kernel, is the bias, is the activation function;

[0029] The convolution output is pooled and the expression is:

[0030]

[0031]

[0032] in, is the pooled output set, For the Convolution output The corresponding pooling output;

[0033] Based on the dependency relationship between different channel components, adaptive channel extraction is performed on the pooling output, and the expression is:

[0034]

[0035]

[0036] in, is the feature set extracted adaptively, is the feature extracted adaptively, is the correlation weight between channels;

[0037] The extracted feature set The input is fed into the iTransformer model, and the long-term dependency features are captured through the multi-scale attention mechanism and memory mechanism to obtain the final output, which is expressed as:

[0038]

[0039] in, is the final output;

[0040] For the non-time series data consisting of weather forecast information of industrial parks and production line production data, graph neural network is used to extract the characteristic information of solar radiation, air pressure, wind speed in weather forecast information and the characteristic information of production line production data. Represents features extracted from non-time series data;

[0041] Features to be extracted from non-time series data With features extracted from time series data To merge, the expression is:

[0042]

[0043] in, is the fused feature matrix;

[0044] The fused feature matrix is ​​input into the fully connected layer, and the energy consumption prediction result of the hybrid deep neural network model is output. The expression is:

[0045]

[0046] in, is the energy consumption prediction result of the hybrid deep neural network model, is the weight matrix, For bias.

[0047] Optionally, the prediction process of the physical empirical model includes:

[0048] According to the empirical formula between equipment efficiency and energy consumption, the equipment efficiency and energy consumption are calculated as follows:

[0049]

[0050] in, is the equipment efficiency energy consumption, It is The power of the device, It is The efficiency of the equipment, It is The running time of the device It is the total loss of the equipment, including transmission loss and heat loss;

[0051] According to the empirical formula between the temperature and energy consumption of the equipment, the equipment temperature energy consumption is calculated as follows:

[0052]

[0053]

[0054] in, is the temperature energy consumption of the refrigeration equipment, is the temperature energy consumption of the heating equipment, and is the coefficient of cooling and heating equipment, is the external ambient temperature, is the set temperature, and is the operating time of cooling and heating equipment;

[0055] According to the empirical formula between the humidity and energy consumption of the equipment, the humidity energy consumption of the equipment is calculated as follows:

[0056]

[0057] in, is the humidity energy consumption of the equipment, is the coefficient of the dehumidification equipment, is the external environment humidity, is to set the humidity, is the operating time of the dehumidification equipment;

[0058] According to the empirical formula between the production load and energy consumption of the equipment, the energy consumption of the equipment production is calculated as follows:

[0059]

[0060] in, Is the equipment production in line with energy consumption, is the energy consumption coefficient of the production equipment, is the production load, It is production time;

[0061] According to the energy consumption of equipment efficiency, equipment temperature, equipment humidity and equipment production compliance, the energy consumption prediction result of the physical empirical model is obtained, and the expression is:

[0062]

[0063] in, It is the energy consumption prediction result of the physical empirical model.

[0064] Optionally, the hybrid deep neural network model and the physical empirical model are fused by an averaging method, a weighted averaging method or a stacking method, and the fusion process includes:

[0065] If the fusion strategy method is a simple average method, the prediction results of different models are averaged, and the average value is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is:

[0066]

[0067] in, is the prediction result of the multi-source and multi-scale comprehensive integration model;

[0068] If the fusion strategy method is the weighted average method, the prediction results of different models are weighted averaged, and the weighted average is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is:

[0069]

[0070] in, represents the weights of the hybrid deep neural network model, represents the weight of the physical empirical model;

[0071] If the fusion strategy method is the stacking method, the prediction results of different models are used as input, a meta-model is trained to combine the prediction results of different models, and the combined results are used as the prediction results of the multi-source and multi-scale comprehensive integration model. The expression is:

[0072]

[0073] in, Represents a cloud model.

[0074] Optionally, the evaluating the predicted energy consumption result according to the composite evaluation factor includes:

[0075] The prediction accuracy, mean square error, mean absolute error, root mean square error and determination coefficient of the multi-source multi-scale comprehensive integrated model are obtained according to the predicted energy consumption results, and the expression is:

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] in, , , , , For different evaluation indicators, For prediction accuracy, is the mean square error, is the mean absolute error, is the root mean square error, is the coefficient of determination, It is The predicted energy consumption results are: For the The actual energy consumption value, is the number of samples of the preprocessed industrial energy consumption multi-source data;

[0082] Determine the weight of each evaluation indicator according to user needs or expert opinions, and the sum of all weights is 1;

[0083] According to different evaluation indicators and the corresponding weights of each evaluation indicator, a composite evaluation factor is constructed, and the expression is:

[0084]

[0085] in, is a composite evaluation factor, The higher the value, the better the predicted energy consumption result. , , , , They are , , , , The weight of .

[0086] To achieve the above-mentioned purpose, the second embodiment of the present application proposes an integrated prediction device for energy consumption of an industrial park based on multi-source and multi-scale information fusion, comprising:

[0087] A data collection module, used to collect multi-source data of industrial energy consumption and perform data preprocessing based on periodic registration on the multi-source data of industrial energy consumption, wherein the multi-source data of industrial energy consumption includes weather forecast information, on-site temperature information of industrial parks, humidity information, power consumption data, production line production data, and physical parameter data of industrial parks;

[0088] A construction module is used to construct a multi-source multi-scale comprehensive integration model, wherein the multi-source multi-scale comprehensive integration model is obtained by fusing a hybrid deep neural network model and a physical empirical model;

[0089] A training module, for training the multi-source and multi-scale comprehensive integration model using an adversarial training strategy for a large amount of unlabeled data and a small amount of labeled data in the multi-source data of industrial energy consumption;

[0090] The prediction and evaluation module is used to input the pre-processed industrial energy consumption multi-source data into the trained multi-source and multi-scale comprehensive integration model to obtain the predicted energy consumption results, and evaluate the predicted energy consumption results according to the composite evaluation factors.

[0091] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0092] The memory stores computer-executable instructions;

[0093] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects above.

[0094] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by a processor, they are used to implement the method as described in any one of the above-mentioned first aspects.

[0095] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0096] (1) Comprehensive use of multi-source and multi-scale information: This application comprehensively utilizes multi-source data such as weather forecast information, industrial park site temperature information, humidity information, power consumption data, production line production data, and industrial park physical parameter data. Compared with the traditional prediction method of a single data source or a single scale information, it can more comprehensively consider the impact of various factors on industrial energy consumption.

[0097] (2) Improving prediction accuracy and stability: Through data preprocessing based on periodic registration, the time series data of various data sources are aligned to a unified time period, and a multi-source and multi-scale comprehensive integration model is established, which includes a variety of prediction methods such as hybrid deep neural network models and physical empirical models. The comprehensive application of these methods can effectively improve the accuracy and stability of predictions.

[0098] (3) Improving model robustness and generalization ability: An adversarial training optimization strategy is adopted to improve the robustness and generalization ability of the model by minimizing the loss function of the model and maximizing the loss function of the adversarial sample, making it more adaptable to noise, interference and abnormal situations.

[0099] (4) Comprehensive evaluation of energy consumption forecast results: By designing a composite evaluation factor and comprehensively considering the impact of multiple evaluation indicators on energy consumption, the accuracy and comprehensiveness of the energy consumption forecast are evaluated. This helps to have a more comprehensive understanding of the energy consumption forecast results and provide more reliable decision support for energy management and planning.

[0100] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0102] Figure 1 It is a flow chart of an integrated prediction method for energy consumption of an industrial park based on multi-source and multi-scale information fusion according to an embodiment of the present application;

[0103] Figure 2 It is a block diagram of an integrated prediction device for energy consumption of an industrial park based on multi-source and multi-scale information fusion according to an embodiment of the present application;

[0104] Figure 3 It is a block diagram of an electronic device. DETAILED DESCRIPTION

[0105] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0106] The following describes, with reference to the accompanying drawings, an integrated prediction method and device for energy consumption of an industrial park based on multi-source and multi-scale information fusion according to an embodiment of the present application.

[0107] Figure 1 is a flow chart of an integrated prediction method for energy consumption of an industrial park based on multi-source and multi-scale information fusion according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0108] Step 101, collect multi-source data on industrial energy consumption, and perform data preprocessing on the multi-source data based on periodic alignment, wherein the multi-source data on industrial energy consumption includes weather forecast information, industrial park site temperature information, humidity information, power consumption data, production line production data, and industrial park physical parameter data.

[0109] In some embodiments of the present application, weather data of the industrial park within a preset future time is obtained from the meteorological department or a third-party meteorological service provider; on-site temperature information and humidity information of the industrial park are collected through the environmental monitoring station in the industrial park; power consumption data is obtained through the electric meters, smart grid management systems or energy management systems in the industrial park; production line production volume data is provided according to the production management system or automated control system in the industrial park, including the output, start-up time and downtime of each production line; physical parameter data of the industrial park is obtained through the equipment efficiency, thermal insulation performance, moisture retention performance and production load fluctuations in the industrial park.

[0110] After collecting multi-source data on industrial energy consumption, it is necessary to further perform periodic alignment on the multi-source data so as to synchronize data at different time scales or different time periods for effective analysis.

[0111] Specifically, the time series of the industrial energy consumption multi-source data is calibrated, and the timestamps of all the industrial energy consumption multi-source data are converted into a unified time format.

[0112] In addition, some other preprocessing processes can also be performed. For example, if there are missing values, they can be filled by interpolation or mean filling; if there are outliers, they need to be identified and removed; data of different magnitudes can be scaled, such as through maximum and minimum normalization or z-score standardization, to make the data comparable on the same scale.

[0113] Through the above data preprocessing steps based on periodic registration, it can be ensured that the time series data of each data source have a consistent time period and timestamp, providing a reliable data foundation for subsequent modeling and analysis.

[0114] Step 102, constructing a multi-source multi-scale comprehensive integration model, wherein the multi-source multi-scale comprehensive integration model is obtained by fusing a hybrid deep neural network model and a physical empirical model.

[0115] In the embodiments of the present application, a hybrid deep neural network model and a physical empirical model are comprehensively utilized to more comprehensively consider the impact of various factors on energy consumption and improve prediction accuracy and reliability.

[0116] (1) The prediction process of the hybrid deep neural network model.

[0117] For the time series data consisting of the temperature information, humidity information, and power consumption data of the industrial park site, the Fourier decomposition method is used to decompose it into narrowband components of multiple channels to reduce its non-stationarity and noise. After Fourier decomposition, each time series data Decomposed into multiple narrowband components , the expression of Fourier decomposition process is:

[0118]

[0119] in, is time series data, are the Fourier coefficients, is the total number of data points in the time series data, It's a point in time.

[0120] Subsequently, a convolutional neural network is used to perform a convolution operation on each decomposed narrowband component to extract local features. The expression is:

[0121]

[0122] in, is the convolution output, is the convolution kernel, is the bias, is the activation function.

[0123] Then, the convolution output is pooled to achieve feature dimensionality reduction, expressed as:

[0124]

[0125]

[0126] in, is the pooled output set, For the Convolution output The corresponding pooling output.

[0127] Furthermore, for the pooled output after dimensionality reduction, the features of different channels are adaptively extracted based on the dependency correlation between components of different channels. The expression is:

[0128]

[0129]

[0130] Among them, among them, is the feature set extracted adaptively, is the feature extracted adaptively, is the correlation weight between channels.

[0131] Finally, the extracted feature set The input is fed into the iTransformer model, and the long-term dependency features are captured through the multi-scale attention mechanism and memory mechanism to obtain the final output, which is expressed as:

[0132]

[0133] in, is the final output.

[0134] In addition, for the non-time series data consisting of weather forecast information of industrial parks and production line production data, graph neural networks are used to extract the characteristic information of solar radiation, air pressure, and wind speed in weather forecast information and the characteristic information of production line production data. Represents features extracted from non-time series data; features extracted from non-time series data With features extracted from time series data To merge, the expression is:

[0135]

[0136] in, is the fused feature matrix.

[0137] Finally, the fused feature matrix is ​​input into the fully connected layer, and the energy consumption prediction result of the hybrid deep neural network model is output, which is expressed as:

[0138]

[0139] in, is the energy consumption prediction result of the hybrid deep neural network model, is the weight matrix, For bias.

[0140] (2) The prediction process of the physical empirical model.

[0141] In order to build a physical empirical model for constraining the energy consumption forecast of industrial parks, it is necessary to introduce some formulas based on physical laws to ensure the reliability of the forecast results. These formulas cover the main physical parameters related to energy consumption, such as temperature, humidity, equipment power, production load, etc.

[0142] In one embodiment of the present application, the actual energy consumption needs to take into account the efficiency and loss of the equipment. Therefore, the efficiency and energy consumption of the equipment can be calculated according to the empirical formula between the efficiency and energy consumption of the equipment. The expression is:

[0143]

[0144] in, is the equipment efficiency energy consumption, It is The power of the device, It is The efficiency of the equipment, It is The running time of the device It is the total loss of the equipment, including transmission loss and heat loss.

[0145] In another embodiment of the present application, the energy consumption of the refrigeration and heating equipment is related to the temperature. Therefore, the temperature energy consumption of the equipment can be calculated according to the empirical formula between the temperature and energy consumption of the equipment. The expression is:

[0146]

[0147]

[0148] in, is the temperature energy consumption of the refrigeration equipment, is the temperature energy consumption of the heating equipment, and is the coefficient of cooling and heating equipment, is the external ambient temperature, is the set temperature, and is the operating time of the cooling and heating equipment.

[0149] Similarly, the empirical formula between the humidity and energy consumption of the equipment can be used to calculate the humidity energy consumption of the equipment. The expression is:

[0150]

[0151] in, is the humidity energy consumption of the equipment, is the coefficient of the dehumidification equipment, is the external environment humidity, is to set the humidity, is the operating time of the dehumidifier.

[0152] Similarly, fluctuations in production load will also affect energy consumption. Therefore, the energy consumption of equipment production can be calculated based on the empirical formula between the equipment's production load and energy consumption. The expression is:

[0153]

[0154] in, Is the equipment production in line with energy consumption, is the energy consumption coefficient of the production equipment, is the production load, It's production time.

[0155] Finally, the above factors are combined, and the energy consumption prediction result of the physical empirical model is obtained according to the equipment efficiency energy consumption, equipment temperature energy consumption, equipment humidity energy consumption and equipment production compliance energy consumption. The expression is:

[0156]

[0157] in, It is the energy consumption prediction result of the physical empirical model.

[0158] In addition, in order to ensure the reliability of energy consumption prediction, physical constraints can also be imposed based on the above industrial park physical parameter data, for example: and add it to the physical empirical model constructed above.

[0159] In some embodiments of the present application, after obtaining the hybrid deep neural network model and the physical empirical model, they are fused to obtain the multi-source and multi-scale comprehensive integrated model, wherein the fusion strategy method includes simple averaging method, weighted averaging method, and stacking method, and by training certain weight parameters, the prediction results of different models can be better combined, thereby improving the accuracy of the overall prediction.

[0160] As a possible implementation method, the prediction results of different models are averaged and the average value is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is:

[0161]

[0162] in, It is the prediction result of the multi-source and multi-scale comprehensive integration model.

[0163] As another possible implementation method, if the fusion strategy method is a weighted average method, the prediction results of different models are weighted averaged, and the weighted average is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is:

[0164]

[0165] in, represents the weights of the hybrid deep neural network model, Represents the weight of the physical empirical model.

[0166] As another possible implementation method, if the fusion strategy method is a stacking method, the prediction results of different models are used as input, a meta-model is trained to combine the prediction results of different models, and the combined results are used as the prediction results of the multi-source multi-scale comprehensive integration model. The expression is:

[0167]

[0168] in, Represents a cloud model.

[0169] It is understandable that the training process of the meta-model is a technology well known to those skilled in the art, and this application will not elaborate on it in detail.

[0170] Step 103, for a large amount of unlabeled data and a small amount of labeled data in the industrial energy consumption multi-source data, an adversarial training strategy is used to train the multi-source and multi-scale comprehensive integration model.

[0171] In an embodiment of the present application, in view of the fact that there are a large amount of unlabeled data and a small amount of labeled data in the historical energy consumption data of an industrial park, in order to make full use of the unlabeled data training model and to mine data feature patterns from historical data as much as possible, an adversarial training strategy is adopted for a large amount of unlabeled data and a small amount of labeled data to enhance the ability of the multi-source and multi-scale comprehensive integration model to mine data features, thereby obtaining a multi-source and multi-scale comprehensive integration model that can effectively extract data features.

[0172] It should be noted that adversarial training is an optimization strategy that aims to improve the robustness and generalization ability of the model by minimizing the loss function of the prediction model while maximizing the loss function of the adversarial sample. In the scenario of energy demand prediction, adversarial training can be used to improve the model's adaptability to noise, interference and abnormal conditions, thereby improving the accuracy and stability of the prediction.

[0173] The optimization strategy for adversarial training is as follows:

[0174] 1. Generate adversarial samples: Use methods such as generative adversarial networks (GANs) or gradient descent to generate adversarial samples, that is, make slight perturbations to the original input.

[0175] 2. Update model parameters: Use adversarial samples to update the parameters of the multi-source and multi-scale comprehensive integration model to minimize the loss function of the multi-scale comprehensive integration model on adversarial samples.

[0176] 3. Adjust the adversarial sample generation strategy: According to the performance of the multi-source and multi-scale comprehensive integration model on adversarial samples, adjust the adversarial sample generation strategy to continuously improve the performance of the multi-source and multi-scale comprehensive integration model on adversarial samples.

[0177] Through adversarial training optimization strategy, the robustness and generalization ability of the energy consumption demand prediction model can be improved, thereby improving the accuracy and stability of the prediction, making it more suitable for complex and changeable practical application scenarios.

[0178] Step 104, input the preprocessed industrial energy consumption multi-source data into the trained multi-source multi-scale comprehensive integration model to obtain predicted energy consumption results, and evaluate the predicted energy consumption results according to the composite evaluation factors.

[0179] After obtaining the predicted energy consumption results, a composite evaluation factor is designed to comprehensively consider the impact of multiple factors on energy consumption and evaluate the accuracy and comprehensiveness of the prediction.

[0180] In one embodiment of the present application, the prediction accuracy, mean square error, mean absolute error, root mean square error and determination coefficient of the multi-source multi-scale comprehensive integrated model are obtained according to the predicted energy consumption result, and the expression is:

[0181]

[0182]

[0183]

[0184]

[0185]

[0186] in, , , , , For different evaluation indicators, For prediction accuracy, is the mean square error, is the mean absolute error, is the root mean square error, is the coefficient of determination, It is The predicted energy consumption results are: For the The actual energy consumption value, is the number of samples of the industrial energy consumption multi-source data after preprocessing.

[0187] Then, the weights of each evaluation index are determined according to user needs or expert opinions, and the sum of all weights is 1.

[0188] Finally, a composite evaluation factor is constructed based on different evaluation indicators and the weights corresponding to each evaluation indicator. The expression is:

[0189]

[0190] in, is a composite evaluation factor, The higher the value, the better the predicted energy consumption result. , , , , They are , , , , The weight of .

[0191] In order to implement the above-mentioned embodiments, the present application also proposes an integrated prediction device for energy consumption of industrial parks based on multi-source and multi-scale information fusion.

[0192] Figure 2 1 is a block diagram of an integrated prediction device 10 for energy consumption of an industrial park based on multi-source and multi-scale information fusion according to an embodiment of the present application, comprising:

[0193] The data collection module 100 is used to collect multi-source data of industrial energy consumption and perform data preprocessing based on periodic registration on the multi-source data of industrial energy consumption, wherein the multi-source data of industrial energy consumption includes weather forecast information, on-site temperature information of industrial parks, humidity information, power consumption data, production line production data, and physical parameter data of industrial parks;

[0194] A construction module 200 is used to construct a multi-source multi-scale comprehensive integration model, wherein the multi-source multi-scale comprehensive integration model is obtained by fusing a hybrid deep neural network model and a physical empirical model;

[0195] A training module 300 is used to train the multi-source and multi-scale comprehensive integration model using an adversarial training strategy for a large amount of unlabeled data and a small amount of labeled data in the multi-source data of industrial energy consumption;

[0196] The prediction and evaluation module 400 is used to input the pre-processed industrial energy consumption multi-source data into the trained multi-source and multi-scale comprehensive integration model to obtain the predicted energy consumption results, and evaluate the predicted energy consumption results according to the composite evaluation factors.

[0197] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0198] Figure 3 A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0199] like Figure 3 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0200] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0201] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above. For example, in some embodiments, the scheme proposed in the present application may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the method shown in the present application in any other appropriate manner (e.g., by means of firmware).

[0202] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0203] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0204] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0205] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0206] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0207] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0208] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0209] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. An integrated prediction method for energy consumption in industrial parks based on multi-source and multi-scale information fusion, characterized in that: include: Collecting multi-source data of industrial energy consumption, and performing data preprocessing based on periodic registration on the multi-source data of industrial energy consumption, wherein the multi-source data of industrial energy consumption includes weather forecast information, on-site temperature information of industrial parks, humidity information, power consumption data, production line production data, and physical parameter data of industrial parks; Constructing a multi-source multi-scale comprehensive integration model, wherein the multi-source multi-scale comprehensive integration model is obtained by fusing a hybrid deep neural network model and a physical empirical model; For a large amount of unlabeled data and a small amount of labeled data in the industrial energy consumption multi-source data, an adversarial training strategy is used to train the multi-source and multi-scale comprehensive integration model; Inputting the preprocessed industrial energy consumption multi-source data into the trained multi-source multi-scale comprehensive integration model to obtain predicted energy consumption results, and evaluating the predicted energy consumption results according to the composite evaluation factors; The prediction process of the hybrid deep neural network model includes: For the time series data consisting of the temperature information, humidity information, and power consumption data of the industrial park site, the Fourier decomposition method is used to decompose it into narrowband components of multiple channels. After Fourier decomposition, each time series data Decomposed into multiple narrowband components , the expression of Fourier decomposition process is: in, is time series data, are the Fourier coefficients, is the total number of data points in the time series data, It’s a point in time; The convolutional neural network is used to perform convolution operations on each narrowband component after decomposition to extract local features. The expression is: in, is the convolution output, is the convolution kernel, is the bias, is the activation function, For narrowband components The number of different convolution kernels; The convolution output is pooled and the expression is: in, is the pooled output set, For the Convolution output The corresponding pooling output; Based on the dependency relationship between different channel components, adaptive channel extraction is performed on the pooling output, and the expression is: in, is the feature set extracted adaptively, is the feature extracted adaptively, is the correlation weight between channels; The extracted feature set The input is fed into the iTransformer model, and the long-term dependency features are captured through the multi-scale attention mechanism and memory mechanism to obtain the final output, which is expressed as: in, is the final output; For the non-time series data consisting of weather forecast information of industrial parks and production line production data, graph neural network is used to extract the characteristic information of solar radiation, air pressure, wind speed in weather forecast information and the characteristic information of production line production data. Represents features extracted from non-time series data; Features to be extracted from non-time series data With features extracted from time series data To merge, the expression is: in, is the fused feature matrix; The fused feature matrix is ​​input into the fully connected layer, and the energy consumption prediction result of the hybrid deep neural network model is output. The expression is: in, is the energy consumption prediction result of the hybrid deep neural network model, is the weight matrix, is bias; The prediction process of the physical empirical model includes: According to the empirical formula between equipment efficiency and energy consumption, the equipment efficiency and energy consumption are calculated as follows: in, is the equipment efficiency energy consumption, It is The power of the device, It is The efficiency of the equipment, It is The running time of the device, It is the total loss of the equipment, including transmission loss and heat loss; According to the empirical formula between the temperature and energy consumption of the equipment, the equipment temperature energy consumption is calculated as follows: in, is the temperature energy consumption of the refrigeration equipment, is the temperature energy consumption of the heating equipment, and is the coefficient of cooling and heating equipment, is the external ambient temperature, is the set temperature, and is the operating time of cooling and heating equipment; According to the empirical formula between the humidity and energy consumption of the equipment, the humidity energy consumption of the equipment is calculated as follows: in, is the humidity energy consumption of the equipment, is the coefficient of the dehumidification equipment, is the external environment humidity, is to set the humidity, is the operating time of the dehumidification equipment; According to the empirical formula between the production load and energy consumption of the equipment, the energy consumption of the equipment production load is calculated as follows: in, is the energy consumption of equipment production load, is the energy consumption coefficient of the production equipment, is the production load, It is production time; According to the energy consumption of equipment efficiency, equipment temperature, equipment humidity and equipment production load, the energy consumption prediction result of the physical empirical model is obtained, and the expression is: in, It is the energy consumption prediction result of the physical empirical model; The hybrid deep neural network model and the physical empirical model are fused by an averaging method, a weighted averaging method or a stacking method, and the fusion process includes: If the fusion strategy method is a simple average method, the prediction results of different models are averaged, and the average value is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is: in, is the prediction result of the multi-source and multi-scale comprehensive integration model; If the fusion strategy method is the weighted average method, the prediction results of different models are weighted averaged, and the weighted average is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is: in, represents the weights of the hybrid deep neural network model, represents the weight of the physical empirical model; If the fusion strategy method is the stacking method, the prediction results of different models are used as input, a meta-model is trained to combine the prediction results of different models, and the combined results are used as the prediction results of the multi-source and multi-scale comprehensive integration model. The expression is: in, Represents a cloud model; The evaluating the predicted energy consumption result according to the composite evaluation factor includes: The prediction accuracy, mean square error, mean absolute error, root mean square error and determination coefficient of the multi-source multi-scale comprehensive integrated model are obtained according to the predicted energy consumption results, and the expression is: in, , , , , For different evaluation indicators, For prediction accuracy, is the mean square error, is the mean absolute error, is the root mean square error, is the coefficient of determination, It is The predicted energy consumption results are: For the The actual energy consumption value, is the number of samples of the preprocessed industrial energy consumption multi-source data; Determine the weight of each evaluation indicator according to user needs or expert opinions, and the sum of all weights is 1; According to different evaluation indicators and the corresponding weights of each evaluation indicator, a composite evaluation factor is constructed, and the expression is: in, is a composite evaluation factor, The higher the value, the better the predicted energy consumption result. , , , , They are , , , , The weight of .

2. The method according to claim 1, characterized in that The collection of multi-source data on industrial energy consumption includes: Obtain weather data for the industrial park within a preset time in the future from the meteorological department or a third-party meteorological service provider; Collect on-site temperature and humidity information of the industrial park through the environmental monitoring station in the industrial park; Obtain power consumption data through electricity meters, smart grid management systems or energy management systems in industrial parks; Provide production line production data based on the production management system or automated control system in the industrial park, including the output, start-up time, and downtime of each production line; The physical parameter data of the industrial park is obtained through the equipment efficiency, thermal insulation performance, moisture retention performance, and production load fluctuations in the industrial park.

3. The method according to claim 2, characterized in that The data preprocessing based on periodic registration for the industrial energy consumption multi-source data includes: The industrial energy consumption multi-source data is time-series calibrated, and the timestamps of all the industrial energy consumption multi-source data are converted into a unified time format.

4. An integrated prediction device for energy consumption in industrial parks based on multi-source and multi-scale information fusion, characterized in that: include: A data collection module, used to collect multi-source data of industrial energy consumption and perform data preprocessing based on periodic registration on the multi-source data of industrial energy consumption, wherein the multi-source data of industrial energy consumption includes weather forecast information, on-site temperature information of industrial parks, humidity information, power consumption data, production line production data, and physical parameter data of industrial parks; A construction module is used to construct a multi-source multi-scale comprehensive integration model, wherein the multi-source multi-scale comprehensive integration model is obtained by fusing a hybrid deep neural network model and a physical empirical model; A training module, for a large amount of unlabeled data and a small amount of labeled data in the industrial energy consumption multi-source data, adopts an adversarial training strategy to train the multi-source and multi-scale comprehensive integration model; A prediction and evaluation module, used for inputting the pre-processed industrial energy consumption multi-source data into the trained multi-source multi-scale comprehensive integration model to obtain predicted energy consumption results, and evaluating the predicted energy consumption results according to the composite evaluation factors; The prediction process of the hybrid deep neural network model includes: For the time series data consisting of the temperature information, humidity information, and power consumption data of the industrial park site, the Fourier decomposition method is used to decompose it into narrowband components of multiple channels. After Fourier decomposition, each time series data Decomposed into multiple narrowband components , the expression of Fourier decomposition process is: in, is time series data, are the Fourier coefficients, is the total number of data points in the time series data, It’s a point in time; The convolutional neural network is used to perform convolution operations on each narrowband component after decomposition to extract local features. The expression is: in, is the convolution output, is the convolution kernel, is the bias, is the activation function, For narrowband components The number of different convolution kernels; The convolution output is pooled and the expression is: in, is the pooled output set, For the Convolution output The corresponding pooling output; Based on the dependency relationship between different channel components, adaptive channel extraction is performed on the pooling output, and the expression is: in, is the feature set extracted adaptively, is the feature extracted adaptively, is the correlation weight between channels; The extracted feature set The input is fed into the iTransformer model, and the long-term dependency features are captured through the multi-scale attention mechanism and memory mechanism to obtain the final output, which is expressed as: in, is the final output; For the non-time series data consisting of weather forecast information of industrial parks and production line production data, graph neural network is used to extract the characteristic information of solar radiation, air pressure, wind speed in weather forecast information and the characteristic information of production line production data. Represents features extracted from non-time series data; Features to be extracted from non-time series data With features extracted from time series data To merge, the expression is: in, is the fused feature matrix; The fused feature matrix is ​​input into the fully connected layer, and the energy consumption prediction result of the hybrid deep neural network model is output. The expression is: in, is the energy consumption prediction result of the hybrid deep neural network model, is the weight matrix, is bias; The prediction process of the physical empirical model includes: According to the empirical formula between equipment efficiency and energy consumption, the equipment efficiency and energy consumption are calculated as follows: in, is the equipment efficiency energy consumption, It is The power of the device, It is The efficiency of the equipment, It is The running time of the device It is the total loss of the equipment, including transmission loss and heat loss; According to the empirical formula between the temperature and energy consumption of the equipment, the equipment temperature energy consumption is calculated as follows: in, is the temperature energy consumption of the refrigeration equipment, is the temperature energy consumption of the heating equipment, and is the coefficient of cooling and heating equipment, is the external ambient temperature, is the set temperature, and is the operating time of cooling and heating equipment; According to the empirical formula between the humidity and energy consumption of the equipment, the humidity energy consumption of the equipment is calculated as follows: in, is the humidity energy consumption of the equipment, is the coefficient of the dehumidification equipment, is the external environment humidity, is to set the humidity, is the operating time of the dehumidification equipment; According to the empirical formula between the production load and energy consumption of the equipment, the energy consumption of the equipment production load is calculated as follows: in, is the energy consumption of equipment production load, is the energy consumption coefficient of the production equipment, is the production load, It is production time; According to the energy consumption of equipment efficiency, equipment temperature, equipment humidity and equipment production load, the energy consumption prediction result of the physical empirical model is obtained, and the expression is: in, It is the energy consumption prediction result of the physical empirical model; The hybrid deep neural network model and the physical empirical model are fused by an averaging method, a weighted averaging method or a stacking method, and the fusion process includes: If the fusion strategy method is a simple average method, the prediction results of different models are averaged, and the average value is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is: in, is the prediction result of the multi-source and multi-scale comprehensive integration model; If the fusion strategy method is the weighted average method, the prediction results of different models are weighted averaged, and the weighted average is used as the prediction result of the multi-source multi-scale comprehensive integration model. The expression is: in, represents the weights of the hybrid deep neural network model, represents the weight of the physical empirical model; If the fusion strategy method is the stacking method, the prediction results of different models are used as input, a meta-model is trained to combine the prediction results of different models, and the combined results are used as the prediction results of the multi-source and multi-scale comprehensive integration model. The expression is: in, Represents a cloud model; The evaluating the predicted energy consumption result according to the composite evaluation factor includes: The prediction accuracy, mean square error, mean absolute error, root mean square error and determination coefficient of the multi-source multi-scale comprehensive integrated model are obtained according to the predicted energy consumption results, and the expression is: in, , , , , For different evaluation indicators, For prediction accuracy, is the mean square error, is the mean absolute error, is the root mean square error, is the coefficient of determination, It is The predicted energy consumption results are: For the The actual energy consumption value, is the number of samples of the preprocessed industrial energy consumption multi-source data; Determine the weight of each evaluation indicator according to user needs or expert opinions, and the sum of all weights is 1; According to different evaluation indicators and the corresponding weights of each evaluation indicator, a composite evaluation factor is constructed, and the expression is: in, is a composite evaluation factor, The higher the value, the better the predicted energy consumption result. , , , , They are , , , , The weight of .

5. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 3 when executed by a processor.

Citation Information

Patent Citations

  • Annual energy consumption and thermal comfort coupling evaluation method for tall and large space building

    CN115795607A

  • Industrial park energy efficiency management and control method based on Internet of Things

    CN116307076A

  • Water area pollution prediction method based on multiple scales and multiple dimensions

    CN118228923A

  • Passenger car production line energy consumption intelligent monitoring and energy efficiency improving method and system

    CN118393962A