A subway integrated monitoring system based on multi-source heterogeneous data

Abstract

The application discloses a subway comprehensive monitoring system based on multi-source heterogeneous data, mainly including the collection and heterogeneity of electrical data, neural network passenger flow prediction, dynamic adjustment of operation diagram according to passenger flow, and the overall design of urban rail transit Web server. The subway comprehensive monitoring system is based on a micro-service framework to realize cloud platform deployment, is data-centered, and is uniformly managed, analyzed and heterogeneous. Various services required for subway operation are designed, including UI interactive interface, intelligent algorithm and the like. The whole server is deployed on the cloud platform to realize cost reduction, resource optimization and data sharing. The overall architecture of the whole platform runs through data collection, reception and processing analysis, intelligent algorithm design, simulation model construction and man-machine application interaction and the like. The designed subway comprehensive monitoring platform is in the form of data flow, the functions and connections of each part are described in detail, the platform has extensibility and reliability, and has important practical application significance.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of urban rail transit integrated monitoring system, and particularly relates to a subway integrated monitoring system based on multi-source heterogeneous data. BACKGROUND

[0002] The construction of urban rail transit improves the efficiency of transportation. In the construction process, China's urban rail transit not only focuses on expanding the scale, but also is committed to promoting technological innovation and introducing intelligent and automated advanced technologies to improve system operation efficiency and service level. The significance of building an urban rail transit integrated monitoring system is as follows: (1) With the progress of technology, the monitoring targets involved in the subway system become more diversified, such as train operation status, equipment health status, passenger flow, energy consumption, etc. The relationship between these targets becomes more complex, and comprehensive information collection and processing are needed. (2) With the rapid development of urban rail transit, in order to better cope with the complex relationship of monitoring targets and improve the operation level, the sharing of resources and information becomes particularly important. The demand for information sharing between various subsystems, different subway lines, and between subway management departments and related institutions has greatly increased. (3) The integrated monitoring system monitors the energy consumption status of each railway line and station, produces various energy reports, energy consumption curves, and related pie charts and column charts, etc., and shares data with relevant managers and operators using information networks, providing appropriate basis for formulating energy-saving strategies for railway transportation operation and management. (4) Based on the collection of current energy consumption data of lines, stations and systems, further analysis, comparison and evaluation are carried out, and targeted energy-saving measures are developed for different passenger flow situations to avoid the blind use of energy-saving equipment and the installation of energy-saving devices. (5) The managers of energy consumption monitoring equipment can monitor the status of various circuits in real time and quickly find energy consumption anomalies. Through comparison analysis and load analysis functions, it is helpful to identify the causes and provide reasonable remedial measures, ultimately achieving the goal of reducing energy consumption and operating costs. (6) The traditional subway integrated monitoring system does not include all the information of non-integrated systems, only part of the equipment operation information is extracted into the system, and the information sharing is insufficient. The use of energy flow and information flow is ignored. The original data is analyzed and heterogeneous in this design, and the simulation total energy consumption data of the whole line and the subway passenger flow information of the next three stations are obtained according to the basic electrical data. The dynamics model, neural network model and optimization energy-saving algorithm use measured data, simulation data and predicted data for error comparison, and continuously correct the prediction accuracy of the model to realize the iteration and upgrading of the model.

[0003] At present, rail transit is developing, its operation environment, working principle and technical indicators have changed a lot. In order to adapt to the new requirements of new technologies such as cloud computing, big data and neural network into urban rail transit, the design of the integrated monitoring system needs to integrate innovative ideas, adjust the ideological structure, and increase the integration and intercommunication of new technologies and practical problems. Aiming at the problems of small integration range, insufficient information sharing, insufficient data utilization and limited subsequent function expansion of the traditional integrated monitoring system, a multi-source heterogeneous data integrated monitoring system with more extensive integration and information sharing is proposed. And the new concept of integrated monitoring system of integrated energy saving optimization system is constructed by combining neural network. The data acquisition equipment installed on the train-ground is used to monitor the catenary voltage, traction current, auxiliary current and substation data, and real-time prediction of passenger flow and energy consumption data. Through online cleaning and analysis of data on the server side, and according to the established neural network model and train simulation model, multi-source heterogeneous data is obtained. Relying on these data, the subway train-road-network information flow and energy flow can be connected, and the optimization scheduling and online diagram adjustment of the subway power supply system can be realized according to the dynamic passenger flow throughout the day. SUMMARY

[0004] The purpose of the present application is to provide a city rail transit energy management micro-service system with good compatibility and expansibility, which can realize real-time collection of overall rail transit energy consumption data, real-time viewing of energy consumption information, remote interactive control of collection equipment and simulation calculation of traction power supply system under the condition of meeting normal demand cost, thereby reducing energy consumption and realizing low-cost operation.

[0005] The technical solution for realizing the purpose of the present application is as follows: according to the voltage and current data measured by the on-board current and voltage sensor, a neural network is established to calculate the current passenger flow based on the train traction power. According to the train power data within ten seconds after starting, which is mainly related to the interval passenger flow, the interval passenger flow (passenger capacity) is predicted. Then the passenger flow (passenger capacity) of the first three stations is input, and the interval passenger flow of the last three stations is output, and a residual neural network is designed for short-term passenger flow prediction. The prediction error is within 3%, and according to the predicted future passenger flow distribution, the operation diagram adjustment for sudden situations is realized.

[0006] A subway integrated monitoring system based on multi-source heterogeneous data, comprising a data acquisition service module, a traction power supply system simulation service module, a passenger flow prediction service module, a dynamic operation diagram adjustment module and a micro-service subway integrated system, and integrating them into a service system through a Web server, wherein:

[0007] The data acquisition service module comprises a distributed data acquisition unit and a data transmission cloud platform server;

[0008] The traction power supply system simulation analysis service module models the metro power supply system and performs power flow calculation and analysis to obtain substation energy consumption data under different operation diagrams;

[0009] The passenger flow prediction service module mainly predicts the metro passenger carrying capacity of the next three stations;

[0010] The dynamic operation diagram adjustment module mainly adjusts the energy-saving operation diagram according to the results of passenger flow prediction to meet the current passenger flow situation.

[0011] The Web server side stores, analyzes and calculates the service data of each module, updates and calculates the neural network model, and integrates all the functions of the monitoring system into a cloud platform.

[0012] Further, the data acquisition service module includes a sensor, an STM32 single-chip microcomputer, an FPGA data processing module, a GPRSDTU data acquisition unit, and a remote server.

[0013] Further, the sensor converts the signals collected from the train and the substation into AD through the STM32 single-chip microcomputer, calculates and heterogenizes the data in the FPGA data processing module, and sends them to the GPRSDTU data acquisition unit in the agreed format; the GPRSDTU data acquisition unit establishes a TCP long connection with the server side for data transmission; the remote server writes the data sent by the GPRSDTU into the database through Socket technology for storage backup, and can actively return instructions to the GPRSDTU data acquisition unit client.

[0014] Further, the traction power supply system simulation analysis service module is based on the modeling and analysis of the traction substation, traction network, and train to establish a whole time-varying power network model of the DC traction system.

[0015] Further, the traction power supply system simulation analysis service uses the Newton-Raphson method with fast convergence speed and good precision to solve the nonlinear equation set obtained by system modeling, and the specific calculation process is as follows:

[0016] 1) Enter the initial simulation data, including train schedule data, interval data determined by train operation simulation calculation, unit resistance of the catenary and running track;

[0017] 2) Initialize the maximum number of iterations and the initial value, including setting the initial voltage value to DC 1500V and the current to 0; and set the simulation time and step size;

[0018] 3) Establish the admittance equation of each node to form the admittance matrix, construct the auxiliary equation to meet the basic requirements of power flow calculation, and write the corresponding nonlinear equation set under the current model.

[0019]

[0020] where Ui is the corresponding node voltage value, i = 1, 2, … n;

[0021] 4) Using the Newton method, the initial value of each variable and the correction value of each variable are substituted into equation (1) to obtain equation (2);

[0022]

[0023] 5) The Newton correction equation is obtained as follows:

[0024]

[0025] Equation (3) is the correction equation for the correction amount After elimination, each correction amount can be solved. Then the initial value is corrected to obtain the correction equation as follows (4);

[0026]

[0027] 6) Iteration is repeated, and the equation set corresponding to the k+1 iteration is:

[0028]

[0029]

[0030] where and are the node voltage values of the k+1 and k iterations, respectively; the difference between the two is the voltage offset;

[0031] 7) Determine whether the catenary current is less than 0 and the voltage of the catenary train node is less than 1800V. If both conditions are met, go to 9); if not, go to 8);

[0032] 8) Determine whether the iteration number exceeds the preset upper limit value. If not, go to 9); if so, correct the initial value of each node and go to 6);

[0033] 9) Save the calculated results and output;

[0034] 10) Determine whether the current simulation time has reached the preset simulation end time. If not, go to 6); if so, the operation is ended and the simulation is completed.

[0035] Further, in step 6), if the node voltage values of two iterations satisfy formula (7), that is, the voltage offset is less than the convergence precision epsilon, it is considered that the node satisfies the convergence condition, and then whether other nodes in the network satisfy is judged, if satisfying, step 9) is entered, otherwise step 7) is entered.

[0036] Further, the passenger flow prediction service module is used to establish an interval passenger flow prediction model of train traction power-passenger flow by using BP neural network and collected actual current and voltage data, and residual neural network is introduced to predict passenger flow of three stations in the future by using the predicted interval passenger flow.

[0037] Further, the dynamic operation diagram adjustment module is used to reduce the waiting time of passengers and traction energy consumption as the optimization target, and a particle swarm algorithm is designed to obtain an optimal energy-saving timetable, and the performance of the timetable is evaluated.

[0038] Further, the micro-service subway comprehensive monitoring system adopts micro-service for module separation between various functions, realizes the connection of data flow and energy flow through an interface database, and deploys the monitoring platform on a cloud platform to realize the expansion of computing resources and storage resources.

[0039] Compared with the prior art, the present application has the following advantages:

[0040] (1) The measured data is fully utilized to realize the heterogeneity of multi-source data, including predicted passenger flow and predicted energy consumption;

[0041] (2) According to the sudden passenger flow, the timetable is adjusted in real time, and intelligent algorithms and prediction models are integrated into the monitoring platform, thereby enriching the functions of the monitoring platform;

[0042] (3) The Netty asynchronous communication technology is adopted to realize a single-channel multi-connection high-concurrency server conveniently and efficiently, so that the server performance is further improved;

[0043] (4) The micro-service design architecture is adopted to solve the server pressure problem caused by complex heterogeneous data, multiple business requirements and large data transmission volume. Virtual computing, storage and network resource configuration functions are realized on the Ali cloud platform. The micro-service design makes each functional module more fine-grained, and the new system can be extracted, split and integrated on the cloud platform. The new system has high safety, good isolation, convenient maintenance and updating, strong load capacity and good scalability. Real-time, heterogeneous data decoupling and intelligence of the comprehensive monitoring platform are realized;

[0044] (5) The traction power supply system simulation analysis is carried in the micro-service system, and the simulation results are visualized on the web page and compared with the measured data, so that the analysis of the measured and simulated data is more intuitive. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The overall structure schematic diagram of the comprehensive monitoring system built by the present application.

[0046] Figure 2 The block diagram of data transmission service in the present application.

[0047] Figure 3 The time-varying power model diagram for simulation of the traction power supply system in the present application.

[0048] Figure 4 The service flow diagram for simulation of the traction power supply system in the present application.

[0049] Figure 5 The technical architecture block diagram of the urban rail energy management micro-service system built by the present application. DETAILED DESCRIPTION

[0050] The present application is further described in detail below in combination with the drawings and specific embodiments.

[0051] The present application designs an urban rail transit comprehensive monitoring system, which combines Figure 1 , and includes the overall design of data remote real-time acquisition service, rail transit traction power supply system simulation analysis service, post-three-station passenger flow prediction, full-line energy consumption simulation service, particle swarm optimization algorithm service and urban rail transit Web server.

[0052] In combination with Figure 2 , the data remote real-time acquisition service is designed at both ends on the basis of the communication service between the terminal and the server, the data acquisition terminal is composed of a sensor, an STM32 single-chip microcomputer and a GPS DTU, and communicates with the data transmission server group through a wireless network.

[0053] The acquisition terminal and the server of the present system are point-to-point communication with frequent data interaction, and the integrity and reliable delivery of data are guaranteed, the number of acquisition terminals is relatively limited, and the TCP / IP long connection mode is adopted, the acquisition terminal is used as a client to establish a connection with the server, and the server maintains the long connection socket by sacrificing a small part of resources, so that the resource loss caused by frequent handshake, waving hand and socket creation of short connection is solved.

[0054] Further, the upper application protocol stack is designed to solve the problem that the TCP transmission protocol is flow-oriented and has no clear message protection boundary, and two ways are adopted:

[0055] 1) Each data packet is encapsulated into a fixed length at the transmitting end, so that the receiving end can naturally segment the data packet by reading the fixed length of data from the receiving buffer each time.

[0056] 2) A separator is provided between data packets as a boundary, so that the receiving end can split different data packets at this boundary.

[0057] After the data acquisition terminal is powered on and starts working, the STM32 acquisition board and the GPS DTU are initialized. The initialization of the STM32 configures the TIM (standard timer module), NVIC (nested vector interrupt controller), RTC (real-time clock), DMA (direct memory access), GPIO (general-purpose input and output) peripherals, and the like. The sampling is triggered by the rising edge of the pwm clock of the timer TIM, the analog-to-digital converter ADC converts the sampling signal, and the data is transferred to the memory by the DMA and processed by the CPU. It is judged whether an interrupt occurs, and if so, the interrupt function is executed to smooth the data.

[0058] Further, a heartbeat packet is set in the DTU to be sent to the server to indicate that the current client is still working normally, and the server can keep the information of the client. The registration packet is sent by the DTU after the connection with the server is successfully established, and is set as the unique device number of the current DTU as its identity. After the configuration is completed, a SIM card is put into the DTU for registration to the network.

[0059] For the data acquisition server, a multiplexed IO mode is adopted to obtain high performance, and an asynchronous event-driven network application framework Netty is used for implementation. The event channel processing flow is as follows:

[0060] 1) First, enter the heartbeat check processor IdleStateHandler under the Netty library package and the custom timeout event logic processor HeartBeatServerHandler to ensure the validity of the TCP connection.

[0061] 2) A custom codec processor MyMessageDecoder inherited from ByteToMessageDecoder is designed. In the processor, mode judgment is first performed: instruction interaction and reception of energy efficiency data mode. For a successfully authenticated context channel, first, the instruction flag is judged. If it is the first round of instruction interaction, that is, the instruction queue set has not been obtained from the corresponding instruction directory and loaded into the program memory, the instruction queue head data is loaded and sent, and the instruction directory is emptied for the next instruction set. Otherwise, the instruction to be executed is obtained from the program memory, and it is judged whether the response to the previous instruction agreed with the data acquisition terminal is correct. If correct, the subsequent instruction is executed. If incorrect or timeout, the response data is re-waited. After all the instructions in the instruction queue are sent, the instruction flag bit is restored. When the current instruction interaction process is completed, the subsequent data is processed according to the energy efficiency data.

[0062] 3) Enter the LoggingHandler under the Netty library package to start the log record of Netty

[0063] 4) Then enter the registered processor ActiveHandler to register the channel according to the unique number of the equipment in the parsed data, and identify and operate the specified channel.

[0064] 5) After registration, enter PresentSaveFileHandler in sequence to store the data in the original format.

[0065] 6) Enter DataFormatHandler and SaveDatabaseHandler to format and store the data into the database.

[0066] The whole model of the DC traction system is established based on modeling and analysis of the traction substation, traction network and train respectively. Figure 3 The relationship between the output voltage of the DC side of the traction substation and the load current is represented by a Thevenin equivalent circuit in which an ideal voltage source is connected in series with an equivalent internal resistance; the running rail and return line in the traction network are the objects of modeling in the return system, and the influence of leakage current is considered; and the train is equivalent to a power source model.

[0067] Further, according to the time-varying power supply system model built, the number of substation nodes in the network is p, the number of online trains is q, and the case of node coincidence is not considered. Each train node or substation node is connected to the catenary and running rail, which corresponds to an electric network node, so the total number of electric network nodes is n=2(p+q). The Newton-Raphson method is used for power flow calculation, combined with Figure 4 , the specific calculation process is as follows:

[0068] 1) Enter the initial data of simulation: train timetable data, interval data determined by train operation simulation calculation, unit resistance of catenary and running rail, etc.

[0069] 2) Initialize the maximum number of iterations and iteration initial value, including setting the initial voltage value to DC 1500V and the current to 0; and set the simulation time and step size;

[0070] 3) Establish the admittance equation of each node to form the admittance matrix, construct the auxiliary equation to meet the basic requirements of power flow calculation, and write the corresponding nonlinear equation group under the current model;

[0071]

[0072] In the formula, Ui is the corresponding node voltage value, i=1,2,…n;

[0073] 4) Using the Newton-Raphson method to solve, the initial value of each variable and the correction value of each variable is substituted into equation (1) to obtain equation (2).

[0074]

[0075] 5) The Newton-Raphson correction equation is obtained as follows:

[0076]

[0077] Equation (3) is a correction equation for the correction amount After elimination, each correction amount is solved. Then the initial value is corrected to obtain the correction equation as follows (equation 4).

[0078]

[0079] 6) Iteration is repeated, and the equation set corresponding to the k+1 iteration is:

[0080]

[0081]

[0082]

[0083] In the equation, and are the node voltage values of the k+1 and k iterations, respectively; the difference between the two is the voltage offset; if the node voltage values of the two iterations satisfy equation (7), i.e., the voltage offset is less than the convergence precision ε, then it is considered that the node satisfies the convergence condition, and it is further determined whether other nodes in the network satisfy the condition; if they do, then step 9) is entered; otherwise, step 7) is entered.

[0084] 7) Since the rectifier set has a rectifier diode, it cannot absorb backflow current, i.e., the catenary current is less than 0; in addition, considering the safety of train travel and the limitation of maximum voltage tolerance, the voltage of the catenary train node cannot be greater than 1800V; if the above conditions are not met, then step 8) is entered; if they are met, then step 9) is entered.

[0085] 8) It is determined whether the iteration number exceeds the preset upper limit value of the iteration number; if it does not, then step 9) is entered; if it does, then the initial value of each node is corrected and step 6) is entered.

[0086] 9) The calculated results are saved and output.

[0087] 10): determine whether the preset simulation end time is reached at the current simulation time, if not, go to 6); if yes, the operation is ended, and the simulation is completed.

[0088] The overall design of the urban rail transit cloud service system adopts a combination of Figure 5 The system architecture can be divided into four parts from top to bottom, namely, a field terminal, a station cloud platform, a center platform and a smart application.

[0089] (1) The field terminal includes an online train terminal and a station terminal. The online train terminal is mainly responsible for the collection and calculation of traction current, braking current, auxiliary power supply current and predicted passenger flow data through the on-board board card. The station terminal is mainly responsible for collecting platform information.

[0090] (2) The station cloud platform is mainly used for scheduling sudden passenger flow and analyzing and utilizing platform information.

[0091] (3) The center platform is mainly used for energy consumption simulation and model training of the collected data by using various neural network models and intelligent algorithms.

[0092] (4) The smart application includes time table adjustment, train capacity arrangement, energy consumption report and early warning and other applications according to the results of the intelligent algorithm.

[0093] Through the micro-service system architecture of the application, the functions of user login, permission management, data monitoring, simulation analysis, statistical analysis and log management are further specifically realized.

Claims

1. A comprehensive subway monitoring system based on multi-source heterogeneous data, characterized in that, This includes a data acquisition service module, a traction power supply system simulation service module, a passenger flow forecasting service module, a dynamic timetable adjustment module, and a microservice-based metro integrated system, which are integrated into a single service system via a web server. The data acquisition service module comprises two parts: a distributed data acquisition unit and a data transmission cloud platform server. The traction power supply system simulation analysis service module models the subway power supply system, performs power flow calculations and analysis, and obtains substation energy consumption data under different operating diagrams. The passenger flow prediction service module mainly predicts the passenger volume of the three subway stations in the future; it uses a BP neural network to establish a passenger flow prediction model of train traction power-passenger flow based on the collected actual current and voltage data, and uses the predicted passenger flow of the interval to introduce a residual neural network to predict the passenger flow of the three stations in the future. The dynamic operation schedule adjustment module mainly adjusts the energy-saving operation schedule according to the passenger flow forecast results to conform to the current passenger flow situation; based on real-time and dynamic passenger flow data, with the optimization goal of reducing passenger waiting time and traction energy consumption, a particle swarm algorithm is designed to solve for the optimal energy-saving timetable, and the performance of the timetable is evaluated. The web server stores, analyzes, and calculates the service data of each module, updates and calculates the neural network model, and centralizes the various functional services of the monitoring system onto a cloud platform.

2. The subway integrated monitoring system based on multi-source heterogeneous data according to claim 1, characterized in that, The data acquisition service module includes sensors, an STM32 microcontroller, an FPGA data processing module, a GPRSDTU data acquisition unit, and a remote server.

3. A subway integrated monitoring system based on multi-source heterogeneous data according to claim 2, characterized in that, The sensors collect signals from the train and substation, perform AD conversion using the STM32 microcontroller, perform data calculation and heterogeneity processing within the FPGA data processing module, and send the data to the GPRS DTU data acquisition unit in a pre-defined format. The GPRS DTU data acquisition unit establishes a long TCP connection with the server for data transmission. The remote server uses Socket technology to write the data sent by GPRSDTU into a database for storage and backup, and can actively return instructions to the GPRSDTU data acquisition unit client.

4. A subway integrated monitoring system based on multi-source heterogeneous data according to claim 1, characterized in that, The aforementioned traction power supply system simulation analysis service module establishes an overall time-varying network model of the DC traction system based on the separate modeling and analysis of the traction substation, traction network, and train.

5. A subway integrated monitoring system based on multi-source heterogeneous data according to claim 1, characterized in that, The aforementioned traction power supply system simulation analysis service uses the Newton-Raphson method, which boasts fast convergence speed and high accuracy, to solve the nonlinear equations obtained from system modeling. The specific calculation steps are as follows: 1) Input the initial simulation data, including train timetable data, section data determined by train operation simulation calculations, and unit resistance of the overhead contact line and running rails; 2) Initialize the maximum number of iterations and initial iteration values, including setting the initial voltage value to DC 1500V and the current to 0; and setting the simulation time and step size; 3) Establish the admittance equations for each node to form the admittance matrix, construct auxiliary equations to meet the basic requirements of power flow calculation, and write the corresponding nonlinear equations under the current model; （1） In the formula, Ui is the corresponding node voltage value, i=1,2,…n; 4) When using the Newton-Lambert method to solve the problem, the initial values ​​of each variable should be... and the correction values ​​of each variable Substituting into equation (1) yields equation (2); （2） 5) The corrected equation of Newton's method is obtained as follows: （3） Equation (3) is for the correction amount The corrected system of equations can be solved by eliminating variables to obtain the various correction quantities. Then, the initial value is corrected to obtain the corrected equation as shown in equation (4). （4） 6) After repeated iterations, the system of equations corresponding to the (k+1)th iteration is: （5） （6） （7） In the formula These are the node voltage values ​​at the (k+1)th and kth iterations, respectively; the difference between them. This is the voltage offset. 7): Determine whether the contact wire current is less than 0 and whether the voltage of the contact wire train node is less than 1800V. If both conditions are met, proceed to 9); otherwise, proceed to 8). 8): Determine if the number of iterations exceeds the preset upper limit. If not, proceed to 9); if so, correct the initial values ​​of each node and proceed to 6). 9): Save and output the calculation results; 10): Determine whether the current simulation time has reached the preset simulation end time. If not, proceed to 6); if it has, the calculation ends and the simulation is complete.

6. A subway integrated monitoring system based on multi-source heterogeneous data according to claim 5, characterized in that, In step 6), if the node voltage values ​​of the two iterations satisfy equation (7), that is, the voltage offset is less than the convergence accuracy... If the node meets the convergence condition, then it is considered to meet the convergence condition. Then it is determined whether other nodes in the network meet the condition. If they do, proceed to step 9; otherwise, proceed to step 7.

7. A subway integrated monitoring system based on multi-source heterogeneous data according to claim 1, characterized in that, The aforementioned microservice-based subway integrated monitoring system uses microservices to separate modules among its various functions, connects data and energy flows through an interface database, and deploys the monitoring platform on a cloud platform to expand computing and storage resources.

Images