WO2023050534A1 - 轨道交通站点设备能耗预测方法、装置、设备和存储介质 - Google Patents

轨道交通站点设备能耗预测方法、装置、设备和存储介质 Download PDF

Info

Publication number
WO2023050534A1
WO2023050534A1 PCT/CN2021/129193 CN2021129193W WO2023050534A1 WO 2023050534 A1 WO2023050534 A1 WO 2023050534A1 CN 2021129193 W CN2021129193 W CN 2021129193W WO 2023050534 A1 WO2023050534 A1 WO 2023050534A1
Authority
WO
WIPO (PCT)
Prior art keywords
energy consumption
equipment
site
target site
data
Prior art date
Application number
PCT/CN2021/129193
Other languages
English (en)
French (fr)
Inventor
李鸿飞
刘文凯
肖中卿
贾建平
Original Assignee
广州新科佳都科技有限公司
广州华佳软件有限公司
佳都科技集团股份有限公司
广东华之源信息工程有限公司
广州佳都城轨智慧运维服务有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 广州新科佳都科技有限公司, 广州华佳软件有限公司, 佳都科技集团股份有限公司, 广东华之源信息工程有限公司, 广州佳都城轨智慧运维服务有限公司 filed Critical 广州新科佳都科技有限公司
Publication of WO2023050534A1 publication Critical patent/WO2023050534A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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

Definitions

  • the embodiments of the present application relate to the field of rail transit, and in particular to a method, device, device, and storage medium for predicting energy consumption of rail transit site equipment.
  • Time series method prediction use node AR model (Autoregressive model, autoregressive model), MA model (moving average model, moving average model), ARMA model (Autoregressive moving average model, autoregressive moving average model) and other time series Predict the energy consumption of equipment in rail transit stations.
  • node AR model Autoregressive model, autoregressive model
  • MA model moving average model, moving average model
  • ARMA model Automatic moving average model, autoregressive moving average model
  • Machine learning and deep learning algorithms for prediction usually use linear regression method, XGBOOST regression method or neural network algorithm (RNN cycle neural network, LSTM long short-term memory network), etc., to train the model by constructing a series of modeling features Fit a trained model to predict the energy consumption of equipment in rail transit stations.
  • linear regression method usually use linear regression method, XGBOOST regression method or neural network algorithm (RNN cycle neural network, LSTM long short-term memory network), etc.
  • Predictive modeling is performed only from characteristic sequence data such as historical energy consumption of sites and equipment, which cannot solve the problem of cold start of forecasting models for new lines and new sites without historical energy consumption data of equipment.
  • Modeling is only based on the historical energy consumption data of equipment at this site, ignoring the connection and influence between sites, such as similar sites and adjacent sites.
  • the embodiment of the present invention provides a method, device, equipment, and storage medium for predicting energy consumption of rail transit station equipment, which solves the problem of low prediction accuracy in the prior art method for predicting the energy consumption of rail transit station equipment technical problem.
  • An embodiment of the present invention provides a method for predicting energy consumption of rail transit site equipment, comprising the following steps:
  • the neural network model is trained according to the historical energy consumption data of the equipment, and the equipment energy consumption prediction model corresponding to each site is obtained;
  • the final equipment energy consumption prediction value of the target site is calculated.
  • the similarity between the target site and other sites is calculated according to the feature vector data, and the specific process of determining similar sites of the target site according to the similarity is:
  • the specific process of calculating the similarity between the target site and other sites is:
  • the neural network model is trained according to the historical energy consumption data of the equipment, and the specific process of obtaining the equipment energy consumption prediction model corresponding to each site is as follows:
  • the LSTM model is trained by using the training label sequence set to obtain the equipment energy consumption prediction model corresponding to each site.
  • the specific process of converting the historical energy consumption sequence into a training label sequence set for each site is as follows:
  • the historical energy consumption sequence is converted into a training label sequence set for each site.
  • the specific process of calculating the final equipment energy consumption prediction value of the target site is as follows:
  • the predicted energy consumption value of the final equipment of the target site is calculated.
  • the specific process of obtaining the historical energy consumption data of equipment at each site is as follows:
  • T is determined according to the granularity of energy consumption prediction and the length of training data.
  • the embodiment of the present invention also provides a rail transit site equipment energy consumption prediction device, including
  • the data acquisition module is used to acquire the eigenvector data of each site and the historical energy consumption data of the equipment;
  • a similar site determination module used to calculate the similarity between the target site and other sites according to the feature vector data, and determine the similar sites of the target site according to the similarity
  • the neural network model training module is used to train the neural network model according to the historical energy consumption data of the equipment to obtain the equipment energy consumption prediction model corresponding to each site;
  • the equipment energy consumption prediction module is configured to combine the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction model of the similar sites to calculate the final equipment energy consumption prediction value of the target site.
  • the similar site determination module includes:
  • the feature vector matrix submodule is used to generate a feature vector matrix according to the feature vector data of each site;
  • a standardization submodule used to standardize the eigenvector matrix to obtain a first eigenvector matrix
  • a similarity calculation submodule configured to calculate the similarity between the target site and the other sites according to the first feature vector matrix
  • a determining submodule configured to determine similar sites to the target site according to the similarity.
  • the similarity calculation submodule is specifically configured to calculate the Euclidean distance between the target site and the other sites according to the first eigenvector matrix, and use the Euclidean distance as the distance between the target site and the other sites similarity.
  • the neural network model training module includes:
  • the energy consumption sequence vector matrix submodule is used to establish the equipment historical energy consumption sequence vector matrix according to the historical energy consumption data of the equipment;
  • the conversion submodule is used to extract the historical energy consumption sequence of each site from the historical energy consumption sequence vector matrix of the equipment, and convert the historical energy consumption sequence into a training tag sequence set for each site;
  • the training sub-module is used to use the training label sequence set to train the LSTM model to obtain the equipment energy consumption prediction model corresponding to each site.
  • the conversion sub-module is used to convert the historical energy consumption sequence into the training label sequence set of each site.
  • the specific process is: the conversion sub-module is used to set the length of the training set feature sequence; according to the length of the training set feature sequence, Convert the historical energy consumption sequence into a training label sequence set for each site.
  • the equipment energy consumption prediction module includes:
  • the first energy consumption prediction value calculation sub-module is used to input the equipment historical energy consumption data of the target site into the equipment energy consumption prediction model of the target site to obtain the first energy consumption prediction value of the target site;
  • the second energy consumption prediction value calculation sub-module is used to input the equipment historical energy consumption data of the similar site into the equipment energy consumption prediction model of the similar site, and obtain the second energy consumption prediction value of the similar site;
  • an average value calculation submodule configured to calculate the average value of the second energy consumption prediction value according to the number of similar sites
  • a weight assignment sub-module configured to assign weights to the average values of the first energy consumption prediction value and the second energy consumption prediction value
  • the predicted value calculation submodule is configured to calculate the final equipment energy consumption predicted value of the target site according to the average value of the weight, the first predicted energy consumption value, and the second predicted energy consumption value.
  • the data acquisition module is specifically configured to acquire the energy consumption data of each site at T historical time nodes, wherein T is determined according to the granularity of energy consumption prediction and the length of training data.
  • the embodiment of the present invention also provides a rail transit site equipment energy consumption prediction device, the device includes: one or more processors; a storage device for storing one or more programs, when the one The one or more programs are executed by the one or more processors, so that the one or more processors realize the method for predicting energy consumption of rail transit site equipment according to the first aspect.
  • a storage medium storing computer-executable instructions, the computer-executable instructions are used to execute the method for predicting energy consumption of rail transit station equipment as described in the first aspect when executed by a computer processor.
  • the embodiment of the present invention obtains the feature vector data and equipment historical energy consumption data of each site; calculates the similarity between the target site and other sites according to the feature vector data, and determines the similar sites of the target site according to the similarity; Train the neural network model with the consumption data to obtain the equipment energy consumption prediction model corresponding to each site; combine the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction model of similar sites to calculate the final equipment energy consumption prediction of the target site value.
  • the embodiment of the present invention takes into account the connection and influence between sites. When calculating the final equipment energy consumption prediction value of the target site, the equipment energy consumption prediction model of similar sites is used.
  • the equipment energy consumption prediction model of the target site By combining the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction model of similar sites to calculate the final equipment energy consumption forecast value of the target site, thereby avoiding the short-term energy consumption fluctuations of the target site, resulting in large error and low accuracy of the final equipment energy consumption forecast value, and improving the The accuracy of prediction; and when the target site is a new site, the equipment energy consumption prediction model based on similar sites can also calculate the final equipment energy consumption prediction value of the target site, even if the new site does not have historical equipment energy consumption data
  • the equipment energy consumption prediction model can also perform cold start, which improves the accuracy of equipment energy consumption prediction for new stations, and solves the existing methods of predicting equipment energy consumption for rail transit stations. Low technical issues.
  • Fig. 1 is a flowchart of a method for predicting energy consumption of rail transit station equipment provided by an embodiment of the present invention.
  • Fig. 2 is a flowchart of another method for predicting energy consumption of rail transit site equipment provided by an embodiment of the present invention.
  • Fig. 3 is a schematic structural diagram of a device for predicting energy consumption of rail transit station equipment provided by an embodiment of the present invention.
  • Fig. 4 is a schematic structural diagram of a rail transit station equipment energy consumption prediction device provided by an embodiment of the present invention.
  • Fig. 1 is a flowchart of a method for predicting energy consumption of rail transit station equipment provided by an embodiment of the present invention.
  • the rail transit station equipment energy consumption prediction method provided by the embodiment of the present invention can be executed by the rail transit station equipment energy consumption prediction device, and the rail transit station equipment energy consumption prediction device can be implemented by software and/or hardware.
  • the rail transit station The device energy consumption prediction device may be composed of two or more physical entities, or may be composed of one physical entity.
  • the equipment energy consumption prediction equipment in rail transit stations can be data processing equipment such as computers, mobile phones, tablets, or intelligent interactive tablets.
  • Step 101 Obtain feature vector data and equipment historical energy consumption data of each site.
  • the eigenvector data is the data that can reflect the equipment characteristics of each site, and the historical energy consumption data of the equipment is the energy consumption data of each equipment in the history of the site.
  • the eigenvector data can be set according to actual needs.
  • the eigenvector data consists of 7 eigenvectors, as shown in Table 1.
  • Step 102 calculate the similarity between the target site and other sites according to the feature vector data, and determine similar sites to the target site according to the similarity.
  • the similarity between the target site and other sites can be calculated according to the feature vector data.
  • the Euclidean distance can be used as the similarity between two sites by calculating the Euclidean distance between the feature vector data of each site. After calculating the Euclidean distance between the target site and other sites, sort the other sites from small to large according to the value of the Euclidean distance, and select the top K other sites from the sequence as similar sites to the target site.
  • Step 103 train the neural network model according to the historical energy consumption data of the equipment, and obtain the equipment energy consumption prediction model corresponding to each site.
  • the neural network model After obtaining the historical energy consumption data of the equipment at each site, input the historical energy consumption data of the equipment into the set neural network model for training, and obtain the trained neural network model corresponding to each site, and each trained The neural network model is used as the equipment energy consumption prediction model for each corresponding site.
  • the neural network model selects the LSTM model, and the LSTM is a time recursive neural network.
  • the open source pytorch framework is used to construct and train the LSTM model.
  • the historical energy consumption data of equipment at each site is input into different LSTM models for training, and finally the equipment energy consumption prediction model corresponding to each site is obtained.
  • Step 104 combining the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction model of similar sites, to calculate the final equipment energy consumption prediction value of the target site.
  • the energy consumption prediction model is used to calculate the final equipment energy consumption prediction value of the target site.
  • the historical equipment energy consumption data of the target site before time t and the historical equipment energy consumption data of similar sites are obtained and input into the corresponding equipment energy consumption prediction model, so that the equipment energy consumption prediction model corresponding to the target site Output the first energy consumption prediction value after time t, the equipment energy consumption prediction model corresponding to each similar site outputs the second energy consumption prediction value after time t, calculate the average value of each second energy consumption prediction value, and then, respectively By assigning weights to the average value of the first predicted energy consumption value and the second predicted energy consumption value and performing weighted summation, the final predicted energy consumption value of the equipment after time t of the target site can be obtained.
  • the embodiment of the present invention calculates the final equipment energy consumption prediction value of the target site by combining the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction model of similar sites, thereby avoiding the
  • the error of energy consumption prediction value is large and the prediction accuracy is low, which improves the accuracy of prediction; and when the target site is a new site, the equipment energy consumption prediction model based on similar sites can also calculate the final equipment energy consumption of the target site.
  • the energy consumption prediction value of the new station enables the cold start of the equipment energy consumption prediction model without the historical energy consumption data of the equipment.
  • the method of predicting the energy consumption of equipment has the technical problem of low prediction accuracy.
  • Fig. 2 is a flow chart of a method for predicting energy consumption of rail transit station equipment provided by another embodiment of the present invention.
  • the method for predicting energy consumption of rail transit station equipment is embodied on the basis of the above-mentioned embodiments.
  • the energy consumption prediction methods of rail transit station equipment include:
  • Step 201 Obtain feature vector data and historical equipment energy consumption data of each site.
  • the specific process of obtaining the historical energy consumption data of equipment of each site is: obtaining the energy consumption data of each site at T historical time nodes, where T is determined according to the granularity of energy consumption prediction and the length of training data.
  • the energy consumption prediction granularity is set to 15 minutes, and the length of the training data is set to 2020-01-01-2020-01-31.
  • Step 202 generate a feature vector matrix according to the feature vector data of each site.
  • the eigenvector data of each site is preprocessed, and then sorted to obtain a eigenvector matrix.
  • the eigenvector matrix X ⁇ R N ⁇ 7 of N rows and 7 columns is obtained after sorting the eigenvector data of each site, Where N represents the number of sites, 7 represents the 7 eigenvectors of the eigenvector data, and the element x ij in the eigenvector matrix represents the jth eigenvector of the i-th site, where N ⁇ i>0, 7 ⁇ j> 0, x 1j is the feature vector data of the target site, x pj is the feature vector data of other sites, where N ⁇ p>1.
  • Step 203 standardize the eigenvector matrix to obtain the first eigenvector matrix.
  • the z-score standard deviation standardization method is used to standardize the eigenvector matrix.
  • Step 204 Calculate the similarity between the target site and other sites according to the first feature vector matrix.
  • the similarity between the target site and other sites can be calculated according to the standardized eigenvectors in the first eigenvector matrix.
  • the specific process of calculating the similarity between the target site and other sites is:
  • the first eigenvector matrix calculate the Euclidean distance between the target site and other sites, and use the Euclidean distance as the similarity between the target site and other sites.
  • Step 205 determine similar sites of the target site according to the similarity.
  • the magnitude of the value, select K similar sites with the smallest value from the distance sequence d, recorded as S k (s 1 , s 2 ,..., s k ), where s q is the number of K similar sites, where K ⁇ q>0.
  • Step 206 according to the historical energy consumption data of the equipment, establish a sequence vector matrix of the historical energy consumption of the equipment.
  • the historical energy consumption sequence vector matrix of the equipment for each site is established according to the historical energy consumption data of the equipment.
  • the energy consumption data of each site at T historical time nodes are obtained, and after sorting, the equipment historical energy consumption sequence vector matrix E ⁇ R N ⁇ T is obtained, where N represents the number of sites, and T Indicates the historical energy consumption data of a certain energy-consuming equipment at the site at time t, where T ⁇ t>0
  • the element e it in the equipment historical energy consumption sequence vector matrix E represents the equipment historical energy consumption data of the i-th site at the t-th time node, where N ⁇ i>0, T ⁇ t>0, e 1t is the target site
  • the historical energy consumption data of equipment at the t-th time node, e pt is the historical energy consumption data of equipment at other sites at the t-th time node, where N ⁇ p>1.
  • Step 207 extracting the historical energy consumption sequence of each site from the equipment historical energy consumption sequence vector matrix, and converting the historical energy consumption sequence into a training label sequence set for each site.
  • the historical energy consumption sequence S e (e i1 ,e i2 ,.... ....,e iT ) ⁇ E, after that, convert the historical energy consumption sequence S e of each site into a training label sequence set.
  • the specific process of converting the historical energy consumption sequence into a training label sequence set for each site is as follows:
  • Step 2071 set the length of the feature sequence of the training set.
  • the length of the characteristic sequence of the training set needs to be set.
  • the length of the characteristic sequence of the training set can be set according to the total length of the historical energy consumption sequence and the actual situation.
  • Step 2072 Convert the historical energy consumption sequence into a training label sequence set for each site according to the length of the characteristic sequence of the training set.
  • x l is the energy consumption sequence feature
  • y l is the energy consumption sequence result label
  • x l and y l constitute the training label sequence set.
  • Step 208 using the training label sequence set to train the LSTM model to obtain the equipment energy consumption prediction model corresponding to each site.
  • the training label sequence set corresponding to each site After obtaining the training label sequence set corresponding to each site, input the training label sequence set of each site into the LSTM model, use the backpropagation algorithm or the gradient descent method to train the LSTM model, and judge the loss function of the LSTM model Whether the error loss of the LSTM model is less than the preset threshold or the number of training times of the LSTM model reaches the preset number of times, if so, stop the training and obtain the trained equipment energy consumption prediction model.
  • Step 209 combining the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction models of similar sites, to calculate the final equipment energy consumption prediction value of the target site.
  • the specific process of calculating the final equipment energy consumption prediction value of the target site from step 2091 to step 209 is as follows:
  • Step 2091 input the historical energy consumption data of the equipment of the target site into the equipment energy consumption prediction model of the target site, and obtain the first energy consumption prediction value of the target site;
  • the historical energy consumption data of equipment at the target site at the 30 time nodes before time t is acquired
  • the historical energy consumption data of the equipment input into the equipment energy consumption prediction model of the target site, so as to obtain the first energy consumption prediction value of the target site at a time node after time t.
  • Step 2092 Input the historical energy consumption data of the equipment at the similar site into the equipment energy consumption prediction model at the similar site to obtain a second energy consumption prediction value at the similar site.
  • Step 2093 Calculate the average value of the second predicted energy consumption according to the number of similar stations.
  • the sum of the second predicted energy consumption values of the K similar sites is divided by K to obtain an average value of the second predicted energy consumption values.
  • Step 2094 assigning weights to the average values of the first energy consumption prediction value and the second energy consumption prediction value respectively.
  • weights are assigned to the average values of the first energy consumption prediction value and the second energy consumption prediction value, wherein the weights are obtained by performing an optimization test. It can be understood that in this embodiment, the first energy consumption The weights of the predicted energy consumption value and the average value of the second predicted energy consumption value are specifically defined.
  • Step 2095 according to the weight, the average value of the first energy consumption prediction value and the second energy consumption prediction value, calculate the final equipment energy consumption prediction value of the target site.
  • e s ( t+1 ) is the predicted value of the final equipment energy consumption of the target site at a time node after time t
  • ⁇ and ⁇ are the weights
  • model s is the equipment energy consumption prediction model of the target site
  • model k is the similar site equipment energy consumption prediction model.
  • the embodiment of the present invention calculates the final equipment energy consumption prediction value of the target site by combining the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction model of similar sites, thereby avoiding the
  • the error of the energy consumption prediction value is large and the accuracy is low, which improves the accuracy of the prediction; and when the target site is a new site, the equipment energy consumption prediction model based on similar sites can also calculate the final equipment energy consumption of the target site
  • the predicted value enables new stations to be cold-started without a prediction model of historical equipment energy consumption data, improves the accuracy of equipment energy consumption prediction for new stations, and solves the problem of equipment energy consumption in rail transit stations in the prior art
  • the method of forecasting has the technical problem of low forecasting accuracy.
  • Figure 3 is a device for predicting energy consumption of rail transit site equipment provided by an embodiment of the present invention, which is characterized in that it includes
  • a data acquisition module 301 configured to acquire feature vector data and equipment historical energy consumption data of each site;
  • Similar site determination module 302 used to calculate the similarity between the target site and other sites according to the feature vector data, and determine the similar sites of the target site according to the similarity
  • the neural network model training module 303 is used to train the neural network model according to the historical energy consumption data of the equipment to obtain the equipment energy consumption prediction model corresponding to each site;
  • the equipment energy consumption prediction module 304 is configured to combine the equipment energy consumption prediction model of the target site and the equipment energy consumption prediction models of similar sites to calculate the final equipment energy consumption prediction value of the target site.
  • the similar site determining module 302 includes:
  • the eigenvector matrix submodule is used to generate the eigenvector matrix according to the eigenvector data of each site;
  • a standardization submodule is used to standardize the eigenvector matrix to obtain the first eigenvector matrix
  • the similarity calculation submodule is used to calculate the similarity between the target site and other sites according to the first feature vector matrix
  • the determination sub-module is used to determine similar sites of the target site according to the similarity.
  • the similarity calculation submodule is specifically configured to calculate the Euclidean distance between the target site and other sites according to the first eigenvector matrix, and use the Euclidean distance as the similarity between the target site and other sites.
  • the neural network model training module 303 includes:
  • the energy consumption sequence vector matrix sub-module is used to establish the equipment historical energy consumption sequence vector matrix according to the equipment historical energy consumption data
  • the conversion sub-module is used to extract the historical energy consumption sequence of each site from the equipment historical energy consumption sequence vector matrix, and convert the historical energy consumption sequence into a training tag sequence set for each site;
  • the training sub-module is used to use the training label sequence set to train the LSTM model to obtain the equipment energy consumption prediction model corresponding to each site.
  • the specific process of converting the historical energy consumption sequence into the training label sequence set of each site by the conversion sub-module is as follows: the conversion sub-module is used to set the length of the training set feature sequence; according to the training set feature Sequence length, which converts the historical energy consumption sequence into a set of training label sequences for each site.
  • the device energy consumption prediction module 304 includes:
  • the first energy consumption prediction value calculation sub-module is used to input the historical energy consumption data of the equipment of the target site into the equipment energy consumption prediction model of the target site to obtain the first energy consumption prediction value of the target site;
  • the second energy consumption prediction value calculation sub-module is used to input the historical energy consumption data of the equipment of the similar site into the equipment energy consumption prediction model of the similar site to obtain the second energy consumption prediction value of the similar site;
  • the average value calculation sub-module is used to calculate the average value of the second energy consumption prediction value according to the number of similar sites
  • a weight assignment sub-module configured to assign weights to the average values of the first energy consumption prediction value and the second energy consumption prediction value
  • the predicted value calculation sub-module is configured to calculate the final predicted energy consumption value of the equipment at the target site according to the weight, the average value of the first predicted energy consumption value and the second predicted energy consumption value.
  • the data acquisition module 301 is specifically configured to acquire energy consumption data of each site at T historical time nodes, where T is determined according to energy consumption prediction granularity and training data length.
  • This embodiment also provides a rail transit station equipment energy consumption prediction device, as shown in Figure 4, a rail transit station equipment energy consumption prediction device, the device includes a processor 400 and a memory 401;
  • the memory 401 is used to store the computer program 402, and transmit the computer program 402 to the processor;
  • the processor 400 is configured to execute the steps in the above embodiment of a method for predicting energy consumption of rail transit site equipment according to instructions in the computer program 402 .
  • the computer program 402 can be divided into one or more modules/units, and one or more modules/units are stored in the memory 401 and executed by the processor 400 to complete the present application.
  • One or more modules/units may be a series of computer program instruction segments capable of accomplishing specific functions, and the instruction segments are used to describe the execution process of the computer program 402 in the rail transit site equipment energy consumption prediction equipment.
  • the energy consumption prediction equipment of rail transit site equipment can be computing equipment such as desktop computers, notebooks, palmtop computers, and cloud servers.
  • the energy consumption prediction device for rail transit site equipment may include, but not limited to, a processor 400 and a memory 401 .
  • Fig. 4 is only an example of rail transit site equipment energy consumption prediction equipment, and does not constitute a limitation to rail transit site equipment energy consumption prediction equipment, and may include more or less components than those shown in the illustration, Or combine certain components, or different components, for example, the energy consumption prediction equipment of rail transit site equipment may also include input and output equipment, network access equipment, bus and so on.
  • the so-called processor 400 may be a central processing unit (Central Processing Unit, CPU), and may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the storage 401 may be an internal storage unit of the equipment energy consumption prediction device for rail transit site equipment, such as a hard disk or memory of the equipment energy consumption prediction equipment for rail transit site equipment.
  • the memory 401 can also be an external storage device of the rail transit station equipment energy consumption prediction device, such as a plug-in hard disk equipped on the rail transit station equipment energy consumption prediction device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (Flash Card), etc.
  • the memory 401 may also include both an internal storage unit of the rail transit site equipment energy consumption prediction device and an external storage device.
  • the memory 401 is used to store computer programs and other programs and data required by the rail transit site equipment energy consumption prediction equipment.
  • the memory 401 can also be used to temporarily store data that has been output or will be output.
  • the disclosed system, device and method can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components can be combined or integrated. to another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
  • a unit described as a separate component may or may not be physically separated, and a component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.
  • the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods in various embodiments of the present invention.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. .
  • the embodiment of the present invention also provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute a method for predicting energy consumption of rail transit site equipment when executed by a computer processor, the method comprising the following steps:
  • the neural network model is trained according to the historical energy consumption data of the equipment, and the equipment energy consumption prediction model corresponding to each site is obtained;
  • the final equipment energy consumption prediction value of the target site is calculated.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Marketing (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Primary Health Care (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Evolutionary Biology (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Educational Administration (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种轨道交通站点设备能耗预测方法、装置、设备和存储介质，包括：获取每个站点的特征向量数据以及设备历史能耗数据；根据特征向量数据计算目标站点与其他站点的相似度，根据相似度确定目标站点的相似站点；根据设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值。本发明考虑了站点与站点之间的联系与影响，通过结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型来计算目标站点的最终设备能耗预测值，解决了目前对轨道交通站点的设备能耗进行预测的方法，存在着预测准确率低的问题。

Description

轨道交通站点设备能耗预测方法、装置、设备和存储介质 技术领域
本申请实施例涉及轨道交通领域，尤其涉及一种轨道交通站点设备能耗预测方法、装置、设备和存储介质。
背景技术
随着我国经济与科学技术的不断发展，许多城市开始大力推进轨道交通的建设，轨道交通给人们的生活带来便利的同时，也促进了城市的经济发展。然而，轨道交通在运行的过程中，会带来巨大的能耗，如何利用轨道交通站点内的能耗指标，进行能耗分析、预测、预警、诊断以及节能管理等，从而促使轨道交通的科学、合理、安全的用能，提高能源管理水平以及经济效益，成为目前亟需考虑的问题。
当前，对轨道交通站点的设备能耗进行预测的技术主要有以下两种方法：
1.时间序列方法预测：利用节点的AR模型(Autoregressive model，自回归模型)、MA模型(moving average model，滑动平均模型)、ARMA模型(Autoregressive moving average model，自回归滑动平均模型)等时间序列对轨道交通站点的设备能耗进行预测。
2.机器学习和深度学习算法进行预测：通常利用线性回归法、XGBOOST回归法或神经网络算法(RNN循环神经网络、LSTM长短期记忆网络)等，通过构造一系列建模的特征对模型进行训练拟合出一个训练好的模型，从而对轨道交通站点的设备能耗进行预测。
然而，上述方法存在以下缺点：
(1)仅仅从站点、设备自身历史能耗等特征序列数据进行预测建模，无法解决新线路新站点没有设备历史能耗数据的预测模型冷启动问题。
(2)仅仅基于本站的设备历史能耗数据进行建模，忽略站点与站点之间的联系与影响，如相似站点、临近站点等。
综上所述，现有技术中对轨道交通站点的设备能耗进行预测的方法，存在着预测准确率低的技术问题。
发明内容
本发明实施例提供了一种轨道交通站点设备能耗预测方法、装置、设备和存储介质，解决了现有技术中对轨道交通站点的设备能耗进行预测的方法，存在着预测准确率低的技术问题。
本发明实施例提供了一种轨道交通站点设备能耗预测方法，包括以下步骤：
获取每个站点的特征向量数据以及设备历史能耗数据；
根据所述特征向量数据计算目标站点与其他站点的相似度，根据所述相似度确定所述目标站点的相似站点；
根据所述设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；
结合所述目标站点的设备能耗预测模型以及所述相似站点的设备能耗预测模型，计算所述目标站点的最终设备能耗预测值。
优选的，所述根据所述特征向量数据计算目标站点与其他站点的相似度，根据所述相似度确定所述目标站点的相似站点的具体过程为：
根据所述每个站点的特征向量数据，生成特征向量矩阵；
标准化所述特征向量矩阵，得到第一特征向量矩阵；
根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的相似度；
根据所述相似度确定所述目标站点的相似站点。
优选的，根据所述第一特征向量矩阵，计算目标站点与其他站点的相似度的具体过程为：
根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的欧式距离，将所述欧式距离作为所述目标站点与所述其他站点的相似度。
优选的，根据所述设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型的具体过程为：
根据所述设备历史能耗数据，建立设备历史能耗序列向量矩阵；
从所述设备历史能耗序列向量矩阵中提取出每个站点的历史能耗序列，将所述历史能耗序列转换为每个站点的训练标签序列集；
使用所述训练标签序列集训练LSTM模型，得到每个站点所对应的设备能耗预测模型。
优选的，所述将所述历史能耗序列转换为每个站点的训练标签序列集的具体过程为：
设定训练集特征序列长度；
根据所述训练集特征序列长度，将所述历史能耗序列转换为每个站点的训练标签序列集。
优选的，结合所述目标站点的设备能耗预测模型以及所述相似站点的设备能耗预测模型，计算所述目标站点的最终设备能耗预测值的具体过程为：
将所述目标站点的设备历史能耗数据输入至所述目标站点的设备能耗预测模型中，得到所述目标站点的第一能耗预测值；
将所述相似站点的设备历史能耗数据输入至所述相似站点的设备能耗预测模型中，得到所述相似站点的第二能耗预测值；
根据所述相似站点的数量计算所述第二能耗预测值的平均值；
分别为所述第一能耗预测值以及所述第二能耗预测值的平均值分配权重；
根据所述权重、所述第一能耗预测值以及所述第二能耗预测值的平均值，计算所述目标站点的最终设备能耗预测值。
优选的，获取每个站点的设备历史能耗数据的具体过程为：
获取每个站点在历史T个时间节点的能耗数据，其中，T根据能耗预测粒度以及训练数据长度确定。
第二方面，本发明实施例还提供了一种轨道交通站点设备能耗预测装置，包括
数据获取模块，用于获取每个站点的特征向量数据以及设备历史能耗数据；
相似站点确定模块，用于根据所述特征向量数据计算目标站点与其他站点的相似度，根据所述相似度确定所述目标站点的相似站点；
神经网络模型训练模块，用于根据所述设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；
设备能耗预测模块，用于结合所述目标站点的设备能耗预测模型以及所述相似站点的设备能耗预测模型，计算所述目标站点的最终设备能耗预测值。
优选的，相似站点确定模块包括：
特征向量矩阵子模块，用于根据所述每个站点的特征向量数据，生成特征 向量矩阵；
标准化子模块，用于标准化所述特征向量矩阵，得到第一特征向量矩阵；
相似度计算子模块，用于根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的相似度；
确定子模块，用于根据所述相似度确定所述目标站点的相似站点。
优选的，相似度计算子模块具体用于根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的欧式距离，将所述欧式距离作为所述目标站点与所述其他站点的相似度。
优选的，神经网络模型训练模块包括：
能耗序列向量矩阵子模块，用于根据所述设备历史能耗数据，建立设备历史能耗序列向量矩阵；
转换子模块，用于从所述设备历史能耗序列向量矩阵中提取出每个站点的历史能耗序列，将所述历史能耗序列转换为每个站点的训练标签序列集；
训练子模块，用于使用所述训练标签序列集训练LSTM模型，得到每个站点所对应的设备能耗预测模型。
优选的，转换子模块用于将历史能耗序列转换为每个站点的训练标签序列集的具体过程为：转换子模块用于设定训练集特征序列长度；根据所述训练集特征序列长度，将所述历史能耗序列转换为每个站点的训练标签序列集。
优选的，设备能耗预测模块包括：
第一能耗预测值计算子模块，用于将所述目标站点的设备历史能耗数据输入至所述目标站点的设备能耗预测模型中，得到所述目标站点的第一能耗预测值；
第二能耗预测值计算子模块，用于将所述相似站点的设备历史能耗数据输入至所述相似站点的设备能耗预测模型中，得到所述相似站点的第二能耗预测值；
平均值计算子模块，用于根据所述相似站点的数量计算所述第二能耗预测值的平均值；
权重分配子模块，用于分别为所述第一能耗预测值以及所述第二能耗预测值的平均值分配权重；
预测值计算子模块，用于根据所述权重、所述第一能耗预测值以及所述第二能耗预测值的平均值，计算所述目标站点的最终设备能耗预测值。
优选的，数据获取模块具体用于，获取每个站点在历史T个时间节点的能耗数据，其中，T根据能耗预测粒度以及训练数据长度确定。
第三方面，本发明实施例还提供了一种轨道交通站点设备能耗预测设备，所述设备包括：一个或多个处理器；存储装置，用于存储一个或多个程序，当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如第一方面所述的轨道交通站点设备能耗预测方法。
第四方面，一种存储计算机可执行指令的存储介质，所述计算机可执行指令在由计算机处理器执行时用于执行如第一方面所述的轨道交通站点设备能耗预测方法。
上述，本发明实施例通过获取每个站点的特征向量数据以及设备历史能耗数据；根据特征向量数据计算目标站点与其他站点的相似度，根据相似度确定目标站点的相似站点；根据设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值。本发明实施例考虑了站点与站点之间的联系与影响，在计算目标站点的最终设备能耗预测值时，利用了相似站点的设备能耗预测模型，通过结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型来计算目标站点的最终设备能耗预测值，从而避免了由于目标站点短期能耗波动导致最终设备能耗预测值误差较大，准确率低的不足，提高了预测的准确性；并且当目标站点为新站点时，基于相似站点的设备能耗预测模型也能够计算出目标站点的最终设备能耗预测值，即使新站点在没有设备历史能耗数据的情况下设备能耗预测模型也能够进行冷启动，提高了对新站点的进行设备能耗预测的准确性，解决了现有技术中对轨道交通站点的设备能耗进行预测的方法，存在着预测准确率低的技术问题。
附图说明
图1为本发明实施例提供的一种轨道交通站点设备能耗预测方法的流程图。
图2为本发明实施例提供的另一种轨道交通站点设备能耗预测方法的流程 图。
图3为本发明实施例提供的一种轨道交通站点设备能耗预测装置的结构示意图。
图4为本发明实施例提供的一种轨道交通站点设备能耗预测设备的结构示意图。
具体实施方式
以下描述和附图充分地示出本申请的具体实施方案，以使本领域的技术人员能够实践它们。实施例仅代表可能的变化。除非明确要求，否则单独的部件和功能是可选的，并且操作的顺序可以变化。一些实施方案的部分和特征可以被包括在或替换其他实施方案的部分和特征。本申请的实施方案的范围包括权利要求书的整个范围，以及权利要求书的所有可获得的等同物。在本文中，各实施方案可以被单独地或总地用术语“发明”来表示，这仅仅是为了方便，并且如果事实上公开了超过一个的发明，不是要自动地限制该应用的范围为任何单个发明或发明构思。本文中，诸如第一和第二等之类的关系术语仅仅用于将一个实体或者操作与另一个实体或操作区分开来，而不要求或者暗示这些实体或操作之间存在任何实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素。本文中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似部分互相参见即可。对于实施例公开的结构、产品等而言，由于其与实施例公开的部分相对应，所以描述的比较简单，相关之处参见方法部分说明即可。
实施例一
图1为本发明实施例提供的一种轨道交通站点设备能耗预测方法的流程图。本发明实施例提供的轨道交通站点设备能耗预测方法可以由轨道交通站点设备能耗预测设备执行，该轨道交通站点设备能耗预测设备可以通过软件和/或硬件的方式实现，该轨道交通站点设备能耗预测设备可以是两个或多个物理实体构成，也可以由一个物理实体构成。例如轨道交通站点设备能耗预测设备可以是电脑、手机、平板或智能交互平板等数据处理设备。
步骤101、获取每个站点的特征向量数据以及设备历史能耗数据。
在本实施例中，首先需要获取每个站点的特征向量数据以及设备历史能耗数据，从而为后续进行计算目标站点的最终设备能耗预测值做好准备。特征向量数据为能够反应每个站点的设备特征的数据，设备历史能耗数据为站点历史上每个设备的能耗数据。特征向量数据可根据实际的需要进行设置，在本实施例中，特征向量数据由7个特征向量组成，如表1所示。
表1
Figure PCTCN2021129193-appb-000001
步骤102、根据特征向量数据计算目标站点与其他站点的相似度，根据相似度确定目标站点的相似站点。
在得到每个站点的特征向量数据后，即可根据特征向量数据，计算出目标站点与其他站点之间的相似度。在一个实施例中，可以通过计算每个站点的特征向量数据之间的欧式距离，将欧式距离作为两个站点之间的相似度。在计算出目标站点和其他站点的欧式距离之后，根据欧式距离的数值，从小到大对其他站点进行排序，从序列中选取出前K个其他站点作为目标站点的相似站点。
步骤103、根据设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型。
在得到每个站点的设备历史能耗数据后，将设备历史能耗数据输入至设定的神经网络模型中进行训练，得到与每个站点对应的训练好的神经网络模型，将每个训练好的神经网络模型作为每个相应站点的设备能耗预测模型。在一个实施例中，神经网络模型选择LSTM模型，LSTM是一种时间递归神经网络， 在本实施例中，利用开源的pytorch框架，对LSTM模型进行构建训练。具体为：预先设置pytorch LSTM模型的网络参数，其中网络隐含层数量hidden_layer_size＝100，训练轮次epochs＝50，损失函数loss_function＝nn.MSELoss()设置为交叉熵损失，优化方法设置成adam优化方法。之后，将每个站点的设备历史能耗数据输入到不同的LSTM模型中进行训练，最终得到每个站点所对应的设备能耗预测模型。
步骤104、结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值。
在得到每个站点所对应的设备能耗预测模型后，选取出目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，来计算目标站点的最终设备能耗预测值。在一个实施例中，获取t时刻之前的目标站点的设备历史能耗数据和相似站点的设备历史能耗数据并输入到相应的设备能耗预测模型中，使得目标站点对应的设备能耗预测模型输出t时刻之后的第一能耗预测值，各个相似站点所对应的设备能耗预测模型输出t时刻之后的第二能耗预测值，计算各个第二能耗预测值的平均值，之后，分别为第一能耗预测值和第二能耗预测值的平均值分配权重并进行加权求和，即可得到目标站点t时刻之后的最终设备能耗预测值。
上述，本发明实施例通过结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型来计算目标站点的最终设备能耗预测值，从而避免了由于目标站点短期能耗波动导致最终设备能耗预测值误差较大，预测准确率低的不足，提高了预测的准确性；并且当目标站点为新站点时，基于相似站点的设备能耗预测模型也能够计算出目标站点的最终设备能耗预测值，使得新站点在没有设备历史能耗数据下设备能耗预测模型也能够进行冷启动，提高了对新站点进行设备能耗预测的准确性，解决了现有技术中对轨道交通站点的设备能耗进行预测的方法，存在着预测准确率低的技术问题。
实施例二
图2为本发明另一个实施例提供的一种轨道交通站点设备能耗预测方法的流程图，该轨道交通站点设备能耗预测方法是在上述实施例的基础上进行具体 化。如图2所示，轨道交通站点设备能耗预测方法包括：
步骤201、获取每个站点的特征向量数据以及设备历史能耗数据。
在一个实施例中，获取每个站点的设备历史能耗数据的具体过程为：获取每个站点在历史T个时间节点的能耗数据，其中，T根据能耗预测粒度以及训练数据长度确定。
示例性的，将能耗预测粒度设置为15分钟，将训练数据的长度设置为2020-01-01～2020-01-31。对于轨道交通线路上的某个站点，获取该站点在2020-01-01～2020-01-31内的每15分钟的能耗数据，则时间节点的个数T＝31*72＝2232。
步骤202、根据每个站点的特征向量数据，生成特征向量矩阵。
本实施例中，在得到每一个站点的特征向量数据后，对每一个站点的特征向量数据进行预处理，之后整理得到一个特征向量矩阵。示例性的，若每个站点的特征向量数据的特征向量如表1所示，对每个站点的特征向量数据进行整理后得到N行7列的特征向量矩阵X∈R N×7，
Figure PCTCN2021129193-appb-000002
其中N表示站点的数量，7表示特征向量数据的7个特征向量，特征向量矩阵中的元素x ij表示第i个站点的第j个特征向量，其中，N≧i＞0，7≧j＞0，x 1j为目标站点的特征向量数据，x pj为其他站点的特征向量数据，其中N≧p＞1。
步骤203、标准化特征向量矩阵，得到第一特征向量矩阵。
由于特征向量数据中每个特征向量的单位不一致，因此，需要对特征向量矩阵进行标准化。在本实施例中，采用z-score标准差标准化方法对特征向量矩阵进行标准化，具体过程为：计算特征向量矩阵X中所有特征向量的均值μ，之后，根据均值μ计算标准差σ，得到标准化转换函数，标准化转换函数x’ ij＝(x ij-μ)/σ，转化后得到的特征矩阵
Figure PCTCN2021129193-appb-000003
其中x’ ij表示标准化后的第i个站点的第j个特征向量，其中，N≧i＞0，7≧j＞0，x’ 1j为目标站点标 准化后的特征向量数据，x’ pj为其他站点标准化后的特征向量数据，其中N≧p＞1。
步骤204、根据第一特征向量矩阵，计算目标站点与其他站点的相似度。
在得到第一特征向量矩阵后，即可根据第一特征向量矩阵中标准化后的特征向量，计算目标站点和其他站点之间的相似度。在一个实施例中，根据第一特征向量矩阵，计算目标站点与其他站点的相似度的具体过程为：
根据第一特征向量矩阵，计算目标站点与其他站点的欧式距离，将欧式距离作为目标站点与其他站点的相似度。
示例性的，对于目标站点，计算其与其他站点之间的欧式距离，具体公式为：
Figure PCTCN2021129193-appb-000004
步骤205、根据相似度确定目标站点的相似站点。
在通过计算目标站点与N-1个其他站点的欧式距离后，即可得到N-1个距离序列d＝(d 2,d 3,......,d N)，根据欧式距离的数值大小，从距离序列d选取出数值最小的K个相似站点，记为S k＝(s 1，s 2,.....,s k)，其中s q为K个相似站点的编号，其中K≧q＞0。
步骤206、根据设备历史能耗数据，建立设备历史能耗序列向量矩阵。
本实施例中，在得到设备历史能耗数据后，根据设备历史能耗数据，建立每个站点的设备历史能耗序列向量矩阵。示例性的，对于N个站点，获取每个站点在历史T个时间节点的能耗数据，进行整理后得到设备历史能耗序列向量矩阵E∈R N×T，其中N表示站点的数量，T表示该站点某能耗设备在t时间节点的设备历史能耗数据，其中，T≧t＞0，
Figure PCTCN2021129193-appb-000005
设备历史能耗序列向量矩阵E中的元素e it表示第i个站点在第t个时间节点的设备历史能耗数据， 其中，N≧i＞0，T≧t＞0，e 1t为目标站点在第t个时间节点的设备历史能耗数据，e pt为其他站点在第t个时间节点的设备历史能耗数据，其中N≧p＞1。
步骤207、从设备历史能耗序列向量矩阵中提取出每个站点的历史能耗序列，将历史能耗序列转换为每个站点的训练标签序列集。
本实施例中，为了对后续的LSTM模型进行训练，在设备历史能耗序列向量矩阵E中，可提取出每一个站点的历史能耗序列S e＝(e i1,e i2,........,e iT)∈E，之后，将每一个站点的历史能耗序列S e转化为训练标签序列集。
在一个实施例中，将历史能耗序列转换为每个站点的训练标签序列集的具体过程为：
步骤2071、设定训练集特征序列长度。
首先需要设定训练集特征序列长度，训练集特征序列长度可根据历史能耗序列的总长度以及实际情况进行设置，在本实施例中，设置训练集特征序列长度train_window＝30。
步骤2072、根据训练集特征序列长度，将历史能耗序列转换为每个站点的训练标签序列集。
在设定了训练集特征序列长度，在设备历史能耗序列向量矩阵E中提取出每一个站点的历史能耗序列S e＝(e i1,e i2,........,e iT)，根据训练集特征序列长度将历史能耗序列转换为训练标签序列集，具体如下：
x 1＝(e i1，e i2，.....，e i30),y 1＝(e i31)
x 2＝(e i2，e i3，.....，e i31),y 2＝(e i32)
.......
x l＝(e il，e il+1，.....，e il+29),y l＝(e il+30)
其中，x l为能耗序列特征，y l为能耗序列结果标签，x l以及y l构成了训练标签序列集。
步骤208、使用训练标签序列集训练LSTM模型，得到每个站点所对应的设备能耗预测模型。
在得到每一个站点对应的训练标签序列集后，将每一个站点的训练标签序列集输入到LSTM模型中，使用反向传播算法或梯度下降法的对LSTM模型进行训练，判断LSTM模型的损失函数的误差损失是否小于预设阈值或LSTM模 型的训练次数达到预设次数，若是，则停止训练，得到训练好的设备能耗预测模型。
步骤209、结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值。
在一个实施例中，结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值的具体由步骤2091-步骤209过程为：
步骤2091、将目标站点的设备历史能耗数据输入至目标站点的设备能耗预测模型中，得到目标站点的第一能耗预测值；
在一个实施例中，获取目标站点在t时刻前30个时间节点的设备历史能耗数据
Figure PCTCN2021129193-appb-000006
将设备历史能耗数据
Figure PCTCN2021129193-appb-000007
输入到目标站点的设备能耗预测模型中，从而得到目标站点在t时刻后一个时间节点的第一能耗预测值。
步骤2092、将相似站点的设备历史能耗数据输入至相似站点的设备能耗预测模型中，得到相似站点的第二能耗预测值。
同理，获取K个相似站点在t时刻前30个时间节点的设备历史能耗数据
Figure PCTCN2021129193-appb-000008
其中，K≧q≧1。将K个相似站点的设备历史能耗数据
Figure PCTCN2021129193-appb-000009
输入到对应的设备能耗预测模型中，从而得到K个相似站点在t时刻后一个时间节点的第二能耗预测值。
步骤2093、根据相似站点的数量计算第二能耗预测值的平均值。
在得到K个相似站点的第二能耗预测值后，将K个相似站点的第二能耗预测值相加后的和除以K，得到第二能耗预测值的平均值。
步骤2094、分别为第一能耗预测值以及第二能耗预测值的平均值分配权重。
在本实施例中，分别为第一能耗预测值以及第二能耗预测值的平均值分配权重，其中，权重通过进行调优测试得到，可理解，在本实施例中，不对第一能耗预测值以及第二能耗预测值的平均值的权重进行具体限定。
步骤2095、根据权重、第一能耗预测值以及第二能耗预测值的平均值，计算目标站点的最终设备能耗预测值。
分别将第一能耗预测值以及第二能耗预测值的平均值乘以各自的权重，之 后进行求和，从而得到目标站点在t时刻后一个时间节点的最终设备能耗预测值，具体公式如下：
Figure PCTCN2021129193-appb-000010
其中，e s( t+1)为目标站点在t时刻后一个时间节点的最终设备能耗预测值，α、β为权重，model s为目标站点的设备能耗预测模型，model k为相似站点的设备能耗预测模型。当目标站点为新站点时，由于缺少目标站点的设备历史能耗数据，因此无法对目标站点的设备能耗预测模型进行训练，α＝0、β＝1，此时，根据相似站点的设备能耗预测模型即可计算出目标站点的t时刻后一个时间节点的最终设备能耗预测值。
上述，本发明实施例通过结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型来计算目标站点的最终设备能耗预测值，从而避免了由于目标站点短期能耗波动导致最终设备能耗预测值误差较大，准确率低的不足，提高了预测的准确性；并且当目标站点为新站点时，基于相似站点的设备能耗预测模型也能够计算出目标站点的最终设备能耗预测值，使得新站点在没有设备历史能耗数据的预测模型也能够进行冷启动，提高了对新站点进行设备能耗预测的准确性，解决了现有技术中对轨道交通站点的设备能耗进行预测的方法，存在着预测准确率低的技术问题。
实施例三
图3为本发明实施例提供的一种轨道交通站点设备能耗预测装置，其特征在于，包括
数据获取模块301，用于获取每个站点的特征向量数据以及设备历史能耗数据；
相似站点确定模块302，用于根据特征向量数据计算目标站点与其他站点的相似度，根据相似度确定目标站点的相似站点；
神经网络模型训练模块303，用于根据设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；
设备能耗预测模块304，用于结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值。
在上述实施例的基础上，相似站点确定模块302包括：
特征向量矩阵子模块，用于根据每个站点的特征向量数据，生成特征向量矩阵；
标准化子模块，用于标准化特征向量矩阵，得到第一特征向量矩阵；
相似度计算子模块，用于根据第一特征向量矩阵，计算目标站点与其他站点的相似度；
确定子模块，用于根据相似度确定目标站点的相似站点。
在上述实施例的基础上，相似度计算子模块具体用于根据第一特征向量矩阵，计算目标站点与其他站点的欧式距离，将欧式距离作为目标站点与其他站点的相似度。
在上述实施例的基础上，神经网络模型训练模块303包括：
能耗序列向量矩阵子模块，用于根据设备历史能耗数据，建立设备历史能耗序列向量矩阵；
转换子模块，用于从设备历史能耗序列向量矩阵中提取出每个站点的历史能耗序列，将历史能耗序列转换为每个站点的训练标签序列集；
训练子模块，用于使用训练标签序列集训练LSTM模型，得到每个站点所对应的设备能耗预测模型。
在上述实施例的基础上，转换子模块用于将历史能耗序列转换为每个站点的训练标签序列集的具体过程为：转换子模块用于设定训练集特征序列长度；根据训练集特征序列长度，将历史能耗序列转换为每个站点的训练标签序列集。
在上述实施例的基础上，设备能耗预测模块304包括：
第一能耗预测值计算子模块，用于将目标站点的设备历史能耗数据输入至目标站点的设备能耗预测模型中，得到目标站点的第一能耗预测值；
第二能耗预测值计算子模块，用于将相似站点的设备历史能耗数据输入至相似站点的设备能耗预测模型中，得到相似站点的第二能耗预测值；
平均值计算子模块，用于根据相似站点的数量计算第二能耗预测值的平均值；
权重分配子模块，用于分别为第一能耗预测值以及第二能耗预测值的平均值分配权重；
预测值计算子模块，用于根据权重、第一能耗预测值以及第二能耗预测值的平均值，计算目标站点的最终设备能耗预测值。
在上述实施例的基础上，数据获取模块301具体用于，获取每个站点在历史T个时间节点的能耗数据，其中，T根据能耗预测粒度以及训练数据长度确定。
实施例四
本实施例还提供了一种轨道交通站点设备能耗预测设备，如图4所示，一种轨道交通站点设备能耗预测设备，设备包括处理器400以及存储器401；
存储器401用于存储计算机程序402，并将计算机程序402传输给处理器；
处理器400用于根据计算机程序402中的指令执行上述的一种轨道交通站点设备能耗预测方法实施例中的步骤。
示例性的，计算机程序402可以被分割成一个或多个模块/单元，一个或者多个模块/单元被存储在存储器401中，并由处理器400执行，以完成本申请。一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述计算机程序402在轨道交通站点设备能耗预测设备中的执行过程。
轨道交通站点设备能耗预测设备可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。轨道交通站点设备能耗预测设备可包括，但不仅限于，处理器400、存储器401。本领域技术人员可以理解，图4仅仅是轨道交通站点设备能耗预测设备的示例，并不构成对轨道交通站点设备能耗预测设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如轨道交通站点设备能耗预测设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器400可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列 (Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
存储器401可以是轨道交通站点设备能耗预测设备的内部存储单元，例如轨道交通站点设备能耗预测设备的硬盘或内存。存储器401也可以是轨道交通站点设备能耗预测设备的外部存储设备，例如轨道交通站点设备能耗预测设备上配备的插接式硬盘，智能存储卡(Smart Media Card,SMC)，安全数字(Secure Digital,SD)卡，闪存卡(Flash Card)等。进一步地，存储器401还可以既包括轨道交通站点设备能耗预测设备的内部存储单元也包括外部存储设备。存储器401用于存储计算机程序以及轨道交通站点设备能耗预测设备所需的其他程序和数据。存储器401还可以用于暂时地存储已经输出或者将要输出的数据。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统，装置和设备的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统，装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本发明各个实施例方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
实施例五
本发明实施例还提供一种包含计算机可执行指令的存储介质，计算机可执行指令在由计算机处理器执行时用于执行一种轨道交通站点设备能耗预测方法，该方法包括以下步骤：
获取每个站点的特征向量数据以及设备历史能耗数据；
根据特征向量数据计算目标站点与其他站点的相似度，根据相似度确定目标站点的相似站点；
根据设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；
结合目标站点的设备能耗预测模型以及相似站点的设备能耗预测模型，计算目标站点的最终设备能耗预测值。
注意，上述仅为本发明实施例的较佳实施例及所运用技术原理。本领域技术人员会理解，本发明实施例不限于这里的特定实施例，对本领域技术人员来说能够进行各种明显的变化、重新调整和替代而不会脱离本发明实施例的保护范围。因此，虽然通过以上实施例对本发明实施例进行了较为详细的说明，但是本发明实施例不仅仅限于以上实施例，在不脱离本发明实施例构思的情况下，还可以包括更多其他等效实施例，而本发明实施例的范围由所附的权利要求范围决定。

Claims (10)

  1. 一种轨道交通站点设备能耗预测方法，其特征在于，包括以下步骤：
    获取每个站点的特征向量数据以及设备历史能耗数据；
    根据所述特征向量数据计算目标站点与其他站点的相似度，根据所述相似度确定所述目标站点的相似站点；
    根据所述设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；
    结合所述目标站点的设备能耗预测模型以及所述相似站点的设备能耗预测模型，计算所述目标站点的最终设备能耗预测值。
  2. 根据权利要求1所述的一种轨道交通站点设备能耗预测方法，其特征在于，所述根据所述特征向量数据计算目标站点与其他站点的相似度，根据所述相似度确定所述目标站点的相似站点的具体过程为：
    根据所述每个站点的特征向量数据，生成特征向量矩阵；
    标准化所述特征向量矩阵，得到第一特征向量矩阵；
    根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的相似度；
    根据所述相似度确定所述目标站点的相似站点。
  3. 根据权利要求2所述的一种轨道交通站点设备能耗预测方法，其特征在于，根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的相似度的具体过程为：
    根据所述第一特征向量矩阵，计算所述目标站点与所述其他站点的欧式距离，将所述欧式距离作为所述目标站点与所述其他站点的相似度。
  4. 根据权利要求1所述的一种轨道交通站点设备能耗预测方法，其特征在于，根据所述设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型的具体过程为：
    根据所述设备历史能耗数据，建立设备历史能耗序列向量矩阵；
    从所述设备历史能耗序列向量矩阵中提取出每个站点的历史能耗序列，将所述历史能耗序列转换为每个站点的训练标签序列集；
    使用所述训练标签序列集训练LSTM模型，得到每个站点所对应的设备能耗预测模型。
  5. 根据权利要求4所述的一种轨道交通站点设备能耗预测方法，其特征在于，所述将所述历史能耗序列转换为每个站点的训练标签序列集的具体过程为：
    设定训练集特征序列长度；
    根据所述训练集特征序列长度，将所述历史能耗序列转换为每个站点的训练标签序列集。
  6. 根据权利要求1所述的一种轨道交通站点设备能耗预测方法，其特征在于，结合所述目标站点的设备能耗预测模型以及所述相似站点的设备能耗预测模型，计算所述目标站点的最终设备能耗预测值的具体过程为：
    将所述目标站点的设备历史能耗数据输入至所述目标站点的设备能耗预测模型中，得到所述目标站点的第一能耗预测值；
    将所述相似站点的设备历史能耗数据输入至所述相似站点的设备能耗预测模型中，得到所述相似站点的第二能耗预测值；
    根据所述相似站点的数量计算所述第二能耗预测值的平均值；
    分别为所述第一能耗预测值以及所述第二能耗预测值的平均值分配权重；
    根据所述权重、所述第一能耗预测值以及所述第二能耗预测值的平均值，计算所述目标站点的最终设备能耗预测值。
  7. 根据权利要求1至权利要求6任一项所述的一种轨道交通站点设备能耗预测方法，其特征在于，获取每个站点的设备历史能耗数据的具体过程为：
    获取每个站点在历史T个时间节点的能耗数据，其中，T根据能耗预测粒度以及训练数据长度确定。
  8. 一种轨道交通站点设备能耗预测装置，其特征在于，包括
    数据获取模块，用于获取每个站点的特征向量数据以及设备历史能耗数据；
    相似站点确定模块，用于根据所述特征向量数据计算目标站点与其他站点的相似度，根据所述相似度确定所述目标站点的相似站点；
    神经网络模型训练模块，用于根据所述设备历史能耗数据对神经网络模型进行训练，得到每个站点所对应的设备能耗预测模型；
    设备能耗预测模块，用于结合所述目标站点的设备能耗预测模型以及所述相似站点的设备能耗预测模型，计算所述目标站点的最终设备能耗预测值。
  9. 一种轨道交通站点设备能耗预测设备，所述设备包括：一个或多个处理器；存储装置，用于存储一个或多个程序，当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如权利要求1-7中任一项所述的轨道交通站点设备能耗预测方法。
  10. 一种存储计算机可执行指令的存储介质，所述计算机可执行指令在由计算机处理器执行时用于执行如权利要求1-7中任一项所述的轨道交通站点设备 能耗预测方法。
PCT/CN2021/129193 2021-09-29 2021-11-08 轨道交通站点设备能耗预测方法、装置、设备和存储介质 WO2023050534A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111154384.9 2021-09-29
CN202111154384.9A CN113850439A (zh) 2021-09-29 2021-09-29 轨道交通站点设备能耗预测方法、装置、设备和存储介质

Publications (1)

Publication Number Publication Date
WO2023050534A1 true WO2023050534A1 (zh) 2023-04-06

Family

ID=78977208

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/129193 WO2023050534A1 (zh) 2021-09-29 2021-11-08 轨道交通站点设备能耗预测方法、装置、设备和存储介质

Country Status (2)

Country Link
CN (1) CN113850439A (zh)
WO (1) WO2023050534A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116128674A (zh) * 2023-04-14 2023-05-16 广州云硕科技发展有限公司 一种基于智能交通的能源数据处理方法及装置
CN116754828A (zh) * 2023-08-21 2023-09-15 济南瑞源智能城市开发有限公司 一种智慧隧道的能耗监测方法、设备及介质
CN118014318A (zh) * 2024-04-09 2024-05-10 国网浙江省电力有限公司宁波供电公司 一种基于多能系统的电力能源需求动态管理方法及装置
CN118070134A (zh) * 2024-04-24 2024-05-24 国网山东省电力公司汶上县供电公司 一种园区内用户用电能耗预测方法及系统
CN118228201A (zh) * 2024-05-23 2024-06-21 陕西高速电子工程有限公司 基于物联网的服务区电能能耗智能化监测方法
CN118690983A (zh) * 2024-08-28 2024-09-24 航天科工广信智能技术有限公司 一种基于多模态的能耗预测与节能方案生成方法及系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114912742A (zh) * 2022-03-25 2022-08-16 江苏联宏智慧能源股份有限公司 一种基于轨道交通车站能耗模型的能耗指标计算方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104442825A (zh) * 2014-11-28 2015-03-25 上海交通大学 一种电动汽车剩余行驶里程的预测方法和系统
CN108564227A (zh) * 2018-04-26 2018-09-21 重庆大学 一种基于时空特征的轨道交通客流量预测方法
CN110008974A (zh) * 2018-11-23 2019-07-12 阿里巴巴集团控股有限公司 行为数据预测方法、装置、电子设备及计算机存储介质
US20210012152A1 (en) * 2018-05-10 2021-01-14 Huawei Technologies Co., Ltd. Data Sequence Prediction Method and Computing Device
CN112686428A (zh) * 2020-12-15 2021-04-20 广州新科佳都科技有限公司 基于地铁线网站点相似性的地铁客流预测方法及装置
CN113159408A (zh) * 2021-04-14 2021-07-23 交控科技股份有限公司 轨道交通站点客流预测方法及装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268424A (zh) * 2014-10-12 2015-01-07 刘岩 一种基于时间序列的地铁能耗综合预测方法
CN107606745B (zh) * 2017-09-27 2019-09-27 南京中灿科技有限公司 地铁空调季逐时环控能耗预测方法
CN108022020B (zh) * 2017-12-15 2020-09-18 东软集团股份有限公司 设备故障管理方法、装置、存储介质和电子设备
CN113111578B (zh) * 2021-04-01 2023-04-07 上海晨翘智能科技有限公司 电力负荷预测方法、装置、计算机设备及存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104442825A (zh) * 2014-11-28 2015-03-25 上海交通大学 一种电动汽车剩余行驶里程的预测方法和系统
CN108564227A (zh) * 2018-04-26 2018-09-21 重庆大学 一种基于时空特征的轨道交通客流量预测方法
US20210012152A1 (en) * 2018-05-10 2021-01-14 Huawei Technologies Co., Ltd. Data Sequence Prediction Method and Computing Device
CN110008974A (zh) * 2018-11-23 2019-07-12 阿里巴巴集团控股有限公司 行为数据预测方法、装置、电子设备及计算机存储介质
CN112686428A (zh) * 2020-12-15 2021-04-20 广州新科佳都科技有限公司 基于地铁线网站点相似性的地铁客流预测方法及装置
CN113159408A (zh) * 2021-04-14 2021-07-23 交控科技股份有限公司 轨道交通站点客流预测方法及装置

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116128674A (zh) * 2023-04-14 2023-05-16 广州云硕科技发展有限公司 一种基于智能交通的能源数据处理方法及装置
CN116754828A (zh) * 2023-08-21 2023-09-15 济南瑞源智能城市开发有限公司 一种智慧隧道的能耗监测方法、设备及介质
CN116754828B (zh) * 2023-08-21 2023-12-01 济南瑞源智能城市开发有限公司 一种智慧隧道的能耗监测方法、设备及介质
CN118014318A (zh) * 2024-04-09 2024-05-10 国网浙江省电力有限公司宁波供电公司 一种基于多能系统的电力能源需求动态管理方法及装置
CN118070134A (zh) * 2024-04-24 2024-05-24 国网山东省电力公司汶上县供电公司 一种园区内用户用电能耗预测方法及系统
CN118228201A (zh) * 2024-05-23 2024-06-21 陕西高速电子工程有限公司 基于物联网的服务区电能能耗智能化监测方法
CN118690983A (zh) * 2024-08-28 2024-09-24 航天科工广信智能技术有限公司 一种基于多模态的能耗预测与节能方案生成方法及系统

Also Published As

Publication number Publication date
CN113850439A (zh) 2021-12-28

Similar Documents

Publication Publication Date Title
WO2023050534A1 (zh) 轨道交通站点设备能耗预测方法、装置、设备和存储介质
WO2020249125A1 (zh) 用于自动训练机器学习模型的方法和系统
CN108764273B (zh) 一种数据处理的方法、装置、终端设备及存储介质
WO2021120677A1 (zh) 一种仓储模型训练方法、装置、计算机设备及存储介质
CN111435463A (zh) 数据处理方法及相关设备、系统
WO2021068563A1 (zh) 样本数据处理方法、装置、计算机设备及存储介质
WO2019214455A1 (zh) 一种数据序列预测方法及计算设备
WO2021004318A1 (zh) 资源数据的处理方法、装置、计算机设备和存储介质
CN113704389A (zh) 一种数据评估方法、装置、计算机设备及存储介质
CN114372532B (zh) 标签标注质量的确定方法、装置、设备、介质及产品
CN110781970A (zh) 分类器的生成方法、装置、设备及存储介质
CN114077912A (zh) 数据预测方法以及数据预测装置
CN112836937A (zh) 基于熵权与bp神经网络技术的洪涝灾害损失评估方法
US20230385597A1 (en) Multi-granularity perception integrated learning method, device, computer equipment and medium
Widodo et al. Model selection using dimensionality reduction of time series characteristics
CN110176309A (zh) 一种用于预测心血管疾病的医疗数据处理方法
CN113139381A (zh) 不均衡样本分类方法、装置、电子设备及存储介质
CN116340831B (zh) 一种信息分类方法、装置、电子设备及存储介质
CN116468967B (zh) 样本图像筛选方法、装置、电子设备及存储介质
CN117252362A (zh) 基于人工智能的调度方法、装置、计算机设备及存储介质
WO2020124977A1 (zh) 生产数据处理方法、装置、计算机设备和存储介质
CN114154564A (zh) 基于异质图的关联度确定方法、装置、电子设备及存储介质
Dai et al. Integrating Multi Spatial-Temporal Information with Globally Attentive Gated Convolutional Network for Transmission Line Icing Tension Prediction
CN116723083B (zh) 一种云服务器在线故障诊断方法及装置
CN112801372B (zh) 数据处理方法、装置、电子设备和可读存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21959096

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE