Under the driving force of global “dual-carbon” energy strategic goals, the scale of electric vehicle cars, as mobile energy storage units, has experienced rapid expansion. It is projected that by 2030, the total battery capacity of electric vehicle cars worldwide will reach 57 billion kWh. The large-scale integration of electric vehicle cars into power grids presents numerous challenges for the operation and regulation of distribution networks. Simultaneously, the construction of a “vehicle-station-road-grid” multi-source information fusion platform (MIFP) offers an effective data infrastructure to balance the interests of multiple entities in deeply integrated power and transportation systems. Such a platform can guide electric vehicle car users to participate in demand response services, thereby promoting the development of green transportation systems within smart cities. Research into simulation methods for electric vehicle car charging behavior within multi-source information fusion scenarios is crucial. It not only aids in analyzing the spatiotemporal distribution patterns of electric vehicle car charging loads but also provides essential technical support for formulating effective guidance strategies for electric vehicle cars.
Early research primarily focused on simulating the impact of electric vehicle car behavior on the time-series load of distribution networks to quantify the grid impact, often lacking analysis of deep power-transportation integration. As research has deepened, the simulation of electric vehicle car behavior and orderly guidance strategies under this integrated framework have become hot topics. Previous studies have explored route optimization using algorithms like A*, considering travel time and energy consumption, and have developed frameworks for evaluating adjustable capacity based on dynamic queue prediction and user characteristics. Others have proposed dynamic routing and charging navigation strategies incorporating real-time traffic information and charging station loads, or built predictive models for dynamic traffic information to devise charging navigation strategies balancing multi-party interests. Research has also involved discrete choice models to depict user charging behavior for spatial load flexibility and Markov dynamic path decision models incorporating traffic congestion penalties. Global demonstrations of “vehicle-grid” interaction platforms indicate that multi-source information fusion and data sharing can enhance the willingness of electric vehicle car users to participate in demand response. With advancements in big data and cloud computing, building more efficient MIFPs has become feasible. Through third-party platform integration and computational processing, the synergistic benefits of multi-entity information fusion can be harnessed to achieve coordinated operation systems that balance diverse interests, fostering the development of smart cities and smart energy.
However, existing research still faces several shortcomings. Firstly, regarding information fusion, most studies focus on vehicle-grid or station-grid interactions within a power-transportation integration context, lacking dynamic interaction among vehicle, station, road, and grid entities simultaneously. Secondly, many studies treat the impact of conventional fuel vehicles on road conditions as background noise, neglecting the interactive influence between fuel vehicles and road conditions. Thirdly, there is a scarcity of research incorporating the concept of virtual State of Charge (SOC) into road network studies to more accurately assess and optimize battery state management for electric vehicle cars during travel and charging. This paper addresses these gaps by proposing an electric vehicle car charging load simulation method that accounts for dynamic multi-source information interaction among vehicles, stations, roads, and the grid. By analyzing multi-source information interaction mechanisms, we construct an indicator system reflecting the integrated operation and multi-entity interests. Combining dynamic traffic information, potential charging queue times, and grid load status, we establish a real-time dynamic optimization model for vehicle routing and charging station selection. The model’s effectiveness is validated through a simulation platform, demonstrating how user behavior choices under multi-source information fusion impact both the power grid and road network, thereby offering theoretical support for building green transportation systems in smart cities.
The core of our simulation method is the EV travel path planning based on a dynamic road network model. The process for simulating electric vehicle car charging behavior under multi-source information fusion involves several key steps, as illustrated in the workflow diagram. Initially, foundational data is input to form a feasible path set. This data includes road network topology and dynamic road network model parameters; vehicle Origin-Destination (OD) pair data, probability distribution parameters for vehicle departure time, return time, and initial SOC of electric vehicle cars; and probability distribution parameters for the grid’s time-series base load. Using the vehicle OD data, an improved Dijkstra algorithm generates feasible path sets for each OD pair.
Next, parameter sampling is performed, and an initial vehicle travel sequence is formed. The Latin Hypercube Sampling method is employed to sample vehicle departure times, return times, initial SOC for electric vehicle cars, and grid load according to their respective probability distributions. These are sorted chronologically to create the initial travel sequence. Subsequently, time-series simulation of electric vehicle car travel based on the dynamic road network model is conducted. The simulation proceeds time-step by time-step according to the initial sequence. For each time step \(t\), the vehicle’s position and SOC at time \(t-1\) are determined, and based on a charging threshold, it is decided whether the electric vehicle car requires charging. The MIFP simulates providing travel options to the vehicle at time \(t-1\) using the vehicle travel path planning method based on the dynamic road network model. The vehicle’s travel state during time period \(t\) is then simulated according to the user’s chosen path and road conditions, charging station loads are recorded, and the simulation checks for completion. If not completed, \(t\) is incremented, and the process repeats. Finally, evaluation metrics are calculated to assess system performance from multiple perspectives.
To evaluate the performance for different stakeholders, specific metrics are defined. For the Traffic Management and Coordination Center (TMCC), the road segment flow deviation metric is used, reflecting the current operational state and load capacity of the traffic network. For the Distribution System Operator (DSO), the active power load deviation metric is applied, characterizing the active load profile within a region. For charging stations, the charging pile utilization rate serves as a key metric, indicating operational efficiency and profit potential. For electric vehicle car users, metrics include average travel time, average charging waiting time, and average path change rate, collectively reflecting the user travel experience.
We employ a graph-theoretic approach to describe the dynamic road network mathematically. The model is represented as:
$$ G = \{X, M, Z(t), K\} $$
where \(X = \{x_{ij} | i \in M, j \in M, i \neq j\}\) is the set of road segments, \(M\) is the set of traffic nodes, \(Z(t) = \{z_{ij}(t) | x_{ij} \in X, t \in K\}\) is the set of road impedances at time \(t\), with \(z_{ij}(t)\) being the impedance for segment \(x_{ij}\) at time \(t\), acting as the edge weight in graph \(G\). \(K = \{t | t=0,1,2,…,T\}\) is the set of time indices over a daily cycle, divided into \(T+1\) time points and \(T\) periods, each of duration \(\Delta t\).
The road impedance \(z_{ij}(t)\) is directional and time-varying, reflecting the travel cost. Considering only travel time, it comprises link impedance and node impedance. The link impedance represents travel time on a segment, and the node impedance represents delay at an intersection, related to traffic signal control. The dynamic road impedance is expressed as:
$$ z_{ij}(t) = R_{ij}(t) + C_j(t) $$
$$ z_{ji}(t) = R_{ji}(t) + C_i(t) $$
where \(R_{ij}(t)\) and \(R_{ji}(t)\) are the link impedances for segments \(x_{ij}\) and \(x_{ji}\) respectively, and \(C_i(t)\) and \(C_j(t)\) are the node impedances for nodes \(i\) and \(j\).
The link impedance \(R_{ij}(t)\) and node impedance \(C_j(t)\) are specifically defined as:
$$ R_{ij}(t) = \begin{cases}
t_{ij}^0 \left[1 + \alpha \left(S_{ij}(t)\right)^\beta \right], & 0 < S_{ij}(t) \leq 1.0 \\
t_{ij}^0 \left[1 + \alpha \left(2 – S_{ij}(t)\right)^\beta \right], & 1.0 < S_{ij}(t) \leq 2.0
\end{cases} $$
$$ C_j(t) = \begin{cases}
\frac{c_j (1 – \lambda_j)^2}{2(1 – \lambda_j S_{ij}(t))} + \frac{9}{10} \left(S_{ij}(t)\right)^2, & 0 < S_{ij}(t) \leq 0.6 \\
\frac{c_j (1 – \lambda_j)^2}{2(1 – \lambda_j)} + \frac{1.5(S_{ij}(t) – 0.6)}{1 – S_{ij}(t)}, & S_{ij}(t) > 0.6
\end{cases} $$
In these formulas, \(S_{ij}(t) = Q_{ij}(t) / C_{ij}\) is the saturation level of segment \(x_{ij}\), where \(Q_{ij}(t)\) is the traffic flow and \(C_{ij}\) is the capacity. \(t_{ij}^0\) is the zero-flow travel time, \(\alpha\) and \(\beta\) are impedance influence factors, and \(c_j\), \(\lambda_j\), and \(q_j\) represent the signal cycle time, green signal ratio, and vehicle arrival rate at node \(j\), respectively.
To accurately depict the state of an electric vehicle car during time-series simulation on the dynamic road network, we characterize each EV using fixed and variable parameter sets. The fixed parameter set \(C_n^c\) and variable parameter set \(C_n^x\) for EV \(n\) are defined as:
$$ C_n^c = \{C_a, D_n^o, T_n^s, T_n^b, P_n^c, B_n^o, B_n^f, B_n^u\} $$
$$ C_n^x = \left\{ D_n(t), D_n^d(t), V_n(t), H_n(t), B_n(t), T_{n,a}^a(t), T_{n,a}^w(t), T_{n,a}^c(t), T_{n,a}^d(t), T_n^d(t) \mid t \in K \right\} $$
The meanings of these parameters are summarized in the table below.
| Fixed Parameter | Meaning | Variable Parameter | Meaning |
|---|---|---|---|
| \(C_a\) | Charging station ID | \(D_n(t)\) | Vehicle position at time \(t\) |
| \(D_n^o\) | Initial vehicle position | \(D_n^d(t)\) | Vehicle destination at time \(t\) |
| \(T_n^s\) | Initial departure time | \(V_n(t)\) | Vehicle speed at time \(t\) |
| \(T_n^b\) | Time charging demand arises | \(H_n(t)\) | Charging path at time \(t\) |
| \(P_n^c\) | Charging power | \(\mathcal{H}_n(t)\) | Travel path at time \(t\) |
| \(B_n^o\) | Initial battery capacity | \(B_n(t)\) | Remaining battery charge at time \(t\) |
| \(B_n^f\) | Total battery capacity | \(T_{n,a}^a(t)\) | Travel time to station \(a\) at \(t\) |
| \(B_n^u\) | Charging threshold | \(T_{n,a}^w(t)\) | Charging wait time at station \(a\) at \(t\) |
| \(T_{n,a}^c(t)\) | Charging time at station \(a\) at \(t\) | ||
| \(T_{n,a}^d(t)\) | Total trip time via station \(a\) at \(t\) | ||
| \(T_n^d(t)\) | Travel time to destination at time \(t\) |
The dynamic coupling between the electric vehicle car’s energy state and road network traffic conditions is captured by the following equations. The speed of EV \(n\) on segment \(x_{ij}\) is:
$$ V_n(t) = \frac{x_{ij}}{z_{ij}(t)} $$
The energy consumption per kilometer \(\Delta e_{ij}(t)\) is:
$$ \Delta e_{ij}(t) = \chi V_n(t) + \delta + \epsilon $$
where \(\chi\), \(\delta\), and \(\epsilon\) are road grade coefficients. The remaining battery charge updates as:
$$ B_n(t) = B_n(t-1) – V_n(t) \cdot \Delta e_{ij}(t) \cdot \Delta t $$
During simulation, if \(B_n(t) < B_n^u\), the electric vehicle car is deemed to have a charging demand, and its charging path is computed.
When an electric vehicle car does not require charging, the travel time from its current position \(D_n(t)\) to destination \(D_n^d(t)\) is:
$$ T_n^d(t) = \sum_{x_{ij} \in \mathcal{H}_n(t)} r_{ij} \cdot z_{ij}(t) $$
where \(\mathcal{H}_n(t)\) is the recommended travel path at time \(t\), and \(r_{ij}\) is a binary variable equal to 1 if segment \(x_{ij}\) is on the path, and 0 otherwise.
When an electric vehicle car requires charging, the travel time from its current position to charging station \(a\) is:
$$ T_{n,a}^a(t) = \sum_{x_{ij} \in H_n(t)} h_{ij} \cdot z_{ij}(t) $$
where \(H_n(t)\) is the recommended charging path, and \(h_{ij}\) is a binary variable indicating path inclusion.
To simulate the dynamic variation of waiting time at charging stations under multi-source dynamic information, we calculate both the actual waiting time upon arrival and the potential queue time for EVs预计 to arrive. Let \(\hat{T}_a^w(t)\) be the charging queue time at station \(a\) at time \(t\) for an EV arriving then. Let \(\tilde{T}_{n,a}^w(t)\) be the potential queue time for EV \(n\)预计 to arrive at station \(a\) at time \(t\). Define \(A_a^c(t)\) as the set of charging piles at station \(a\) that are occupied by charging EVs at time \(t\), with cardinality \(|A_a^c(t)|\). Let \(A_{a,g}^w(t)\) be the set of EVs waiting in queue for charging pile \(g\) at station \(a\). Let \(N_a\) be the total number of charging piles at station \(a\). Let \(n(i)\) denote the EV编号 arriving before EV \(n\) in the queue, with \(I_n\) being the total number of EVs ahead. The expressions are:
$$ \hat{T}_a^w(t) = \min_{g} \left[ T_g^{c,a}(t) + \sum_{i \in A_{a,g}^w(t)} T_{i}^{c}(t) \right] $$
$$ \tilde{T}_{n,a}^w(t) = \sum_{i=1}^{I_n} \frac{T_{n(i)}^{c}(t)}{N_a – |A_a^c(t)| + 1} $$
where \(T_g^{c,a}(t)\) is the remaining charging time for the EV currently using pile \(g\) at station \(a\). The actual charging wait time for EV \(n\) at station \(a\) at time \(t\) is then:
$$ T_{n,a}^w(t) = \max\left[ \left( \hat{T}_a^w(t) + \tilde{T}_{n,a}^w(t) – T_{n,a}^a(t) \right), 0 \right] $$
The required charging time is:
$$ T_n^c(t) = \frac{B_n^f – B_n(t)}{P_n^c \cdot \eta} $$
where \(\eta\) is charging efficiency. The total trip time for EV \(n\) via station \(a\) at time \(t\) is the sum of travel time to the station, wait time, charging time, and travel time from the station to the destination:
$$ T_{n,a}^d(t) = T_{n,a}^a(t) + T_{n,a}^w(t) + T_{n,a}^c(t) + T_{n,a}^{d’}(t) $$
where \(T_{n,a}^{d’}(t)\) is the travel time from station \(a\) to the final destination.
The charging load at distribution network node \(a\) (where a charging station is located) is aggregated as:
$$ P_a(t) = \sum_{i=1}^{N_a(t)} P_i(t) $$
where \(N_a(t)\) is the number of electric vehicle cars charging at station \(a\) during period \(t\), and \(P_i(t)\) is the charging power of the \(i\)-th EV. The MIFP预测 charging load for a future period \(t’\) is:
$$ \hat{P}_a(t’) = \sum_{i=1}^{N_a(t’)} \hat{P}_i(t’), \quad t’ = t+\tau, t+2\tau, \ldots $$
with the prediction step \(\tau\) calculated based on the travel time along the planned path:
$$ \tau = \max_{x_{ij} \in H_n(t)} \left( h_{ij} \cdot z_{ij}(t) \right) $$
The path optimization model aims to minimize a composite cost for the electric vehicle car user. During normal travel, minimizing travel time is the primary concern. However, when charging is needed, minimizing the total time to complete charging and reach the destination becomes paramount. The objective function is:
$$ \min W = \varphi_1 T_n^d(t) + \varphi_2 \sum_{a \in C_a} T_{n,a}^d(t) $$
where \(\varphi_1\) and \(\varphi_2\) are binary parameters: \(\varphi_1=1, \varphi_2=0\) if no charging is needed; \(\varphi_1=0, \varphi_2=1\) if charging is required. The optimization is subject to constraints. The charging path constraint ensures the electric vehicle car has sufficient battery to reach the recommended charging station:
$$ B_n(t) \geq \sum_{x_{ij} \in H_n(t)} x_{ij} \cdot \Delta e_{ij}(t) $$
The charging station power constraint prevents overload:
$$ P_a(t) \leq P_a^{\text{max}} \quad \forall a, t $$
To enhance accuracy and avoid errors due to battery capacity limits during the optimization of charging-discharging control coefficients, we introduce the concept of Virtual State of Charge (Virtual SOC). Virtual SOC is defined as a variable used to pre-calculate the SOC of an electric vehicle car after a planned charging-discharging action is completed. It allows for computation after the user selects optimization weights but before the electric vehicle car executes the plan, yielding optimized results for charging capacity and cost. The Virtual SOC after completing charging-discharging behavior in each time period is calculated as:
$$ S_{v,m,t}^{\text{oc}} = S_{m,t-1}^{\text{oc}} + \frac{x_{i,t}^p \cdot P_m^{cd}}{C_b} $$
where \(S_{v,m,t}^{\text{oc}}\) is the Virtual SOC for the electric vehicle car at location \(m\) in period \(t\), \(S_{m,t-1}^{\text{oc}}\) is the actual SOC at the previous period, \(x_{i,t}^p\) is the charging-discharging control coefficient for particle \(i\) in period \(t\) during optimization, \(P_m^{cd}\) is the charging-discharging power at location \(m\), and \(C_b\) is the battery capacity of the electric vehicle car.
During the iterative process of path planning or optimization algorithms, the charging-discharging control coefficients need to be adjusted based on the Virtual SOC to satisfy battery capacity constraints. The adjustment process includes: (1) If the Virtual SOC in a given period exceeds the upper limit \(SOC^{\text{max}}\) and the period is during peak or flat rate, the coefficient is修正 to discharge at rated power; if during off-peak, it is修正 to idle. (2) If the Virtual SOC falls below the lower limit \(SOC^{\text{min}}\), regardless of the period type, the coefficient is修正 to charge at rated power. This ensures path planning not only considers real-time traffic and charging station status but also incorporates the anticipated battery state post-charging, meeting后续 travel demands and battery health requirements.
The model is solved using a dynamic Dijkstra algorithm for EV path planning. The process involves: 1) Real-time synchronization of multi-source dynamic information (EV, traffic network, power grid, charging station status). 2) Dynamic planning of travel paths for electric vehicle cars without charging需求, minimizing travel time based on dynamic road impedances, with continuous monitoring and re-planning from the current node if the shortest path changes due to traffic conditions. 3) Dynamic planning of travel paths for electric vehicle cars with charging需求, minimizing total trip time to generate a charging path. This process incorporates the Virtual SOC concept for pre-evaluation and incorporates real-time monitoring of traffic flow, charging station queue times, and charging times. If any parameter causes the original path to become non-optimal, re-planning occurs. This ensures optimal routing considering real-time conditions and battery state expectations.
Evaluation metrics reflecting multi-entity interests are calculated as follows. The road segment flow deviation \(S\) for the TMCC is:
$$ S = \frac{1}{T} \sum_{t=1}^{T} \sum_{x_{ij} \in X} \left( S_{ij}(t) – S_{ij}^{\text{av}} \right)^2 $$
$$ S_{ij}^{\text{av}} = \frac{1}{T} \sum_{t=1}^{T} S_{ij}(t) $$
The active power load deviation \(D_a\) for the DSO at node \(a\) is:
$$ D_a = \frac{1}{T} \sum_{t=1}^{T} \left( P_a(t) – P_a^{\text{av}} \right)^2 $$
$$ P_a^{\text{av}} = \frac{1}{T} \sum_{t=1}^{T} P_a(t) $$
The charging pile utilization rate \(\gamma_a\) for charging station \(a\) is:
$$ \gamma_a = \frac{1}{T} \sum_{t=1}^{T} \frac{n_a^g(t)}{N_a^g} $$
where \(n_a^g(t)\) is the number of charging piles in use at station \(a\) at time \(t\), and \(N_a^g\) is the total number of piles at station \(a\). For electric vehicle car users, the average travel time \(\mu\), average charging wait time \(\kappa\), and average path change rate \(\rho\) are:
$$ \mu = \frac{1}{N} \sum_{n=1}^{N} W_n $$
$$ \kappa = \frac{1}{N_c} \sum_{n=1}^{N_c} T_n^w $$
$$ \rho = \frac{1}{H} \sum_{n=1}^{N} \sum_{m=1}^{M_n} l_{n,m} $$
Here, \(N\) is the total number of vehicles, \(N_c\) is the number of electric vehicle cars requiring charging in a day, \(W_n\) is the total travel time for vehicle \(n\), \(T_n^w\) is the charging wait time for EV \(n\), \(H\) is the total number of trips for all vehicles, \(M_n\) is the number of trips for vehicle \(n\), and \(l_{n,m}\) is a binary variable indicating whether the path changed during trip \(m\) for vehicle \(n\).
For case study analysis, a coupled road-power network simulation system is established. The power grid model is the IEEE 33-node system, and the road network is a 29-node regional traffic network. The coupled system includes four functional zones: office area, commercial area, residential area 1, and residential area 2. Four charging stations are deployed, each equipped with 10 charging piles of 80 kW power. Key parameters for the simulation are summarized in the following tables.
| Segment ID | Length (km) | Zero-Flow Time \(t_{ij}^0\) (min) | Capacity \(C_{ij}\) (veh/h) |
|---|---|---|---|
| 1-2 | 1.5 | 2.0 | 1200 |
| 2-3 | 2.0 | 2.5 | 1000 |
| … | … | … | … |
| Road Grade | \(\chi\) | \(\delta\) | \(\epsilon\) |
|---|---|---|---|
| Highway | 0.05 | 0.1 | 0.15 |
| Arterial | 0.08 | 0.15 | 0.2 |
| Local | 0.12 | 0.2 | 0.25 |
The system involves 3000 vehicles with an electric vehicle car penetration rate of 40%. Each electric vehicle car has a battery capacity of 40 kWh, charging efficiency \(\eta = 0.9\), and initial SOC following a normal distribution \(N(0.5, 0.15)\). Initial departure and return times are set based on common commuting patterns. The simulation周期 is 24 hours from 00:00 to the next 00:00, with a time interval \(\Delta t = 1\) minute, resulting in 1440 time periods.
To demonstrate the effectiveness and superiority of the proposed method, we compare it with two benchmark approaches: a static simulation method (Method 1) and an existing dynamic simulation method from literature (Method 2). Our proposed method is denoted as the Proposed Method. The comparison focuses on metrics like路段流量分布, user travel time, charging station load, and grid active load deviation.
Analysis from the road network perspective shows significant differences in road segment flow distributions under the three methods. During peak hours, Method 1 results in severe congestion on certain roads, leading to highly uneven traffic flow and large segment flow deviation. In contrast, both Method 2 and the Proposed Method引导 users to dynamically change paths based on changing traffic conditions, leading to more balanced traffic flow. The segment flow deviation values are: Method 1: \(9.13 \times 10^4\); Method 2: \(4.37 \times 10^4\); Proposed Method: \(4.84 \times 10^4\). Compared to Method 1, Method 2 reduces deviation by 52.13%, and the Proposed Method reduces it by 46.99%. This indicates that both dynamic methods effectively alleviate congestion during peak hours by guiding vehicles to sub-optimal paths, avoiding severe congestion on specific segments, and distributing vehicles more evenly across the network. The slightly higher deviation in the Proposed Method compared to Method 2 is attributed to its additional consideration of user-centric metrics, but the overall traffic distribution remains相似 and does not significantly impair user travel.
The impact on electric vehicle car user experience is also evaluated. Under the static method (Method 1), the average daily travel time per user is 82.46 minutes, and the average charging wait time for electric vehicle cars needing charge is 7.571 minutes. In contrast, Method 2 and the Proposed Method reduce average travel time to 70.2 minutes and 67.7 minutes, improving efficiency by 14.87% and 17.9%, respectively. Average charging wait time is reduced to 7.057 minutes and 6.034 minutes, improving by 6.79% and 20.3%. The Proposed Method shows a more significant reduction in average charging wait time compared to Method 2, primarily because it accounts for the potential queue time for electric vehicle cars预计 to arrive at charging stations. Furthermore, the average path change rates are 36.38% for Method 2 and 39.85% for the Proposed Method, indicating that path changes based on actual conditions are more frequent in our method, further optimizing travel efficiency. These results highlight the advantages of the Proposed Method in reducing travel time, shortening charging waits, and increasing routing flexibility, thereby enhancing overall transportation system efficiency and user experience for electric vehicle car owners.
From the power grid perspective, due to the predictive nature of the Proposed Method, some prediction errors exist. The charging load and its prediction error distribution for Residential Area 1 are shown graphically; similar distributions apply to other areas. Prediction errors mainly occur during morning and evening peak travel hours. This is because during peaks, many electric vehicle cars may simultaneously choose a particular charging path, leading to increased charging wait times and potential changes in the chosen path, thereby generating prediction errors.
The Proposed Method, by optimizing the logic for electric vehicle car users selecting charging stations based on multi-entity information fusion, achieves a more balanced load distribution. The core innovation lies in considering the total of travel time, wait time, and charging time, rather than relying solely on physical distance to the nearest station. This optimization strategy leads users to prefer stations with lower current utilization and shorter wait times, preventing sudden load surges at stations simply because they are the closest. Through dynamic optimization, the charging demand of electric vehicle cars is分散 across different stations and time periods, resulting in a more uniform charging load profile. A comparison of total charging load over time between the methods visually demonstrates this flattening effect achieved by the Proposed Method.
To further analyze charging load distribution, the charging pile utilization rates for each station under the three methods are compared quantitatively.
| Method | Station 1 | Station 3 | Station 13 | Station 22 |
|---|---|---|---|---|
| Method 1 | 30.03 | 49.60 | 10.10 | 10.27 |
| Method 2 | 28.40 | 34.67 | 16.33 | 20.60 |
| Proposed Method | 28.21 | 31.75 | 18.66 | 21.28 |
The table reveals significant differences. In Method 1, electric vehicle car users concentrate on specific stations. Method 2 offers more flexibility in station choice. The Proposed Method, by considering both dynamic traffic information and potential charging wait times, provides a more comprehensive approach, leading to the most balanced utilization across stations.
The advantage of the Proposed Method is also evident in its ability to optimize grid load. A comparison of active power load deviation for the distribution network nodes hosting charging stations under the three methods is presented below.
| Method | Station 1 Node | Station 3 Node | Station 13 Node | Station 22 Node |
|---|---|---|---|---|
| Method 1 | 6.989×10⁴ | 6.302×10⁴ | 1.407×10⁵ | 2.489×10⁵ |
| Method 2 | 2.341×10⁴ | 4.824×10⁴ | 1.472×10⁵ | 1.647×10⁵ |
| Proposed Method | 2.266×10⁴ | 4.143×10⁴ | 1.170×10⁵ | 1.329×10⁵ |
Compared to Method 1 and Method 2, the Proposed Method results in a smoother load profile over time, demonstrating its significant effectiveness in optimizing grid load for systems with high electric vehicle car penetration.
In conclusion, within the context of vehicle-station-road-grid multi-source information fusion, this paper comprehensively considers electric vehicle car travel time and potential charging time. We employ a dynamic Dijkstra path search algorithm to recommend optimal paths for electric vehicle cars and introduce a Virtual SOC model for real-time battery state assessment and constraint. The simulation captures the dynamic travel process of electric vehicle cars and the impact of their charging load on both the distribution network and the traffic network. Through comparative simulation, the following key conclusions are drawn:
1) The dynamic Dijkstra path search algorithm under multi-source information fusion not only provides navigation and recommended travel paths for all vehicle users but also recommends the optimal charging path for electric vehicle car users based on dynamic road information, potential charging station times, and Virtual SOC pre-evaluation. This significantly reduces the total trip time for electric vehicle car users requiring charging.
2) By guiding electric vehicle car users in selecting charging stations, the method enables real-time dynamic optimization of traffic flow on various road segments, alleviating pressure on congested roads. Compared to static methods, the proposed approach markedly improves the overall traffic condition of the road network.
3) The proposed method dynamically updates the potential charging time at target stations and utilizes the Virtual SOC concept for finer control over charging behavior. This facilitates guiding electric vehicle car users to charge during off-peak periods,合理 controls the number of electric vehicle cars entering stations, shortens in-station wait times, improves charging station utilization rates, and flattens fluctuations in the charging load profile.
Despite these advancements, the model establishment and analysis are based on simulation data, and the electric vehicle car type is assumed homogeneous. Furthermore, only charging at public stations is considered, ignoring home charging scenarios for residential electric vehicle cars. Therefore, future work will involve validation using measured data, developing more refined guidance strategies for different types of electric vehicle cars considering their varied charging behaviors and travel characteristics, and further exploring the application of the Virtual SOC model in more complex scenarios. Additionally, incorporating residential charging behavior for electric vehicle cars will be a crucial extension.