An electrified on-demand mobility dispatching method, apparatus and system
By using a bipartite graph matching model and a state transition network flow model, the order allocation and charging dispatch of the E-MOD fleet are coordinated, which solves the problems of spatial and temporal scale differences and information barriers between the transportation system and the power system, and realizes low-carbon emission E-MOD fleet scheduling.
Patent Information
- Application Number
- CN202211235118.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-10-10
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2042-10-10
AI Technical Summary
Existing research has failed to fully coordinate travel demand in transportation systems with the integration of renewable energy in power systems, and has neglected the spatial and temporal scale differences and information barriers between E-MOD fleet scheduling and power system dispatching, leading to carbon emission problems.
A bipartite graph matching model is used to calculate order allocation and short-distance fleet relocation, generating vehicle-order level matching results. A state transition network flow model is used to optimize charging dispatch, coordinate the scheduling of fleet operators and charging infrastructure operators, and reduce carbon emissions.
It enables efficient coordination of E-MOD fleet services-rebalancing-charging deployment under different spatiotemporal scales and privacy protection, reducing total carbon emissions and making it suitable for real-world applications.
Smart Images

Figure CN115577938B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electrified on-demand mobility, and in particular to an electrified on-demand mobility scheduling method, apparatus and system. Background Technology
[0002] In the transportation sector, the emergence of the sharing economy has completely redefined the patterns of daily mobility, thus transforming associated carbon emissions. As algorithmic scheduling enables on-demand mobility (MOD) platforms to serve more consumers with fewer fleets, the sharing economy is also anticipated as a driver of the low-carbon transformation in the transportation sector. In the power grid sector, the injection of renewable energy sources, such as solar and wind power, is also reshaping electricity production. While the carbon intensity of electricity is decreasing, production is becoming more unstable and intermittent. Simultaneously, due to the rapid development of electric vehicles (EVs), carbon emission processes have become increasingly intertwined.
[0003] With the growth of the sharing economy and renewable energy, carbon emissions from the transportation and power sectors are highly dependent on scheduling algorithms. Under the trend of vehicle electrification, charging demand will further couple the operation of power and transportation systems, making vehicle-grid integrated scheduling frameworks for coordinating the service and charging processes of E-MOD fleets crucial. Recent works have jointly optimized ordering services, vehicle rebalancing, and charging scheduling. Rossi et al. proposed a coupled network model that jointly captures the interaction between the automated on-demand moving fleet and the distribution network (PDN). This model is further extended in [reference needed] to investigate the benefits of coordination between E-MOD operators and PDN operators, and to avoid potential overloads or voltage drops caused by E-MOD operations. Based on similar network flow modeling of E-MOD fleets, Lu et al. described the interaction between multiple charging station operators and E-MOD operators as a non-cooperative Stackelberg game and solved the game using a deep reinforcement learning (DRL) algorithm. Furthermore, [reference needed] discusses battery swapping options for E-MOD systems, and [reference needed] further examines the service quality pricing strategy for swapping stations. Beyond network flows, other methods such as model predictive control, deep reinforcement learning, and agent-based simulation are employed. Iacobucci et al. proposed a two-layer model predictive control model that optimizes charging and travel services in parallel. Liang et al. described joint decision-making as a partially observable Markov decision process, which they approximate by combining DRL with binary linear programming. Qian et al. proposed a shadow price deep reinforcement learning framework for online dispatching of E-MOD fleets, combining tightly coupled power and transportation network models with a data-driven model (DRL) algorithm. Zhang et al. constructed an agent-based smart charging framework to identify the benefits of effective shared electric vehicle (SEV) charging management for peak shaving and renewable energy integration. Melendez et al. developed a robust mixed-integer linear programming model for the optimal operation of cyber-physical systems comprising assemblies of autonomous SEV fleets and charging hubs.
[0004] While some current research has discussed the interactions between E-MOD fleets and power system or charging infrastructure operators, a comprehensive discussion of emissions reductions related to scheduling issues involving both travel demand in the transportation system and the integration of renewable energy in the power system has not been conducted. Furthermore, existing research often overlooks the different spatiotemporal scales of E-MOD fleet scheduling and power system dispatching, as well as the information barriers between them. Summary of the Invention
[0005] To address the problems existing in the prior art, this invention provides an electrified on-demand mobility dispatching method, apparatus, and system. The electrified on-demand mobility dispatching method of this invention optimizes fleet service-rebalancing-charging allocation and reduces carbon emissions.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] This invention provides an electrified on-demand mobility scheduling method, comprising:
[0008] Obtain passenger orders;
[0009] For the passenger orders, order allocation and short-distance fleet relocation are calculated based on a bipartite graph matching model, generating vehicle-order level matching results;
[0010] Based on the vehicle-order level matching results, the origin and destination information of all vehicles are aggregated into the flow level between regions, which will be uploaded to the charging infrastructure operator as the operational boundary conditions of the flow level.
[0011] The process involves obtaining a flow-level scheduling plan generated by the charging infrastructure operator, decomposing the flow-level scheduling into vehicle-level plans, and generating a final order service-vehicle rebalancing-charging plan for each vehicle. The flow-level scheduling plan is generated by the charging infrastructure operator by combining operational boundary conditions for multiple time periods, solving a state transition network flow model for charging dispatch involving emissions, determining the charging intervals and areas of the fleet throughout the day, and the initial SOC range of EVs for mobility services in each scheduling period.
[0012] As a further improvement of the present invention, the step of calculating order allocation and short-distance fleet relocation based on a bipartite graph matching model for the passenger order, and generating vehicle-order level matching results, includes:
[0013] Individual sets of journeys T, each journey T i ∈T by tuple It means that among them It's the pickup time. It is the receiving location. It is the time to get off the bus, and It's the drop-off point; for T i ,T j ∈T, satisfying e ij Defined as two journeys T i and T j Energy consumption between, where t ij From to Rebalancing time;
[0014] e0 is defined as the fixed startup energy consumption used to schedule another SEV during this period, which represents the energy consumption distributed in each period, such as air conditioning and vehicle production energy.
[0015] In each time period, the order allocation and short-distance fleet relocation are calculated by the bipartite graph matching model, giving the minimum energy consumption vehicle-order matching sequence that satisfies all journeys.
[0016] As a further improvement of the present invention, in each time period, the order allocation and short-distance fleet relocation are calculated by a bipartite graph matching model to provide a minimum energy consumption vehicle-order matching sequence that satisfies all journeys, including:
[0017] An E-MOD service scheduling model is constructed with the goal of meeting all mobility demands while minimizing total energy consumption. Total energy consumption includes the total energy consumption of launching all SEVs and the total energy consumption wasted between any two journeys. The total energy consumption wasted between any two journeys is modeled as a weighted vehicle-sharing network G. V = (V, E, W), whose elements are defined as follows:
[0018] VSN node: The set of nodes V corresponding to the journey set T;
[0019] VSN edges: include directed edges (v i ,v j The set of edges E of ) satisfies
[0020] VSN weights: Weighting function W(v i ,v j ) = e ij This represents the energy consumption corresponding to each edge;
[0021] The energy-minimizing schedule is calculated using a bipartite graph matching model, including:
[0022] The energy-minimizing scheduling problem is transformed into a problem in a bipartite graph G. B The minimum weighted perfect bipartite graph matching problem on (P∪Q,F,S) is defined by the following elements:
[0023] BM nodes: As a bipartite graph, its node set consists of two subsets of nodes: P = {A1, ..., A2} |V|}∪{C1,…C |V|} and Q={B1,…B |V|}∪{D1,…D |V|}, where A i B i Indicates journey T i The end and the beginning, and C i D iSEV i The start and end of operations during this scheduling period. |V| represents the vehicle-sharing network G. V The size of the node set;
[0024] BM edges: The edge set F comprises multiple subsets of edges representing different types of vehicle-travel order matching.
[0025] F = {(A i B j )∣(v i ,v j )∈E}
[0026] ∪{(C i B i |i=1,2,…,|V|}
[0027] ∪{(A i D j |i,j=1,2,…,|V|}
[0028] ∪{(C i D i |i=1,2,…,|V|}
[0029] {(A i B j )∣(v i ,v j )∈E}: After journey i, the same SEV serves journey j;
[0030] {(C i B i |i=1,2,…,|V|}:Journey i is SEV during this time period i The first order assigned;
[0031] {(A i D j |i,j=1,2,…,|V|}:Journey i is SEV during this time period j The last order assigned;
[0032] {(C i D i |i=1,2,…,|V|}:SEV during this time period i Not serving any orders;
[0033] BM weights: The weight function S represents the energy consumption corresponding to each edge subset;
[0034] S((A i B j ))=eij
[0035] S((C i B i ))=e0
[0036] S((A i D j ))=0
[0037] S((C i D i ))=0
[0038] It includes the operational and startup energy consumption in the first two types of edges;
[0039] The objective is to calculate the minimum ∑S(F) opt The minimum weighted bipartite graph is perfectly matched. A perfect match means that every node in the left-hand side P matches one node in the right-hand side Q.
[0040] As a further improvement of the present invention, the perfect match means that each node of the left P matches one node of the right Q, and further includes:
[0041] If edge (C) i D i If it is not in a perfect match, then it represents SEV. i If at least one order is assigned, the size of the operating fleet during this period also needs to be calculated.
[0042] As a further improvement of the present invention, the step of aggregating the origin and destination information of all vehicles into a flow level between aggregated regions based on the vehicle-order level matching results includes:
[0043] After calculating vehicle-order matching for all scheduling periods, the locations of all SEVs are aggregated at the regional level, specifically:
[0044] Service sequences are aggregated to the vehicle flow level based on space-time transitions and energy consumption; aggregated flows are treated as equivalent origin-destination pairs, with each origin-destination pair consisting of tuples. It means that o k and d k These are the request start area and end area, respectively; and It refers to the request start time interval and end time interval, and c k It is the battery's SOC (State of Charge) consumed; λ k It is the number of SEVs to be transferred; it receives aggregated service scheduling information at the stream level and uses it as the boundary condition for fleet operation.
[0045] As a further improvement of the present invention, the method for constructing the state transition network flow model includes:
[0046] The map is divided into multiple quadrilateral regions based on longitude and latitude to aggregate service scheduling to traffic flow, and a directed graph G is established. r =(N r A r ), to model the road network, where N r Let A represent the set of nodes representing the region and A r Let r represent the set of edges, and let r denote the initial road network.
[0047] The time-varying characteristics and battery charge levels of the E-MOD vehicle fleet are modeled, and the directed graph of the road network is extended to a space-time SOC graph and defined as a state transition network STN G. g =(N g A g ), where the subscript g indicates an extended network; the state transition network flow model specifically includes:
[0048] 1) STN nodes and injection are as follows:
[0049] State node set N g Represents the space-time SOC state of a SEV; each node I∈N g Corresponding to tuple I = (i I ,t I ,c I ), where i I ∈N r It is the initial region node, t I It is discrete time and c I It is a discrete charging level; using To represent a flow of groups of vehicles with the same origin and destination during the same time period, o k and d k These are the request start area and request end area of the stream, respectively; and These are related start and end times; c k It is the battery SOC consumed and λ k It is necessary to provide this scheduling F k The number of SEVs served is determined by the fleet operator's BM algorithm; the CIO decides how many are used for F... k Service Injection The group; each injection represents the type of SEV with the same SOC amount at departure; and To indicate the beginning and end of a day in I∈N g The number of EVs in the state;
[0050] 2) STN edge and flow are as follows:
[0051] In the After the SEV is injected into the STN, its state transitions along the timeline; edge set A g The SOC represents the spatial state transition relationship between time intervals, while the edge flow represents the number of SEVs in such transitions; the edge set can be divided into three subsets A according to physical meaning. S A R and A C ;
[0052] Edge (I,J)∈A S This represents the state transitions corresponding to the service ordering process, generated based on the aggregated fleet transitions from the origin to the destination; service flow. This represents the number of SEVs assigned to the order service transfer corresponding to request k;
[0053] b) Rebalancing - empty edges are as follows:
[0054] Edge (I,J)∈A R This represents the state transitions during the rebalancing process for SEVs that have no ordered service tasks, including idle edges where the region and SOC remain unchanged; service flow. This is expressed as the number of SEVs allocated for such rebalancing or residence transfers;
[0055] c) The charging and discharging sides are as follows:
[0056] Edge (I,J)∈A C This represents the state transition corresponding to the charging process of the SEV; service flow. This represents the number of SEVs to which such charge or discharge transfer is allocated.
[0057] As a further improvement of the present invention, solving the state transition network flow model for charging dispatch involving emissions includes:
[0058] The goal of charging infrastructure operators is to minimize the total carbon emissions caused by E-MOD fleets while satisfying service scheduling, by treating all sideflows, start-to-end service injections, and start-to-end SEV injections as decision variables; the carbon emissions of EVs are linked to the electricity they use, expressed as:
[0059]
[0060] Where P Cha Indicates the rated charging / discharging power, and δ I Indicates time t I Emission coefficient at the location;
[0061] The constraints of the network flow model are as follows:
[0062]
[0063]
[0064]
[0065]
[0066]
[0067] Where constraint (2) represents the continuity constraint of the passenger service flow, which links the passenger service flow to the start-end service injection; constraint (3) represents the start service flow and end service flow that complete the travel request; constraint (4) represents the flow conservation constraint for all types of transfer flows; function 1 x This represents an indicator function for the Boolean variable x = {true, false}, that is, if x is true, then 1. x =1, and if x is false, then 1 x =0; constraint (5) is a SOC constraint, which indicates that when the SOC of a vehicle in node I does not satisfy road c ij When the required SOC is reached, both the service flow and the rebalancing flow are restricted to 0 to avoid infeasible allocation results; constraint (6) represents the charging pile limit in charging station n.
[0068] As a further improvement of the present invention, in the ordering service-vehicle rebalancing-charging planning, the ordering service is to match vehicles with travel requests that have specific origin and destination nodes and start times; the vehicle rebalancing is a vehicle rebalancing associated with the transportation sector to pre-position vehicles to other high-demand areas for future orders; and the charging is charging coupled with charging infrastructure and the power sector to determine the charging location and time for each SEV.
[0069] An electrified on-demand mobility dispatching device, comprising:
[0070] The acquisition module is used to acquire passenger orders;
[0071] The generation module is used to calculate order allocation and short-distance fleet relocation based on a bipartite graph matching model for the passenger orders, and generate vehicle-order level matching results;
[0072] The aggregation module is used for vehicle-order level matching results. It aggregates the origin and destination information of all vehicles into a flow level between aggregated regions, and uploads it to the charging infrastructure operator as the flow level operation boundary conditions.
[0073] The de-aggregation module is used to obtain the flow-level schedule generated by the charging infrastructure operator, de-aggregate the flow-level schedule into the vehicle level, and generate the final order service-vehicle rebalancing-charging plan for each vehicle. The flow-level schedule is generated by combining the operating boundary conditions of multiple time periods to solve the state transition network flow model for charging dispatch involving emissions, determining the charging intervals and areas of the fleet throughout the day, and the initial SOC range of EVs for mobility services in each scheduling period.
[0074] An electrified on-demand mobility dispatching system, comprising:
[0075] The E-MOD operator module is used to acquire passenger orders; for the passenger orders, it calculates order allocation and short-distance fleet relocation based on a bipartite graph matching model, and generates vehicle-order level matching results; the vehicle-order level matching results aggregate the origin and destination information of all vehicles into a flow level between aggregated regions, and upload it to the charging infrastructure operator module as the flow level operating boundary conditions;
[0076] The charging infrastructure operator module is used to solve the state transition network flow model for charging dispatch involving emissions by combining the operating boundary conditions of multiple time periods, to determine the charging intervals and areas of the fleet throughout the day, as well as the initial SOC range of EVs for mobility services in each scheduling period, and to generate flow-level scheduling; the E-MOD operator module deaggregates the flow-level scheduling into vehicle-level, and generates the final order service-vehicle rebalancing-charging plan for each vehicle.
[0077] Compared with the prior art, the beneficial effects of the present invention are reflected in:
[0078] This invention proposes an electrified on-demand mobility scheduling method for carbon-limited scheduling of E-MOD fleets, coordinating fleet operators and charging infrastructure operators. It comprehensively considers various scheduling spatiotemporal scales, computational efficiency, and privacy issues associated with E-MOD operation and charging dispatch management. In the E-MOD operator module, an efficient bipartite graph matching model is used to calculate accurate order service scheduling at the vehicle order matching level on the vehicle-sharing network. The scheduling results for each time period are aggregated to the region-stream level and uploaded to the charging infrastructure operator. Based on the aggregated operational boundary conditions, the charging infrastructure operator implements a multi-commodity network flow model to optimize fleet service-rebalancing-charging dispatch, aiming to reduce carbon emissions. Then, the flow-level dispatch results are de-aggregated to allow the E-MOD operator to complete the scheduling sequence for each vehicle. A case study of New York City in the real world is conducted to verify the effectiveness of the proposed scheduling framework in different scenarios. By incorporating vehicle-network interaction, it is demonstrated that the fleet releases near-zero emissions. Simultaneously, the framework of this invention avoids excessive information flow between the transportation and power sectors, protecting privacy and security.
[0079] Furthermore, this invention provides a coordinated scheduling framework for electrified on-demand mobility systems and power systems in the transportation sector to reduce emissions. It considers the different scheduling spatiotemporal scales, computational efficiency, and privacy issues between fleet operators and charging infrastructure operators, and can be applied to real-world scenarios. An efficient bipartite graph matching model is proposed for vehicle order matching-level scheduling, and a state transition network flow model is established for flow-level charging dispatch that integrates vehicle-to-grid (V2G) communication. Attached Figure Description
[0080] Figure 1 The diagram illustrates the problem overview and scheduling framework, where (a) is the problem setting overview, (b) is the multi-time-scale scheduling, (c) is the multi-spatial-scale scheduling, and (d) is the optimization model summary.
[0081] Figure 2 Here is an illustrative example of matching from a shared network to a bipartite graph, where (a) is an example setup, (b) is a possible solution, (c) is a bipartite graph transformation, and (d) is a matching solution.
[0082] Figure 3 This is a state transition network flow model;
[0083] Figure 4 The data represents travel demand in New York City, where (a) is OD demand and (b) is zoning and FCS location.
[0084] Figure 5The components of electricity supply and carbon emission factors are shown, where (a) represents electricity demand and renewable energy production; and (b) represents carbon emissions and average emission factor curves.
[0085] Figure 6 To optimize the computational efficiency of the scheduling framework, (a) represents bipartite graph matching, and (b) represents network flow.
[0086] Figure 7 The scheduling results are SEVs in different states, where (a) represents scenario 1 and (b) represents scenario 2.
[0087] Figure 8 Comparison of charging allocation and carbon emissions between two scenarios; where (a) is the carbon emission coefficient, (b) is the charging power (Scenario 1); (c) is the cumulative carbon emissions (Scenario 1); (d) is the charging and discharging power (Scenario 2); and (e) is the cumulative carbon emissions (Scenario 2).
[0088] Figure 9 This is a schematic diagram of the electrified on-demand mobility scheduling method of the present invention;
[0089] Figure 10 This is a schematic diagram of the structure of the electrified on-demand mobility dispatching device of the present invention;
[0090] Figure 11 This is a schematic diagram of an electronic device structure according to the present invention. Detailed Implementation
[0091] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.
[0092] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.
[0093] Explanation of related terms
[0094]
[0095] Electrified Mobility on Demand (E-MOD) systems will play a crucial role in future urban transportation systems, shifting carbon emissions from the transportation sector to the electricity sector. To meet the daily charging needs of E-MOD fleets, charging infrastructure operators (CIOs) will provide charging navigation and energy management services. This invention proposes a carbon-limited dispatching framework for E-MOD fleets to coordinate fleet operators and charging infrastructure operators, comprehensively considering the various scheduling spatiotemporal scales, computational efficiency, and privacy issues associated with E-MOD operation and charging dispatch management.
[0096] like Figure 9 As shown, the first objective of this invention is to provide an electrified on-demand mobility scheduling method, applied to E-MOD operators, comprising:
[0097] S1, retrieve passenger orders;
[0098] S2, for the passenger order, calculate the order allocation and short-distance fleet relocation based on the bipartite graph matching model, and generate vehicle-order level matching results;
[0099] S3, based on the vehicle-order level matching results, aggregates the origin and destination information of all vehicles into the flow level between regions, and uploads it to the charging infrastructure operator as the flow level operation boundary conditions;
[0100] S4, obtain the flow-level scheduling plan generated by the charging infrastructure operator, deaggregate the flow-level scheduling into vehicle-level, and generate a final order service-vehicle rebalancing-charging plan for each vehicle; the flow-level scheduling plan is generated by the charging infrastructure operator by combining the operating boundary conditions of multiple time periods, solving the state transition network flow model for charging dispatch involving emissions, determining the charging intervals and areas of the fleet throughout the day, and the initial SOC range of EVs for mobility services in each scheduling period, and generating the flow-level scheduling plan.
[0101] In the E-MOD operator module, an efficient bipartite graph matching model is used to calculate accurate order service scheduling at the vehicle order matching level on the vehicle-sharing network. The scheduling results for each time period are clustered at the region-flow level and uploaded to the charging infrastructure operator. Based on the aggregated operation boundary conditions, the charging infrastructure operator implements a multi-commodity network flow model to optimize fleet service-rebalancing-charging dispatch, aiming to reduce carbon emissions. Then, the flow-level dispatch results are de-aggregated to allow the E-MOD operator to complete the scheduling sequence for each vehicle.
[0102] To reduce carbon emissions from E-MOD fleets, this invention proposes a two-tiered scheduling framework for E-MOD. In this work, daily orders are distributed across multiple scheduling periods (e.g., 1 hour). A scheduling period can also be divided into several intervals (e.g., 15 minutes) for charging scheduling. Within each scheduling period, the E-MOD operator calculates accurate order allocation and fleet rebalancing via a bipartite graph matching model that provides vehicle-order level matching. Then, the origin and destination information of all SEVs is aggregated into a flow level between aggregated regions, which is then uploaded as operational boundary conditions to the charging infrastructure operator. The charging infrastructure operator then solves a network flow model for emissions-related charging dispatch by combining operational boundary conditions from multiple time periods. This model determines the charging intervals and regions of the fleet throughout the day, as well as the initial SOC range for EVs used for mobility services in each scheduling period. Finally, the E-MOD fleet operator deaggregates flow-level dispatching into vehicle-level dispatching and generates a final service-rebalancing-charging plan for each vehicle. During this process, heterogeneous scheduling of spatiotemporal scales, computational efficiency, and privacy issues were comprehensively considered.
[0103] This invention proposes a two-tiered scheduling framework for emissions-related E-MOD fleets. Typically, E-MOD operators face three types of decisions: ordering services (matching vehicles with travel requests that have specific origin and destination nodes and start times), vehicle rebalancing associated with the transportation sector (pre-positioning vehicles to other high-demand areas for future orders), and charging coupled with charging infrastructure and the power sector (determining the charging location and time for each SEV). Inspired by current research on electric vehicle aggregation, this invention attempts to provide models with appropriate accuracy for tasks in different systems to handle their heterogeneous spatiotemporal scales. In the transportation sector, this invention establishes a computationally efficient bipartite graph matching model for accurate order allocation and fleet rebalancing. In the power sector, this invention uses a transfer network flow model for emissions-related charging allocation and energy management, utilizing the aggregated traffic scheduling results above as flow-level operational boundary conditions. During this process, accurate travel order information for each customer is maintained by the fleet operator, protecting customer privacy in the transportation sector and avoiding the difficulties of charging allocation in the power sector.
[0104] On the other hand, from an implementation perspective, most interdisciplinary studies have overlooked the spatiotemporal scale differences and information barriers between E-MOD fleet scheduling and power system dispatching. If this invention links ordering services and vehicle rebalancing to the E-MOD fleet, and simultaneously links charging scheduling to the power system, then this invention reveals that the scheduling optimization of the two systems actually operates on very different temporal and spatial scales: In transportation systems, order allocation and fleet rebalancing typically rely on precise origin and destination information for each order as input. For example, pick-up times are typically on the minute level, and origin-destination addresses are typically on the 100m level. Therefore, graph-based models such as bipartite graph matching have been proposed as computationally efficient solutions at the level of matching millions of orders with vehicles. However, these models are difficult to use to describe state transitions such as SOC (State of Charge) changes and flexible choices of charging time and location. In contrast, in power systems, charging dispatching is typically calculated on a coarser scale, where decision variables can be made at a temporal precision of 15 minutes and a spatial precision of a 1000m × 1000m area. Therefore, mathematical programming methods such as extended network flow, typically used in the above interdisciplinary research, can only handle a few thousand aggregated origin-endpoint (OD) pairs as input and can only provide dispatch results at the aggregated flow level. Thus, directly using existing models at the same spatiotemporal scale in coordinated scheduling can lead to enormous computational burdens or biases. In addition to the spatiotemporal scale differences, customer privacy must also be considered; detailed order data should not be exchanged between E-MOD fleet operators and charging infrastructure operators.
[0105] The problem investigated in this invention is the low-carbon electrified transportation system for future cities, in which the current fleet of cruising taxis is replaced by an on-demand mobility (MOD) system comprising shared electric vehicles (SEVs). The objective of this invention is to limit the total carbon emissions (i.e., eliminate the carbon footprint) caused by the E-MOD platform by deploying fleet scheduling and SEV charging.
[0106] This invention is in Figure 1(a) outlines the E-MOD system scheduling process. SEVs are centrally scheduled by E-MOD fleet operators to meet travel demand (i.e., matching SEVs with travel orders from passengers). Then, based on the service scheduling of the E-MOD operators, the Charging Infrastructure Operator (CIO) provides charging dispatch management for the E-MOD fleet, aiming to achieve carbon reduction and increase renewable energy injection. The location and size of charging stations are also given. The emission factor of electricity (corresponding to emissions per unit of electricity load) varies based on 15-minute intervals, according to renewable energy production and power system scheduling. This invention also considers two scenarios of SEV-grid interaction. One scenario models the current situation where EVs are charged solely by the grid. The other scenario is a V2G scenario where SEVs can feed their energy back into the grid.
[0107] This invention proposes, as follows Figure 1 The scheduling framework shown comprehensively considers multi-space time scales, computational efficiency, and privacy issues.
[0108] This invention aims to coordinate the scheduling of green SEV charging and ordering services. Therefore, the framework of this invention includes two modules: an ordering service scheduling module and a charging / discharging module. In terms of time, daily orders are distributed across multiple scheduling periods for vehicle-order matching, but a scheduling period can be further divided into several charging intervals. For example, as... Figure 1 As shown in (b), the present invention can set the length of the scheduling period to 1 hour, but the charging allocation interval to 15 minutes. The ordering service scheduling module determines the number of SEVs required for service in each period and their time-space transfer scheduling. Then, the information of the vehicles in service is used for the SEV charging arrangement in each interval. The detailed process is summarized below.
[0109] Module 1: E-MOD Operator: Order Service Dispatch (Vehicle Level)
[0110] Within each time period (1 hour), the E-MOD operator calculates accurate order allocation and short-distance fleet relocation using the proposed bipartite graph matching model, which is... Figure 1 The arrow in (b) indicates this. Figure 1 As shown on the left side of (c), this calculation gives the passenger service sequence for the SEV.
[0111] In this study, the invention assumes that scheduling is based on accurate information at the single-order level, represented as a separate set of journeys T. Each journey T i ∈T by tuple It means that among them It's the pickup time. It is the receiving location. It is the time to get off the bus, and This is the drop-off point. For T... i ,T j ∈T, satisfying This invention will e ij Defined as two journeys T i and T j Energy consumption between, where t ij From to The rebalancing time. This invention also defines e0 as the fixed startup energy consumption used to schedule another SEV during this time period, representing energy consumption distributed across each time period, such as air conditioning and vehicle production energy. For example... Figure 1 As shown in (d), this scheduling aims to examine the minimum energy consumption vehicle-order matching sequence that satisfies all journeys.
[0112] Module 2: CIO: Charge Dispatch (Flow Level)
[0113] After vehicle-order matching calculations for all 24 scheduling periods, the locations of all SEVs are aggregated to the regional level. For example... Figure 1 As shown on the right side of (b), service sequences are aggregated to the vehicle flow level based on space-time transitions and energy consumption. Aggregated flows are treated as equivalent origin-destination pairs (OD pairs), a method widely used in flow-level traffic management studies. Each OD pair consists of tuples. It means that o k and d k These refer to the request start area and end area, respectively. and It refers to the request start time interval and end time interval, and c k This is the battery's SOC (State of Charge). Then, λ k This refers to the number of SEVs being transferred. In this way, the CIO only receives aggregated service scheduling information at the stream level, using it as a boundary condition for fleet operation, which greatly reduces the computational burden and avoids privacy issues.
[0114] like Figure 1 As indicated by the red dashed arrow in (b), this aggregated service scheduling information is combined into fleet operation boundary conditions. The CIO will then implement emissions-related charging allocation based on the entire day, determining the initial State of Charge (SOC) of SEVs deployed to carry passengers in each time period throughout the day, as well as the fleet charging time and location. This calculation is based on an established state transition network flow model that integrates the spatiotemporal transitions of SEVs with SOC changes, and sets minimizing fleet carbon emissions as its objective (see [link to relevant documentation]). Figure 1(d)). Based on the CIO charging scheduling results, E-MOD fleet operators will also deaggregate the flow-level scheduling into vehicle-level scheduling signals, which determine the complete service-rebalancing-charging sequence for each SEV.
[0115] An E-MOD service scheduling model for allocating orders to SEVs was established. While this invention aims to limit carbon emissions from E-MOD systems, it also seeks to find an order-service scheduling method that minimizes total energy consumption in each time period. Furthermore, this invention attempts to determine the lowest carbon-emission charging strategy that achieves energy consumption across all time periods.
[0116] The energy-minimizing scheduling model of this invention is an extension of the minimum fleet model in [the original text]. For example... Figure 2 As shown in (a), the service scheduling problem can be formulated as searching for a scheduling plan that minimizes total energy consumption while satisfying all mobility demands. Total energy consumption comprises two parts: the total energy consumption for launching all SEVs and the total energy wasted between any two journeys. The latter can be modeled as a weighted vehicle-sharing network G. V = (V, E, W), whose elements are defined as follows:
[0117] VSN node: The set of nodes V corresponding to the journey set T.
[0118] VSN edges: include directed edges (v i ,v j The set of edges E of ) satisfies
[0119] VSN weights: Weighting function W(v i ,v j ) = e ij This represents the energy consumption corresponding to each edge.
[0120] Figure 2 (a) illustrates an illustrative example of the energy-minimizing scheduling of the present invention. Arrows represent three journeys, which are three VSN nodes in a shared network. Dashed lines are VSN edges, indicating the shortest path between the end of one journey and the start of the next. The corresponding energy consumption as a VSN weight is marked next to the corresponding VSN edge. The scheduling solution includes a group of paths and the corresponding number of SEVs, where each path represents one SEV scheduled in this time period. The total energy of the solution includes the total weights covered by the paths and the total startup energy consumption of all SEVs. For example, Figure 2 (b) presents a possible solution that leases two SEVs to serve three journeys at a cost of 2+2e0. It's important to remember that e0 represents the startup energy consumption per vehicle.
[0121] In general, finding the energy-minimizing scheduling problem is NP-hard. This invention proposes a bipartite graph matching algorithm to solve the energy-minimizing scheduling problem. The algorithm in this invention is an evolution of the algorithm in [previous work]. This invention first transforms the energy-minimizing scheduling problem into a problem within a bipartite graph G. B The minimum weighted perfect bipartite matching (BM) problem on (P∪Q,F,S) (e.g.) Figure 2 (c) shows that its elements are defined as follows:
[0122] 1) BM nodes: As a bipartite graph, its node set consists of two subsets of nodes: P = {A1, ..., A2} |V|}∪{C1,…C |V|} and Q={B1,…B |V|}∪{D1,…D |V|}, where A i B i Indicates journey T i The end and the beginning, and C i D i SEV i The start and end of operations during this scheduling period. |V| represents the vehicle-sharing network G. V The size of the node set.
[0123] 2) BM edges: The edge set F consists of four subsets of edges, representing different types of vehicle-travel order matching:
[0124] F = {(A i B j )∣(v i ,v j )∈E}
[0125] ∪{(C i B i |i=1,2,…,|V|}
[0126] ∪{(A i D j |i,j=1,2,…,|V|}
[0127] ∪{(C i D i |i=1,2,…,|V|}
[0128] a){(A i B j )∣(v i ,v j )∈E}: After journey i, the same SEV serves journey j.
[0129] b){(C iB i )|i=1,2,…,|V|}:Journey i is the first order assigned to SEVi during this time period.
[0130] c){(A i D j )|i,j=1,2,…,|V|}:The last order assigned to SEVj by journey i during this time period.
[0131] d){(C i D i |i=1,2,…,|V|}: SEVi does not serve any orders during this period.
[0132] 3) BM weights: The weight function S represents the energy consumption corresponding to each edge subset.
[0133] S((A i B j ))=e ij
[0134] S((C i B i ))=e0
[0135] S((A i D j ))=0
[0136] S((C i D i ))=0
[0137] This includes the operational and startup energy consumption in the edges of the first two types. Note that the value of e0 can also be adjusted for each SEV node C. i They are different, taking into account the initial location of the SEV and additional energy consumption caused by traffic congestion.
[0138] Based on the above BM, the scheduling problem is equivalent to finding the minimum ∑S(F) opt The minimum weighted bipartite graph is perfectly matched. A perfect matching is defined as every node in the left side P matching a node in the right side Q. Any perfect matching in this bipartite graph corresponds to a subscription service scheduling solution. Note that if edge (C...) i D i If it is not a perfect match, it means that SEVi has been assigned at least one order. The size of the operating fleet during this period can then be calculated.
[0139] Figure 2 (c) shows Figure 2(a) is an illustrative example of a bipartite graph transformation, where the first side indicates operational energy consumption and the second side indicates fixed startup energy consumption. Figure 2 (d) corresponds to Figure 2 (b) is a scheduling solution that means SEV1 serves journeys 1 and 3, SEV2 serves journey 2, and SEV3 remains idle during this period.
[0140] For large-scale solutions, this problem can be solved in polynomial time O(n^2). 3 The Kuhn-Munkres algorithm (also known as the KM algorithm) is used to solve the problem, based on which the SEV's order service sequence and corresponding energy consumption can be obtained.
[0141] This invention establishes a charging dispatch model. The design of charging dispatch is constrained by the need to meet mobility demands. Therefore, the charging dispatch model considers the aggregated flow-level service dispatch from fleet operators as its boundary condition.
[0142] The interval length is set to 15 minutes, which is a refinement of the 1-hour scheduling period assumed in this invention. To aggregate service scheduling to traffic flow, the map is divided into multiple quadrilateral regions based on longitude and latitude. A directed graph G is then constructed. r =(N r A r ), to model the road network, where N r Let A represent the set of nodes representing the region and A r Let r represent the set of edges. The subscript r indicates the initial road network.
[0143] To model the time-varying characteristics and battery charge levels of the E-MOD vehicle fleet, the directed graph of the road network was extended to a space-time SOC graph and defined as a state transition network (STN) G. g =(N g A g (The subscript g denotes an extended network). The model is based on a generation and expansion mechanism, and also incorporates the discharge and idle selection of SEVs. Here, the invention first gives the definitions of network topology, injection, and flow. Also... Figure 3 The network is shown in the image.
[0144] 1) STN nodes and injection
[0145] State node set N g This represents the space-time SOC state of the SEV. Each node I∈N g Corresponding to tuple I = (i I ,t I ,c I ), where i I ∈N rIt is the initial region node, t I It is discrete time and c I It is a discrete charging level. For example, in Figure 3 In this diagram, triangular nodes represent areas with charging stations, while circular nodes represent normal areas. This invention uses... This represents a flow of vehicle transfer scheduling groups that have the same origin and destination during the same time period. Here, o k and d k These are the request start area and request end area of the stream, respectively. and These are the associated start and end times. k It is the battery SOC consumed and λ k It is necessary to provide this scheduling F k The number of SEVs served is determined by the fleet operator's BM algorithm. The CIO must decide how many SEVs to use for the fleet. k Service Injection The group. Each injection represents the type of SEV with the same SOC at departure. Figure 3 In this invention, gray arrows indicate the components injected into the STN. The number of cars that complete their service can be represented as the negative end-of-service injection leaving the STN. exist Figure 3 In this invention, the departure from STN is drawn using gray arrows. This invention also assumes that the CIO makes optimal decisions regarding the initial and final states of the daily SEV. Here, this invention uses... and To indicate the beginning and end of a day in I∈N g The number of EVs in the state.
[0146] 2) STN edges and flows
[0147] In the After the SEV is injected into the STN, its state transitions along the time horizon. Edge set A g The edge flow represents the spatial state transition relationship between time intervals, while the edge flow represents the number of spatially-oriented variables (SEVs) in such transitions. The edge set can be partitioned into three subsets A based on physical meaning. S A R and A C .
[0148] a) Service side
[0149] Edge (I,J)∈A SThis represents the state transition corresponding to the service ordering process, generated based on the aggregated fleet transitions from the origin to the destination. The characteristics of this edge are a change in region, an increase in time, and a decrease in SOC. Service Flow This represents the number of SEVs assigned to such order service transfers corresponding to request k. Figure 3 In this invention, gray lines with arrows are used to represent service edges.
[0150] b) Rebalancing - Free Edges:
[0151] Edge (I,J)∈A R This represents the state transition during the rebalancing process of SEVs that have no ordered service tasks, characterized by a changed location and a decreased SOC. Furthermore, considering that some SEVs may sometimes be idle and remain stationary, it also includes idle edges where the region and SOC remain unchanged. Service Flow This represents the number of SEVs allocated for such rebalancing or residence transfers. Figure 3 In this invention, a gray dashed line with an arrow is used to represent the rebalancing-free edge.
[0152] c) Charging / Discharging Side
[0153] Edge (I,J)∈A C This represents the state transition corresponding to the charging process of a SEV, distinguished by the unchanged position and the increasing SOC. Simultaneously, considering the V2G process, a decreasing SOC can also be added to the discharging side. Service Flow This indicates the number of SEVs allocated for such charge or discharge transfer. Figure 3 In this invention, blue dashed lines with arrows are used to represent the charging and discharging edges.
[0154] Based on STN, the energy management and charging allocation problem of CIO can be represented as a multi-commodity network flow model, in which the following objectives and constraints are defined.
[0155] 1) Objective
[0156] The goal of charging infrastructure operators is to minimize the total carbon emissions caused by E-MOD fleets while satisfying service scheduling requirements, by treating all sidestream and start-to-end service injections and start-to-end SEV injections as decision variables. The carbon emissions of an EV are related to the electricity it uses, and can be represented as follows:
[0157]
[0158] Where P Cha Indicates the rated charging / discharging power, and δ I Indicates time t IThe emission factor (mTCO2 / MWh) is used here. Since the charging load may only account for a limited portion of the total electrical load, the present invention uses the average emission factor, which can be obtained based on historical hourly load data and corresponding hourly emission data.
[0159] 2) Constraints
[0160] The constraints of the network flow model are listed below:
[0161]
[0162]
[0163]
[0164]
[0165]
[0166] Constraint (2) indicates the order service flow that satisfies the continuity constraint, which associates the order service flow with the start-end service injection. Constraint (3) represents the start and end service flows that complete the trip request. Constraint (4) represents the flow conservation constraint for all kinds of transition edge flows. Function 1 x This represents an indicator function for the Boolean variable x = {true, false}, that is, if x is true, then 1. x =1, and if x is false, then 1 x =0. Constraint (5) is a SOC constraint, which indicates that when the SOC of a vehicle in node I does not satisfy road c ij When the required SOC is reached, both the service flow and the rebalancing flow are restricted to 0 to avoid infeasible allocation results. Constraint (6) represents the charging pole limit in charging station n.
[0167] The network flow problem described above is a high-dimensional linear programming problem that can be solved using commercial solvers such as GUROBI that employ barrier methods. It's important to note that using sparse matrix modeling in MATLAB can significantly reduce memory usage for data storage and the time required to model the optimization problem. Pruning techniques during the modeling process, such as eliminating unnecessary transition edges, can also reduce the problem's scale.
[0168] The above model provides a flow-level fleet dispatching scheme, but it does not specify the specific task of each SEV, thus protecting the privacy of E-MOD fleets and reducing the computational burden on CIOs. From the perspective of E-MOD fleet operators, this invention provides a heuristic de-aggregation algorithm to obtain detailed dispatching information for each SEV.
[0169] For each time period, vehicles are allocated based on their current location, power status, and the time period to satisfy the state transition plan obtained from the network flow model. In other words, each state node is enumerated in chronological order, and the edges for assigning vehicles to nodes satisfy flow constraints. Then, E-MOD operators can implement a bipartite graph matching algorithm for more accurate scheduling calculations.
[0170] Example
[0171] This invention establishes large-scale numerical cases based on real-world data to illustrate the effectiveness of the proposed model. The bipartite graph matching model and its solution are coded in C++. Furthermore, a transition network flow model is built in MATLAB R2019b and solved using Gurobi 9.0.0. Experiments were primarily run on a laptop computer with an Intel Core i7-9750 CPU @ 2.60GHz and 16GB of RAM.
[0172] On the transportation side, travel demand was collected from New York City taxi data in 2014, and order information included pick-up time, drop-off time, pick-up location (origin), drop-off location (destination), and travel distance. All time information was at the second level, and location information was at the 1e-3 level of latitude and precision. The sampled pick-up-drop-off information (i.e., OD pairs) was... Figure 4 As can be seen in (a), it covers the study area between [40.64, 40.80] and [-74.02, -73.86].
[0173] For the bipartite graph matching model, accurate data is distributed across each scheduling period (1 hour) based on pick-up time and input into the model. Considering 436,875 travel orders for the entire day of January 10, 2014, the scheduling model was solved 24 times. Furthermore, it is assumed that an E-MOD fleet of 9,355 vehicles serves approximately 2% of all travel orders.
[0174] For the network flow model, the study region is divided into 16 regions, each with an area of 0.04° × 0.04°, such as... Figure 4 As shown in (b). Furthermore, assume that eight fast-charging stations, each with a capacity of 500 charging piles and an average of 75kW, are located in areas 4, 6, 7, 8, 9, 11, 14, and 16. Set the SEV's battery capacity to 75kWh and the power consumption to 0.2kWh / km. Assume that each SEV starts with 60% SOC and ends with over 50% SOC throughout the day.
[0175] On the electricity side, to represent power systems with a high proportion of renewable energy, historical load data and renewable energy production data were obtained from the California ISO. Figure 5 The graph is presented in (a). As can be seen, electricity load peaks at night, but photovoltaic (PV, solar) output, which constitutes the majority of renewable energy, reaches its maximum at midday (approximately 40% of real-time electricity). Average emission factors (grid emissions at 15-minute intervals) were calculated using two types of data (see [reference]). Figure 5 (b) The emission coefficients ranged from 0.2817 mTCO2 / MWh to 0.3711 mTCO2 / MWh during the nighttime hours and decreased to 0.2276–0.3443 mTCO2 / MWh during the daytime hours, indicating that the city has high solar energy injection.
[0176] Based on different approaches to developing fleet charging flexibility, this invention establishes the following research scenarios.
[0177] Scenario 1: Controlled charging allocation
[0178] In this scenario, the invention assumes that the CIO provides charging and navigation services for the E-MOD fleet. While meeting the needs of travelers, the invention further optimizes the in-flight charging allocation of E-MODs by reducing the overall carbon emissions caused by the fleet through the CIO.
[0179] Scenario 2: Controlled charge and discharge regulation
[0180] Given the localized peaks in travel demand, a significant proportion of SEVs may be idle during off-peak periods, demonstrating substantial vehicle-to-grid (V2G) capacity and the potential to be developed as mobile energy storage units. Through charge / discharge scheduling, E-MOD fleets can provide greater energy flexibility to the power system. For example, the fleet can absorb excess photovoltaic power generation at midday and release it during evening peak hours, thereby increasing the overall renewable energy utilization rate. Therefore, in this scenario, the CIO acts as an intermediary to optimize the charging and discharging of SEVs throughout the day.
[0181] Based on the above case settings, this invention first tested the computational efficiency of the scheduling framework. The bipartite graph matching problem for all 24 scheduling periods was solved in parallel, demonstrating... Figure 6(a) The presented solution time. This invention shows that the solution time is positively correlated with the size of the order (corresponding to the number of nodes in the bipartite graph). For a grouped travel set with a sampling rate of 25%, the highest solution time is 1273s, which is acceptable on a personal computer. For the complete dataset (the results are presented above), the model was solved on an Intel(R) Xeon(R) Platinum 8360H CPU @ 3.00GHz and 1TB of memory, and completed within a highest solution time of 71641s, which is consistent with O(n) 3 The time complexity is consistent with that of traditional computing. With the support of high-performance computers, the computation will be much faster in real-world applications.
[0182] For the CIO side, Figure 6 (b) shows the modeling time of the transfer network and the solution time of the network flow model, all computationally performed in MATLAB and GUROBI. The time consumption was compared on sampled and full datasets. This invention shows that modeling time accounts for the dominant proportion and increases with the size of the dataset, while the solution time remains almost constant (100s). Because modeling can be performed primarily offline, this computational efficiency is usable for real-world applications. It should be noted that if the time division precision and SOC interval are too fine (to replace bipartite graph matching), the memory consumed during the solution process will exceed the limits of a personal computer, further demonstrating the significance of fusing the two algorithms in this invention.
[0183] Figure 7 (a) Presents the number of SEVs in different states (service, charging, rebalancing, and idle) during each 15-minute interval in Scenario 1. The coordinated scheduling of this invention primarily involves charging the SEV batteries between 13:00 and 15:00, corresponding to the "greenest" phase with the lowest emission coefficient. Because... Figure 7 The invention demonstrates that only half or less of the fleet is in service during most hours, so in addition to limiting carbon emissions, the coordinated dispatching of this invention also limits the size of the fleet on the road. Therefore, although this invention aims to minimize energy consumption rather than the size of the dispatched fleet, the method of this invention is also efficient for managing the operational proportion of the fleet. Furthermore, dispatching is efficient from the perspective of matching passengers with SEVs, as only a small proportion of SEVs are in a rebalancing state throughout the day. Unlike Scenario 1, in Scenario 2, SEVs can feed energy back into the grid. Whether or not SEVs are allowed to feed energy back into the grid will significantly impact SEV dispatching and total carbon emissions from the transportation and power sectors. Figure 7 (b) The number of cars in different states at each 15-minute interval in Scenario 2 was summarized, and its relationship with... Figure 7 (a) is fundamentally different. In Figure 7 In (b), SEVs are frequently charged and discharged to reduce carbon emissions during their idle periods. SEV charging is primarily concentrated during the day, while most discharging occurs at night. In this way, the E-MOD fleet functions much like a distributed energy storage system that transfers clean solar energy from daytime to nighttime, reducing fleet redundancy and increasing the flexibility of the power system. Simultaneously, the possibility of vehicle-to-grid connectivity alters the matching between passengers and SEVs. Furthermore, the number of vehicles undergoing rebalancing is higher in scenario 2 than in scenario 1, suggesting a trade-off between facilitating clean energy injection and improving the energy efficiency of vehicles on the road.
[0184] The present invention then compared the charging allocation and carbon emissions between two scenarios. For scenario 1, respectively in Figure 8 (a) Figure 8 (b) Figure 8 (c) presents the time-varying carbon coefficient, charging power, and cumulative carbon emissions. As mentioned above, charging time is scheduled during the carbon emission trough with the highest solar energy injection. The cumulative carbon emissions associated with SEV charging time gradually increase to 56.3 kg CO2 at the end of the day.
[0185] For scenario 2, in Figure 8 (d) and Figure 8 (e) presents the electricity generated during charging and discharging and the cumulative carbon emissions. Figure 8 In (d), some columns appear to have the same height, constrained by the capacity of the fast charging stations. It can also be noted that a small percentage of SEVs also charge at night. This is due to fluctuations in the emission coefficient curve; SEVs can achieve carbon reduction through alternating charging and discharging, similar to an arbitrage process. The reader may also note that some intervals involve simultaneous charging and discharging decisions, which can be interpreted as differences in the energy consumption of the SEVs. SEVs in some areas must charge for mobility services, while SEVs in other areas can reinject energy into the grid to reduce carbon emissions. Finally, the coordinated scheduling of this invention can limit carbon emissions. When SEVs can reinject energy into the grid, the total emissions of E-MOD can approach carbon neutrality. Figure 8As shown in (e), the cumulative carbon emissions are negative before interval 58 (approximately 15:00), representing the emission reduction effect of E-MOD charging scheduling. After that, emissions increase to positive values and end at 3.86 kg CO2 by interval 96 (24:00), which is approximately 6% of the emissions in scenario 1. In other words, the E-MOD fleet achieves near-zero emissions through coordinated scheduling between fleet operators and charging infrastructure operators. With larger fleet sizes, the fleet can even exhibit negative emissions, which will be presented in the next section. Furthermore, if spatial differences in the electricity emission coefficient are included, the E-MOD fleet, as a mobile storage unit, can demonstrate greater spatial flexibility, possessing the ability to mitigate the reduction of renewable energy due to transmission constraints.
[0186] A case study of New York City was conducted to validate the effectiveness of the proposed dispatching framework in various scenarios. By incorporating vehicle-to-grid (V2G) technology, near-zero emissions from the fleet were demonstrated. Furthermore, the framework avoids excessive information flow between the transportation and power sectors, thus protecting privacy and security.
[0187] like Figure 10 As shown, the present invention also provides an electrified on-demand mobility dispatching device, applied to E-MOD operators, comprising:
[0188] The acquisition module is used to acquire passenger orders;
[0189] The generation module is used to calculate order allocation and short-distance fleet relocation based on a bipartite graph matching model for the passenger orders, and generate vehicle-order level matching results;
[0190] The aggregation module is used for vehicle-order level matching results. It aggregates the origin and destination information of all vehicles into a flow level between aggregated regions, and uploads it to the charging infrastructure operator as the flow level operation boundary conditions.
[0191] The de-aggregation module is used to obtain the flow-level schedule generated by the charging infrastructure operator, de-aggregate the flow-level schedule into the vehicle level, and generate the final order service-vehicle rebalancing-charging plan for each vehicle. The flow-level schedule is generated by combining the operating boundary conditions of multiple time periods to solve the state transition network flow model for charging dispatch involving emissions, determining the charging intervals and areas of the fleet throughout the day, and the initial SOC range of EVs for mobility services in each scheduling period.
[0192] The present invention also provides an electrified on-demand mobility dispatching system, comprising:
[0193] The E-MOD operator module is used to acquire passenger orders; for the passenger orders, it calculates order allocation and short-distance fleet relocation based on a bipartite graph matching model, and generates vehicle-order level matching results; the vehicle-order level matching results aggregate the origin and destination information of all vehicles into a flow level between aggregated regions, and upload it to the charging infrastructure operator module as the flow level operating boundary conditions;
[0194] The charging infrastructure operator module is used to solve the state transition network flow model for charging dispatch involving emissions by combining the operating boundary conditions of multiple time periods, to determine the charging intervals and areas of the fleet throughout the day, as well as the initial SOC range of EVs for mobility services in each scheduling period, and to generate flow-level scheduling; the E-MOD operator module deaggregates the flow-level scheduling into vehicle-level, and generates the final order service-vehicle rebalancing-charging plan for each vehicle.
[0195] like Figure 11 As shown, a third objective of the present invention is to provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the electrification on-demand mobility scheduling method.
[0196] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the electrified on-demand mobility scheduling method.
[0197] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0198] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0199] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0200] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
Claims
1. An electrified on-demand mobility scheduling method, characterized in that, include: Obtain passenger orders; For the passenger orders, order allocation and short-distance fleet relocation are calculated based on a bipartite graph matching model, generating vehicle-order level matching results; Based on the vehicle-order level matching results, the origin and destination information of all vehicles are aggregated into a flow level between regions, which is then uploaded to the charging infrastructure operator as the flow level's operational boundary conditions. This aggregation ensures that the charging infrastructure operator only receives aggregated service scheduling information at the flow level, thereby avoiding excessive information flow between the transportation and power sectors and protecting privacy. The process involves obtaining a flow-level scheduling plan generated by the charging infrastructure operator, and then de-aggregating the flow-level schedule into a vehicle-level plan. This de-aggregation includes: the E-MOD operator implementing a bipartite graph matching algorithm for scheduling computation to generate a final order service-vehicle rebalancing-charging plan for each vehicle; the de-aggregation of the flow-level schedule into a vehicle-level plan includes: enumerating each state node in chronological order, and assigning vehicles to nodes along edges to satisfy flow constraints; the flow-level scheduling plan is generated by the charging infrastructure operator by combining operational boundary conditions across multiple time periods to solve a state transition network flow model for charging dispatch involving emissions, determining the charging intervals and areas of the fleet throughout the day, and the initial SOC range of EVs for mobility services in each scheduling period. The method for constructing the state transition network flow model includes: The map is divided into multiple quadrilateral regions based on longitude and latitude to aggregate service scheduling to traffic flow, thus creating a directed graph. To model the road network, in which Represents the set of nodes representing the region and Let r represent the set of edges, and let r denote the initial road network. The time-varying characteristics and battery charge levels of the E-MOD vehicle fleet are modeled, and the directed graph of the road network is extended to a space-time SOC graph and defined as a state transition network (STN). The subscript g indicates an extended network; the state transition network flow model specifically includes: 1) STN nodes and injection are as follows: State Node Set Represents the space-time SOC state of the SEV; each node Corresponding to tuple ,in It is the initial region node. It is discrete time and It is a discrete charging level; using This represents a flow of groups of vehicles with the same origin and destination during the same time period. and These are the request start area and request end area of the stream, respectively; and These are related start and end times; It is the battery SOC consumed; It is necessary to schedule this. The number of SEVs served is determined by the fleet operator's BM algorithm; the CIO decides how to allocate them for... Service Injection { The group; each injection represents the type of SEV with the same SOC amount at departure; and To indicate the beginning and end of the day The number of EVs in the state; 2) STN edge and flow are as follows: In the After the SEV is injected into the STN, its state transitions along the timeline; edge set The edge flow represents the spatial state transition relationship between time intervals, while the edge flow represents the number of spatially-oriented variables (SEVs) in such transitions; the edge set can be divided into three subsets according to its physical meaning. , and ; a) Edge This represents the state transitions corresponding to the service ordering process, generated based on the aggregated fleet transitions from the origin to the destination; service flow. This is represented as corresponding to the request. The number of SEVs assigned to the order service transfer; b) Rebalancing - empty edges are as follows: side This represents the state transitions during the rebalancing process for SEVs that have no ordered service tasks, including idle edges where the region and SOC remain unchanged; service flow. This is expressed as the number of SEVs allocated for such rebalancing or residence transfers; c) The charging and discharging sides are as follows: side This represents the state transition corresponding to the charging process of the SEV; service flow. This represents the number of SEVs to which such charge or discharge transfer is allocated; Solving the state transition network flow model for charging dispatch involving emissions includes: The goal of charging infrastructure operators is to minimize the total carbon emissions caused by E-MOD fleets while satisfying service scheduling, by treating all sideflows, start-to-end service injections, and start-to-end SEV injections as decision variables; the carbon emissions of EVs are linked to the electricity they use, expressed as: in Indicates the rated charging / discharging power, and Indicates time Emission coefficient at the location; The constraints of the network flow model are as follows: Where constraint (2) indicates the passenger service flow that satisfies the continuity constraint, which links the passenger service flow to the start-end service injection; constraint (3) represents the start service flow and end service flow that complete the travel request; constraint (4) represents the flow conservation constraint for all kinds of transfer edge flows; function Represents Boolean variables The indicator function, that is, if If true, then , and if If it is false, then Constraint (5) is a SOC constraint, which indicates when node The SOC of the vehicles in the road does not meet the requirements. When the required SOC is reached, both the service flow and the rebalancing flow are restricted to 0 to avoid infeasible allocation results; constraint (6) indicates that the charging station The charging pole is limited in the range.
2. The electrified on-demand mobility scheduling method according to claim 1, characterized in that, For the passenger order, the system calculates order allocation and fleet short-distance relocation based on a bipartite graph matching model, generating vehicle-order level matching results, including: Individual travel collections Every journey From tuples It means that among them It's the pickup time. It is the receiving location. It is the time to get off the bus, and It is the drop-off point; for ,satisfy ,Will Defined as two journeys and The energy consumption between them, of which, From to Rebalancing time; Will Defined as the fixed startup energy consumption used to schedule another SEV during this period, it represents the energy consumption distributed in each period, such as air conditioning and vehicle production energy. In each time period, the order allocation and short-distance fleet relocation are calculated by the bipartite graph matching model, giving the minimum energy consumption vehicle-order matching sequence that satisfies all journeys.
3. The electrified on-demand mobility scheduling method according to claim 2, characterized in that, In each time period, order allocation and short-distance fleet relocation are calculated using a bipartite graph matching model, providing a minimum energy consumption vehicle-order matching sequence that satisfies all journeys, including: An E-MOD service scheduling model is constructed with the goal of meeting all mobility demands while minimizing total energy consumption. Total energy consumption includes the total energy consumption of launching all SEVs and the total energy consumption wasted between any two journeys. The total energy consumption wasted between any two journeys is modeled as a weighted vehicle-sharing network. Its elements are defined as follows: VSN node: corresponds to the journey set Node set ; VSN edges: including directed edges edge set ,satisfy ; VSN Weights: Weighting Function This represents the energy consumption corresponding to each edge; The energy-minimizing schedule is calculated using a bipartite graph matching model, including: The energy-minimizing scheduling problem is transformed into a bipartite graph problem. The minimum weighted perfect bipartite graph matching problem on the graph is defined by the following elements: BM nodes: As a bipartite graph, its node set consists of two subsets of nodes: and ,in Indicates journey The end and the beginning, and SEV The start and end of operations during this scheduling period; Indicates a vehicle-sharing network The size of the node set; BM edges: set of edges Includes multiple subsets of edges, representing different types of vehicle-travel order matching: On the journey Then the same SEV will be used for the journey. Serve; :journey It is during this period that SEV is The first order assigned; :journey SEV during this period The last order assigned; During this period, SEV Not serving any orders; BM weights: weighting function This represents the energy consumption corresponding to each edge subset; It includes the operational and startup energy consumption in the first two types of edges; The objective calculation has a minimum The minimum weighted bipartite graph is perfectly matched. A perfect match indicates that the left side is the left side. Each node and the right side A node match.
4. The electrified on-demand mobility scheduling method according to claim 3, characterized in that, The perfect match refers to the left side. Each node and the right side A node match also includes: If the edge If it is not in a perfect match, then it represents SEV. If at least one order is assigned, the size of the operating fleet during this period also needs to be calculated.
5. The electrified on-demand mobility scheduling method according to claim 1, characterized in that, The process of aggregating the origin and destination information of all vehicles into a flow level between aggregated regions based on the vehicle-order level matching results includes: After calculating vehicle-order matching for all scheduling periods, the locations of all SEVs are aggregated at the regional level, specifically: Service sequences are aggregated to the vehicle flow level based on space-time transitions and energy consumption; aggregated flows are treated as equivalent origin-destination pairs, with each origin-destination pair consisting of tuples. It means that among them and These are the request start area and end area, respectively; and It refers to the request start time interval and end time interval, and It is the battery SOC consumed; It is the number of SEVs to be transferred; it receives aggregated service scheduling information at the stream level and uses it as the boundary condition for fleet operation.
6. The electrified on-demand mobility scheduling method according to claim 1, characterized in that, In the ordering service-vehicle rebalancing-charging planning, ordering service matches vehicles with travel requests that have specific origin and destination nodes and start times; vehicle rebalancing is vehicle rebalancing associated with the transportation sector to pre-position vehicles to other high-demand areas for future orders; and charging is charging coupled with charging infrastructure and the power sector to determine the charging location and time for each SEV.
7. An electrified on-demand mobility scheduling device, based on the electrified on-demand mobility scheduling method according to any one of claims 1 to 6, characterized in that, include: The acquisition module is used to acquire passenger orders; The generation module is used to calculate order allocation and short-distance fleet relocation based on a bipartite graph matching model for the passenger orders, and generate vehicle-order level matching results; The aggregation module is used for vehicle-order level matching results. It aggregates the origin and destination information of all vehicles into a flow level between aggregated regions, and uploads it to the charging infrastructure operator as the flow level operation boundary conditions. The de-aggregation module is used to obtain the stream-level schedule generated by the charging infrastructure operator, de-aggregate the stream-level schedule into the vehicle level, and generate the final order service-vehicle rebalancing-charging plan for each vehicle. The flow-level scheduling is achieved by combining the operational boundary conditions of multiple time periods to solve the state transition network flow model for charging dispatch involving emissions, thereby determining the charging intervals and areas of the fleet throughout the day, as well as the initial SOC range of EVs for mobility services in each scheduling period, and generating flow-level scheduling.
8. An electrified on-demand mobility scheduling system, based on the electrified on-demand mobility scheduling method according to any one of claims 1 to 6, characterized in that, include: The E-MOD operator module is used to acquire passenger orders; for the passenger orders, it calculates order allocation and short-distance fleet relocation based on a bipartite graph matching model, and generates vehicle-order level matching results. The vehicle-order level matching results aggregate the origin and destination information of all vehicles into a flow level between aggregated regions, which will be uploaded to the charging infrastructure operator module as the operational boundary conditions of the flow level. The charging infrastructure operator module is used to solve the state transition network flow model for charging dispatch involving emissions by combining the operating boundary conditions of multiple time periods, to determine the charging intervals and areas of the fleet throughout the day, as well as the initial SOC range of EVs for mobility services in each scheduling period, and to generate flow-level scheduling; the E-MOD operator module deaggregates the flow-level scheduling into vehicle-level, and generates the final order service-vehicle rebalancing-charging plan for each vehicle.
Citation Information
Patent Citations
On-demand travel system charging and rebalancing combined scheduling method based on fluid model
CN111915146A
