A method and system for predicting energy consumption of a low-carbon construction production line

Abstract

This invention discloses a method and system for predicting energy consumption in low-carbon construction production lines. The invention utilizes ARIMA and artificial neural networks to design and develop a hybrid energy consumption prediction model for short-term energy forecasting in energy systems. First, electricity consumption data is classified using an ARIMA model. Then, the results obtained from the ARIMA model are used as one of the inputs to an artificial neural network model. The artificial neural network model simultaneously considers factors affecting electricity consumption, such as solar power generation, operating hours, and production volume. By integrating the correlations between these factors, a predicted value is output. Finally, the prediction values ​​from both the ARIMA model and the artificial neural network are combined to obtain the final energy consumption prediction result.

Description

Technical Field

[0001] This invention belongs to the field of energy forecasting, specifically relating to a method and system for forecasting the energy consumption of a low-carbon construction production line. Background Technology

[0002] Energy consumption is one of the main contributors to carbon emissions. When fossil fuels such as coal, oil, and natural gas are used as energy sources during the design, construction, and operation of low-carbon construction production lines, large amounts of greenhouse gases such as carbon dioxide (CO2) are released. Therefore, energy consumption forecasting is a crucial step, helping various industries to rationally plan and manage energy consumption, reducing energy waste and carbon emissions. Traditional production processes often suffer from opaque energy consumption and inaccurate energy consumption forecasts. To address these issues, advanced technologies and methods are needed to achieve energy consumption forecasting for low-carbon construction production lines.

[0003] Existing energy consumption prediction methods largely rely on traditional forecasting techniques, such as statistical methods and mathematical models. Traditional statistical methods typically use historical data and statistical models for energy consumption prediction. For example, time series analysis can be used to analyze and predict historical energy consumption data. This method assumes that future energy consumption patterns will be similar to past patterns, but it may have limitations when facing complex energy consumption changes and nonlinear relationships. Traditional energy consumption models are mathematical models built based on physical principles and empirical laws. These models use the physical parameters and characteristics of production equipment, such as material thermal conductivity and equipment energy efficiency, to predict energy consumption. Common energy consumption models include heat transfer models, lighting models, and air conditioning models. However, these models require accurate parameters and input data, and need appropriate adjustments and validation for different building and equipment types.

[0004] These traditional methods have played a role in predicting energy consumption in production lines, but they also have some limitations. They may lack accuracy, real-time performance, and personalization, and they cannot leverage advanced data analytics and intelligent technologies to improve the accuracy and efficiency of energy consumption prediction.

[0005] The accuracy of energy consumption forecasting largely depends on reliable and sufficient data. However, acquiring and processing large amounts of energy consumption data can be challenging. Incomplete, inaccurate, or inconsistent data can lead to biases and inaccuracies in forecasts. Furthermore, energy consumption forecasting involves multiple variables and complex interrelationships, such as equipment energy efficiency and climate conditions. The interactions and uncertainties among these variables make energy consumption forecasting complex and difficult. Traditional methods and models may have limited forecasting accuracy and may fail to capture complex nonlinear relationships and energy consumption variation patterns. Therefore, in some cases, forecast results may deviate significantly from actual energy consumption. Traditional methods typically rely on human judgment and intervention, lacking intelligent and automated features. This makes the energy consumption forecasting process more time-consuming and labor-intensive, while also increasing the risk of human error. Summary of the Invention

[0006] The purpose of this invention is to address the problems existing in the prior art by providing a method and system for predicting energy consumption in low-carbon construction production lines. By incorporating solar power generation data, working days, product production data, and electricity consumption data into the prediction, and combining the linear part of the ARIMA (Autoregressive Integrated Moving Average Model) model with the nonlinear part of the artificial neural network, the error rate between actual and predicted data is minimized, thereby improving the accuracy and reliability of energy prediction.

[0007] According to one aspect of the present invention, a method for predicting the energy consumption of a low-carbon construction production line is provided, comprising: Acquire data on the production line's power consumption, solar power generation, number of working days, and product production within a preset time period; The acquired electricity consumption data, solar power generation data, working days and product production data are input into the trained energy consumption prediction hybrid model, and the energy consumption prediction results are output. The training of the energy consumption prediction hybrid model includes: Obtain past time period data on power consumption, solar power generation, working days, and product production data of the production line, construct a training dataset, and perform extended Dickie-Fowler unit root test on it; The processed electricity consumption data is input into the autoregressive moving average model to obtain the linear predicted value and nonlinear residual of the electricity consumption data. The nonlinear residuals output by the autoregressive moving average model are combined with solar power generation data, working days and product production data within the same time period and input into the artificial neural network model. The nonlinear least squares algorithm is used to perform optimization calculations to determine the minimum value of the sum of squares of the nonlinear function and output the nonlinear prediction value. Based on the linear and nonlinear prediction values, the energy consumption prediction results are obtained, and the mean square error is used to evaluate the energy consumption prediction results to obtain a trained hybrid energy consumption prediction model.

[0008] As a further technical solution, the training of the energy consumption prediction hybrid model also includes: determining the parameters of the autoregressive moving average model based on the processed electricity consumption data using correlation functions and the generalized Dickey-Fowler unit root test.

[0009] As a further technical solution, the parameter determination of the autoregressive moving average model also includes: Acquire meteorological and environmental data for the same time period as the power consumption data; The seasonal parameters and exogenous variables of the autoregressive moving average model are determined using the meteorological and environmental data, resulting in the modified autoregressive moving average model.

[0010] As a further technical solution, the modified autoregressive moving average model is as follows: , in, Let β represent the exogenous variable, and let β represent the coefficient of the exogenous variable, indicating the effect of the exogenous factor on the exogenous variable. The linear effect, ARIMA terms represent the uncorrected autoregressive moving average model; seasonal terms represent seasonal parameters. This represents the white noise error term.

[0011] As a further technical solution, the method also includes: The objective function for constructing a hybrid energy consumption prediction model is: , Where T represents the number of time periods. Actual power consumption data for each time period t This refers to the predicted electricity consumption data for each time period t obtained from the energy consumption prediction hybrid model.

[0012] As a further technical solution, the formula for the relationship between input data and prediction data in an artificial neural network is: , in, These are the connection coefficients of neuron i in the input layer of an artificial neural network model. These are the connection coefficients between the input layer and hidden layer in an artificial neural network model, corresponding to the connection between neurons i and j. Solar power generation data for each time period t. The number of working days in each time period t. Product production data for each time period t. The prediction error is the power consumption data for each time period t obtained from the artificial neural network model.

[0013] According to one aspect of the present invention, an energy consumption prediction system for a low-carbon construction production line is provided, comprising: The input module is used to obtain data on the production line's power consumption, solar power generation, number of working days, and product production within a preset time period. The prediction module is used to input the acquired electricity consumption data, solar power generation data, working days and product production data into the trained energy consumption prediction hybrid model and output the energy consumption prediction results. The training of the energy consumption prediction hybrid model includes: Obtain past time period data on power consumption, solar power generation, working days, and product production data of the production line, construct a training dataset, and perform extended Dickie-Fowler unit root test on it; The processed electricity consumption data is input into the autoregressive moving average model to obtain the linear predicted value and nonlinear residual of the electricity consumption data. The nonlinear residuals output by the autoregressive moving average model are combined with solar power generation data, working days and product production data within the same time period and input into the artificial neural network model. The nonlinear least squares algorithm is used to perform optimization calculations to determine the minimum value of the sum of squares of the nonlinear function and output the nonlinear prediction value. Based on the linear and nonlinear prediction values, the energy consumption prediction results are obtained, and the mean square error is used to evaluate the energy consumption prediction results to obtain a trained hybrid energy consumption prediction model.

[0014] According to one aspect of the present invention, an energy consumption prediction device for a low-carbon construction production line is provided, comprising: at least one processor, at least one memory, and a communication interface; wherein the processor, memory, and communication interface communicate with each other; the memory stores program instructions executed by the processor, and the processor invokes the program instructions to execute the energy consumption prediction method for a low-carbon construction production line.

[0015] According to one aspect of the present invention, a non-transitory computer read storage medium is provided, the non-transitory computer read storage medium storing computer instructions that cause the computer to execute the energy consumption prediction method for a low-carbon construction production line.

[0016] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention improves the accuracy of energy consumption prediction by combining the advantages of ARIMA model and ANN artificial neural network, providing important support for achieving optimal management decisions. ARIMA model can capture the linear trend and seasonality of time series data, while artificial neural network can handle nonlinear relationships and complex patterns. By combining these two methods, it is possible to more accurately predict short-term changes in energy consumption.

[0017] (2) Based on the ARIMA model, this invention introduces meteorological data and other environmental factors for training, adjusts and corrects future energy consumption, and improves the accuracy of energy consumption prediction.

[0018] (3) By providing accurate short-term energy forecasts, this invention helps to optimize energy planning and allocation. Energy suppliers and energy managers can make reasonable energy purchases and allocations based on the forecast results to meet actual needs, avoid energy waste, and ensure the stability and sustainability of energy supply.

[0019] (4) This invention supports sustainable energy management. By accurately predicting energy consumption, energy managers can better integrate renewable energy sources, such as solar and wind energy, to maximize the use of clean energy and reduce dependence on traditional energy sources, which helps to reduce environmental impact and promote sustainable development. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the energy consumption prediction method for a low-carbon construction production line according to an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the training process of the energy consumption prediction hybrid model according to an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the energy consumption prediction system for a low-carbon construction production line according to an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] This invention aims to address the problems of inaccurate and inefficient energy consumption forecasting in low-carbon construction production lines. Existing energy management methods cannot provide accurate energy consumption forecasts and lack real-time monitoring capabilities, resulting in an inability to adjust and optimize energy control strategies in a timely manner. Therefore, a method is needed to accurately predict the energy consumption of low-carbon construction production lines to achieve precise control and optimization of energy consumption. Effective energy forecasting and optimization are crucial for reducing carbon emissions from production lines. By establishing an energy forecasting system and analyzing energy consumption in real time, energy waste can be identified and corrected. Simultaneously, taking energy-saving measures, such as optimizing equipment operating parameters, improving production processes, and implementing energy recovery and utilization, can minimize energy use and waste.

[0025] This invention addresses the issue of low accuracy in traditional energy consumption forecasting methods by providing a hybrid energy consumption forecasting model to improve the reliability and accuracy of energy consumption prediction. This helps companies or organizations building low-carbon production lines to better predict energy consumption, achieving more efficient energy use and reducing carbon emissions. The application of this method can provide valuable information for enterprises in energy planning and decision-making, thereby promoting sustainable development and environmental protection.

[0026] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This invention proposes an energy consumption prediction method for low-carbon production line construction, aiming to minimize the error rate between actual and predicted data. Based on a developed hybrid energy consumption prediction model, real data (electricity consumption, solar power generation, production volume, and working days) are obtained from the actual production line to predict short-term / medium-term electricity consumption data.

[0028] like Figure 1 As shown, this embodiment of the invention provides an energy consumption prediction method for low-carbon construction production lines, including: acquiring power consumption data, solar power generation data, working days and product production data of the production line within a preset time period; inputting the acquired power consumption data, solar power generation data, working days and product production data into a trained energy consumption prediction hybrid model, and outputting energy consumption prediction results.

[0029] The training of the hybrid energy consumption prediction model includes: acquiring past power consumption data, solar power generation data, working days, and product production data for the production line; constructing a training dataset and performing extended Dickie-Fowler unit root tests on it; inputting the processed power consumption data into an autoregressive moving average model (hereinafter referred to as the ARIMA model) to obtain linear predicted values ​​and nonlinear residuals of the power consumption data; inputting the nonlinear residuals output by the autoregressive moving average model, combined with solar power generation data, working days, and product production data for the same time period, into an artificial neural network model, using a nonlinear least squares algorithm for optimization calculation to determine the minimum value of the sum of squares of the nonlinear function, and outputting nonlinear predicted values; obtaining energy consumption prediction results based on the linear and nonlinear predicted values, and evaluating the energy consumption prediction results using mean square error to obtain the trained hybrid energy consumption prediction model.

[0030] This invention utilizes an ARIMA model and an artificial neural network to design and develop a hybrid energy consumption prediction model for short-term energy forecasting. Specifically, energy consumption data within a preset time period is predicted using the ARIMA method. The ARIMA prediction is then used as one of the inputs to the artificial neural network model. Factors influencing energy consumption, such as solar power generation, working hours, production volume, and past electricity consumption data, are considered. After assessing the correlations between these factors, different numbers of neurons and different training algorithms are studied and tested to design the optimal system.

[0031] The goal of this invention's hybrid energy consumption prediction model is to improve the accuracy and reliability of energy prediction by combining the linear part of the ARIMA model with the nonlinear part of an artificial neural network. Simultaneously, by considering multiple factors and using different training algorithms, the optimal model configuration is found to achieve accurate short-term energy prediction.

[0032] In the energy consumption prediction method for low-carbon construction production lines provided in this invention, as a preferred embodiment, the ARIMA model configuration is determined based on the non-stationary nature of time series data. First, the difference order is determined by performing a unit root (ADF) test on the data or observing autocorrelation and partial autocorrelation plots to ensure data stationarity. Second, for data with seasonal variations, appropriate seasonal parameters (P, D, Q) are selected for seasonal adjustment. Furthermore, by experimenting with different combinations of (p, d, q) and comparing their AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) values, the optimal model parameter configuration is found to achieve the lowest information criterion value, thereby improving the accuracy of model predictions.

[0033] In the energy consumption prediction method for low-carbon construction production lines provided in this invention, as a preferred embodiment, the artificial neural network part adopts a multilayer perceptron (MLP) structure, which includes an input layer, multiple hidden layers, and an output layer. Each hidden layer uses the ReLU activation function to enhance the model's ability to handle nonlinear problems. The depth and width of the network (i.e., the number of hidden layers and the number of neurons per layer) are adjusted according to the model's performance on the validation dataset to avoid overfitting while maintaining sufficient learning capacity. The network is trained using the Adam optimizer, whose automatic learning rate adjustment makes training more stable. During training, early stopping is used to prevent overfitting; that is, if the error on the validation set does not improve after several consecutive epochs, training is stopped.

[0034] In the energy consumption prediction method for low-carbon construction production lines provided in this invention, as a preferred embodiment, in feature selection, in addition to the output of the ARIMA model, other factors affecting short-term energy consumption are also considered, such as temperature, humidity, and weekday / holiday identification. All feature data are standardized before being input into the neural network, i.e., the mean of each feature is subtracted and divided by its standard deviation, thereby accelerating the convergence of the neural network and reducing training time. Furthermore, exploratory data analysis is used to identify and handle outliers or missing data to ensure data quality.

[0035] In the energy consumption prediction method for low-carbon construction production lines provided in this invention, as a preferred embodiment, the model training is divided into multiple stages. Each stage utilizes cross-validation to optimize parameters and prevent the model from overfitting on specific samples. Furthermore, by evaluating the model's performance on independent test sets, its generalization ability to unknown data can be verified. Performance metrics such as mean squared error (MSE) and absolute percentage error (MAPE) are used to quantify the accuracy and reliability of the model's predictions.

[0036] In the energy consumption prediction method for low-carbon construction production lines provided in this invention, as a preferred embodiment, the training of the energy consumption prediction hybrid model is as follows: Figure 2 As shown, this includes data acquisition and preprocessing, data division into linear and nonlinear components, and model evaluation.

[0037] Data acquisition and preprocessing include: Step 1.1: Collect power consumption data of the production line over past time periods.

[0038] Step 1.2: Collect indirectly related power generation data. For example, solar power generation data. The addition of indirect data is mainly because the energy output of solar panels is directly affected by weather conditions, cloud distribution, wind, and other meteorological factors. In this way, the model can indirectly incorporate meteorological factors into the electricity consumption data set.

[0039] Step 1.3 involves performing an extended Dickey-Fuller unit root test on the relevant electricity consumption dataset to attempt to understand the overall structure of the dataset. Specifically, the extended Dickey-Fuller unit root test is first performed on the electricity consumption dataset to check whether the data is stable, i.e., whether data characteristics such as mean and variance change over time. After completing the stability test, more information is explored in the dataset, such as trends, seasonal variations, and cyclical fluctuations, to comprehensively understand the behavior and characteristics of the data for better prediction or decision support.

[0040] It's important to clarify that attempting to understand the overall structure of a dataset refers to "determining" its overall structure through data exploration, which is typically not a one-step process. Data exploration is an iterative and multi-method process; through continuous analysis and validation, the complex structure of the data can be gradually revealed and understood. Each analysis may provide a perspective on a particular aspect of the data, but a combination of multiple tools and methods is needed to achieve a more comprehensive understanding of the dataset's overall structure.

[0041] The process of dividing data into linear and nonlinear components includes: Step 2.1 involves inputting historical electricity consumption data into the ARIMA model to decompose the linear and nonlinear components of the dataset (i.e., residuals). The ARIMA model is defined by three important parameters (p, d, and q), all of which are positive integers. Here, p represents the order of the autoregressive model (time lag), d represents the degree of differencing based on historical data, and q represents the order of the moving average model.

[0042] Step 2.2: Use the autocorrelation function to find the correlation p and q values ​​that are close to the confidence interval. Determine the d value based on the results of the generalized Dickey-Fuller unit root test.

[0043] The specific steps are as follows: 1) Determine p and q using the autocorrelation function (ACF) and partial autocorrelation function (PACF). The order *p* of the autoregressive (AR) component is determined by the partial autocorrelation function (PACF), which shows the direct relationship between each lag and the current value, eliminating the influence of intermediate lags. If the PACF approaches zero after order *p*, then *p* can be considered as that order. For the moving average (MA) component, the order *q* is determined by the autocorrelation function (ACF). The ACF shows the correlation between the time series and its past values; if the ACF rapidly decays to near zero after order *q*, then *q* is a suitable order.

[0044] 2) Determine d using the generalized Dickey-Fuller test. The differencing order *d* is the number of differencing operations required to bring a time series to a stationary state. The generalized Dickey-Fuller (GDF) test is used to detect the presence of a unit root in the series, i.e., to determine whether the series is non-stationary. If the test rejects the null hypothesis of a unit root, the series is considered stationary, and *d* can be zero or a smaller value. If the null hypothesis is not rejected, the differencing order needs to be increased.

[0045] The mathematical expressions for the above two steps are as follows: , Where L represents the lag operator, Represents the autoregressive coefficient. Represents the moving average coefficient. This is the white noise error term, where i and j represent the i-th and j-th samples, and t represents time.

[0046] As a preferred embodiment, meteorological data and other environmental factors are used to adjust and correct future energy consumption to improve forecast accuracy. The SARIMAX model is an extension of the ARIMA model, incorporating seasonal adjustments and exogenous variables, making it particularly suitable for complex time series data that need to handle seasonal fluctuations and external influencing factors. Based on the previously determined basic ARIMA parameters (p, d, q), seasonal parameters (P, D, Q) and a value S defining the seasonal cycle are added; for example, S might be set to 12 for monthly data.

[0047] In the SARIMAX model, the introduction of exogenous variables allows the model to consider factors that directly affect the target variable but are not entirely explained by changes in the time series itself, such as temperature, humidity, and special days. By incorporating these exogenous variables into the model, energy consumption trends significantly influenced by external conditions can be predicted more accurately. The introduction is as follows: , in These are the coefficients of exogenous variables, representing the influence of exogenous factors on the exogenous variables. The linear effect.

[0048] In step 2.3, the predicted values ​​generated by the ARIMA model are treated as linear components and sent to the end of the system. The nonlinear residuals (the difference between the true and predicted values) continue to be processed in the system as one of the input data for the artificial neural network model. (Formula) ,in, Actual power consumption data for each time period t For nonlinear power consumption data obtained from the ARIMA model for each time period t, This provides linear power consumption data for each time period t obtained from the ARIMA model.

[0049] In the hybrid model approach, the final prediction dataset is generated using a linear dataset obtained from the ARIMA model and prediction results obtained from an artificial neural network. The formula is as follows: ,in, To predict electricity consumption data for each time period t obtained from the hybrid model, Predict electricity consumption data for each time period t obtained from an ANN artificial neural network model.

[0050] Specifically, past electricity consumption data is used to obtain the linear and nonlinear components of the ARIMA model. The following formula describes the prediction obtained by subtracting the predicted data from the actual dataset and using it as one of the inputs to the artificial neural network.

[0051] The formula is as follows: Where t is the time interval (t: 1, ..., T), and p and q are ARIMA model constants. Here, represents the autocorrelation coefficient of the ARIMA model, and y represents the actual electricity consumption data for each time period t. These are the moving average coefficients of the ARIMA model. This is for model error. At this stage, other input data directly or indirectly related to electricity consumption (weekday and solar power generation data) are also added to the artificial neural network model.

[0052] Step 2.4: Train the input information using an artificial neural network and perform optimization calculations using the Levenberg-Marquardt algorithm to find the minimum value of the sum of squares of the nonlinear function.

[0053] The ANN (Artificial Neural Network), or ANN model, comprises an input layer, hidden layers, and an output layer. The input layer of the ANN model receives various types of data, including time-series data preprocessed by the ARIMA model (primarily handling linear trends and seasonal factors) and other nonlinear factors that may affect energy consumption, such as meteorological data like temperature and humidity. After analysis by the ARIMA model, the main linear components and seasonal patterns have been removed. The remaining residual information (i.e., the nonlinear components in the time-series data) and direct external factors constitute the input of the ANN.

[0054] In the hidden layers of an ANN model, the input data is passed to one or more hidden layers, which consist of numerous neurons. Each neuron can process information through an activation function. Within the hidden layers, the network learns complex patterns and nonlinear relationships in the input data through weighted connections. Each neuron performs a weighted summation of the inputs from the previous layer, and then transforms these input signals through a nonlinear activation function to extract and learn hidden features and patterns in the data.

[0055] In the output layer of an ANN model, processed information flows through the hidden layers before reaching the network's output layer. In the output layer, the network calculates a prediction of future energy consumption based on the input data and the features learned from the hidden layers. The output is typically a continuous numerical value representing the predicted energy consumption.

[0056] Preferably, the ANN model continuously adjusts the weights between neurons through backpropagation and other optimization techniques (such as gradient descent) to minimize the difference between the predicted output and the actual data, thereby improving the model's prediction accuracy. In this way, the network can better learn and adapt to nonlinear relationships in the data, providing more accurate predictions of energy consumption.

[0057] The objective function of the energy consumption prediction hybrid model is: .

[0058] The formula relating input data and prediction data in an artificial neural network is: ,in, These are the connection coefficients of neuron i in the input layer of the ANN model. These are the connection coefficients between the input and hidden layers in an ANN model, corresponding to the connections between neurons i and j. Solar power generation data for each time period t. The number of working days in each time period t. Product production data for each time period t. H represents the prediction error of electricity consumption data for each time period t obtained from the ANN model, and H is the number of neurons. The training objective of the model is set to have an R² value close to 1, using connection weights between the neural networks, to reveal the relationship between the input and predicted data.

[0059] The model evaluation includes: The mean squared error (MSE) is used to evaluate the prediction results. After the model is built and predictions are made, the accuracy of the predictions needs to be evaluated. MSE is a commonly used statistical measure to assess the average magnitude of the difference between the model's predicted values ​​and the actual observed values. The formula for calculating MSE is as follows: , in, It is the model's prediction of the i-th observation. These are the actual observed values, and n is the total number of observed values. MSE provides a quantitative metric to evaluate the magnitude of the deviation between model predictions and actual values.

[0060] Furthermore, based on the modeling results, future energy consumption trends are predicted.

[0061] The various embodiments of this invention are implemented through programmed processing using a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of this invention are encapsulated into various modules. Based on this reality, and building upon the above embodiments, this invention provides an energy consumption prediction system for a low-carbon construction production line. This device is used to execute the energy consumption prediction method for a low-carbon construction production line in the above method embodiments. See also... Figure 3 The device includes: an input module for acquiring power consumption data of the production line within a preset time period; and a prediction module for inputting the acquired power consumption data into a trained hybrid energy consumption prediction model and outputting an energy consumption prediction result. The training of the hybrid energy consumption prediction model includes: constructing a power consumption data sample to be trained; inputting the power consumption data sample to be trained into an autoregressive moving average model for training, determining the parameters of the autoregressive moving average model, and obtaining a trained autoregressive moving average model; inputting the nonlinear power consumption data prediction value output by the autoregressive moving average model, combined with solar power generation data, working days, and product production data within the same time period, into an artificial neural network model for training, determining the minimum value of the sum of squares of the nonlinear function, and obtaining a trained artificial neural network model; and obtaining an energy consumption prediction result based on the linear power consumption data prediction value output by the autoregressive moving average model and the linear power consumption data prediction value output by the artificial neural network model.

[0062] The energy consumption prediction system for the low-carbon construction production line provided in this embodiment of the invention adopts... Figure 3 Several modules in the system combine the linear part of the ARIMA model with the nonlinear part of the artificial neural network to minimize the error rate between actual and predicted data, thereby improving the accuracy and reliability of energy forecasting.

[0063] It should be noted that the device embodiments provided by the present invention, in addition to implementing the methods in the above method embodiments, are also used to implement the methods in other method embodiments provided by the present invention. The difference lies only in the setting of corresponding functional modules. Their principles are basically the same as those of the above device embodiments provided by the present invention. Anyone skilled in the art, based on the above device embodiments and referring to the specific technical solutions in other method embodiments, can obtain corresponding technical means and technical solutions constituted by combining technical features, and improve the device in the above device embodiments while ensuring the practicality of the technical solutions, to obtain corresponding device-type embodiments for implementing the methods in other method-type embodiments. For example: Based on the above-described device embodiments, as an optional embodiment, the energy consumption prediction system for a low-carbon construction production line provided in this embodiment of the invention further includes training the hybrid energy consumption prediction model, which further includes: The evaluation module is used to evaluate the accuracy of the energy consumption prediction results using mean square error. After the accuracy of the energy consumption prediction results meets the preset requirements, the trained energy consumption prediction hybrid model is output.

[0064] Based on the above-described device embodiments, as an optional embodiment, the energy consumption prediction system for a low-carbon construction production line provided in this embodiment of the invention further includes the training of the autoregressive moving average model, which further includes: The sample data acquisition module is used to acquire meteorological data and environmental data samples within the same time period as the power consumption data sample; The model training module is used to input the acquired sample data into the trained autoregressive moving average model for training, determine the coefficients of seasonal parameters and exogenous variables, and obtain the corrected autoregressive moving average model.

[0065] Based on the above-described device embodiments, as an optional embodiment, the energy consumption prediction system for a low-carbon construction production line provided in this embodiment of the invention further includes the training of the autoregressive moving average model, which further includes: The first determining module is used to determine the order of the autoregressive part of the autoregressive moving average model using the partial autocorrelation function; The second determining module is used to determine the order of the moving average part of the autoregressive moving average model using the autocorrelation function. The third determination module is used to determine the difference order of the autoregressive moving average model using the generalized Dickey-Fowler test.

[0066] Based on the above-described device embodiments, as an optional embodiment, the energy consumption prediction system for a low-carbon construction production line provided in this embodiment of the invention further includes the training of the artificial neural network model, which further includes: The optimization calculation module is used to perform optimization calculations using the Levenberg-Marquardt algorithm to determine the minimum value of the sum of squares of the nonlinear function.

[0067] The method in this embodiment of the invention is implemented using an electronic device; therefore, it is necessary to introduce the relevant electronic device. For this purpose, embodiments of the present invention provide an electronic device, such as... Figure 4 As shown, the electronic device includes: at least one processor, a communication interface, at least one memory, and a communication bus, wherein the at least one processor, the communication interface, and the at least one memory communicate with each other via the communication bus. The at least one processor invokes logical instructions stored in the at least one memory to execute all or part of the steps of the methods provided in the foregoing method embodiments.

[0068] Furthermore, when the logical instructions in at least one of the aforementioned memories are implemented as software functional units and sold or used as independent products, they are stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, is embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (a personal computer, server, or network device) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks—various media for storing program code.

[0069] Based on the same inventive concept as the foregoing embodiments, this embodiment of the invention also provides a non-transitory computer read storage medium, which stores computer instructions that cause the computer to execute the energy consumption prediction method for a low-carbon construction production line.

[0070] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, located in one place, or distributed across multiple network units. The purpose of this embodiment is achieved by selecting some or all of the modules according to actual needs. Those skilled in the art will understand and implement this without any inventive effort.

[0071] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0072] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for energy consumption prediction of a low carbon construction production line, characterized by, include: Acquire data on the production line's power consumption, solar power generation, number of working days, and product production within a preset time period; The acquired electricity consumption data, solar power generation data, working days and product production data are input into the trained energy consumption prediction hybrid model, and the energy consumption prediction results are output. The training of the energy consumption prediction hybrid model includes: Obtain past time period data on power consumption, solar power generation, working days, and product production data of the production line, construct a training dataset, and perform extended Dickie-Fowler unit root test on it; The processed electricity consumption data is input into the autoregressive moving average model to obtain the linear predicted value and nonlinear residual of the electricity consumption data. The nonlinear residuals output by the autoregressive moving average model are combined with solar power generation data, working days and product production data within the same time period and input into the artificial neural network model. The nonlinear least squares algorithm is used to perform optimization calculations to determine the minimum value of the sum of squares of the nonlinear function and output the nonlinear prediction value. Based on the linear and nonlinear prediction values, the energy consumption prediction results are obtained, and the mean square error is used to evaluate the energy consumption prediction results to obtain a trained hybrid energy consumption prediction model.

2. The method of claim 1, wherein, The training of the energy consumption prediction hybrid model also includes: determining the parameters of the autoregressive moving average model based on the processed electricity consumption data using correlation functions and the generalized Dickey-Fowler unit root test.

3. The method of claim 2, wherein the energy consumption of the low- carbon construction production line is predicted based on the energy consumption of each of the plurality of construction processes. The parameter determination of the autoregressive moving average model also includes: Acquire meteorological and environmental data for the same time period as the power consumption data; The seasonal parameters and exogenous variables of the autoregressive moving average model are determined using the meteorological and environmental data, resulting in the modified autoregressive moving average model.

4. The method of claim 3, wherein the energy consumption prediction of the low- carbon construction production line is based on the energy consumption of the low- carbon construction production line. The modified autoregressive moving average model is as follows: ， where, represents exogenous variables, β represents the coefficient of exogenous variables, represents the linear effect of exogenous factors on ARIMA terms represent the autoregressive moving average model before correction; seasonal terms represent seasonal parameters, represents white noise error term.

5. The method of claim 1, wherein the low-carbon construction production line is a building construction production line. The method further includes: The objective function for constructing a hybrid energy consumption prediction model is: ， wherein T is the number of time periods, is actual power consumption data for each time period t, is predicted power consumption data for each time period t obtained from the energy consumption prediction hybrid model.

6. The method of claim 1, wherein, The formula relating input data and prediction data in an artificial neural network is: ， wherein, is a connection coefficient for input layer neuron i in the artificial neural network model, is a connection coefficient between the input layer and the hidden layer in the artificial neural network model, corresponding to a connection between neurons i and j, is solar power generation data for each time period t, is the number of weekdays for each time period t, is product production data for each time period t, is a power consumption data prediction error for each time period t obtained from the artificial neural network model.

7. A system for predicting energy consumption of a low carbon construction production line, characterized in that, include: The input module is used to obtain data on the production line's power consumption, solar power generation, number of working days, and product production within a preset time period. The prediction module is used to input the acquired electricity consumption data, solar power generation data, working days and product production data into the trained energy consumption prediction hybrid model and output the energy consumption prediction results. The training of the energy consumption prediction hybrid model includes: Obtain past time period data on power consumption, solar power generation, working days, and product production data of the production line, construct a training dataset, and perform extended Dickie-Fowler unit root test on it; The processed electricity consumption data is input into the autoregressive moving average model to obtain the linear predicted value and nonlinear residual of the electricity consumption data. The nonlinear residuals output by the autoregressive moving average model are combined with solar power generation data, working days and product production data within the same time period and input into the artificial neural network model. The nonlinear least squares algorithm is used to perform optimization calculations to determine the minimum value of the sum of squares of the nonlinear function and output the nonlinear prediction value. Based on the linear and nonlinear prediction values, the energy consumption prediction results are obtained, and the mean square error is used to evaluate the energy consumption prediction results to obtain a trained hybrid energy consumption prediction model.

8. An energy consumption prediction device for a low-carbon construction production line, characterized by, include: The system comprises at least one processor, at least one memory, and a communication interface; wherein the processor, memory, and communication interface communicate with each other; the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to execute the energy consumption prediction method for a low-carbon construction production line as described in any one of claims 1 to 6.

9. A non-transitory computer readable storage medium, comprising: The non-transitory computer read storage medium stores computer instructions, which cause the computer to execute the energy consumption prediction method for a low-carbon construction production line as described in any one of claims 1 to 6.

Images