Uncontrolled charging of large-scale electric vehicles (EVs) can affect the safe and economic operation of power systems, especially at the distribution level. The centralized EVs charging optimization methods require complete information of physical appliances and using habits, which will cause problems of high dimensionality and communication block. Given this, an ant-based swarm algorithm (ASA) is proposed to realize the EVs charging coordination at the transformer level, which can overcome the drawbacks of centralized control method. First, the EV charging load model is developed, and the charging management structure based on swarm intelligence is presented. Second, basic data of the EV using habit is sampled by the Monte Carlo method, and the ASA is applied to realize the load valley filling. The load fluctuation and the transformer capacity are also considered in the algorithm. Finally, the charging coordination of 500 EVs under a 12.47 KV transformer is simulated to demonstrate the validity of the proposed method.
As a new effective means of alleviating the energy crisis, reducing environmental pollution and global warming, EVs have more incomparable advantages than conventional cars and become the focus of governments, automakers, and energy companies now [
If widely used, EVs will aggregately contribute a new large load to the power grid. Large-scale integration of EVs will pose new challenges to the safe and economic operation of the power system [
Most distributed energy resources, such as small wind turbines and roof-top photovoltaic panels, have the characteristics of random and intermittent. It is very difficult to dispatch the traditional generators to balance their power. If appropriate charging strategies are adopted, EVs can also be used to improve the performance of distributed energy resources. So it is very important to do some deep research on the EV charging model and the charging control algorithm.
Smart grid revolutionizes the current electric power infrastructure by integrating with advanced communication and information technologies which can provide efficient, reliable and safe energy automation service with two-way communication and power flows [
Most current research and applications on charging management are based on direct centralized and hierarchical centralized methods. Swarm intelligence realizes the overall intelligence by the simple interaction between the agents and is very suitable for the distributed complex adaptive system (CAS). The power distribution system integrated with large scale of EVs is a CAS in fact. So the decentralized swarm intelligence technique is an ideal management/coordination solution for EV charging which is more flexible and adaptive than the top-down centralized or hierarchical control. The proposed charging coordination method in this paper is a decentralized swarm algorithm.
The remainder of this paper is organized as follows. Section
Recently, many research efforts and studies about the EV charging dispatch have been reported in some literatures [
In [
In [
In [
In [
In [
In [
In the related literature, there are also several studies on swarm intelligence of ants, which simulate insect behavior of ants. It is suitable for the distributed circumstances and has been widely applied to the solution of TSP, production scheduling, WSN routing, and other complex optimization problems [
The basis of the research of EV charging algorithm is to build appropriate EV load model. First we need to analyze the key influence factor of the EV charging to get better forecast and control of the EV load, which mainly include the physical characteristics of EV charging and EV using habit.
The physical characteristics mainly include the EV type, battery capacity, and charging power.
The public service cars’ daily travel miles are longer than private cars, and charging one time per day cannot satisfy their actual need. So they need the method of fast charging or battery replacement.
The private cars are flexible to use, and they are in stop status in 90 percent time of one day, which is very convenient for the charging. In the future, the private cars will be the main part of the EVs. And there are some official survey data about the national household travel of the United States, which is about the private cars and can be used as the basis of the simulation. So in this paper, we mainly consider the charging of private cars.
Table
Comparison of different charging modes of EVs.
Charging mode | Slow charge | Fast charge |
---|---|---|
Time/h | 5~10 | 1/2~1 |
Current/A | 15 | 150~400 |
Voltage/V | 220 | 220 |
Applicable location | Home | Charging station |
The using habits mainly include the travel time, travel miles, charging place, and charging time. The user’s drive and travel habits are also important to the EV charging. Most of the researches neglect the statistic of daily travel miles and only assume the basic information of SOC, start charging time, and so forth, which is differ from the reality.
The Federal Highway Administration of the Department of Transportation of the United States made a survey about the national household travel survey in 2009 and released the result [
Frequency histogram and normal distribution density function of the start charge time of EVs.
Frequency histogram and lognormal distribution density function of the daily travel miles of EVs.
There are several management structures of efficient EV charging in recent literatures, and the representative structures include the centralized control, hierarchical control, and decentralized control.
A lot of literatures use the vertical dispatch scheme, which is very common in the power system control, to realize the management of the EVs charging.
Although the centralized dispatch scheme is conducive to get the global optimal solution in theory, it is very hard to implement in the power system with large-scale EVs using the technologies of today. For example, EVs will add too many variables to the central optimization problem which will lead to the high dimensionality, and centralized management needs high computation and communication resources to deal with large amount of information.
Some literatures use the hierarchical scheme to resolve the problem of the pure centralized control. The core idea is to divide the power system into two layers or three layers according to the voltage level and divide the distribution system into many zones. The EVs dispatching problem is divided into the transmission system and several distribution system dispatching problems in this scheme. This scheme can alleviate the pressure of the dispatch center of the transmission to a certain extent.
In this scheme, EVs are characterized as agents with a certain level of autonomy taking decisions based on their local and global environment. Local environment includes EV owner’s preferences and charging infrastructure parameters, while the global environment includes the transmission/distribution network conditions and energy market conditions.
Swarm intelligence stemmed from the mimic of the living colony such as ant, bird, and fish in nature, which shows unparalleled excellence in swarm than in single in food seeking or nest building. Drawing inspiration from this, researches design many algorithms simulating colony living, such as ant colony algorithm, particle swarm optimization algorithm, artificial bee colony algorithm, and artificial fish colony algorithm, which shows excellent performance in dealing with complex optimization problems [
Swarm intelligence has the ability to manage complex systems of interacting individuals through minimal communication with only local neighbors to produce a global behavior, which typically do not follow commands from a centralized leader [
Because of the high dimensionality and communication demand of the centralized control, new methods such as the artificial swarm intelligence are worthy of exploring for the solution of charging coordination of large-scale EVs. Next we will use the ASA to realize the valley filling at the transformer level.
Notation description.
Notation | Description |
---|---|
|
Battery capacity of the EV |
|
Rated charging power of the EV |
|
Start state of charge (SOC) of the EV |
|
End state of charge (SOC) of the EV |
|
Real state of charge (SOC) of the EV |
|
Charging efficiency of the charging device for the EV |
|
Plug-in grid time of the EV |
|
Plug-off grid time of the EV |
|
Daily travel miles of the EV |
|
Energy need when the EV |
|
Real charge time of the EV |
|
Maximum load power of the transformer EV |
|
Charging energy of the EV |
|
Forecast load of the transformer at the time |
|
Real load of the transformer at the time |
|
Total load of the transformer at the time |
|
Pheromone used to guide the charging of the EV |
|
Maximum EV charging valley power of the day before. |
For the
The SOC constraint is defined by
The charging time constraint is defined by
Considering the upper limit of the transformer capacity, the total load with the
After careful investigation and analysis, the idea of multirobot cooperation based on ant colony algorithm can be applied into the EV swarm intelligent charging control [
The cooperation methods of multirobot system include the centralized control, distributed control, and the hybrid control, which integrate the first two methods effectively and are used extensively now. Based on this and integrating the centralized control, hierarchical control, and decentralized control methods of EV charging illustrated in many literatures, a hybrid control structure is proposed as shown in Figure
EVs swarm coordinated charging structure.
It includes four layers typically (maybe more according to the system scale), which is the control center, substation, transformer, and EVs in turn. The first three layers are using the traditional centralized control structure from top to down which is widely used in the power grid now. The third layer only controls the fourth layer partly, and the fourth layer uses the distributed control structure. The EVs can exchange information with others to make a decision independently which can get not only the flexibility but also the adaptability. Generally, the EVs in one line can form a dynamic union, which can be called a subsidiary swarm, to realize the coordinated charging with others.
The communication is the basis of the information share and task cooperate between EVs and mainly includes the point-to-point communication, broadcast, and group communication. Wireless communication such as WSNs is suitable for the communication of ant-based swarm charging system because of its distributed characteristic.
The multirobot system is a distributed system, and the advantage of it is that a robot can either work alone or cooperate with others. The relationships among the robots are equality. The robot can exchange information with others through communication and can make a decision independently. Different tasks were given different quantity of pheromone in the ACA used in multirobot system, which was used to attract the robots to choose which task should be accomplished with priority. The more difficult tasks will possess higher pheromone amount than the easier ones, so that the robots will choose to participate in the difficult tasks with priority.
The ant-based charging plan algorithm (Figure
Ant-based charging planning algorithm flow.
Setting up the EVs’ parameters. Including the EV number (ant population), the use habit of EV such as the plug-in and plug-off grid time (can help to get the usable charging time, assume that the car plug-in the grid when comes home and plug-off the grid when leaves home), the daily travel miles and the energy need when traveling 100 KM (can help to compute the SOC of the battery), the charging efficiency, the rated charging power, and so forth.
Initializing the pheromone. Compute the pheromone according to the day-ahead load forecast data of the transformer and other parameters using (
Setting up the maximum iterations
Loop for iterating in turn of the plug-in time of the EVs. Each ant decides the suitable charging time segment according to its parameters, and the pheromone of other ants is released.
Compare the total load
Update the pheromone at the charging time segment for other EVs to use.
Generate the charging scheduling of all EVs.
The charging planning pheromone of the
where
Assume that
For each EV as an individual agent, its charging instant will be arranged in the moment at which the pheromone is low. Suppose that
Using the quick sort algorithm to list
Choose the first
where
There are
Because
And there are
Ant-based charging adjustment algorithm flow for the load fluctuation.
Get the corresponding parameters of the EVs plug in power grid that mainly include the EV number, the charging status of each EV, and other parameters of EV itself.
Get the real load of the current time segment. The algorithm can get the latest load in process of the simulation time.
Compute the pheromone needs to compensate according to the real load, the forecast load, and the load fluctuation sum using (
Choose the different charging control method according to the load fluctuation. If the real load is larger than the forecast load, let the EVs charging at the current time segment stop charge, and choose another suitable time segment to charge considering the constraints of its own. If the real load is less than the forecast load, let the EVs charging at the later time segment move to the current time segment to charge.
Loop for iterating in turn of the plug-in time of the EVs.
Compare the total load
Update the pheromone of the load fluctuation for other plug-in EVs.
Quit the iteration when the fluctuation compensation object is finished.
Update the charging scheduling of all EVs as the time process.
The charging adjustment pheromone of the
Based on the charging load model described in Section
Our simulation environment is set to a 12.47 KV distribution transformer in a residential area, and the initial data of the original forecast load are simulated according to the actual loads of weather zones report published by the Electric Reliability Council of Texas (ERCOT). The rated capacity of the transformer
So the maximum load power of the transformer in this paper is 1615 KW. The 24 hours load data of one day is the basis of the simulation.
Considering the private cars that are mainly charged in the night, the dispatching period is set from today’s 12 o’clock to next day’s 12 o’clock and the dispatch time segment is one hour. The curves of the original load and real load with fluctuation are shown in Figure
Original forecast load and real load curve of 12.47 KV transformer.
Simulation parameters.
Parameter | Value |
---|---|
Battery capacity | 20 Kwh |
Rated charging Power | 3 Kw/h |
Charging efficiency | 90% |
Daily travel miles of EV | Monte Carlo sampling |
Energy need per 100 KM | 15 Kwh |
Start state of charge (SOC) | Computed by ( |
End state of charge (SOC) | ≥90% |
Plug-in grid time | Monte Carlo sampling |
Plug-off gird time | Monte Carlo sampling |
Total EV number | Adjustable |
EV number of each iteration | Adjustable |
Load fluctuation compensate proportion by EV | Adjustable |
Our simulation includes four parts: (1) getting the load curve with EVs of the free charge mode using the previous parameters; (2) getting the load curve with EVs of the cooperated charge mode using an ant-based algorithm to realize the self-organized charging in load valley; (3) getting the load curves with EVs using ASA and PSO to make a comparison; (4) getting the load curve with EVs using charging adjustment algorithm to realize the load fluctuation compensate.
In the free charge mode, the EVs charge in rated power immediately when plug in the power grid and left when the battery is full. The whole charging process is out of control and regulation. The simulation results based on the parameters in Table
Load curve of 500 and 1000 EVs free charge.
As displayed in Figure
In the coordination charge mode, the simulation of EVs to realize the planned charging at the valley time segment is just as follows.
From Figure
Load curve of 500 EVs coordinated charge by iterating one car each time at different hours.
In fact, there are more than one car plug-in grid at the same time. So it is necessary to deal with a group of cars at each communication in order to get better compute speed and lower communication cost. From Figure
Load curve of 500 EVs coordinated charging by iterating different numbers of cars each time.
From Table
Comparisons between the free charge and coordinated charge.
Item | EV number | Original load | Free charge mode | Coordinated charge mode |
---|---|---|---|---|
Peak (KW) |
|
|
|
|
Valley (KW) |
|
|
|
|
Peak valley difference (KW) |
|
|
|
|
Particle swarm optimization (PSO) algorithm was developed by Kennedy and Eberhart in 1995 [
PSO has the characters of high efficiency and simplicity and has been successfully applied in many areas. But it also maybe falls into the local minimum in dealing with the discretization problems.
Next we used two methods of PSO to optimize the EVs’ charging schedule problem. The basic parameters of EVs are same to the parameters used in the ASA, and the optimization objects are the load valley and the minimum fluctuation in the valley time segment.
According to the real charge time, which is calculated by the charging energy and rated charging power, the EVs can be divided into 7 types. According to the EV numbers can get the object valley filling time segments, which are the 22 o’clock of first day to 10 o’clock of the second day if there are 500 EVs. Then we take the one type of EV start charging in one valley filling time segment as one dimension of the solution, which can reduce the solution dimensions to 84.
The optimal solution search space of the second method is smaller than the first one and has better convergence rate. Next we will use the second method to make a comparison with the ASA.
From Figure
Load curve of 500 EVs coordinated charge in ASA and PSO.
From Table
Performance comparisons between the ASA and PSO algorithms with 500 EVs charging.
Item | ASA | PSO (iterate 500 times) | PSO (iterate 1500 times) |
---|---|---|---|
Program execution time (S) | 0.17 | 0.46 | 1.15 |
Load fluctuation in valley (KW) | 32.33 | 52.59 | 32.37 |
Based on the simulation results, we can see that the ASA is suitable for the EVs charging optimization.
In this section we evaluated the performance of the ant-based EV charging adjustment algorithm for the load fluctuation response we proposed in this paper. The real load with fluctuation can be computed from (
Different types of load curve of 500 EVs charging adjustment for response fluctuation at 18 o’clock.
Different types of load curve of 500 EVs charging adjustment for response fluctuation at 24 o’clock.
Different types of load curve of 500 EVs charging adjustment for response fluctuation at 6 o’clock of the next day.
Different types of load curve of 500 EVs charging adjustment for response fluctuation at 12 o’clock of the next day.
From Figure
From Figure
From Figures
In this paper, a swarm algorithm for charging coordination of EVs at the transformer level is presented. The advantages of the proposed algorithm can be summarized as follows. (i) It does not need the centralized decision of the upper level. (ii) It can respond effectively to the transformer constraint and the load fluctuation. (iii) Its computation burden is relatively low, thus suitable for large-scale application. This paper is our first step in the swarm coordination algorithm in the EVs charging. In the future, we will extend the swarm algorithm to realize some more complicated applications such as the coordinated dispatch of EVs charging and wind power.
This work was supported by the National High-tech R&D (863) Program of China (Grant no. 2012AA050803), National Natural Science Foundation of China (Grant no. 51007058), Research Fund for the Doctoral Program (for Ph.D. Supervisor) of Higher Education of China (Grant no. 20120073110020), and SMC Excellent Young Faculty Program of Shanghai Jiao Tong University.