Method for real-time entering passenger flow volume anomaly detection of urban rail transit AFC system

Abstract

The invention provides a method for real-time entering passenger flow volume anomaly detection of an urban rail transit AFC system. First an improved a small data volume method is adopted to calculate Lyapunov exponents of a historical entering passenger flow data time series, the chaotic characteristic of the sequence is verified, then a C-C method is utilized to determine the time delay and optimal embedded dimension of the chaotic time series, phase-space reconstruction is performed on the original time series, a model training test sample set is determined, a large-range grid search method is adopted to optimize model parameters, the optimized model is utilized to predict the entering passenger flow volume of each time period, then the entering amount of each time period in a training sample is utilized to predict the distribution characteristic of a residual series, and finally a confidence interval of entering passenger flow volume predicted residual of each time period at a certain confidence level is determined, thereby determining a threshold upper limit and a threshold lower limit of entering passenger flow volume anomaly detection of a future time period. The method provided by the invention effectively reduces the range of anomaly detection of entering passenger flow volume, and reduces the false alarm rate of data anomaly detection.

Description

technical field [0001] The invention relates to a real-time abnormal detection method for inbound passenger flow of an urban rail transit AFC system, which belongs to the intelligent technology of urban rail transit. Background technique [0002] The real-time passenger flow data information in the urban rail transit system is crucial to the improvement of the service capability of the subway system. However, due to the diversity of equipment suppliers in the AFC system and the complexity of the real-time data transmission process, the real-time passenger flow data obtained from the AFC system cannot fully reflect the actual operation situation. There is a large difference between the inbound passenger flow and the actual inbound passenger flow. In order to ensure the quality of the real-time passenger flow data, it is necessary to perform anomaly detection and error correction processing on the real-time passenger flow data. By setting a reasonable threshold for the passen...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: In order to solve the problem that the threshold range calculated by the existing threshold-based real-time passenger flow data anomaly detection method is not good enough for data anomaly detection, the present invention provides an urban rail transit AFC system for real-time station entry Passenger flow anomaly detection method, this method determines the model training test sample set by verifying the chaotic characteristics of the sequence, and then uses the distribution characteristics of the residual sequence of the inbound volume in each period in the training sample to determine the inbound passenger flow in the future period Upper threshold and lower threshold for anomaly detection

Images