Multi-purpose optimized elevator intelligent scheduling method and system

Abstract

The invention discloses a multi-purpose optimized elevator intelligent scheduling method and system, and the method comprises the steps: achieving the prediction of an elevator passenger flow direction mode through the construction of a flow direction transfer tensor and an improved density peak clustering algorithm, and adjusting a clustering result when a passenger behavior suddenly changes; a direction inertia factor is introduced, a direction switching decision rule and a local direction stability adjusting mechanism are combined, and the elevator running direction is dynamically optimized; passenger priority factors are defined, comprehensive priority scores of passengers are calculated, a passenger gradient optimization scheduling model is constructed based on the comprehensive priority scores, and the priorities of the passengers are dynamically adjusted; a multi-elevator non-cooperative game model is constructed, and task allocation among multiple elevators is realized by defining an elevator revenue function and a strategy space and utilizing Nash equilibrium solution. According to the method, the scheduling strategy can be adaptively adjusted in a complex traffic mode, so that the response speed is higher, the resource utilization rate is better, and the passenger waiting time is shorter.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator dispatching, and in particular to a multi-purpose optimized elevator intelligent dispatching method and system. Background Art

[0002] In modern high-rise buildings, the elevator system, as a core transportation hub, carries the vertical transportation needs of a large number of people. However, due to the dynamic changes in passenger flow and the unpredictability of individual travel patterns, efficient elevator scheduling has become a very challenging problem. Especially in crowded building environments such as office buildings, shopping malls, and residential buildings, elevator scheduling not only needs to meet basic transportation needs, but also needs to achieve optimal scheduling decisions when passenger demand is highly concentrated and directional, so as to reduce the average waiting time of passengers, reduce system energy consumption, and improve overall operating efficiency.

[0003] Passenger flow in a building may show significant dynamic characteristics at different time periods. For example, during rush hour, a large number of passengers may request elevators from certain specific floors in a short period of time, while during non-peak hours, passengers' travel patterns are more dispersed and the demand for going up and down stairs changes randomly. In addition, different types of buildings have different load patterns on the elevator system. For example, in commercial buildings, passengers mainly move from lower floors (such as the lobby) to higher floors in the morning, while the opposite is true during lunch time or after get off work, resulting in a significant change in the direction of passenger flow. This dynamic characteristic of passenger flow requires the elevator scheduling algorithm to adapt to the changing demand patterns at different time periods to avoid system congestion or resource waste due to improper strategies.

[0004] Existing elevator dispatching algorithms, such as the directional optimization dispatching method based on genetic algorithms, can improve dispatching efficiency to a certain extent, but they often have difficulty in quickly adapting to changing traffic demands when faced with complex passenger flow patterns. For example, genetic algorithms usually rely on preset fitness functions and directional optimization strategies, but in the case of sudden changes in dynamic directional demands, their search efficiency may be greatly reduced, and it may even be difficult to find the optimal dispatching solution in a short period of time. At the same time, although traditional rule-based elevator dispatching methods (such as first-come-first-served, group control partition dispatching, etc.) are simple to calculate, they often cannot take into account the global optimality, which can easily lead to uneven use of elevator resources, thereby increasing passenger waiting time and system energy consumption.

[0005] Therefore, in order to effectively respond to the changing passenger flow direction demands in high-rise buildings, it is necessary to design an intelligent elevator dispatching method that can adaptively adjust the dispatching strategy under complex traffic patterns to meet the ever-changing elevator transportation needs with higher response speed, better resource utilization and shorter passenger waiting time. Summary of the invention

[0006] In order to solve the above problems, the present invention provides a multi-purpose optimized elevator intelligent scheduling method and system.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0008] On the one hand, the present invention discloses a multi-purpose optimized elevator intelligent dispatching method, comprising:

[0009] Step 1: By constructing the flow transfer tensor and the improved density peak clustering algorithm, combined with the adaptive update mechanism and time series prediction, the flow pattern of elevator passengers can be predicted, and the clustering results can be adjusted when the passenger behavior changes suddenly;

[0010] Step 2: Introduce the directional inertia factor, combine the direction switching decision rule and the local directional stability adjustment mechanism, dynamically optimize the elevator running direction, and adapt to the fluctuation of passenger flow in different time periods through adaptive adjustment of the directional inertia factor;

[0011] Step 3: Define the passenger priority factor, calculate the comprehensive priority score of the passenger, and build a passenger gradient optimization scheduling model based on it to dynamically adjust the passenger priority;

[0012] Step 4: Construct a multi-elevator non-cooperative game model, define the elevator profit function and strategy space, and use Nash equilibrium solution to achieve task allocation among multiple elevators.

[0013] Further: Step 1 comprises:

[0014] Collect passenger call requests, elevator operation status and historical traffic data to form a high-dimensional data set, and organize it into a time series sample set through a sliding time window mechanism to extract short-term dynamic change characteristics;

[0015] A third-order flow transfer tensor is constructed based on historical data, where the first dimension represents the starting floor, the second dimension represents the target floor, the third dimension represents the time step, and the tensor elements represent the passenger flow;

[0016] The improved density peak clustering algorithm is used to extract the flow pattern, including calculating the local density, the minimum distance and selecting the cluster center to identify the main passenger flow pattern;

[0017] An adaptive update mechanism is introduced to detect sudden changes in passenger flow by calculating anomaly indicators, and future flow patterns are predicted based on time series modeling.

[0018] Further: Step 2 comprises:

[0019] Considering the target floor weight of the currently loaded passengers, the demand weight of the waiting passengers, the current number of passengers on the elevator and the number of waiting passengers, as well as the running direction of the elevator, the directional inertia factor is defined, and the strength of the direction maintenance is adjusted through the inertia control parameter, thereby quantifying the tendency of the elevator to maintain the current running direction;

[0020] According to the directional inertia factor, a decision rule for direction switching is established. When the directional inertia factor is lower than the preset threshold, the current directional load factor is lower than the load threshold, or the demand of high-priority passengers in the opposite direction increases, the elevator is allowed to switch direction; otherwise, the elevator maintains the current running direction to reduce direction reversal, energy consumption and time loss;

[0021] A local directional stability adjustment mechanism is introduced to determine whether the directional distribution is unbalanced by calculating the global directional imbalance index. If unbalanced, the directional inertia factors of some elevators are forcibly adjusted to balance the number of up and down elevators and prevent all elevators from running in the same direction.

[0022] Further: Step 3 includes:

[0023] Considering the waiting time factor, target floor factor, direction consistency factor and group effect factor, the passenger priority score is defined. The waiting time factor measures the impact of the passenger's waiting time, the target floor factor considers the impact of the target floor on transportation efficiency, the direction consistency factor reflects the matching degree between the passenger's request direction and the elevator's running direction, and the group effect factor optimizes the request response for the same target floor.

[0024] Based on the calculated priority scores, passenger requests are sorted and optimally allocated according to the current elevator status. The optimal elevator is allocated to the passenger by solving the optimization function, taking into account passenger priority, additional elevator stop time, and load balancing, ensuring that high-priority passengers are picked up first and optimizing the elevator stop order.

[0025] A priority adjustment strategy is designed to address the dynamic changes in passenger waiting time. If a passenger’s waiting time exceeds the threshold, their priority score will be forcibly increased. During peak hours, a buffer factor is introduced to prevent elevator resources from being over-occupied by high-priority passengers, ensuring that low-priority passengers still have the opportunity to be picked up when the elevator is close to full, thereby balancing fairness and overall transportation efficiency.

[0026] Further: Step 4 includes:

[0027] Each elevator is defined as a game participant, and its state vector includes the current floor, running direction, number of passengers and load rate; the passenger request state is composed of the target floor, direction and priority score; the elevator revenue function takes into account the passenger priority score, load balancing, additional stop time and energy consumption, and quantifies the elevator scheduling revenue through the revenue function;

[0028] The initial allocation is based on the priority of the passengers and the principle of the nearest available elevator. Each elevator calculates its own profit function. If it is found that the current task allocation leads to a decrease in profit, it requests to adjust the allocation. Through multiple rounds of profit calculation and strategy update, the Nash equilibrium is solved, so that each elevator has no intention to adjust the task allocation individually under the equilibrium strategy, thereby achieving collaborative optimization among multiple elevators.

[0029] A load balancing correction factor is introduced to prevent some elevators from being overloaded. When the elevator load or priority is unbalanced, the task reallocation mechanism is triggered and the optimal strategy is recalculated to maximize the global benefit. During peak hours, a stratified diversion strategy is adopted to prioritize high-rise passengers to specific elevators, reduce the number of stops, further improve the overall scheduling efficiency of the system and reduce energy consumption.

[0030] On the other hand, the present invention discloses a multi-purpose optimized elevator intelligent dispatching system, comprising:

[0031] Dynamic prediction and clustering analysis module: By constructing the flow transfer tensor and the improved density peak clustering algorithm, combined with the adaptive update mechanism and time series prediction, it can predict the flow pattern of elevator passengers and adjust the clustering results when the passenger behavior changes suddenly;

[0032] Self-balancing scheduling strategy module: introduces directional inertia factor, combines direction switching decision rules and local directional stability adjustment mechanism, dynamically optimizes the elevator running direction, and adapts to passenger flow fluctuations in different time periods through adaptive adjustment of directional inertia factor;

[0033] Passenger priority gradient optimization scheduling module: defines passenger priority factors, calculates the comprehensive priority scores of passengers, and builds a passenger gradient optimization scheduling model based on this to dynamically adjust passenger priorities;

[0034] Multi-elevator collaborative scheduling module: Construct a multi-elevator non-cooperative game model, define the elevator profit function and strategy space, and use Nash equilibrium solution to achieve task allocation among multiple elevators.

[0035] Compared with the prior art, the present invention has the following technical advances:

[0036] The demand prediction algorithm based on adaptive flow clustering of the present invention can adaptively analyze the passenger flow patterns during peak hours and non-peak hours through real-time data-driven high-dimensional feature extraction, combined with time series modeling and dynamic flow clustering mechanism, and predict the flow direction in the future time period in advance. This capability enables the scheduling algorithm to make optimal scheduling adjustments before passenger requests are generated, thereby reducing the elevator resource mismatch problem caused by sudden changes in passenger flow direction and improving the system's forward-looking scheduling ability. The self-balancing scheduling algorithm driven by directional inertia introduces directional inertia factors and passenger flow trend analysis to ensure that the elevator can meet the current passenger needs while minimizing unnecessary direction switching, reducing the number of additional stops and scheduling delays caused by direction changes. At the same time, the present invention combines the load balancing mechanism to dynamically adjust the running direction of each elevator, so that the system can still maintain stable and efficient operation during peak hours, avoiding the problem of reduced efficiency caused by sudden changes in passenger flow direction.

[0037] The passenger priority gradient optimization algorithm dynamically assigns priority to each passenger through multi-dimensional modeling (including waiting time, distance to destination, group characteristics, historical behavior patterns, etc.), and introduces a gradient optimization mechanism to ensure that the elevator gives priority to high-priority passengers. This mechanism not only reduces the waiting time of emergency passengers or high-rise passengers, but also avoids the imbalanced allocation of elevator resources that may be caused by the traditional "first come, first served" strategy, improving the fairness and service quality of the overall system.

[0038] The multi-elevator collaborative scheduling algorithm based on game equilibrium uses non-cooperative game modeling to treat multiple elevators as intelligent game entities, so that the elevators can adaptively adjust task allocation through revenue optimization strategies, thereby achieving global optimal collaborative scheduling. Compared with the traditional group control method based on fixed rules, this algorithm can dynamically balance the elevator load during peak hours, preventing some elevators from being overloaded due to too many high-priority tasks while other elevators are running empty, ultimately reducing the overall system energy consumption and improving elevator utilization.

[0039] The present invention adopts a parallel computing architecture and combines it with a deep learning model to accelerate computing, so that the algorithm can process large-scale passenger data and optimize decisions at the millisecond level, ensuring efficient real-time scheduling capabilities under complex traffic patterns. At the same time, the present invention can adapt to different passenger flow characteristics in different building environments (office buildings, shopping malls, residential buildings, etc.), making the elevator dispatching system more versatile and scalable.

[0040] In summary, the present invention solves the problems of low efficiency, frequent direction switching, uneven task distribution, and excessive computational overhead of traditional elevator scheduling algorithms in complex passenger flow environments through a multi-level collaborative scheduling strategy of demand forecasting, directional inertia optimization, passenger priority modeling, and game equilibrium optimization. It can realize intelligent, efficient, and low-energy multi-purpose optimization scheduling in complex high-rise building elevator systems, ensuring adaptive dynamic optimization under different passenger flow modes, thereby providing a smoother and more comfortable elevator riding experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0042] In the attached picture:

[0043] Figure 1 is a flow chart of the present invention;

[0044] Figure 2 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0045] The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0046] Example 1

[0047] like Figure 1 As shown, the present invention discloses a multi-purpose optimized elevator intelligent dispatching method, comprising the following steps:

[0048] Specifically, step 1 includes:

[0049] 1.1 Data collection and preprocessing

[0050] First, a high-dimensional dataset containing passenger call requests, elevator operation status, and historical flow data is constructed, where the data dimensions include but are not limited to:

[0051] Time characteristics (time t, day of the week w, weekday / holiday identifier h)

[0052] Floor characteristics (starting floor f s 、Destination floor f d )

[0053] Elevator status (current floor f c , current direction d c 、Number of passengers p )

[0054] Passenger request (request time t r , waiting time t w )

[0055] The data set is organized by a sliding time window mechanism, that is, the data is divided according to the time window ΔT (for example, 5 minutes) to form a time series sample set X = {X 1 ,X 2 ,...,X n} to ensure that the model can effectively extract short-term dynamic change features.

[0056] 1.2 Constructing the flow transfer tensor

[0057] Based on historical data, construct the third-order flow transfer tensor T(f s ,f d ,t), where:

[0058] The first dimension represents the starting floor f s (i.e. the floor from which the passenger enters the elevator);

[0059] The second dimension represents the target floor f d (i.e. the floor the passenger expects to arrive at);

[0060] The third dimension represents the time step t (i.e. the time window to which the data belongs).

[0061] Tensor element T i,j,k Indicates that at time t k Inside, from floor f i To floor f j The passenger flow is:

[0062]

[0063] Among them, N k represents the time window t k The number of all passenger requests in is an indicator function that takes the value 1 when the condition is met and 0 otherwise.

[0064] This flow direction transfer tensor provides the elevator flow direction distribution within each time window, providing basic data for subsequent clustering analysis.

[0065] 1.3 Dynamic Density Peak Clustering

[0066] Since the passenger flow distribution in the elevator system has the characteristic of dynamic cluster center change, it is necessary to use an improved density peak clustering algorithm to extract the flow pattern.

[0067] 1.3.1 Calculation of local density

[0068] Define floor flow point P i=(f s ,f d ) local density ρ i for:

[0069]

[0070] Among them, d ij Indicates the flow direction point P i and P j The Euclidean distance between them, K(·) is the kernel function, d c It is the density radius threshold, which is usually adjusted dynamically through statistical methods to adapt it to the flow direction distribution in different time windows.

[0071] 1.3.2 Calculating the minimum distance

[0072] Define the minimum distance δ from each flow point to a higher density point i :

[0073]

[0074] If i is the largest among all points, then let δ i =max j d ij .

[0075] 1.3.3 Selecting cluster centers

[0076] Considering the density ρ and the minimum distance δ, select the points that meet the following conditions as the cluster center C k :

[0077] C k = argmax(ρ i .δ i )

[0078] Each cluster center represents a major passenger flow pattern.

[0079] 1.4 Adaptive Update Mechanism

[0080] Since passenger flow has dynamic characteristics, it is necessary to introduce an adaptive update mechanism to adjust the clustering results when passenger behavior changes suddenly.

[0081] 1.4.1 Calculation of abnormality

[0082] Define the abnormality index A t Measure the degree of change in flow direction between the current time window t and the past m windows:

[0083]

[0084] When A tWhen it exceeds the preset threshold τ, it indicates that the passenger flow has changed suddenly and needs to be re-clustered.

[0085] 1.4.2 Adaptive Clustering Adjustment

[0086] If A t >τ, the following update strategy is executed:

[0087] 1. Add a new cluster center: If the local density ρ of the new flow point is higher than the mean of the currently known cluster centers, add the point as a new center.

[0088] 2. Merge similar clusters: calculate the current cluster center C k If the similarity is higher than the threshold, they are merged.

[0089] 3. Adjust the density radius d c : If the new flow points are too dispersed, increase d c , otherwise reduce d c .

[0090] 1.5 Future time window flow prediction

[0091] To predict the flow pattern in the future time window t+1, time series modeling is used:

[0092]

[0093] Among them, α is an adaptive weighting parameter, which is adjusted according to the stationarity of historical data.

[0094] The final predicted flow pattern will serve as the input for elevator dispatching decisions and guide the optimization of subsequent dispatching strategies.

[0095] This step achieves accurate prediction of elevator passenger flow patterns through technical means such as flow transfer tensor construction, density peak clustering, adaptive update and time series prediction, and can make rapid adjustments when the flow direction changes suddenly, providing key input data for the subsequent directional inertia-driven self-balancing scheduling algorithm.

[0096] Specifically, step 2 includes:

[0097] 2.1 Definition and calculation of directional inertia factor

[0098] Since the frequent changes of direction of the elevator during operation may cause additional energy consumption and time loss, it is necessary to introduce a directional inertia factor to measure the tendency of the elevator to maintain the current running direction. Suppose the directional inertia factor I of the kth elevator in the time window t is k (t) is:

[0099]

[0100] in:

[0101] W i is the target floor weight of the currently loaded passengers, d i is the target direction of passenger i (up / down);

[0102] U j is the demand weight of waiting passengers, d j the passenger's desired direction;

[0103] n k is the number of passengers currently carried by the elevator, m k is the number of passengers waiting for the elevator;

[0104] d k The current running direction of the elevator (up / down);

[0105] λ 1 ,λ 2 It is the inertia control parameter, which controls the strength of direction maintenance.

[0106] If I k (t)>0, then keep the current running direction; if I k (t)≤0, direction adjustment is allowed.

[0107] 2.2 Inertia-driven direction decision mechanism

[0108] After the directional inertia factor is calculated, it is necessary to establish a decision rule for direction switching to avoid unnecessary direction reversal of the elevator due to local flow changes.

[0109] 2.2.1 Direction switching decision conditions

[0110] Assume the current elevator running direction is d k , direction switching is allowed when one of the following conditions is met:

[0111] 1. The directional inertia factor is lower than the directional flip threshold θ d :

[0112] I k (t)≤θ d

[0113] Among them, θ d Dynamically adjusted by historical flow patterns.

[0114] 2. Current direction load factor L k Below the load threshold θ L :

[0115]

[0116] Among them, Ck is the maximum capacity of elevator k.

[0117] 3. The demand for high-priority passengers in the opposite direction increases significantly, namely:

[0118]

[0119] Among them, α is the direction adjustment weight parameter.

[0120] If none of the above conditions are met, the elevator maintains the current running direction to reduce unnecessary direction reversal.

[0121] 2.3 Self-balancing mechanism: local directional stability adjustment

[0122] In order to further improve the stability of the elevator system and prevent all elevators from running in the same direction, causing passengers in the opposite direction to wait for a long time, a local directional stability adjustment mechanism is proposed.

[0123] 2.3.1 Local directional balance calculation

[0124] Assume that the entire elevator system consists of M elevators and define the global directional imbalance index D imb for:

[0125]

[0126] If |D imb (t)|>θ imb (Directional imbalance threshold), the directional inertia factors of some elevators are forced to be adjusted to balance the number of elevators going up and down, that is:

[0127]

[0128] Among them, β is the local direction adjustment strength, which ensures that the overall direction of the system tends to be balanced.

[0129] 2.4 Directional inertia dynamic adaptive adjustment

[0130] Since the passenger flow direction fluctuates in different time periods, it is necessary to design a dynamic adaptive adjustment mechanism of the directional inertia factor to adapt to different flow patterns.

[0131] 2.4.1 Directional inertia weight self-adjustment

[0132] During peak hours, passenger flow has a strong unidirectional trend, so it is necessary to increase the inertia factor λ 1 To maintain the direction; during off-peak hours, the passenger flow is more uniform, so the inertia factor λ needs to be reduced 1 To increase the flexibility of direction adjustment, the specific adjustment methods are as follows:

[0133] λ 1(t+1)=λ 1 (t)+γ·(D imb (t)-D imb (t-1))

[0134] Among them, γ is the adaptive adjustment rate.

[0135] 2.4.2 Adaptive adjustment of directional inertia threshold

[0136] Direction flip threshold θ d It needs to be adjusted adaptively according to the flow mode. The adjustment method is:

[0137]

[0138] Among them, η is the adjustment proportional coefficient, which ensures that when the directional imbalance is serious, the flexibility of directional flipping is improved, and when the directional balance is maintained at a higher directional inertia.

[0139] This step realizes intelligent scheduling of the elevator direction through directional inertia factor calculation, direction switching decision, local directional self-balancing mechanism and dynamic adaptive adjustment of directional inertia, so that the elevator can maintain a stable and efficient operation state when the passenger flow changes drastically, and dynamically balance the directional distribution of the elevator on a global scale, thereby reducing unnecessary direction reversals and improving overall transportation efficiency.

[0140] Specifically, step 3 includes:

[0141] After the demand prediction algorithm based on adaptive flow clustering calculates the passenger flow pattern and the directional inertia-driven self-balancing scheduling algorithm determines the elevator running direction, there is still the problem that different passengers have different urgency levels for elevator services. For example, passengers who have been waiting for a long time during peak hours should be picked up by the elevator first, while passengers on lower floors may affect the overall scheduling efficiency due to high-frequency requests. Therefore, it is necessary to design a passenger priority gradient optimization algorithm to ensure the overall elevator transportation efficiency while taking into account the fairness of different passengers.

[0142] 3.1 Definition and calculation of passenger priority factors

[0143] In the current time window t, for any passenger request p i , define its comprehensive priority score P i (t), which is determined by multiple factors including waiting time, target floor, direction consistency, group effect, etc.:

[0144] P i (t) = ω 1 ·W i +ω 2 ·D i +ω 3 ·Ai +ω 4 ·G i

[0145] in:

[0146] W i : Waiting time factor, indicating passenger p i The waiting time affects the weight;

[0147] D i : Target floor factor, considering passenger p i The impact of the destination floor on system efficiency;

[0148] A i : Direction consistency factor, which reflects the matching degree between the passenger's requested direction and the current elevator running direction;

[0149] G i : Group effect factor, considering the overall optimization effect when there are multiple requests for the same target floor;

[0150] ω 1 ,ω 2 ,ω 3 ,ω 4 is the dynamically adjusted weight parameter.

[0151] 3.1.1 Waiting time factor W i

[0152] The waiting time factor is used to measure the waiting time of passengers in the current time window t. The impact on priority is calculated as:

[0153]

[0154] Among them, max(T w ) is the maximum waiting time of all current passenger requests, ensuring that passengers with longer waiting times have higher priority.

[0155] 3.1.2 Target floor factor D i

[0156] The destination floor factor is used to measure whether the passenger's destination floor affects the overall transportation efficiency and is defined as:

[0157]

[0158] in, For passengers i The target floor, f c is the floor where the current elevator is located, and H is the total number of floors in the building. Passengers whose target floors are farther away have a higher priority to reduce the number of stops along the way.

[0159] 3.1.3 Directional consistency factor A i

[0160] The direction consistency factor is used to measure the passenger's requested direction d i Is it consistent with the current elevator running direction? k Consistent, defined as:

[0161]

[0162] Among them, α is the penalty coefficient for direction mismatch, which is usually set to 0<α<1 to ensure that passenger requests with the same direction are given priority.

[0163] 3.1.4 Group effect factor G i

[0164] If multiple passengers make requests in similar directions on the same floor, they should be responded to first to reduce the overall number of stops and improve system throughput. The group effect factor is defined as:

[0165]

[0166] in, Indicates the target floor The number of requests, max(N f ) is the number of passengers who requested the most target floors in the current time window.

[0167] 3.2 Passenger Priority Gradient Scheduling Strategy

[0168] Based on the calculated passenger priority scores, a passenger gradient optimization scheduling model is constructed to optimize the overall elevator stop order while ensuring that high-priority passengers are picked up as quickly as possible.

[0169] 3.2.1 Passenger Priority Sorting

[0170] For all passenger requests within the current time window, P = {P 1 ,P 2 ,...,P N}, sorted by priority score from high to low:

[0171] P sorted =sort(P,descending)

[0172] Among them, passengers with high priority scores will be assigned to the appropriate elevator first.

[0173] 3.2.2 Passenger dispatch and allocation

[0174] For the sorted passenger requests P sorted, according to the current state of the elevator, the optimal allocation is performed. Assume that the system contains M elevators, and each elevator k has a current state vector:

[0175]

[0176] in, is the current floor of the elevator, d k is the current running direction, C k is the current number of passengers, L k is the load factor.

[0177] Define the objective optimization function J k Measuring passenger p i The dispatch benefit after being assigned to elevator k is:

[0178] J k =λ 1 ·P i -λ 2 ·Δt k -λ 3 ·C k

[0179] in:

[0180] λ 1 Controlling the impact of passenger priority;

[0181] λ 2 Control penalties for extra elevator stop time;

[0182] λ 3 Control load balancing.

[0183] By solving:

[0184]

[0185] Select the optimal elevator k * To pick up passengers i .

[0186] 3.3 Dynamic adjustment mechanism of passenger priority

[0187] As passengers' waiting time continues to increase, passenger priority scores must be dynamically adjusted to prevent some passengers from not being picked up for a long time due to low priority.

[0188] 3.3.1 Dynamic adjustment of waiting time

[0189] If a passenger p i The waiting time exceeds the set threshold T max , then its priority score is forced to increase:

[0190] P i(t+1)=P i (t)+δW i

[0191] Among them, δ is the priority compensation coefficient, which ensures that passengers who wait for a long time will not be continuously ignored.

[0192] 3.3.2 Priority Buffer Mechanism

[0193] During peak hours, in order to avoid all elevator resources being occupied by high-priority passengers in a short period of time, a buffer factor β is set:

[0194] P′ i =P i ·(1-β·C k )

[0195] Among them, C k is the load ratio of the current elevator, and β controls the impact of the load on the priority, so that low-priority passengers still have the opportunity to be picked up when the elevator is close to full load.

[0196] This step ensures that the elevator system can improve the overall transportation efficiency while taking into account fairness and reduce the long waiting time of passengers due to low priority through passenger priority factor calculation, priority gradient sorting, optimal elevator allocation and dynamic adjustment mechanism, laying the foundation for the final energy consumption optimization and multi-objective scheduling.

[0197] Specifically, step 4 includes:

[0198] After the passenger priority gradient optimization algorithm calculates the priority of each passenger and determines the running direction of a single elevator in combination with the directional inertia driven self-balancing scheduling algorithm, the elevator system still faces the challenge of coordinated scheduling of multiple elevators, that is, how to reasonably allocate tasks among multiple elevators when passenger requests are concentrated to minimize passenger waiting time and energy consumption while avoiding inefficient competition for elevator resources. Therefore, this step proposes a multi-elevator coordinated scheduling algorithm based on game equilibrium, constructs a non-cooperative game model between multiple elevators, and solves the equilibrium solution to achieve the optimal state of overall elevator scheduling.

[0199] 4.1 Construction of multi-elevator dispatching game model

[0200] Assume that there are M elevators and N unassigned passenger requests in the system. Define each elevator E k As a participant in the game, its goal is to maximize scheduling benefits while ensuring its own load balance.

[0201] 4.1.1 Elevator status definition

[0202] Each elevator E k The state vector S kIt consists of the following four key factors:

[0203]

[0204] in:

[0205] Current floor;

[0206] d k : Current running direction;

[0207] C k : Current number of passengers;

[0208] L k : Current load rate, defined as Among them C max It is the maximum load capacity of the elevator.

[0209] 4.1.2 Passenger Request Status Definition

[0210] For any unassigned passenger request p i , whose status consists of the target floor, direction, and priority:

[0211]

[0212] in:

[0213] The passenger's destination floor;

[0214] d i : Passenger requests directions;

[0215] P i : The passenger’s priority score (calculated by the passenger priority gradient optimization algorithm).

[0216] 4.1.3 Definition of Elevator Revenue Function

[0217] Each elevator E k The profit function J k It is determined by the passenger priority score, elevator load balance, elevator extra stop time, and elevator energy consumption, and is defined as follows:

[0218]

[0219] in:

[0220] λ 1 ,λ 2 ,λ 3 ,λ 4 is the weight parameter;

[0221] P kis the set of assigned passengers for the current elevator k;

[0222] Δt k,i Pick up passenger P for elevator k i Additional stop time required;

[0223] L k For the current elevator load rate, ensure the task load balance between different elevators;

[0224] E k It is the current energy consumption of the elevator (which can be calculated based on the floor moving distance and elevator acceleration).

[0225] 4.2 Game Solution for Multi-Elevator Task Allocation

[0226] 4.2.1 Policy Space Definition

[0227] Each elevator E k The optional strategy set A k The request set R of all unassigned passengers is R = {R 1 ,R 2 ,...,R N},Right now:

[0228] A k = {R i |R i Not yet assigned}

[0229] During the dispatching process, each elevator needs to autonomously select the optimal passenger task to maximize its own revenue. k .

[0230] 4.2.2 Non-cooperative game between elevators

[0231] Define the strategy optimization problem between elevators as a non-cooperative game, where each elevator E k As a rational player, his strategy A k The choice affects the revenue of other elevators.

[0232] The game process is as follows:

[0233] 1. Initial allocation: According to the priority sorting of the passenger priority gradient optimization algorithm, the highest priority requests are initially allocated according to the principle of the nearest available elevator;

[0234] 2. Revenue calculation: Each elevator calculates its own revenue function J_k;

[0235] 3. Strategy update: If an elevator finds that its current task allocation leads to reduced revenue, it will request other elevators to adjust the allocation;

[0236] 4. Equilibrium solution: Through multiple rounds of profit calculation and strategy adjustment, a Nash equilibrium solution is found, that is, a state in which all elevators are unwilling to adjust task allocation.

[0237] Nash equilibrium conditions:

[0238]

[0239] That is: for each elevator, if in the equilibrium strategy Under this condition, any single change in its own strategy will not increase the profit, and the system reaches a stable state.

[0240] 4.3 Multi-elevator collaborative optimization mechanism

[0241] 4.3.1 Load Balancing Mechanism

[0242] To prevent some elevators from being overloaded due to high priority tasks, a load balancing correction factor γ is defined k :

[0243] J′ k =J k -γ k L k

[0244] in:

[0245] γ k =α(L k -L avg )

[0246] L avg is the average load factor of all elevators;

[0247] α is a dynamically adjusted parameter to ensure that elevators with loads exceeding the mean have fewer tasks assigned to them.

[0248] 4.3.2 Task redistribution mechanism

[0249] When an elevator reaches its load limit or the priority balance is unbalanced, the task reallocation mechanism is triggered and the optimal strategy is recalculated to maximize the global benefit.

[0250] 4.3.3 Peak congestion optimization mechanism

[0251] During peak hours, if the load of all elevators is higher than the threshold L max , a stratified diversion strategy is adopted to assign high-level passengers to specific elevators first to reduce the number of stops and improve transportation efficiency.

[0252] This step designs a multi-elevator collaborative scheduling algorithm based on game equilibrium theory. By constructing a non-cooperative game model, global optimization is performed in multiple dimensions such as profit function optimization, load balancing, and task redistribution. This allows the elevator system to efficiently respond to passenger needs while reducing unnecessary resource competition, ultimately improving overall scheduling efficiency and reducing energy consumption.

[0253] Example 2

[0254] Dynamic prediction and clustering analysis module: By constructing the flow transfer tensor and the improved density peak clustering algorithm, combined with the adaptive update mechanism and time series prediction, it can predict the flow pattern of elevator passengers and adjust the clustering results when the passenger behavior changes suddenly;

[0255] Self-balancing scheduling strategy module: introduces directional inertia factor, combines direction switching decision rules and local directional stability adjustment mechanism, dynamically optimizes the elevator running direction, and adapts to passenger flow fluctuations in different time periods through adaptive adjustment of directional inertia factor;

[0256] Passenger priority gradient optimization scheduling module: defines passenger priority factors, calculates the comprehensive priority scores of passengers, and builds a passenger gradient optimization scheduling model based on this to dynamically adjust passenger priorities;

[0257] Multi-elevator collaborative scheduling module: Construct a multi-elevator non-cooperative game model, define the elevator profit function and strategy space, and use Nash equilibrium solution to achieve task allocation among multiple elevators.

[0258] The module in the embodiment 2 is used to implement the function in the embodiment 1. This embodiment can be implemented by a system, the system includes a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the multi-purpose optimized elevator intelligent scheduling method according to the embodiment 1 of the present application is implemented. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose settings and functions are known in the art, so they are not repeated here.

[0259] In the present application, the aforementioned memory may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, system, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in this application may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0260] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of protection of the claims of the present invention.

Claims

1. A multi-purpose optimized elevator intelligent dispatching method, characterized in that: include: Step 1: By constructing the flow transfer tensor and the improved density peak clustering algorithm, combined with the adaptive update mechanism and time series prediction, the flow pattern of elevator passengers can be predicted, and the clustering results can be adjusted when the passenger behavior changes suddenly; Step 2: Introduce the directional inertia factor, combine the direction switching decision rule and the local directional stability adjustment mechanism, dynamically optimize the elevator running direction, and adapt to the fluctuation of passenger flow in different time periods through adaptive adjustment of the directional inertia factor; Step 3: Define the passenger priority factor, calculate the comprehensive priority score of the passenger, and build a passenger gradient optimization scheduling model based on it to dynamically adjust the passenger priority; Step 4: Construct a multi-elevator non-cooperative game model, define the elevator profit function and strategy space, and use Nash equilibrium solution to achieve task allocation among multiple elevators.

2. A multi-purpose optimized elevator intelligent dispatching method according to claim 1, characterized in that: The step 1 comprises: Collect passenger call requests, elevator operation status and historical traffic data to form a high-dimensional data set, and organize it into a time series sample set through a sliding time window mechanism to extract short-term dynamic change characteristics; Based on historical data, a third-order flow transfer tensor is constructed, whose first dimension represents the starting floor, the second dimension represents the target floor, the third dimension represents the time step, and the tensor elements represent the passenger flow; The improved density peak clustering algorithm is used to extract the flow pattern, including calculating the local density, the minimum distance and selecting the cluster center to identify the main passenger flow pattern; An adaptive update mechanism is introduced to detect sudden changes in passenger flow by calculating anomaly indicators, and future flow patterns are predicted based on time series modeling.

3. A multi-purpose optimized elevator intelligent dispatching method according to claim 2, characterized in that: The step 2 comprises: Considering the target floor weight of the currently loaded passengers, the demand weight of the waiting passengers, the current number of passengers on the elevator and the number of waiting passengers, as well as the running direction of the elevator, the directional inertia factor is defined, and the strength of the direction maintenance is adjusted through the inertia control parameter, thereby quantifying the tendency of the elevator to maintain the current running direction; According to the directional inertia factor, a decision rule for direction switching is established. When the directional inertia factor is lower than the preset threshold, the current directional load factor is lower than the load threshold, or the demand of high-priority passengers in the opposite direction increases, the elevator is allowed to switch direction; otherwise, the elevator maintains the current running direction to reduce direction reversal, energy consumption and time loss; A local directional stability adjustment mechanism is introduced to determine whether the directional distribution is unbalanced by calculating the global directional imbalance index. If unbalanced, the directional inertia factors of some elevators are forcibly adjusted to balance the number of up and down elevators and prevent all elevators from running in the same direction.

4. A multi-purpose optimized elevator intelligent dispatching method according to claim 3, characterized in that: The step 3 comprises: Considering the waiting time factor, target floor factor, direction consistency factor and group effect factor, the passenger priority score is defined. The waiting time factor measures the impact of the passenger's waiting time, the target floor factor considers the impact of the target floor on transportation efficiency, the direction consistency factor reflects the matching degree between the passenger's request direction and the elevator's running direction, and the group effect factor optimizes the request response for the same target floor. Based on the calculated priority scores, passenger requests are sorted and optimally allocated according to the current elevator status. The optimal elevator is allocated to the passenger by solving the optimization function, taking into account passenger priority, additional elevator stop time, and load balancing, ensuring that high-priority passengers are picked up first and optimizing the elevator stop order. A priority adjustment strategy is designed to address the dynamic changes in passenger waiting time. If a passenger’s waiting time exceeds the threshold, their priority score will be forcibly increased. During peak hours, a buffer factor is introduced to prevent elevator resources from being over-occupied by high-priority passengers, ensuring that low-priority passengers still have the opportunity to be picked up when the elevator is close to full, thereby balancing fairness and overall transportation efficiency.

5. A multi-purpose optimized elevator intelligent dispatching method according to claim 4, characterized in that: The step 4 comprises: Each elevator is defined as a game participant, and its state vector includes the current floor, running direction, number of passengers and load rate; the passenger request state is composed of the target floor, direction and priority score; the elevator revenue function takes into account the passenger priority score, load balancing, additional stop time and energy consumption, and quantifies the elevator scheduling revenue through the revenue function; The initial allocation is based on the priority of the passengers and the principle of the nearest available elevator. Each elevator calculates its own profit function. If it is found that the current task allocation leads to a decrease in profit, it requests to adjust the allocation. Through multiple rounds of profit calculation and strategy update, the Nash equilibrium is solved, so that each elevator has no intention to adjust the task allocation individually under the equilibrium strategy, thereby achieving collaborative optimization among multiple elevators. A load balancing correction factor is introduced to prevent some elevators from being overloaded. When the elevator load or priority is unbalanced, the task reallocation mechanism is triggered and the optimal strategy is recalculated to maximize the global benefit. During peak hours, a stratified diversion strategy is adopted to prioritize high-rise passengers to specific elevators, reduce the number of stops, further improve the overall scheduling efficiency of the system and reduce energy consumption.

6. A multi-purpose optimized elevator intelligent dispatching system, characterized in that: include: Dynamic prediction and clustering analysis module: By constructing the flow transfer tensor and the improved density peak clustering algorithm, combined with the adaptive update mechanism and time series prediction, it can predict the flow pattern of elevator passengers and adjust the clustering results when the passenger behavior changes suddenly; Self-balancing scheduling strategy module: introduces directional inertia factor, combines direction switching decision rules and local directional stability adjustment mechanism, dynamically optimizes the elevator running direction, and adapts to passenger flow fluctuations in different time periods through adaptive adjustment of directional inertia factor; Passenger priority gradient optimization scheduling module: defines passenger priority factors, calculates the comprehensive priority scores of passengers, and builds a passenger gradient optimization scheduling model based on this to dynamically adjust passenger priorities; Multi-elevator collaborative scheduling module: Construct a multi-elevator non-cooperative game model, define the elevator profit function and strategy space, and use Nash equilibrium solution to achieve task allocation among multiple elevators.

Images