[Paper Review] Planning of Fast Charging Infrastructure for Electric Vehicles in a Distribution System and Prediction of Dynamic Price

K. Victor Sam Moses Babu, Pratyush Chakraborty+1 more|arXiv (Cornell University)|Jan 17, 2023

Electric Vehicles and Infrastructure18 citations

TL;DR

The paper optimally places five 1000 kW fast-charging EVCIs in a 33-bus distribution system using a modified multi-objective PSO to minimize power loss and voltage deviation, and uses ARIMA to predict dynamic EV charging prices.

ABSTRACT

The increasing number of electric vehicles (EVs) has led to the need for installing public electric vehicle charging stations (EVCS) to facilitate ease of use and to support users who do not have the option of residential charging. The public electric vehicle charging infrastructures (EVCIs) must be equipped with a good number of EVCSs, with fast charging capability, to accommodate the EV traffic demand, which would otherwise lead to congestion at the charging stations. The location of these fast-charging infrastructures significantly impacts the distribution system (DS). We propose the optimal placement of fast-charging EVCIs at different locations in the distribution system, using multi-objective particle swarm optimization (MOPSO), so that the power loss and voltage deviations are kept at a minimum. Time-series analysis of the DS and EV load variations are performed using MATLAB and OpenDSS. We further analyze the cost benefits of the EVCIs under real-time pricing conditions and employ an autoregressive integrated moving average (ARIMA) model to predict the dynamic price. The simulated test system without any EVCI has a power loss of 164.36 kW and squared voltage deviations of 0.0235 p.u. Using the proposed method, the results obtained validate the optimal location of 5 EVCIs (each having 20 EVCSs with a 50kWh charger rating) resulting in a minimum power loss of 201.40 kW and squared voltage deviations of 0.0182 p.u. in the system. Significant cost benefits for the EVCIs are also achieved, and an R-squared value of dynamic price predictions of 0.9999 is obtained. This would allow the charging station operator to make promotional offers for maximizing utilization and increasing profits.

Motivation & Objective

• Minimize power loss in the distribution system.
• Minimize voltage deviation across buses while placing EVCIs.
• Demonstrate time-varying load effects and dynamic pricing on EVCI economics.
• Predict dynamic EVCI prices to enable promotional strategies and revenue optimization.

Proposed method

• Formulate EVCI placement as a two-objective problem with decision variables being the locations of 5 EVCIs (each 1000 kW) on IEEE 33-bus system buses 2..Nbus; substation not allowed.
• Use a modified multi-objective particle swarm optimization (MOPSO) to generate non-dominated Pareto fronts for power loss and squared voltage deviation.
• Incorporate a Forward-Backward Sweep (FBS) distribution load flow within each MOPSO evaluation to compute objectives.
• Maintain personal best archives and a Pareto front with crowding distance to limit archive size.
• Select the best-compromise solution using fuzzy membership-based normalization across objectives.
• Include a time-varying load analysis with random EV arrivals and states of charge, using OpenDSS/MATLAB for time-series simulation.
• Model dynamic pricing where EVCI price is EVCI_price = R_f + R_t with fixed R_f and time-based R_t distinguishing peak, normal, and off-peak periods; predict prices using ARIMA (p,d,q) to forecast daily prices.

Experimental results

Research questions

• RQ1 Can five 1000 kW EVCIs be optimally located in a 33-bus distribution system to minimize power losses and voltage deviations?
• RQ2 How do time-varying loads and stochastic EV arrivals affect optimal EVCI siting and system performance?
• RQ3 Can ARIMA-based price prediction provide reliable forecasting to support dynamic pricing and profitability of EVCIs?
• RQ4 What is the economic impact (cost/revenue) of dynamic pricing for EVCIs under realistic grid prices and EV traffic patterns?

Key findings

• Optimal placement found at buses 8, 15, 16, 17, and 18 yields a power loss of 201.40 kW and a squared voltage deviation of 0.0182 p.u. in the 33-bus system.
• Compared to the base case without EVCIs (164.36 kW loss and 0.0235 p.u. deviation), installing 5 EVCIs reduces voltage deviation and increases loss modestly, indicating a more balanced voltage profile.
• MOPSO converged in 51 iterations per run with a best-compromise solution matching the iterative search result (201.40 kW loss and 0.0182 p.u. deviation).
• The lowest bus voltage achieved is 0.9172 p.u. at bus 18; the minimum line power loss is 0.0065 kW at line 32.
• ARIMA-based price predictions achieve an R-squared value of 0.9999 for dynamic EVCI price forecasting.
• Day-one profit for each EVCI ranges from 136 to 168 (in unspecified units) under the presented pricing model.