Building energy consumption data generation and parallel prediction method and device based on generative adversarial network and medium

Abstract

The invention provides a building energy consumption data generation and parallel prediction method and device based on a generative adversarial network, and a medium, and belongs to the technical field of building energy consumption prediction. The method comprises the following steps: collecting actual building energy consumption data as a training sample, including time sequence data and physical attributes of a building; constructing and training a generative adversarial network model; inputting the actual building energy consumption data into the generative adversarial network model to generate enhanced building energy consumption data; the actual building energy consumption data and the enhanced building energy consumption data are mixed to obtain mixed data, the mixed data are used for training a machine learning model, a building energy consumption parallel prediction model is constructed, and a future building energy consumption value is predicted; and evaluating the performance of the prediction model on the test set, and adjusting the model structure or increasing the number of training rounds to improve the prediction precision. According to the method, the problem of insufficient building energy consumption data is solved, the precision and reliability of building energy consumption prediction are remarkably improved, and powerful support is provided for building energy consumption management.

Description

Technical Field

[0001] The invention relates to a method, device and medium for generating and parallel predicting building energy consumption data based on a generative adversarial network, and belongs to the technical field of building energy consumption prediction. Background Art

[0002] Building energy consumption data is a key component of the massive operating data of buildings, directly reflecting the distribution and changing trends of building energy consumption. Building energy consumption data analysis is of great significance for achieving the goal of smart building construction, optimizing building energy control, and participating in grid demand-side response. Building energy consumption prediction, that is, predicting the future energy consumption distribution of buildings, plays an important role in building energy-saving assessment, multi-energy scheduling, and energy-saving strategy formulation.

[0003] At present, building energy consumption prediction based on machine learning mainly uses supervised learning methods to achieve building energy consumption prediction through regression models. However, due to the simple structure, traditional machine learning methods often cannot fully explore data features, resulting in low accuracy in building energy consumption prediction. This limitation is particularly evident in building energy consumption prediction, because the changes in building energy consumption data are affected by many factors, such as weather, user behavior, building structure characteristics, etc. These factors make it difficult for simple models to capture the complex relationship between data.

[0004] In practical applications, the collection of building energy consumption data still faces many challenges. Due to unstable transmission lines, sensor failures, or system testing, buildings often face a large amount of data missing in the early stages of operation or after energy-saving renovations. This insufficient data information will seriously affect the performance of the constructed model. Especially when processing one-dimensional time series data, the data distribution characteristics are complex. Traditional data generation methods often estimate the distribution of data points one by one and generate data within the distribution range. This method ignores the connection between data points, resulting in the generated data being difficult to reflect the true law of energy consumption changes.

[0005] Therefore, finding an effective method to solve the problem of insufficient building energy consumption data and improve the accuracy of building energy consumption prediction has become a hot topic in current research. Generative adversarial network (GAN), as a new type of deep learning model, has the ability to discover the potential distribution of data and generate data with similar distribution. It has achieved good application results in many fields such as biomedicine and image recognition. This model provides a new solution to the problem of insufficient energy consumption data in buildings, that is, to generate realistic building energy consumption data through GAN, which is then used to train the building energy consumption prediction model to improve the prediction accuracy. Summary of the invention

[0006] The purpose of the present invention is to provide a method, device and medium for generating and parallel predicting building energy consumption data based on a generative adversarial network. The powerful data generation capability of the GAN model is used to solve the problem of insufficient building energy consumption data. At the same time, the generated enhanced data is used to build a more accurate energy consumption prediction model, providing strong support for building energy consumption management.

[0007] In order to achieve the above object, the present invention is implemented through the following technical solutions: A method for generating building energy consumption data and parallel prediction based on a generative adversarial network, comprising: Collect actual building energy consumption data as training samples, including time series data and physical properties of buildings; Construct and train a generative adversarial network model, in which the generator is used to generate enhanced building energy consumption data, and the discriminator is used to distinguish between the actual building energy consumption data and the enhanced building energy consumption data generated by the generator; Input the actual building energy consumption data into the generative adversarial network model to generate enhanced building energy consumption data; The actual building energy consumption data and the enhanced building energy consumption data are mixed to obtain mixed data, which is used to train the machine learning model, build a parallel prediction model for building energy consumption, and predict future building energy consumption values; Evaluate the performance of the prediction model on the test set, and adjust the model structure or increase the number of training rounds to improve the prediction accuracy. Preferably, the time series data includes electricity consumption, water consumption, gas usage and building-related environmental parameters, and the building-related environmental parameters include temperature, humidity and light intensity; The physical properties of the building include building area, building materials, and building purpose.

[0008] Preferably, the generator of the generative adversarial network model transitions from a simple linear generator to a complex nonlinear generator; the nonlinear generator includes an autoencoder and a variational autoencoder; the generator maps the potential correlation of the energy consumption time series to the data space and performs data transformation by increasing randomness; The discriminator transitions from a multi-layer perceptron to a deep neural network, which includes an Inception module and a ResNet module; the discriminator distinguishes between generated data and original data, and reversely updates the weight of the generator based on the distinction result.

[0009] Preferably, during the training process of the generative adversarial network model, the generated data is filtered by setting an average power factor threshold, and when the average power factor of the generated data is greater than the threshold, it is included in the training.

[0010] Preferably, generating enhanced building energy consumption data comprises the following specific steps: Obtain building energy consumption data, including time series data and physical properties of buildings; Data encoding: Encode the energy consumption data and input it into the generator model established in the model construction phase; use the encoded data to train the generator, and use the generator and discriminator to generate energy consumption data; Generate augmented data and predict augmented data based on the generator model.

[0011] Preferably, the generative adversarial network model sets a filter to remove data sequences with large differences. The filter compares the generated data sequence with the mean sequence of the original data and obtains MAPE. When the MAPE is greater than a set threshold, the generated data sequence is discarded.

[0012] Preferably, the building energy consumption parallel prediction model is constructed using BPNN, ELM or SVR algorithm.

[0013] Preferably, accuracy, precision, and recall are used to evaluate the performance of the prediction model.

[0014] A device for generating building energy consumption data and parallel prediction based on a generative adversarial network comprises a processor and a memory storing program instructions, wherein the processor is configured to execute the method for generating building energy consumption data and parallel prediction based on a generative adversarial network when running the program instructions.

[0015] A computer-readable storage medium stores a computer program, which is executed by a processor to generate building energy consumption data and parallel prediction method based on a generative adversarial network.

[0016] The advantages of the present invention are: The present invention builds an advanced and usable model. The generator part gradually transitions from a simple linear model to a complex nonlinear model, such as an autoencoder, a variational autoencoder (VAE), etc., and introduces an attention mechanism, a gating mechanism, etc. to improve the quality of generated data and the stability of the model. The discriminator part also gradually transitions from a multi-layer perceptron (MLP) to a deep neural network, such as an Inception module, a ResNet module, etc., which enhances the classification and judgment capabilities of the model.

[0017] In the process of data generation and training, the present invention adopts a unique data filtering method. By setting a filter to remove data sequences with large distribution differences from the original data sequence, the specific method is to compare the mean sequence of the generated data sequence with the original data and calculate the MAPE (mean absolute percentage error). When the MAPE is greater than a certain threshold, the generated data sequence is discarded.

[0018] While the model is being trained, the generated data is filtered and an average power factor coefficient (APCC) threshold is set. Only when the APCC of the generated data is greater than the threshold is it included in the training, thereby improving the quality of the training data and the performance of the model.

[0019] By exploring the data distribution of the original data set and generating enhanced data similar to the original data distribution, the problem of insufficient building energy consumption data is effectively solved. Compared with the traditional prediction model, the present invention uses GAN (Generative Adversarial Network) to train a high-quality generative model without a large amount of real labeled data. By exploring the data distribution of the original data set and generating enhanced data similar to the original data distribution, the problem of insufficient building energy consumption data is effectively solved, and the laws and trends of building energy consumption are better captured, thereby improving the prediction accuracy.

[0020] In summary, the present invention not only solves the problem of insufficient building energy consumption data by introducing the GAN model and combining it with advanced data processing and model building techniques, but also significantly improves the accuracy and reliability of building energy consumption prediction, providing strong support for building energy consumption management. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0022] Figure 1 It is a schematic diagram of the process of the present invention.

[0023] Figure 2 Schematic diagram of the building energy consumption data generation and prediction structure of the present invention DETAILED DESCRIPTION The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] like Figure 1 As shown, a building energy consumption data generation and parallel prediction method based on a generative adversarial network includes: S1: Collect actual building energy consumption data as training samples, including time series data and physical properties of buildings; S2: Build and train a generative adversarial network model, where the generator is used to generate enhanced building energy consumption data, and the discriminator is used to distinguish between the actual building energy consumption data and the enhanced building energy consumption data generated by the generator; S3: Input the actual building energy consumption data into the generative adversarial network model to generate enhanced building energy consumption data; S4: Mix the actual building energy consumption data and the enhanced building energy consumption data to obtain mixed data, which is used to train the machine learning model, build a parallel prediction model for building energy consumption, and predict future building energy consumption values; S5: Evaluate the performance of the prediction model on the test set and adjust the model structure or increase the number of training rounds to improve the prediction accuracy. As a refinement of the above embodiment, in step S1, time series data of actual building energy consumption and physical property data of the building are collected from multiple reliable sources (such as smart meters, building management systems, etc.). The time series data includes electricity consumption, water consumption, gas usage, and environmental parameters related to the building, and the environmental parameters related to the building include temperature, humidity, and light intensity. The physical properties of the building include building area, building materials, and building purpose.

[0025] Data preprocessing: including data cleaning (removing missing values ​​and outliers), standardization / normalization, and feature engineering (such as building derived features) to ensure data quality and improve model training results.

[0026] As a refinement of the above embodiment, the generative adversarial network model in step S2 is composed of a generator and a discriminator. A generative adversarial network (GAN) is constructed using a neural network structure such as a convolutional neural network (CNN), a recurrent neural network (RNN), and a long short-term memory network (LSTM). The generator part starts with a simple linear generator and gradually transitions to a complex nonlinear generator, such as an autoencoder, a variational autoencoder (VAE), etc. At the same time, attention mechanisms, gating mechanisms, etc. are introduced to improve the quality of generated data and the stability of the model. The discriminator part starts with a multi-layer perceptron (MLP) and gradually transitions to a deep neural network, such as an Inception module, a ResNet module, etc.

[0027] After the model is built, it is trained. During the training process, the generator and the discriminator update the weights alternately. For each training iteration, the generator part takes the original building energy consumption data as input to generate a batch of new data samples, and then the discriminator classifies them to determine whether these samples are from the real world. According to the feedback from the discriminator, the generator adjusts its parameters to optimize its ability to generate data so that it is closer to the real data distribution. At this stage, the cross entropy loss function, binary cross entropy loss function, and Andrew Ng loss function are mainly used to train the model. The loss value of the model and the quality of the generated data are regularly monitored so that the structure and hyperparameters of the model can be adjusted in time to improve the quality of the generated data and the performance of the model.

[0028] After the generator generates a noise sequence, it inputs it into the discriminator and optimizes the quality of its generated data based on the feedback information of the discriminator. The discriminator is responsible for evaluating the effectiveness of the generated data, classifying it according to its authenticity and accuracy, and optimizing the weight of the generator based on the classification results.

[0029] It should be noted that the training of the generative adversarial network model is an iterative training process, that is, the generator is first trained to update the weights, and then the discriminator is trained to update the weights, so that the generator and discriminator are continuously iterated and optimized, so that the generator can generate more realistic data.

[0030] Optionally, the generative adversarial network model building process is as follows: Use the latent distribution of the generated sequence to build a convolutional neural network (CNN) or a recurrent neural network (RNN) or a variational autoencoder (VAE) to explore the latent distribution of the energy consumption time series and encode it into a latent vector to construct a generator; build another CNN or RNN or VAE as a discriminator to recognize the generated data; Select the binary cross entropy function as the loss function; Determine a threshold value that is related to the mean and standard deviation of the original energy consumption data. For the discriminator and the generator, the weight parameters are updated alternately in each training iteration cycle. During training, the generator and the discriminator influence each other. The loss value is optimized by adjusting the hyperparameters of the generator and discriminator to obtain more realistic generated data; when the generator reconstructs the data, unreasonable generated data is filtered out.

[0031] As a refinement of the above embodiment, in the training of the generative adversarial network, the validity and accuracy of the generated data directly affect the accuracy of the final prediction model. In order to generate high-quality data, an additional filter module is set between the generator and the discriminator. The filter compares the data generated by the generator with the mean of the original data, and filters out invalid data with large errors based on the comparison result. Only when the average power factor of the generated data is greater than a certain threshold will it be used for training, thereby improving the generalization ability of the energy consumption data set, and then improving the performance of the energy consumption prediction model.

[0032] Specifically, in the training process of the GAN model, we set an average power factor (APF) threshold as a standard for data filtering. The average power factor is an indicator of the effective use of electrical energy by equipment or systems in the power system. For building energy consumption data, it reflects the energy conversion efficiency of energy-consuming equipment over a period of time.

[0033] After each data generation, we calculate the average power factor of the generated data. This calculation can be based on the average value of the entire data series, or it can be calculated in segments and then averaged, depending on the characteristics of the data and analysis requirements.

[0034] If the average power factor of the generated data is greater than or equal to the threshold we set, then the data is considered to comply with physical laws and practical constraints, and will therefore be included in the training set for further training and optimization of the subsequent GAN model.

[0035] On the contrary, if the average power factor of the generated data is less than the set threshold, then these data may reflect unreasonable energy consumption patterns or outliers, and therefore will be filtered out and will not participate in the training of the model.

[0036] In this way, we can effectively control the quality of generated data and ensure that they conform to actual physical laws and constraints, thereby improving the accuracy and reliability of the GAN model in predicting building energy consumption. At the same time, this also provides a strong guarantee for the stability and generalization ability of the model in practical applications.

[0037] As a refinement of the above embodiment, in step S3, GAN is used to explore the data distribution of the original data set, and enhanced data similar to the original data distribution is generated, that is, "small data" is converted into "big data".

[0038] Generate enhanced building energy consumption data, including: S301: Obtain building energy consumption data, including time series data and physical properties of the building; S302: Data encoding, encoding the energy consumption data and inputting it into the generator model established in the model building phase; using the encoded data to train the generator, and using the generator and the discriminator to generate energy consumption data; S303: Generate enhanced data, and predict the enhanced data according to the generator model.

[0039] Optional, assuming there is The energy consumption data series of real buildings in the past 24 hours are as follows: , In the formula, t represents the number of days (days); n represents the number of sampling data per day, Represents a series of energy consumption data for a building.

[0040] The distribution of daily energy consumption data of buildings has strong similarity. In the data generation stage, the daily energy consumption data sequence is input into GAN, and GAN is used to discover the potential distribution of daily energy consumption data, and to generate an enhanced data sequence similar to the distribution of the original daily energy consumption data. The generation of enhanced data can be regarded as the conversion from noise data to generated data achieved by GAN. In the actual data generation process, some data generated by GAN have a great difference in distribution from the original data sequence, and the use of these data with large differences may reduce the prediction accuracy of the building energy consumption model. In order to solve this problem, a filter is set to remove data sequences with large differences. The filter compares the generated data sequence with the mean sequence of the original data and obtains the MAPE. When the MAPE is greater than a certain threshold (for example, 10% or adjusted according to specific needs), the generated data sequence is discarded. After filtering the generated data, the enhanced data is obtained, which is mixed with the original data to train the machine learning regression model (such as BPNN, ELM, SVR, etc.) to realize the construction of the parallel prediction model of building energy consumption.

[0041] MAPE is a commonly used indicator to measure the difference between predicted values ​​and actual values, especially suitable for time series data.

[0042] The formula for calculating MAPE is as follows: , in, is the original data sequence at time point The actual value of Is to generate data series at the same time point The predicted value of is the total number of data points.

[0043] Specifically, suppose GAN generates The time series data of building energy consumption for 24 hours is shown in the following formula: , In the formula, Represents the enhanced building energy consumption data series.

[0044] The essence of using GAN to generate parallel energy consumption data is to explore the potential inherent distribution of the original energy consumption data. By adding random factors to generate new data, it is actually equivalent to using GAN to build a virtual energy consumption system parallel to the real building. In principle, the energy consumption data generated by GAN is infinite and non-repetitive.

[0045] In the energy consumption parallel prediction stage based on hybrid data, The mixed energy consumption time series of the day is converted into a one-dimensional mixed data time series as shown in the following formula: , For simplicity, the above formula is expressed as: , In the formula, K=n×( ), this mixed data sequence will be used to build a machine learning prediction model.

[0046] As a refinement of the above embodiment, the actual building energy consumption data and the enhanced building energy consumption data in step S4 are mixed and used to train machine learning models (such as BPNN, ELM, SVR, etc.) to build a parallel prediction model for building energy consumption, that is, to discover "small knowledge and small rules" from "big data".

[0047] The p data before time k in the mixed data sequence are used to predict the energy consumption value at time k, that is, using To predict The prediction formula is as follows: , In the formula, Represents a prediction model implemented by a machine learning algorithm, such as BPNN, ELM, SVR, etc.

[0048] As a refinement of the above embodiment, in the performance evaluation phase in step S5, in order to verify the validity of the generated data and the accuracy of the prediction model, it is necessary to perform a quality evaluation on the generated data and verify the prediction results. The performance of the model is evaluated on the test set, and a variety of evaluation indicators, such as accuracy, precision, recall, etc., are used to comprehensively evaluate the effect of the model. If the model performs poorly, the structure of the model will be adjusted or the number of training rounds will be increased as needed.

[0049] The embodiment of the present disclosure also provides a device for generating building energy consumption data and parallel prediction based on a generative adversarial network, including a processor and a memory. Optionally, the device may also include a communication interface and a bus. The processor, the communication interface, and the memory may communicate with each other through the bus. The communication interface may be used for information transmission. The processor may call the logic instructions in the memory to execute the method for generating building energy consumption data and parallel prediction based on a generative adversarial network of the above embodiment.

[0050] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0051] The memory, as a computer-readable storage medium, can be used to store software programs and computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present disclosure. The processor executes the program instructions / modules stored in the memory to perform functional applications and data processing, that is, to implement the building energy consumption data generation and parallel prediction method based on the generative adversarial network in the above embodiment.

[0052] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory may include a high-speed random access memory and may also include a non-volatile memory.

[0053] An embodiment of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to execute the above-mentioned building energy consumption data generation and parallel prediction method based on a generative adversarial network.

[0054] The computer-readable storage medium mentioned above may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0055] The technical solution of the embodiment of the present disclosure can be embodied in the form of a software product, which is stored in a storage medium and includes one or more instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiment of the present disclosure. The aforementioned storage medium may be a non-transient storage medium, including: a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, and other media that can store program codes, or a transient storage medium.

[0056] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A building energy consumption data generation and parallel prediction method based on generative adversarial network, characterized in that: include: Collect actual building energy consumption data and preprocess the data as training samples, including time series data and physical properties of buildings; Construct and train a generative adversarial network model, in which the generator is used to generate enhanced building energy consumption data, and the discriminator is used to distinguish between the actual building energy consumption data and the enhanced building energy consumption data generated by the generator; Input the actual building energy consumption data into the generative adversarial network model to generate enhanced building energy consumption data; The actual building energy consumption data and the enhanced building energy consumption data are mixed to obtain mixed data, which is used to train the machine learning model, build a parallel prediction model for building energy consumption, and predict future building energy consumption values; Evaluate the performance of the prediction model on the test set, and adjust the model structure or increase the number of training rounds to improve the prediction accuracy.

2. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 1 is characterized in that: The time series data includes electricity consumption, water consumption, gas usage and building-related environmental parameters, wherein the building-related environmental parameters include temperature, humidity and light intensity; The physical properties of the building include building area, building materials, and building purpose.

3. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 1 is characterized in that: The generator of the generative adversarial network model transitions from a simple linear generator to a complex nonlinear generator; the nonlinear generator includes an autoencoder and a variational autoencoder; the generator maps the potential correlation of the energy consumption time series to the data space and performs data transformation by increasing randomness; The discriminator transitions from a multi-layer perceptron to a deep neural network, which includes an Inception module and a ResNet module; the discriminator distinguishes between generated data and original data, and reversely updates the weight of the generator based on the distinction result.

4. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 1 is characterized in that: During the training process of the generative adversarial network model, the generated data is filtered by setting the average power factor threshold. When the average power factor of the generated data is greater than the threshold, it is included in the training.

5. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 3 is characterized in that: Generate enhanced building energy consumption data, including: Obtain building energy consumption data, including time series data and physical properties of buildings; Data encoding: Encode the energy consumption data and input it into the generator model established in the model construction phase; use the encoded data to train the generator, and use the generator and discriminator to generate energy consumption data; Generate augmented data and predict augmented data based on the generator model.

6. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 1 is characterized in that: The generative adversarial network model sets a filter to remove data sequences with large differences. The filter compares the generated data sequence with the mean sequence of the original data and obtains the MAPE. When the MAPE is greater than the set threshold, the generated data sequence is discarded.

7. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 1 is characterized in that: The building energy consumption parallel prediction model is constructed using BPNN, ELM or SVR algorithm.

8. The method for generating building energy consumption data and parallel prediction based on a generative adversarial network according to claim 1 is characterized in that: The performance of the prediction model is evaluated using accuracy, precision, and recall.

9. A device for generating building energy consumption data and parallel prediction based on a generative adversarial network, comprising a processor and a memory storing program instructions, characterized in that: The processor is configured to execute the building energy consumption data generation and parallel prediction method based on a generative adversarial network as described in any one of claims 1-8 when running the program instructions.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the method described in any one of claims 1 to 8 is implemented.

Images