This paper identifies peak periods by establishing a profit threshold model, comprehensively considers bus line resource allocation and passenger waiting time, establishes a multi-objective optimization model, optimizes the bus dispatching plan through an artificial immune algorithm, and predicts passenger flow by establishing a gradient boosting tree model. Thus predicting "peak" and "off-peak" periods.
Considering that citizens' travel is not "even", "peak" needs to be characterized as a period when there are more passengers. In order to facilitate processing, we will divide a period of time Δ T \varDelta TThe difference between the total operating revenue and the total operating cost within ΔT is not less than a given "threshold" period. The key to the problem is how to establish the relationship between time, operating income and cost, and the selection of thresholds. This group established a net profit threshold model and introduced the concept of traffic index as the basis for threshold selection, thereby reasonably defining "peak" and "flat peak" ” basis for division. \par
Net profit threshold model
Taking into account the cost of operating a bus, there are many factors, including variable costs over a period of time C ~ \widetilde{C}C
——Fuel consumption, wear factor, etc., and fixed cost C ‾ \overline{C}C——Driver wages and car purchase expenses. Here we consider that in order to distinguish the peak period or the off-peak period, this group first divides the departure time into a number of close intervals, and observes whether the net profit of each divided interval reaches the threshold: Ω = [ T
0 , T 1 ] ∪ [ T 1 , T 2 ] ∪ ⋯ ∪ [ TN − 1 , TN ] \varOmega =\left[ T_0,T_1 \right] \cup \left[ T_1,T_2 \right] \cup \cdots \ cup \left[ T_{N-1},T_N \right]Oh=[T0,T1]∪[T1,T2]∪⋯∪[TN−1,TN]
The length of the divided interval is:Δ T = T n − T n − 1 \varDelta T=T_n-T_{n-1}ΔT=Tn−Tn−1
Book IIThe driving time of flight i is recorded as:[ ti , ti + ttotal ] [t_i ,t_i +t_{total}][ti,ti+ttotal] , so each shift is in the sampling division area[ T n , T n + 1 ] \left[ T_{n},T_{n+1} \right][Tn,Tn+1The cost consumed on ] is cni c_{n}^{i}cni:
c n i = { [ min ( T n + 1 , t i + 1 ) − max ( T n , t i ) ] ⋅ v ⋅ f T n ⩽ t i + 1 a n d T n + 1 ⩾ t i 0 T n > t i + 1 o r T n + 1 < t i c_{n}^{i}=\begin{cases} \left[ \min \left( T_{n+1},t_{i+1} \right) -\max \left( T_n,t_i \right) \right] \cdot v\cdot f\,\, \qquad T_n\leqslant t_{i+1}\,\,and\,\,T_{n+1}\geqslant t_i\,\, \\ 0 \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \quad \,\, T_n>t_{i+1}\,\,or\,\, T_{n+1}<t_i\\ \end{cases} cni={
[min(Tn+1,ti+1)−max(Tn,ti)]⋅v⋅fTn⩽ti+1andTn+1⩾ti0Tn>ti+1orTn+1<ti
Among them, Part IIThe starting time of bus i is ti t_iti:
t i = T 0 + ( i − 1 ) Δ t i − 1 t_i=T_0+\left( i-1 \right) \varDelta t_{i-1} ti=T0+(i−1)Δti−1
Then the sampling division area [ T n , T n + 1 ] \left[ T_{n},T_{n+1} \right][Tn,Tn+1The total variable cost consumed on ]
is: Δ C ~ N = ∑ i = 1 M c N i \varDelta \widetilde{C}_N=\sum_{i=1}^M{c_{N}^{ i}}DC
N=i=1∑McNi
Therefore, the sampling division area [ T n , T n + 1 ] [ T_{n},T_{n+1}][Tn,Tn+1] The total cost CCconsumedC is:
Δ C = Δ C ~ + Δ C ‾ \varDelta C=\varDelta \widetilde{C}+\varDelta \overline{C}Δ C=DC
+DC
Considering that the influencing factor of income is passenger arrival rate, let jjThe arrival rate of passengers at j stations over time isrj r_jrj, assuming rj r_jrjObeying the uniform distribution (person/minute), the total income in the sampling divided area $ \left[ T_{n},T_{n+1} \right]$ is: Δ I = P ⋅ ∑ Δ t Δ t ⋅
rj \varDelta I=P\cdot \sum_{\varDelta t}{\varDelta t \cdot r_j }ΔI _=P⋅Δt∑Δt⋅rj
Model establishment for question 2
Assume that the passenger arrival rate follows a uniform distribution:
rj ∼ U ( tj − 1 , tj ) r_j \sim U\left( t_{j-1},t_j \right)rj∼U(tj−1,tj)
The average waiting time of passengers is:
t ˉ = ti − ti − 1 2 \bar{t}=\frac{t_i-t_{i-1}}{2}tˉ=2ti−ti−1
Book IIThe departure time of vehicle i isT i T_iTi, No. iiCar i arrives (or leaves)jjThe time at station j is: T ij T_i^jTij则:
T i j = { T i + ∑ n = 2 j d n j > 1 T i j = 1 T_{i}^j =\begin{cases} T_i+\sum_{n=2}^j{d_n}\,\,\quad \qquad j>1 \\ T_i\,\, \qquad \qquad \quad \qquad j=1\\ \end{cases} Tij={
Ti+∑n=2jdnj>1Tij=1
Taking into account the customer's time interests and service satisfaction, the waiting timeout rate WT WT is introducedW T represents the convenience of the client:
WT = The total number of people who have waited for more than 15 minutes. The total number of people who boarded the bus during the scheduling period. WT=\frac{\text{The total number of people who have waited for more than} 15\text{minutes}}{\text {Total number of people boarding the bus during the dispatch period}}WT=The total number of people boarding the bus during the dispatch periodThe total number of people waiting for more than 15 minutes
即为:
W T = ∑ T i j ∑ h = 1 h i j { W i j ( h ) ∣ W T i j ( h ) ⩾ 15 } ∑ T i j P i ( T i j ) WT=\frac{\sum\limits_{T_{i}^j}{\sum\limits_{h=1}^{h_{ij}}{\left\{ W_{i}^j \left( h \right) |WT_{ij}\left( h \right) \geqslant 15 \right\}}}}{\sum\limits_{T_{i}^j}{P_i\left( T_{i}^j \right)}} WT=Tij∑Pi(Tij)Tij∑h=1∑hij{
Wij(h)∣WTij(h)⩾15}
In the formula W ij ( h ) W_{i}^j (h)Wij( h ) PerformanceiiCar i arrives atjjthWhen station j , hhhas already been waited for on this station.h bus, but the number of people who have not yet boarded the bus, assuming that the waiting time of these people is:
WT ij ( h ) = T ij − T i − hj , W ij ( 0 ) = U ij WT_{i}^j\left ( h \right) =T_{i}^j-T_{ih}^j,W_{i}^j \left( 0 \right) =U_{ij}WTij(h)=Tij−Ti−hj,Wij(0)=Uij
In the formula U ij U_{i}^jUijExpressed in T i − 1 j T_{i-1}^jTi−1j到 T i j T_{i}^j TijDuring the time period, the number of new people waiting for the bus at the station:
WW ij ( 0 ) = ∫ T i − 1 j T ijui ( t ) dt WW_{i}^{j}\left( 0 \right) =\int_{ T_{i-1}^{j}}^{T_{i}^{j}}{u_i\left( t \right) dt}WWij(0)=∫Ti−1jTijui(t)dt
设 h i j h_{ij} hijrepresents the kthk cars arriveat jjAt station j , the number of waiting vehicles for the longest waiting passengers at that station is theniiCar i leavesjjThe number of passengers on the bus at station j
can be expressed as: P i ( T ij ) = { aij + ∑ r = 0 hij W ij ( r ) , hij ∗ = 0 , M , hij ∗ > 0. P_i\left( T_ {i}^{j} \right) =\begin{cases} a_{ij}+\sum_{r=0}^{h_{ij}}{W_{i}^{j}}\left( r \ right) ,& \qquad h_{ij}^{*}=0,\\ M,& \qquad h_{ij}^{*}>0.\\ \end{cases}Pi(Tij)={
aij+∑r=0hijWij(r),M,hij∗=0,hij∗>0.
inside aij a_{ij}aijPart IICar i arrives atjjthAfter station j , after the passengers get off the bus, the number of remaining passengers on the bus:
aij = max { ( P k ( T ij − 1 ) − D j ) , 0 } a_{ij}=\max \left\{ \ left( P_k\left( T_{i}^{j-1} \right) -D_j \right) \text{,}0 \right\}aij=max{
(Pk(Tij−1)−Dj), 0 }
whereD j D_jDjFor in jjThe number of passengers getting off at stop j .
AssumeMMM is the maximum number of people the vehicle can carry, then the number of people that can be accommodated on the vehicle at this time is recorded as:
bij = M − aij b_{ij}=M-a_{ij}bij=M−aij
Based on the queuing principle of first come first boarding, at k + 1 k + 1k+1 car arrives atjjWhen station j is reached, the maximum number of waiting times for vehicles hij ∗ h_{ij}^{*}hij∗:
h i j ∗ = max { h ∣ ∑ r = 0 h i j W i j ( r ) ⩽ b i j } , h_{ij}^{*}=\max \left\{ h|\sum_{r=0}^{h_{ij}}{W_{ij}\left( r \right)}\leqslant b_{ij} \right\} , hij∗=max⎩
⎨
⎧h∣r=0∑hijWij(r)⩽bij⎭
⎬
⎫,