CN111882033A - A Keras-based active and passive carbon emission prediction method for regional civil aviation - Google Patents
A Keras-based active and passive carbon emission prediction method for regional civil aviationDownload PDFInfo
- Publication number
- CN111882033A CN111882033ACN202010679105.XACN202010679105ACN111882033ACN 111882033 ACN111882033 ACN 111882033ACN 202010679105 ACN202010679105 ACN 202010679105ACN 111882033 ACN111882033 ACN 111882033A
- Authority
- CN
- China
- Prior art keywords
- model
- active
- civil aviation
- passive
- function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Translated fromChinese
Description
Translated fromChinese技术领域technical field
本发明涉及一种基于Keras的区域民航主被动碳排放预测方法，属于民航碳排放预测领域。The invention relates to a Keras-based active and passive carbon emission prediction method for regional civil aviation, which belongs to the field of civil aviation carbon emission prediction.
背景技术Background technique
民航碳排放来源广泛，其中航空器是排放的主体。因此，减少航空器碳排放是实现民航绿色发展的核心任务。而进行减碳行动之前，需要对民航碳排放的情况有详细的了解，这就要求有关部门建立民航碳排放清单，系统掌握民航碳排放的现状与趋势。考虑到民航飞机运行跨区域的特殊性以及有效提升减排措施的针对性，将某区域内民航碳排放划分为主动碳排放(即在区域内起降的航班LTO(Landing and Take-off，着陆与起飞)及CCD(Climb、Cruise and Descent，爬升巡航及进近)阶段在该区域内产生的碳排放)与被动碳排放(即其余飞越该区域航班的CCD阶段的碳排放)，可提出更合理且有针对性的减排目标，以达到理想的减排效果。因此，区域民航主被动碳排放的预测可以为更精确地开展民航碳减排工作提供支撑。Civil aviation has a wide range of sources of carbon emissions, of which aircraft are the main source of emissions. Therefore, reducing aircraft carbon emissions is the core task of realizing the green development of civil aviation. Before carrying out carbon reduction actions, it is necessary to have a detailed understanding of the carbon emissions of civil aviation, which requires relevant departments to establish a carbon emission inventory of civil aviation and systematically grasp the status and trends of carbon emissions of civil aviation. Considering the particularity of the operation of civil aviation aircraft across regions and the pertinence of effective emission reduction measures, the carbon emissions of civil aviation in a certain region are divided into active carbon emissions (that is, flights that take off and land in the region, LTO (Landing and Take-off, landing). and take-off) and CCD (Climb, Cruise and Descent, climb cruise and approach) phase in the carbon emissions generated in the area) and passive carbon emissions (that is, the remaining flights over the area CCD phase carbon emissions), can propose more Reasonable and targeted emission reduction targets to achieve the desired emission reduction effect. Therefore, the prediction of active and passive carbon emissions of regional civil aviation can provide support for more accurate carbon emission reduction of civil aviation.
通常预测以线性回归、神经网络、随机森林等方法最为常见，线性回归方法需要在使用前判断输入与输出是否为线性关系，有较高的局限性；神经网络可以较好地弥补这一缺陷，常见的BP(Back Propagaion，误差逆传播算法)神经网络虽然可以实现预测但其拟合速度慢、精度不够高，使得预测结果无法满足理想标准；随机森林方法涉及参数较复杂并且模型训练和预测速度比较慢。基于Keras框架的神经网络能够较好地满足民航碳排放预测时精度、速度、主被动排放区分等要求。Keras是基于Theano(机器学习库)的一个深度学习框架，也是一个高层神经网络API(Application Programming Interface，应用程序接口)，它的设计参考Torch(深度学习框架)，使用Python(计算机程序设计语言)编写进一步对TensorFlow(基于数据流编程的符号数学系统)进行了封装。Usually, linear regression, neural network, random forest and other methods are the most common prediction methods. Linear regression method needs to judge whether the input and output are linear before use, which has high limitations; neural network can better make up for this defect, Although the common BP (Back Propagaion, error back propagation algorithm) neural network can achieve prediction, its fitting speed is slow and the accuracy is not high enough, so that the prediction results cannot meet the ideal standard; the random forest method involves more complex parameters and the speed of model training and prediction. slower. The neural network based on the Keras framework can better meet the requirements of accuracy, speed, and distinction between active and passive emissions in the prediction of civil aviation carbon emissions. Keras is a deep learning framework based on Theano (machine learning library) and a high-level neural network API (Application Programming Interface). Its design refers to Torch (deep learning framework) and uses Python (computer programming language). The writing further encapsulates TensorFlow (a symbolic math system based on dataflow programming).
发明内容SUMMARY OF THE INVENTION
为解决现有区域民航排放预测未区分主被动碳排放、预测精度低、计算速度慢等不足，本发明提出了一种基于Keras的区域民航主被动碳排放预测方法，确定涵盖区域地理位置、社会经济、民航运输等影响区域民航主被动碳排放的综合指标体系，收集并预处理区域内各项指标具体数值建立数据集，并将数据集划分为训练集和测试集，基于Keras框架构建神经网络模型，运用Python调用Keras框架的API构建损失函数和优化函数，将神经网络模型在训练集上训练，并在测试集上进行测试，以此优化并确定模型各参数，最终输出并保存神经网络的结构及权重。依据未来区域各指标数据，使用上述模型即可进行未来的区域主被动碳排放预测。In order to solve the shortcomings of existing regional civil aviation emission forecasts that do not distinguish between active and passive carbon emissions, low prediction accuracy, and slow calculation speed, the present invention proposes a Keras-based regional civil aviation active and passive carbon emission prediction method, which determines the geographical location, social A comprehensive index system that affects the active and passive carbon emissions of regional civil aviation, such as economy and civil aviation transportation, collects and preprocesses the specific values of various indicators in the region to establish a data set, divides the data set into training set and test set, and builds a neural network based on the Keras framework Model, use Python to call the API of the Keras framework to build the loss function and optimization function, train the neural network model on the training set, and test it on the test set, so as to optimize and determine the parameters of the model, and finally output and save the neural network. structure and weight. Based on the data of various indicators in the future region, the above model can be used to predict the future active and passive carbon emissions in the region.
本发明为解决其技术问题采用如下技术方案：The present invention adopts following technical scheme for solving its technical problem:
一种基于Keras的区域民航主被动碳排放预测方法，包括以下步骤：A Keras-based active and passive carbon emission prediction method for regional civil aviation, including the following steps:
(1)确定涵盖区域地理位置、社会经济、民航运输方面影响区域民航主被动碳排放的综合指标体系；(1) Determine a comprehensive index system covering regional geographic location, social economy, and civil aviation transportation that affects the active and passive carbon emissions of regional civil aviation;
(2)收集并预处理区域内各项指标具体数值建立数据集，并将数据集划分为训练集和测试集；(2) Collect and preprocess the specific values of various indicators in the area to establish a data set, and divide the data set into a training set and a test set;
(3)基于Keras框架构建包含输入层、全连接层和输出层的神经网络模型，设置每层的神经元个数及激活函数；(3) Build a neural network model including an input layer, a fully connected layer and an output layer based on the Keras framework, and set the number of neurons and activation function of each layer;
(4)运用Python调用Keras框架的API构建损失函数和优化函数，将神经网络模型在训练集上训练，并用测试集完成模型参数的优化，最终输出并保存神经网络的结构及权重；(4) Use Python to call the API of the Keras framework to construct the loss function and optimization function, train the neural network model on the training set, and use the test set to complete the optimization of model parameters, and finally output and save the structure and weight of the neural network;
(5)依据未来区域的各指标数据，使用训练完成的神经网络模型进行未来的区域主被动碳排放预测。(5) According to the index data of the future region, use the trained neural network model to predict the active and passive carbon emissions of the future region.
步骤(1)中所述影响区域民航主被动碳排放的综合指标体系包括地理位置、社会经济、民航运输和主被动排放标签，所述地理位置指标包括区域中心经纬度坐标和区域面积，所述社会经济指标包括人口、GDP总量和人均GDP，所述民航运输指标包括区域内民用机场数量、起降架次、旅客吞吐量和货邮吞吐量，所述主被动碳排放标签包括主动排放标签和被动排放标签，所述主动排放标签是指在区域内起降的航班起降及巡航阶段在该区域内产生的碳排放区分标签，所述被动排放标签是指飞越该区域航班的CCD阶段的碳排放区分标签。In step (1), the comprehensive index system for the active and passive carbon emissions of civil aviation in the affected area includes geographic location, social economy, civil aviation transportation, and active and passive emission labels. The geographic location index includes the latitude and longitude coordinates of the regional center and the area of the area. The economic indicators include population, total GDP and per capita GDP, the civil aviation transportation indicators include the number of civil airports in the region, the number of take-offs and landings, passenger throughput and cargo and mail throughput, and the active and passive carbon emission labels include active emission labels and passive carbon emission labels. Emission label, the active emission label refers to the carbon emission distinction label generated in the area during the take-off, landing and cruise phase of the flight taking off and landing in the area, and the passive emission label refers to the carbon emission during the CCD phase of the flight over the area Distinguishing labels.
步骤(2)的具体过程如下：The specific process of step (2) is as follows:
收集区域各指标对应数据建立数据集，采用标准归一化的方式对数据集进行预处理：The data set corresponding to each indicator in the collection area is established, and the data set is preprocessed by standard normalization:
其中，x*为处理后的样本数据，x为原始样本数据，μ为样本数据的均值，δ为样本数据的标准差；Among them, x* is the processed sample data, x is the original sample data, μ is the mean value of the sample data, and δ is the standard deviation of the sample data;
将数据集划分为训练集和测试集，划分的方法为：The data set is divided into training set and test set, and the division method is as follows:
train_test_spilt(*arrays,test_size,random_state)train_test_spilt(*arrays,test_size,random_state)
其中，train_test_spilt()为数据集划分函数，*arrays为数据集，test_size为测试集样本数占数据集总数的比例，且test_size∈[0,1]，random_state为随机种子。Among them, train_test_spilt() is the data set division function, *arrays is the data set, test_size is the ratio of the number of test set samples to the total number of data sets, and test_size∈[0,1], random_state is the random seed.
步骤(3)中所述基于Keras框架构建包含输入层、全连接层和输出层的神经网络模型具体过程如下：The specific process of constructing a neural network model including an input layer, a fully connected layer and an output layer based on the Keras framework described in step (3) is as follows:
基于Keras框架使用model＝Sequential()构建序贯模型，在模型中搭建神经网络的方式为：Based on the Keras framework, use model=Sequential() to build a sequential model. The way to build a neural network in the model is:
model.add(Dense(units,activation))model.add(Dense(units,activation))
其中，model.add()为创建神经网络层函数，Dense()为创建全连接层函数，units为结点数，activation为激活函数；Among them, model.add() is the function to create the neural network layer, Dense() is the function to create the fully connected layer, units is the number of nodes, and activation is the activation function;
设置输入层及全连接层的激活函数为“relu”，其形式为：Set the activation function of the input layer and the fully connected layer to "relu", and its form is:
f(x)＝max(0,x)f(x)=max(0,x)
其中，max()为最大值函数，x为神经元输入值；Among them, max() is the maximum value function, and x is the input value of the neuron;
因此，构建包含输入层、全连接层和输出层的神经网络方式为：Therefore, the way to construct a neural network including an input layer, a fully connected layer and an output layer is:
其中，input_shape为输入层输入张量的大小，None为不设置激活函数。Among them, input_shape is the size of the input tensor of the input layer, and None means that the activation function is not set.
步骤(4)的具体过程如下：The specific process of step (4) is as follows:
神经网络构建完成后，需要训练模型并计算损失值，配置学习过程的方式为：After the neural network is constructed, it is necessary to train the model and calculate the loss value. The way to configure the learning process is as follows:
model.compile(loss,optimizer,metrics)model.compile(loss, optimizer, metrics)
其中，model.compile()为配置函数，loss为损失函数，optimizer为优化器，metrics表示评估模型在训练和测试时的性能指标；Among them, model.compile() is the configuration function, loss is the loss function, optimizer is the optimizer, and metrics represent the performance indicators of the evaluation model during training and testing;
设置损失函数为“mean_squared_error”，设置优化器为“Adam”优化算法，设置metrics为均方误差“metrics.mae”；Set the loss function to "mean_squared_error", set the optimizer to the "Adam" optimization algorithm, and set the metrics to the mean squared error "metrics.mae";
参数配置完成后，使用model对象进行训练，训练模型的方式为：After the parameter configuration is completed, use the model object for training. The way to train the model is:
model.fit(x_train,y_train,validation_data,epochs,verbose,batch_size)model.fit(x_train,y_train,validation_data,epochs,verbose,batch_size)
其中，model.fit()为训练函数，x_train和y_train为训练集的输入和输出，validation_data为验证集，epochs表示训练总轮数，verbose为训练过程的展示选项；batch_size表示一次训练中所取的样本数；Among them, model.fit() is the training function, x_train and y_train are the input and output of the training set, validation_data is the validation set, epochs represents the total number of training rounds, verbose is the display option of the training process; batch_size represents the data taken in one training Number of samples;
训练结束后，通过输出的拟合评价结果来判断拟合效果，其评价结果输出方式为：After the training, the fitting effect is judged by the output fitting evaluation result. The output method of the evaluation result is:
model.evaluate(x_test,y_test,verbose)model.evaluate(x_test,y_test,verbose)
其中，model.evaluate()为评价输出函数，x_test和y_test为测试集输入和输出数据，verbose为展示选项；Among them, model.evaluate() is the evaluation output function, x_test and y_test are the input and output data of the test set, and verbose is the display option;
通过输出的评价指标判断是否拟合以及拟合效果是否满足要求；当拟合结果未满足要求时，调整模型的参数设置，进行神经网络模型的优化，直至拟合结果满足要求；Judge whether the fitting and the fitting effect meet the requirements through the output evaluation index; when the fitting results do not meet the requirements, adjust the parameter settings of the model and optimize the neural network model until the fitting results meet the requirements;
神经网络完成拟合后，即获取神经网络的结构以及任一层上各结点的权重值并输出保存，保存网络结构及权重的方式为：After the neural network is fitted, the structure of the neural network and the weight value of each node on any layer are obtained and saved. The way to save the network structure and weight is:
model.to_json(path)model.to_json(path)
model.save_weights(path)model.save_weights(path)
其中model.to_json()为模型结构保存函数，model.save_weights()为权重保存函数，path表示保存路径。Where model.to_json() is the model structure saving function, model.save_weights() is the weight saving function, and path represents the saving path.
步骤(5)所述预测方式为：The prediction method described in step (5) is:
model.predict(xfuture)model.predict(xfuture )
其中model.predict()为预测函数，xfuture为未来的某区域输入数据，通过设置xfuture中主被动排放标签即预测该区域未来主动碳排放或被动碳排放的总量。Among them, model.predict() is the prediction function, and xfuture is the input data for a certain area in the future. By setting the active and passive emission label in xfuture , the total amount of active carbon emission or passive carbon emission in the area in the future can be predicted.
本发明的有益效果如下：The beneficial effects of the present invention are as follows:
(1)碳排放预测指标体系中纳入与空间信息相关的指标，考虑空间因素可使得最终的预测结果更加具有合理性与可解释性。(1) The indicators related to spatial information are included in the carbon emission prediction index system, and the consideration of spatial factors can make the final prediction results more reasonable and interpretable.
(2)为能够精确、有效地制定和实施排放决策，将区域排放预测结果划分为主动排放和被动排放，提升碳排放预测结果的丰富性、准确性，有助于制定与实施更加精准的减排策略。(2) In order to accurately and effectively formulate and implement emission decisions, the regional emission prediction results are divided into active emission and passive emission, so as to improve the richness and accuracy of carbon emission prediction results, and help to formulate and implement more accurate emission reduction measures. Arrangement strategy.
(3)提出基于Keras框架的神经网络预测模型，充分利用Keras框架的高度模块化，便于搭建和调试，并且代码结构简明清晰执行效率高，能够快速实现拟合并得到预测结果。(3) A neural network prediction model based on the Keras framework is proposed, making full use of the high modularity of the Keras framework, which is easy to build and debug, and the code structure is concise and clear, and the execution efficiency is high, which can quickly achieve fitting and obtain prediction results.
附图说明Description of drawings
图1是本发明实施方式的方法流程图。FIG. 1 is a flow chart of a method according to an embodiment of the present invention.
图2是本发明实施过程中训练总轮数对应的损失值图。FIG. 2 is a graph of loss values corresponding to the total number of training rounds during the implementation of the present invention.
图3是本发明实施的预测结果验证图。FIG. 3 is a prediction result verification diagram of the implementation of the present invention.
具体实施方式Detailed ways
下面结合附图对本发明创造做进一步详细说明。The present invention will be further described in detail below with reference to the accompanying drawings.
本发明实施方式所述方法的流程如图1所示，包括下列步骤：The process flow of the method according to the embodiment of the present invention is shown in Figure 1, and includes the following steps:
步骤(1)，确定涵盖区域地理位置、社会经济、民航运输方面影响区域民航主被动碳排放的综合指标体系。具体包括以下步骤：Step (1): Determine a comprehensive index system covering regional geographic location, social economy, and civil aviation transportation that affects the active and passive carbon emissions of regional civil aviation. Specifically include the following steps:
步骤(A)，确定地理位置指标。地理位置指标为表明该区域的明显特征，其中包括区域中心坐标经纬度(度)和区域面积(平方公里)；In step (A), the geographic location index is determined. Geographical location indicators are the obvious features of the area, including the latitude and longitude of the center of the area (degrees) and the area of the area (square kilometers);
步骤(B)，确定社会经济指标。社会经济指标主要反映该区域的发展水平，与该区域人口、产业、文化等相关。社会经济指标考虑选取各区域不同年份人口(人)、GDP总量(亿元)与人均GDP指标(元)；In step (B), socioeconomic indicators are determined. Socioeconomic indicators mainly reflect the development level of the region, and are related to the region's population, industry, and culture. Socio-economic indicators consider the selection of population (person), total GDP (100 million yuan) and per capita GDP (yuan) in different regions in different years;
步骤(C)，确定民航运输指标。民航运输指标在一定程度上表明该区域民航发展情况，同时间接表明该区域民航排放水平。民航运输指标包括民用机场数量(个)、起降架次(架次)、旅客吞吐量(人次)和货邮吞吐量(吨)；Step (C), determine the civil aviation transportation index. The civil aviation transportation indicators indicate the development of civil aviation in the region to a certain extent, and at the same time indirectly indicate the emission level of civil aviation in the region. Civil aviation transportation indicators include the number of civil airports (number), the number of take-offs and landings (number of flights), passenger throughput (person-time) and cargo and mail throughput (tons);
步骤(D)，确定主被动碳排放标签。通常在区域内起降的航班LTO及CCD阶段在该区域内产生的碳排放称为该区域的主动排放，而飞越该区域的航班CCD阶段的碳排放称为被动排放。为能够自主选择预测主动排放或被动排放，选择添加主动排放标签(1)与被动排放标签(0)用以区分数据。In step (D), the active and passive carbon emission labels are determined. Usually, the carbon emissions generated in the LTO and CCD stages of flights that take off and land in the region are called active emissions in the region, and the carbon emissions in the CCD stage of flights that fly over the region are called passive emissions. In order to be able to choose to predict active emission or passive emission, choose to add active emission label (1) and passive emission label (0) to distinguish data.
步骤(2)，收集并预处理区域内各项指标具体数值建立数据集，并将数据集划分为训练集和测试集。具体包含以下步骤：Step (2), collecting and preprocessing specific values of various indicators in the area to establish a data set, and dividing the data set into a training set and a test set. Specifically includes the following steps:
步骤(a)，收集区域各指标的具体数值，其中碳排放结果划分为主动排放与被动排放，由此建立数据集，并存储为逗号分隔符(.csv)文件位于file_path路径下。In step (a), the specific values of each indicator in the area are collected, wherein the carbon emission results are divided into active emission and passive emission, thereby establishing a data set and storing it as a comma-separated (.csv) file under the path of file_path.
将数据集导入python中进行处理，读取方式为：Import the dataset into python for processing, and the reading method is:
pd.read_csv(file_path)pd.read_csv(file_path)
其中：pd.read_csv()表示逗号分隔符文件读取方法；Among them: pd.read_csv() indicates the comma-separated file reading method;
步骤(b)，为便于快速收敛神经网络，选择采用标准归一化的方式统一数据的量纲，对数据集进行预处理：Step (b), in order to facilitate the rapid convergence of the neural network, the standard normalization method is selected to unify the dimension of the data, and the data set is preprocessed:
其中，x*为处理后的样本数据，x为原始样本数据，μ为样本数据的均值，δ为样本数据的标准差。Among them, x* is the processed sample data, x is the original sample data, μ is the mean of the sample data, and δ is the standard deviation of the sample data.
预处理完成后，将数据集划分为训练集和测试集，划分的方法为：After the preprocessing is completed, the dataset is divided into training set and test set. The division method is as follows:
train_test_spilt(*arrays,test_size,random_state)train_test_spilt(*arrays,test_size,random_state)
其中，train_test_spilt()为数据集划分函数，*arrays为样本数据，test_size为测试集样本数占数据集总数的比例，且test_size∈[0,1]，random_state为随机种子，随机种子设置为常数可以保证多次运行结果保持一致。Among them, train_test_spilt() is the data set division function, *arrays is the sample data, test_size is the ratio of the number of test set samples to the total number of data sets, and test_size∈[0,1], random_state is the random seed, and the random seed can be set as a constant. Ensure that the results of multiple runs are consistent.
步骤(3)，基于Keras框架构建包含输入层、全连接层和输出层的神经网络。Step (3), construct a neural network including an input layer, a fully connected layer and an output layer based on the Keras framework.
具体包括以下步骤：Specifically include the following steps:
全连接层的每一个结点都与上一层的所有结点相连接，用于综合上步所提取到的特征。基于Keras框架使用model＝Sequential()构建序贯模型，在模型中搭建神经网络的方式为：Each node of the fully connected layer is connected to all the nodes of the previous layer to synthesize the features extracted in the previous step. Based on the Keras framework, use model=Sequential() to build a sequential model. The way to build a neural network in the model is:
model.add(Dense(units,activation))model.add(Dense(units,activation))
其中，model.add()为创建神经网络层函数，Dense()为创建全连接层函数，units为结点数，activation为激活函数。Among them, model.add() is the function to create the neural network layer, Dense() is the function to create the fully connected layer, units is the number of nodes, and activation is the activation function.
并将输入层及全连接层的激活函数设置为“relu”，其形式为：And set the activation function of the input layer and the fully connected layer to "relu", in the form of:
f(x)＝max(0,x)f(x)=max(0,x)
其中，max()为最大值函数，x为神经元输入值；Among them, max() is the maximum value function, and x is the input value of the neuron;
最终构建神经网络的方式如下所示：The final way to build the neural network is as follows:
其中，input_shape为输入层输入张量的大小，None为不设置激活函数。Among them, input_shape is the size of the input tensor of the input layer, and None means that the activation function is not set.
创建完成后使用model.add(Dropout)，在前向传播的过程中，通过设置Dropout数值来改变训练过程中各全连接层保留每个神经元的概率，让某个神经元的激活值以一定的概率停止工作，可以使模型泛化性更强，因为它不会过于依赖某些局部的特征，由此减轻或避免过拟合的现象；After the creation is completed, use model.add(Dropout), in the process of forward propagation, by setting the Dropout value to change the probability that each fully connected layer retains each neuron during the training process, so that the activation value of a neuron is set to a certain value. The probability of stopping work can make the model more generalizable, because it will not rely too much on some local features, thereby reducing or avoiding the phenomenon of overfitting;
步骤(4)，运用Python调用Keras框架的API构建损失函数和优化函数，将神经网络模型在训练集上训练，并用测试集完成模型参数的优化，最终输出并保存最终的神经网络的结构及权重。具体步骤如下：Step (4), use Python to call the API of the Keras framework to construct the loss function and optimization function, train the neural network model on the training set, and use the test set to optimize the model parameters, and finally output and save the final neural network structure and weight . Specific steps are as follows:
步骤(I)，对搭建完成的神经网络进行训练并计算损失值，配置学习过程的方式为：In step (1), the neural network that is built is trained and the loss value is calculated, and the mode of configuring the learning process is:
model.compile(loss＝'mean_squared_error',optimizer＝'adam',metrics＝[metrics.mae])model.compile(loss='mean_squared_error',optimizer='adam',metrics=[metrics.mae])
其中，model.compile()为配置函数，loss为损失函数，mean_squared_error为均方根误差，adam为自适应矩估计算法，optimizer为优化器，metrics表示评估模型在训练和测试时的评价指标，mae为平均绝对误差。Among them, model.compile() is the configuration function, loss is the loss function, mean_squared_error is the root mean square error, adam is the adaptive moment estimation algorithm, optimizer is the optimizer, metrics indicates the evaluation index of the evaluation model during training and testing, mae is the mean absolute error.
选择将均方误差“mean_squared_error”设置为损失函数，其形式为：Choose to set the mean squared error "mean_squared_error" as the loss function of the form:
其中，MSE为均方根误差，
选择将优化器设置为自适应矩估计(Adam)优化算法。Adam算法可以计算每个参数的自适应学习率，参数更新公式如下：Choose to set the optimizer to the Adaptive Moment Estimation (Adam) optimization algorithm. The Adam algorithm can calculate the adaptive learning rate of each parameter, and the parameter update formula is as follows:
其中，mt和vt分别是梯度中第一时刻和第二时刻的估计值，
评价指标metrics选择设置为均方误差“metrics.mae”，其结果不参与训练过程之中。The evaluation index metrics selection is set to mean square error "metrics.mae", and the results are not involved in the training process.
步骤(II)，损失函数定义完成后，使用model对象进行训练，训练模型的参数设置方式为：Step (II), after the definition of the loss function is completed, use the model object for training, and the parameter setting method of the training model is as follows:
model.fit(x_train,y_train,val_data＝(x_val,y_val),epochs,verbose＝1,batch_size)model.fit(x_train,y_train,val_data=(x_val,y_val),epochs,verbose=1,batch_size)
其中，model.fit()为训练函数，x_train和y_train为训练集区域的综合指标值与对应区域的主被动排放，val_data为验证集，其中包含数据集划分出的x_val与y_val，epochs表示训练的总轮数，verbose＝1表示训练过程中输出日志信息，batch_size表示一次训练中所取的样本数大小。Among them, model.fit() is the training function, x_train and y_train are the comprehensive index values of the training set area and the active and passive emissions of the corresponding area, val_data is the validation set, which includes x_val and y_val divided by the data set, and epochs represents the training The total number of rounds, verbose=1 indicates that log information is output during the training process, and batch_size indicates the size of the number of samples taken in one training.
训练结束后，通过输出网络拟合评价结果判断拟合效果，其输出方式为：After the training, the fitting effect is judged by outputting the network fitting evaluation results. The output method is:
model.evaluate(x_val,y_val,verbose＝0)model.evaluate(x_val,y_val,verbose=0)
其中，model.evaluate()为评价输出函数，x_val和y_val为测试集输入和输出数据，verbose＝0表示不在标准输出流输出日志信息；Among them, model.evaluate() is the evaluation output function, x_val and y_val are the input and output data of the test set, and verbose=0 means that the log information is not output in the standard output stream;
通过输出的评价指标判断是否拟合以及拟合效果的优劣，调整神经网络构建与模型训练配置中的参数设置，以此完成神经网络模型的优化。Through the output evaluation index to judge whether the fitting and the quality of the fitting effect, adjust the parameter settings in the neural network construction and model training configuration, so as to complete the optimization of the neural network model.
步骤(III)，获取训练完成的神经网络的结构以及任一层上各结点的权重值并进行输出保存。保存网络结构及权重的方式为：In step (III), the structure of the trained neural network and the weight value of each node on any layer are obtained, and the output is saved. The way to save the network structure and weights is:
model.to_json(path)model.to_json(path)
model.save_weights(path)model.save_weights(path)
其中，model.to_json()为模型结构保存函数，可将模型结构保存为json(JavaScript Object Notation，JS对象标记)文件，model.save_weights()为权重保存函数，可将权重矩阵保存为h5文件，path表示保存路径。Among them, model.to_json() is the model structure saving function, which can save the model structure as a json (JavaScript Object Notation, JS object mark) file, and model.save_weights() is the weight saving function, which can save the weight matrix as an h5 file. path represents the save path.
对应加载模型结构和权重的方式为：The corresponding way to load the model structure and weights is:
model＝model_from_json(open(path,'r').read())model=model_from_json(open(path,'r').read())
model.load_weights(path)model.load_weights(path)
其中，model为模型对象，open()为寻找路径下的模型文件函数，read()为模型文件读取函数，'r'为只读方式，model_from_json()为加载模型结构函数，model.load_weights()为加载权重函数，path表示保存路径。Among them, model is the model object, open() is the function to find the model file under the path, read() is the function to read the model file, 'r' is the read-only mode, model_from_json() is the function to load the model structure, model.load_weights( ) is the load weight function, and path represents the save path.
步骤(5)，依据未来区域各指标数据，使用训练完成的神经模型进行未来的区域主被动碳排放预测。其预测方式为：Step (5), using the trained neural model to predict future regional active and passive carbon emissions according to the index data of the future region. Its prediction method is:
model.predict(xfuture)model.predict(xfuture )
其中，model.predict()为预测函数，xfuture为经过标准归一化处理后未来区域的输入数据。Among them, model.predict() is the prediction function, and xfuture is the input data of the future area after standard normalization.
具体以全国各省份民航主被动碳排放预测为例，选取全国31个省(市/自治区)(不包含港澳台)作为区域，设定时间跨度为2007-2016年共10年时间，收集10年间各省份各指标的具体数值，其中碳排放结果划分为主动排放与被动排放，由此建立维度为12共计620条样本的数据集，并存储为逗号分隔符文件(.csv)位于file_path路径下。Specifically, taking the active and passive carbon emission forecast of civil aviation in various provinces in the country as an example, select 31 provinces (municipalities/autonomous regions) (excluding Hong Kong, Macao and Taiwan) as regions, set a time span of 10 years from 2007 to 2016, and collect data for 10 years. The specific values of each indicator in each province, in which the carbon emission results are divided into active emission and passive emission, thus establishing a dataset with a dimension of 12 and a total of 620 samples, and storing it as a comma-separated file (.csv) under the path of file_path.
将数据集文件导入python中进行处理，读取方式为：Import the dataset file into python for processing. The reading method is:
pd.read_csv(file_path)pd.read_csv(file_path)
将预处理完成后的数据集划分为训练集和测试集，划分的方法为：The data set after preprocessing is divided into training set and test set, and the division method is as follows:
train_test_spilt(*arrays,test_size＝0.3,random_state＝2020)train_test_spilt(*arrays,test_size=0.3,random_state=2020)
其中，*arrays为数据集，test_size为测试集样本数占数据集总数的比例，且test_size∈[0,1]，这里test_size＝0.3表示数据集样本划分为30％测试集(186个样本)与70％训练集(434个样本)，random_state为随机种子，将随机种子设置为常数(这里为2020)可以保证多次运行结果保持一致。Among them, *arrays is the data set, test_size is the ratio of the number of test set samples to the total number of data sets, and test_size∈[0,1], where test_size=0.3 means that the data set samples are divided into 30% test set (186 samples) and 70% of the training set (434 samples), random_state is the random seed, and setting the random seed to a constant (here, 2020) can ensure consistent results across multiple runs.
构建包含输入层、三层全连接层和输出层的神经网络模型并设置各层的神经元个数分比为50、35、20、15、1，最终构建神经网络的方式如下所示：Build a neural network model including an input layer, a three-layer fully connected layer and an output layer, and set the number of neurons in each layer to 50, 35, 20, 15, and 1. The final way to build a neural network is as follows:
创建完成后使用model.add(Dropout)，在前向传播的过程中，通过设置Dropout数值来改变训练过程中各全连接层保留每个神经元的概率，让某个神经元的激活值以一定的概率停止工作，可以使模型泛化性更强，因为它不会过于依赖某些局部的特征，由此减轻或避免过拟合的现象。将输入层以及三层全连接层概率分别设置为0.5、0.5、0.25、0.1，通过函数model.summary()输出模型各层的参数状况，具体如表1所示；After the creation is completed, use model.add(Dropout), in the process of forward propagation, by setting the Dropout value to change the probability that each fully connected layer retains each neuron during the training process, so that the activation value of a neuron is set to a certain value. The probability of stopping work can make the model more generalizable because it does not rely too much on some local features, thereby reducing or avoiding the phenomenon of overfitting. Set the probability of the input layer and the three-layer fully connected layer to 0.5, 0.5, 0.25, and 0.1 respectively, and output the parameter status of each layer of the model through the function model.summary(), as shown in Table 1;
表1Table 1
表中Layer为神经网络层名，包含网络层对应Dropout参数名称，Output Shape为网络层输出大小，以张量形式呈现，Param#为网络层所包含参数的总个数。In the table, Layer is the name of the neural network layer, including the name of the Dropout parameter corresponding to the network layer, Output Shape is the output size of the network layer, which is presented in the form of a tensor, and Param# is the total number of parameters contained in the network layer.
使用model对象进行训练，训练模型的具体参数设置方式为：Use the model object for training, and the specific parameters of the training model are set as follows:
model.fit(x_train,y_train,val_data＝(x_val,y_val)，model.fit(x_train,y_train,val_data=(x_val,y_val),
epochs＝800,verbose＝1,batch_size＝64)epochs=800, verbose=1, batch_size=64)
其中，model.fit()为训练函数，x_train和y_train为训练集区域的综合指标值与对应区域的主被动排放，val_data为验证集，其中包含数据集划分出的验证集x_val与y_val，epochs＝800表示训练的总轮数为800轮，verbose＝1表示训练过程中输出日志信息，batch_size＝64表示一次训练中所取的样本数为64。训练完成后，通过可视化的方式输出损失值与训练总轮数关系图像以及预测值与实际值分布图像，可以清晰观察到神经网络的拟合情况，如图2、图3所示；训练结束后，通过输出的拟合评价结果判断拟合效果，损失值越小，说明预测模型描述实验数据具有更好的精确度，经过多次参数调整得到测试集损失大小为14.9848。Among them, model.fit() is the training function, x_train and y_train are the comprehensive index values of the training set area and the active and passive emissions of the corresponding area, val_data is the validation set, which includes the validation set x_val and y_val divided by the data set, epochs= 800 indicates that the total number of rounds of training is 800 rounds, verbose=1 indicates that log information is output during the training process, and batch_size=64 indicates that the number of samples taken in one training is 64. After the training is completed, the image of the relationship between the loss value and the total number of training rounds and the distribution image of the predicted value and the actual value can be outputted visually, and the fitting of the neural network can be clearly observed, as shown in Figure 2 and Figure 3; , the fitting effect is judged by the output fitting evaluation results. The smaller the loss value, the better the accuracy of the prediction model in describing the experimental data. After multiple parameter adjustments, the loss size of the test set is 14.9848.
通过训练完成的神经网络，即可进行未来某省份主被动排放预测。以2035年广东省为例，收集各个输入指标对应的预测值，部分指标对应的数据如表2所示；Through the trained neural network, the active and passive emission prediction of a province in the future can be carried out. Taking Guangdong Province in 2035 as an example, the forecast values corresponding to each input indicator are collected, and the data corresponding to some indicators are shown in Table 2;
表2Table 2
将表2数据存储至xfuture数组中，预测函数使用的具体方式为：Store the data in Table 2 into the xfuture array, and the specific method used by the prediction function is:
result＝pd.DataFrame({'pollution':model.predict(xfuture).reshape(1,-1)[0])result=pd.DataFrame({'pollution':model.predict(xfuture ).reshape(1,-1)[0])
其中：pd.DataFrame表示表格创建方式，pollution为表头名称，reshape为数据重塑方式；Among them: pd.DataFrame represents the table creation method, pollution is the table header name, and reshape is the data reshaping method;
由此得到result中存储的预测结果，其结果具体为：2035年广东省主动排放为133.86×105吨，被动排放为77.01×105吨。整个拟合过程耗时14.58s，预测均方误差为0.0020。From this, the prediction results stored in the result are obtained, and the specific results are as follows: in 2035, the active emission of Guangdong Province will be 133.86×105 tons, and the passive emission will be 77.01×105 tons. The whole fitting process takes 14.58s, and the mean square error of prediction is 0.0020.
综上，本发明的方法设计一种基于Keras区域民航主被动碳排放的预测方法。综合考虑多方面与民航碳排放相关的因素，将民航碳排放划分为主动排放与被动排放，确定涵盖区域地理位置、社会经济、民航运输等影响区域民航主被动碳排放的综合指标体系；基于Keras框架完整的模块以及简洁的API构建神经网络，可以为研究人员解决参数设置等问题，通过数据集训练拟合能够得到较高精度的区域主被动碳排放预测结果，提升了碳排放预测结果的丰富性与准确性，为更精准地开展民航碳减排工作提供理论支撑。To sum up, the method of the present invention designs a prediction method based on the active and passive carbon emissions of civil aviation in Keras region. Comprehensively considering various factors related to civil aviation carbon emissions, divide civil aviation carbon emissions into active emissions and passive emissions, and determine a comprehensive index system covering regional geographic location, social economy, civil aviation transportation, etc. that affect regional civil aviation active and passive carbon emissions; Based on Keras The complete module of the framework and the concise API to build a neural network can solve problems such as parameter setting for researchers. Through data set training and fitting, high-precision regional active and passive carbon emission prediction results can be obtained, which improves the richness of carbon emission prediction results. It provides theoretical support for more accurate carbon emission reduction work in civil aviation.
Claims (6)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010679105.XACN111882033B (en) | 2020-07-15 | 2020-07-15 | Keras-based regional civil aviation main passive carbon emission prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010679105.XACN111882033B (en) | 2020-07-15 | 2020-07-15 | Keras-based regional civil aviation main passive carbon emission prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111882033Atrue CN111882033A (en) | 2020-11-03 |
| CN111882033B CN111882033B (en) | 2024-04-05 |
Family
ID=73151769
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010679105.XAActiveCN111882033B (en) | 2020-07-15 | 2020-07-15 | Keras-based regional civil aviation main passive carbon emission prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111882033B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112446170A (en)* | 2020-11-06 | 2021-03-05 | 山东大学 | Method for establishing regional ocean bottom temperature thermal infrared remote sensing monitoring and forecasting model |
| CN112668802A (en)* | 2021-01-05 | 2021-04-16 | 广东工业大学 | Construction carbon emission prediction method based on design parameters |
| CN113592194A (en)* | 2021-08-23 | 2021-11-02 | 北京科技大学 | Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect |
| CN113610268A (en)* | 2021-06-28 | 2021-11-05 | 天津大学 | Carbon emission prediction method based on residential area space form |
| CN113642864A (en)* | 2021-07-30 | 2021-11-12 | 南京航空航天大学 | Analysis method for space-time balance of carbon emission of airport passenger aircraft |
| CN113902166A (en)* | 2021-09-07 | 2022-01-07 | 南方电网大数据服务有限公司 | Big Data Fusion Carbon Emission Governance, Predictive Analysis Model Method |
| CN113919574A (en)* | 2021-10-14 | 2022-01-11 | 天津大学 | Prediction method of heavy metal emissions from waste incineration plant flue gas based on Keras |
| CN114998075A (en)* | 2022-06-10 | 2022-09-02 | 长安大学 | Transportation carbon emission calculation method and system and storable medium |
| CN117575175A (en)* | 2024-01-15 | 2024-02-20 | 国网浙江省电力有限公司 | Carbon emission evaluation method, device, electronic equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975799A (en)* | 2016-06-01 | 2016-09-28 | 广东电网有限责任公司电力科学研究院 | Method and system for calculating carbon emissions |
| CN106529732A (en)* | 2016-11-18 | 2017-03-22 | 国网山东省电力公司电力科学研究院 | Carbon emission efficiency prediction method based on neural network and random frontier analysis |
| CN107909524A (en)* | 2016-11-28 | 2018-04-13 | 中国铁路总公司 | A kind of environmental benefit computational methods of high-speed railway operation phase carbon emission reduction |
| CN108846526A (en)* | 2018-08-08 | 2018-11-20 | 华北电力大学 | A kind of CO2 emissions prediction technique |
| CN109190211A (en)* | 2018-08-17 | 2019-01-11 | 中国科学院地理科学与资源研究所 | A method of appraising and deciding area alignment space |
| CN110378383A (en)* | 2019-06-19 | 2019-10-25 | 江苏大学 | A kind of picture classification method based on Keras frame and deep neural network |
- 2020
- 2020-07-15CNCN202010679105.XApatent/CN111882033B/enactiveActive
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975799A (en)* | 2016-06-01 | 2016-09-28 | 广东电网有限责任公司电力科学研究院 | Method and system for calculating carbon emissions |
| CN106529732A (en)* | 2016-11-18 | 2017-03-22 | 国网山东省电力公司电力科学研究院 | Carbon emission efficiency prediction method based on neural network and random frontier analysis |
| CN107909524A (en)* | 2016-11-28 | 2018-04-13 | 中国铁路总公司 | A kind of environmental benefit computational methods of high-speed railway operation phase carbon emission reduction |
| CN108846526A (en)* | 2018-08-08 | 2018-11-20 | 华北电力大学 | A kind of CO2 emissions prediction technique |
| CN109190211A (en)* | 2018-08-17 | 2019-01-11 | 中国科学院地理科学与资源研究所 | A method of appraising and deciding area alignment space |
| CN110378383A (en)* | 2019-06-19 | 2019-10-25 | 江苏大学 | A kind of picture classification method based on Keras frame and deep neural network |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112446170A (en)* | 2020-11-06 | 2021-03-05 | 山东大学 | Method for establishing regional ocean bottom temperature thermal infrared remote sensing monitoring and forecasting model |
| CN112668802A (en)* | 2021-01-05 | 2021-04-16 | 广东工业大学 | Construction carbon emission prediction method based on design parameters |
| CN113610268A (en)* | 2021-06-28 | 2021-11-05 | 天津大学 | Carbon emission prediction method based on residential area space form |
| CN113642864A (en)* | 2021-07-30 | 2021-11-12 | 南京航空航天大学 | Analysis method for space-time balance of carbon emission of airport passenger aircraft |
| CN113642864B (en)* | 2021-07-30 | 2023-12-05 | 南京航空航天大学 | An analysis method for the spatial and temporal balance of carbon emissions from airport passenger aircraft |
| CN113592194A (en)* | 2021-08-23 | 2021-11-02 | 北京科技大学 | Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect |
| CN113902166A (en)* | 2021-09-07 | 2022-01-07 | 南方电网大数据服务有限公司 | Big Data Fusion Carbon Emission Governance, Predictive Analysis Model Method |
| CN113919574A (en)* | 2021-10-14 | 2022-01-11 | 天津大学 | Prediction method of heavy metal emissions from waste incineration plant flue gas based on Keras |
| CN114998075A (en)* | 2022-06-10 | 2022-09-02 | 长安大学 | Transportation carbon emission calculation method and system and storable medium |
| CN117575175A (en)* | 2024-01-15 | 2024-02-20 | 国网浙江省电力有限公司 | Carbon emission evaluation method, device, electronic equipment and storage medium |
| CN117575175B (en)* | 2024-01-15 | 2024-03-29 | 国网浙江省电力有限公司 | Carbon emission evaluation method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111882033B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111882033B (en) | Keras-based regional civil aviation main passive carbon emission prediction method | |
| CN111401601B (en) | Delay propagation-oriented flight take-off and landing time prediction method | |
| CN107844915A (en) | A kind of automatic scheduling method of the call center based on traffic forecast | |
| CN113657671B (en) | A flight delay prediction method based on ensemble learning | |
| CN113643571A (en) | An Airspace Network Optimization Method Based on Flight Regularity Objective | |
| CN110210648B (en) | A method for predicting strategic flow in controlled airspace based on gray long-short-term memory network | |
| CN114943356B (en) | Short-time demand integrated prediction method for airport arrival passenger to take taxi | |
| CN111192090A (en) | Seat allocation method and device for flight, storage medium and electronic equipment | |
| CN111353625B (en) | Method, device, computer equipment and storage medium for predicting net point quantity | |
| CN112215416B (en) | Intelligent planning inspection route system and method | |
| Liang et al. | An improved adaptive parallel genetic algorithm for the airport gate assignment problem | |
| CN117707742B (en) | Load prediction-based energy consumption balance job scheduling method and system | |
| CN112435077A (en) | Sales forecasting method | |
| CN112926809B (en) | Flight flow prediction method and system based on clustering and improved xgboost | |
| CN114723142A (en) | A multi-objective land use simulation system and method based on non-dominated sorting genetic algorithm and FLUS model | |
| CN115543626A (en) | Image Simulation Method of Power Defect Using Load Balanced Scheduling of Heterogeneous Computing Resources | |
| CN117076887A (en) | A method and system for operating status prediction and health assessment of pumping station units | |
| CN109787821A (en) | A kind of Large-scale Mobile customer traffic consumption intelligent Forecasting | |
| CN114781704B (en) | Flight delay prediction method based on outbound flight guarantee flow | |
| CN111080023A (en) | Aircraft carrier-based aircraft takeoff scheduling planning method based on brain storm optimization | |
| CN111126827A (en) | Input-output accounting model construction method based on BP artificial neural network | |
| CN110909946B (en) | An optimization method of flight planning based on highway transfer | |
| Song et al. | Calibration of agent-based model using reinforcement learning | |
| CN116663834A (en) | Aviation passenger scheduling method, device, computer equipment and storage medium | |
| Zeng et al. | You Only Train Once: A highly generalizable reinforcement learning method for dynamic job shop scheduling problem |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |