Intelligent scheduling table generation method based on multi-source data analysis of buses

A method for guiding the dispatching of vehicles, which generates an intelligent scheduling table based on multi-source data analysis of buses, the method comprising following steps:

• defining LC as the number of all lines owned by a bus company, representing a serial number of one of lines with Lk, then 1≤k≤LC, representing the number of all vehicles in the line Lk by BC, representing a serial number of one of bus vehicles in the line Lk by Bn, then 1≤n≤BC, representing upgoing and downgoing of lines by ud, wherein ud is 0 for downgoing, 1 for upgoing, and representing a serial number of one of bus stops in the line Lk by St, wherein a serial number range of up-going stops is 1≤t≤su, su is the number of all up-going stops in the line Lk, sp is the number of all down-going stops in the line Lk, a serial number range of down-going stops is su+1≤t≤su+sp;
• indicating the dth day from the current day by d, 1≤d≤D, and indicating the maximum number of days with D from a selected time range to the current day;
• recording electronic swiping times of one of bus vehicles numbered as Bn in the line Lk passing though the stop St in ith shift on the dth day as Ci, the number of shifts of this bus vehicle as sc, electronic swiping times of the bus vehicle numbered as Bn in the line Lk at the stop St on the dth day as Wt, electronic swiping times of all vehicles in the line Lk at the stop St on the dth day as Vt, electronic swiping times of the bus vehicle numbered as Bn in the line Lk at all stops on the dth day as Vn, electronic swiping times of all vehicles in the line Lk at all stops on the dth day as Vd, total cash charge of the bus vehicle numbered as Bn in the line Lk at all stops on the dth day as Mn, total cash charge of all stops in the line Lk on the dth day with Md, and the ticket price of the line Lk as Y yuan, the number of cash charges at all stops in the line Lk on the dth day as Rd, cash times at the stop St of the line Lk on the dth day as Rt, a cashier coefficient at each stop of the line Lk on the dth day as λt, and the number of passengers boarding at the stop St of the line Lk on the dth day as Pt;
• step 1
• firstly acquiring boarding cashier data of buses, including is card, UnionPay online, UnionPay offline, electronic cash and cash data, then acquiring real-time operation shift data, vehicle real-time GPS, actual cycle time and historical operation shift data in a bus dispatching system;
• step 2
• converting the boarding cashier data into passenger flow data of the number of people boarding and dropping off through a passenger flow reverse compensation algorithm;
• step 3
• analyzing the passenger flow data of boarding and dropping off at each stop of each line obtained in step 2 to obtain morning, midday and evening peak time periods, and a passenger flow and a cross-section passenger flow of each line;
• step 4
• analyzing the morning, midday and evening peak time periods, line busy situation and line passenger flow data obtained in step 3 to obtain the morning, midday and evening peak time periods of the each line, corresponding the morning, midday and evening peak time periods to three starting time and ending time respectively, and dividing non-morning, midday and evening peak time periods according to a bus time length θ and obtaining a line number, starting time and ending time of the intelligent scheduling table, so that each line makes departure plans according to different peak time;
• step 5
• obtaining the number of allocated buses, planned departure shifts, planned circle time, upgoing and downgoing in the intelligent scheduling table by a scheduling optimization algorithm through using the historical operation shift data obtained in step 1, the morning, midday and evening peak time periods and the passenger flow of each line obtained in step 3, combined with the line number, starting time and ending time of the intelligent scheduling table obtained in step 4;
• step 6
• obtaining an initial departure interval through a departure interval calculation algorithm through using the line number, the starting time and the ending time obtained in step 4, the number of allocated buses, the planned departure shifts, the planned circle time, the upgoing and downgoing in the intelligent scheduling table obtained in step 5 and the cross-section passenger flow obtained in step 3;
• step 7
• calculating a real-time interval distance Lp of two adjacent vehicles and actual circle time Tg of each bus through using the vehicle real-time GPS and the actual circle time obtained in step 1, wherein when the Lp and the Tg are respectively greater than a maximum allowable threshold or less than a minimum allowable threshold, it is judged that the vehicles drive abnormally, and information of vehicle large interval or traffic jam is judged;
• step 8
• adjusting departure interval time through a real-time departure interval adjustment algorithm by using the real-time operation shift data obtained in step 1, the real-time cross-section passenger flow, the real-time morning, midday and evening peak periods and real-time line passenger flow obtained in step 3, the number of allocated buses, the planned departure shifts, the planned circle time, the upgoing and downgoing in the intelligent scheduling table obtained in step 5, the initial departure interval obtained in step 6 and the information of vehicle large interval or traffic jam obtained in step 7;
• and
• step 9
• uploading the real-time departure interval obtained in step 8 to the bus dispatching system and directing the bus vehicles to dispatch in real time;
• wherein in step 2, the passenger flow reverse compensation algorithm comprises following specific steps:
• A1:
• calculating the electronic swiping times
• W
• t
• =
• ∑
• i
• =
• 1
• SC
• C
• i
• of the bus vehicle numbered as Bn in the line Lk at the stop St on the dth day, the electronic swiping times
• V
• t
• =
• ∑
• n
• =
• 1
• BC
• W
• t
• of all vehicles in the line Lk at the stop St on the dth day, the electronic swiping times
• V
• n
• =
• ∑
• su
• +
• sp
• t
• =
• 1
• W
• t
• of the bus vehicle numbered as Bn in line Lk at all stops on the dth day, the electronic swiping times
• V
• d
• =
• ∑
• n
• =
• 1
• BC
• V
• n
• of all vehicles in the line Lk at all stops on the dth day, the total cash charges Mn of bus vehicle numbered as Bn in the line Lk at all stops on the dth day, total cash charges
• M
• d
• =
• ∑
• n
• =
• 1
• BD
• M
• n
• in the line Lk at all stops on the dth day, the number of cash charges
• R
• d
• =
• ∑
• n
• =
• 1
• BC
• M
• n
• Y
• at all stops in the line Lk on the dth day, and the cashier coefficient
• λ
• t
• =
• V
• t
• V
• d
• =
• ∑
• n
• =
• 1
• BC
• W
• t
• ∑
• n
• =
• 1
• BC
• V
• n
• at each stop in the line Lk on the dth day;
• A2:
• calculating cash times Rt=Rd*λt of the line Lk at the stop St on the dth day and the number of boarding passengers Pt=Vt+Rt of the line Lk at the stop St on the dth day according to the cashier times Rd and cashier coefficient λt;
• in step 5, the scheduling optimization algorithm comprises following specific steps:
• B1:
• using weighted arithmetic mean respectively on the historical cycle time and historical departure shifts of each vehicle within time periods of the starting time and ending time to obtain the historical average cycle time and historical average shifts for upgoing and downgoing of each line according to the historical operation shift data, combined with the line number, starting time and ending time in the intelligent scheduling table obtained in step 4, wherein the historical average cycle time is the planned cycle time and the historical average shifts are the planned departure shifts;
• B2:
• predicting the line passenger flow at the execution time of the scheduling table through a machine learning algorithm LightGBM according to the line passenger flow, calculating the passenger flow of each line at the starting time and the ending time in the result of step 4, respectively calculating running times of each vehicle in each time period=(ending time−starting time)/2*the planned cycle time of the upgoing or the downgoing, and the number of allocated buses in each line at the starting time and ending time=the passenger flow in this time period/(single-ride average passenger capacity*running times of each bus in each time period), wherein the single-ride average passenger capacity is obtained from the passenger flow data of each line;
• in step 6, the departure interval calculation algorithm specifically comprises following specific steps:
• C1:
• actual departure shifts in each time period=passenger flow in this time period/single-ride average passenger capacity;
• C2:
• deviation threshold=actual departure shifts/planned departure shifts in each time period, wherein if the deviation threshold is within the normal range, operation time per shift=(ending time−starting time)/actual departure shifts, otherwise, initial departure interval=(ending time−starting time)/2*(actual departure shifts+planned departure shifts);
• C3:
• predicting the cross-section passenger flow of the at execution time of the scheduling table through the machine learning algorithm LightGBM according to the cross-section passenger flow, with the initial congestion coefficient=the maximum cross-section passenger flow/number of passengers in each time period at the execution time, adjustment time of congestion departure interval=actual departure interval*congestion coefficient*congestion adjustment parameters, wherein the congestion adjustment parameters are set according to the actual situation of buses;
• C4:
• when the congestion coefficient>normal threshold coefficient, initial departure interval time=operation time per shift−adjustment time of congestion departure interval, and when the congestion coefficient<normal threshold coefficient, initial departure interval time=operation time per shift+adjustment time of congestion departure interval, wherein the normal threshold coefficient is set according to the actual situation of buses;
• in step 8, the real-time departure interval adjustment algorithm specifically comprises following steps:
• D1:
• calculating the congestion coefficient=the maximum cross-section passenger flow/number of passengers at the real-time stop, the adjustment time of congestion departure interval=initial departure interval*congestion coefficient*congestion adjustment parameters, where initial values of the congestion adjustment parameters are set according to the actual situation of buses;
• the large interval or traffic jam coefficient=the maximum distance between two buses/the normal distance threshold, the large interval or traffic jam adjustment time=initial departure interval*large interval or traffic jam coefficient*large interval or traffic jam adjustment parameters, wherein initial values of large interval or traffic jam adjustment parameters are set based on the actual situations of buses;
• D2:
• when the congestion coefficient>the normal threshold coefficient, with the normal threshold coefficient set according to the actual situation of buses, adjusted departure interval time=initial departure interval-adjustment time of congestion departure interval;
• when the congestion coefficient<the normal threshold coefficient, the adjusted departure interval time=initial departure interval+adjustment time of congestion departure interval;
• D3:
• the large interval or traffic jam coefficient>the normal threshold coefficient, with the normal threshold coefficient set according to the actual situation of buses, the adjusted departure interval time=initial departure interval-large interval or traffic jam adjustment time;
• when the congestion coefficient<the normal threshold coefficient, the adjusted departure interval time=initial departure interval+large interval or traffic jam adjustment time;
• D4:
• calculating whether the congestion coefficient reaches the normal threshold every 1 minute, and training the congestion adjustment parameters, large interval or traffic jam adjustment parameters by using a deep neural network algorithm until optimal congestion adjustment parameters, large interval or traffic jam adjustment parameters are obtained;
• D5:
• using the congestion adjustment parameters, the large interval or the traffic jam adjustment parameters obtained in D4 to update the congestion adjustment parameters, the large interval or the traffic jam adjustment parameters in D1 to obtain a final real-time departure interval.