real time determination of rechargeable batteries\' type and the state of charge via cascade correlation neural network.

I.
Brief introduction although batteries seem simple, they are non-linear and complex systems due to their physical and chemical structures.
With the development of technology, the use of batteries is increasing.
It is important to estimate the state of charge (SOC)
Use the battery accurately in the battery management system to use the battery effectively.
Estimate the SoC of the battery by mathematical and electrical methods;
Mathematical and electro-chemical methods include complex equations that must be redesigned for other types of batteries.
The electrical method is easy to calculate and the user can develop the battery model by looking at the Battery Data sheet or measuring the battery parameters.
A satisfactory battery model can be obtained using the data set generated by the electrical method.
Various SoC estimation methods are proposed in the literature using experimental data sets.
By monitoring the battery voltage, current, electrical impedance spectrum, etc. , the data set used in the battery model can be obtained. and parameters.
By measuring parameters while charging, discharging, or in a steady state, data collection is possible.
The methods known at present mainly include neural network, fuzzy logic, filter and radial basis function neural network (RBFNN).
Artificial neural network (ANN)
Since it does not have complex mathematical equations and has high working accuracy, the method is easy to establish [1]-[7].
It is difficult to develop fuzzy rules and membership functions as well as fuzzy outputs.
A large amount of data and expert knowledge is needed to develop fuzzy systems [8]-[10].
The filter calculation is complex, and the conditional independence of the measurement error is required [11]--[15].
RBFNN is easy to build, but it will be slow when the data set is large [16]. In [17]
A model was developed to estimate the available capacity of lead-acid batteries used in electric vehicles.
High accuracy was obtained using ANN. In [18]
The SoC of NiMH battery is estimated using a three-layer forward neural network.
This method is used to estimate that the error rate of SoC is below 5%. In [2]
The back propagation neural network is applied to the SoC estimation of NiMH batteries for electric vehicles.
The SoC of the battery can be estimated at the steady state after charging, discharging and charging.
The open circuit voltage of the battery is used as the input parameter of the neural network.
The simulation results show that this method is suitable for hybrid vehicles. In [19]a three-
The hierarchical back propagation neural network is used to estimate the SoC of high-power NiMH batteries.
Apply five input parameters to the neural network;
These are battery discharge currents, total amps-
Hours, open voltage of the battery, time-
The average open-circuit voltage and the time-dependent average open-circuit voltage are twice.
The data set is obtained during the battery from full charge to full discharge.
Training on the application of the Levenberg Marquart algorithm.
The simulation and the measured results are compared to verify the performance of the artificial neural network.
After 10 minutes, the SoC of the battery can be estimated with an error rate of less than 5%.
Knowing the percentage of the battery\'s remaining energy can give the user an idea of how long the battery will continue to run without charging.
It is important to charge and discharge the battery in the correct form to prevent fire and explosion;
In addition, the correct use of the battery provides users with higher efficiency and longer life.
On the other hand, improper use will reduce the life of the battery, and the defective battery will cause chemical pollution in essence.
In this study, an experimental device was developed to monitor the electrical parameters of the battery.
Specialized software has been designed to systematically save measurement data and determine the type and SoC of rechargeable batteries online.
The software is also able to stop the experiment when the battery exceeds the voltage, current or temperature boundary.
Cascade related neural networks (CCNN)
Used to determine the type and SoC of the battery when discharging the battery under constant load.
Generate data sets using the battery\'s end voltage, current, and power data. Lead acid (Pb), Lithium Ion (Li-Ion)
Lithium polymer (LiPo)
Nickel and cadmium (NiCd)
Nickel Metal hydrogen (NiMH)
Rechargeable batteries were used in the experiment. The Watt-
Select the hourly value of a very similar experimental battery;
This is to successfully determine the type of battery with almost the same performance.
The difference between this study and other academic studies is the idea of determining the type of battery and the form of charge, discharge.
Experimental setup and database structure can be an example of someone who is committed to monitoring battery behavior.
The purpose of this study is to determine the type and SoC of rechargeable batteries with high precision through CCNN.
There are many applications to determine the SoC of the battery, but it is estimated that the battery type is a new study.
It is estimated by CCNN that battery is another innovation in this study. Pb, Li-Ion, Li-
Po, NiCd and NiMH rechargeable batteries were used in the experiment. II.
At present, rechargeable batteries are part of our daily life, and all wireless devices running using electric energy get this energy from the batteries.
In modern times, portability is important, which in turn increases the importance of batteries.
Battery usage in a country is proportional to the use of technology.
There are many different types of rechargeable batteries in this study, Pb, Li-Ion, Li-
Rechargeable batteries for Po, NiCd, NiMH are used.
Pb batteries are suitable for applications that do not require battery weight and size.
So they are cheap.
These batteries are mainly used for vehicles, medical equipment, electric seats for the disabled, and emergency lighting spotlights and uninterrupted power supplies. Li-
More stable and lightweight ion batteries;
The organic electrolyte provides an actual battery voltage above 4 V.
They have a high energy density and provide simple applications without the need to connect several batteries in the [series]20]-[22].
They are used by laptops, mobile phones, music players and more digital portable devices.
LiPo battery is a rechargeable battery that continues to be used from Li-
Ion battery technology
Therefore, LiPo batteries have a high energy density depending on volume and weight;
They use a large area.
They are used in electric vehicles, laptops and many electronic applications.
The most important feature of the NiCd batteries is that they do not lose capacity while maintaining capacity, essentially, it has the same capacity two weeks after the last charging time.
NiCd batteries are used in a single or grouped form in drill bits, measuring instruments, etc.
The fast charging of these types of batteries reduces their service life.
With standard charging, the average life of the NiCd battery is 5 years.
The NiMH battery has a higher energy density compared to the NiCd battery, but less charge times.
For laptops, mobile phones, cameras, toys, etc.
There are some memory effects on NiMH batteries. III.
Materials and Methods A.
The experimental device in the electric battery test device determines the shape of the experimental device according to the measurement parameters.
If the internal resistance of the battery is measured, the internal resistance meter is used, and if the current is measured, the current sensor is used, and if the voltage is measured, the voltage sensor must be used.
The battery must be charged using a charger and the load must be used to discharge the battery.
If the temperature parameters are necessary, the temperature sensor must be used in the system.
A computer or embedded system can process the collected data.
By collecting data from electrical measurements, a successful battery model can be obtained.
In this study, the open circuit voltage, current, power, load, ambient temperature and battery temperature were all measured during the charging and discharging of the battery.
The measuring device for this study is given in Figure 11.
Use the Imax B8 charging device to charge the battery and discharge the battery array 3711A programmable DC load device.
A circuit is designed to select a charger or load from the software. A LTS25-
The circuit also includes NP current sensors, LV25P voltage sensors, and LM35 temperature sensors.
The circuit can be connected with three batteries and experimental batteries can be selected from the software.
There are also contacts on the circuit that control the charger button.
The contacts on this circuit are controlled by digital I/O on Advantech USB
4716 data collection (DAQ)card.
Output of K
The thermocouple is connected through this circuit to the digital I/O of the data acquisition card.
Programmable DC load connected to PC via array 3312 Seri-
USB port converter.
All batteries that define their identity are stuck with square code.
The Perkon Spider SP400 Square reader is used to read the code.
The device is connected to the computer via a USB port.
The network camera is used to observe the experimental settings.
When discharging, the current, voltage, load and power parameters are obtained from the load device, while charging the battery, the current and voltage parameters are measured by the sensor, and transmitted to the computer through the data acquisition card.
When charging, the battery load and power parameters are calculated using voltage and current data. Ttec 6 V 1.
3 Ah Pb battery, Panasonic cgr18. 0cg 3. 7 V 2. 2 Ah Li-
Ion battery, Power Xtra x864055 3. 7 V 2 Ah Li-
Po battery, AA Portable companyCD-SC2200P 3. 6 V 2.
2 Ah NiCd battery and GP211AFH 3 in Jinfeng group. 6 V 2.
1 Ah NiMH battery was used in the experiment.
Table 1 gives the technical information of these batteries.
The maximum percentage difference in the mean Wh value is 4,15%.
The capacity of the battery is very similar, which makes it difficult to determine the type of battery.
Although the capacity of these selected batteries is similar, they have different charging types and charging current. B.
CCNN is a cascade-related neural network developed by Fahlman in 1990.
CCNN is a supervised learning algorithm.
CCNN starts with the smallest network and then automatically trains and adds new hidden units one by one to create a multipleLayer structure.
The CCNN architecture has several advantages over the existing algorithms: It learns very quickly, and the network determines its own size and topology, even if the training set changes, it also retains the structure it builds itself. It doesn\'t need any support.
Propagation of error signal through network connection [23].
An untrained cascade network is a blank sheet of paper;
It has no hidden unit.
The output weight of the Cascade-related network is trained before a solution is found or the progress is stagnant.
If a hierarchical network is enough, then the training is complete.
The weight of the hidden nerve element is static;
Once they start training, they will not be touched again.
The features they identify are permanently projected into the memory of the network.
Retaining the direction of the hidden neurons allows Cascade correlation to accumulate experience after initial training.
Few neural network architectures allow this. If a back-
The propagation network is retrained and after adding two hidden units it \"forgets\" its initial state.
Vertical lines are added to all incoming activations.
Box connections are frozen and X connections are trained repeatedly.
CCNN combines two ideas: The first is a cascading architecture in which hidden units are added only one at a time and will not change after adding.
The second is to learn the algorithm and create and install a new hidden unit.
For each new hidden unit, the algorithm tries to maximize the correlation between the new unit output and the network residual error signal. IV.
Special software to determine the battery type and SOC developed a graphical user interface in C language in Visual Studio 2010 software for monitoring the condition of the battery and saving the measurement data to the database, to determine the type and SoC of the battery.
Users can add new batteries to the database.
The user selects the test battery, the duration of the experiment, the sampling time, and selects the charge or discharge battery.
An experimental code is automatically generated when All adjustments are completed.
Create a table in the database with a name for this code, and the measurement data is saved to this table.
The measurement data curve can be seen online.
Measurements previously saved to the database can be listed.
The battery can be plugged into charging-
The discharge circuit is safe, because during the experiment, the battery is controlled if the battery reaches the critical limit value of voltage, current and temperature.
If one of the values is reached, the software automatically shuts down the system and generates an alarm.
The rest time between charge and discharge is adjustable.
Users can generate and standardize data sets to identify CCNN\'s battery and save it in Excel format.
CCNN identifies that the input variables of the battery are voltage, current, power, voltage drop angle and current drop angle.
In order to determine the voltage and current drop angle, the battery must discharge within the determined time.
This app has been selected for 400 seconds.
It is possible to generate and normalize the data set used to train CCNN to determine the battery SoC.
Determine SoC values based on measurement data.
The input variables of this CCNN are voltage, current, power and time :[
Non-reproducible mathematical expressions], (1)TA = [n. Sum up (i=1)][g. sub. i+1]+ [g. sub. i]/2 x ([t. sub. i+1]-[t. sub. i]), (2)IA = [m. Sum up (i=1)][g. sub. i+1]+ [g. sub. i]/2 x ([t. sub. i+1]-[t. sub. i]), (3)
SOC = QC/QMax x 100 ,(4)SOC = TA-IA/TA x 100. (5)
The SoC of the battery can be determined from the current curve when the battery is discharged.
From full charge to full discharge, the area under the current curve represents 100% SOC. In (1)
Equation of total area (TA)
The battery curve from full charge to full discharge is given.
In this equation I represent the current value and t the time.
In the software, the integral can be determined by ladder method;
As described in, apply this method (2)
Where g is the current value of the I th ,[g. sub. i]
1 is the current value of the I th time; [t. sub. i]
Is the time value of 1 time ,[t. sub. i+]
1 is the I Time value;
N is the quantity measured.
For mth measurement data, region (IA)
Before this time, it can be based on (3)
, IA indicates the capacity of the battery.
The SoC of the battery can be determined by dividing the remaining capacity of the battery by the full capacity of the battery, as described in (4).
In this equation, QC is the remaining capacity of the battery, and QMax is the maximum capacity of the battery.
Therefore, the rate of the remaining capacity of the battery can be from (5).
In this equationTA-IA)
Give the remaining capacity of the battery, TA give the maximum capacity of the battery [25]. V.
Experimental studies to obtain data sets for Determining battery type and SoC the battery is first fully charged and then fully discharged at constant load. 3 [ohm], 5 [ohm], and 10 [ohm]
Use a constant load value.
All the experiments were done with a healthy battery at ambient temperature.
This experimental data is used as training data for CCNN to determine the type of battery and SoC. VI.
Determining the type of battery and SOC through CCNN has a lot of research on estimating the battery SoC, but estimating the type of battery is a new study.
In the future, the use of electric vehicles will increase, and the importance of batteries will increase.
Users will not wait at the toll station.
Instead, they will replace the battery pack in this case.
Therefore, a software that determines the battery type and SoC and provides information on how to charge and use the battery pack will be very useful.
Based on this idea, we initially tried to determine the type of battery in Matlab.
There is a battery block in the Simulink that supports multiple types of rechargeable batteries.
By using a full charge of voltage and current to a full discharge value and using the CCNN method, we have successfully determined the type of battery.
This method is then studied in practical application.
Figure 1 gives the architecture of the CCNN used to determine the battery type. 3.
The architecture has five inputs, one hidden layer and five outputs.
The input value is current, voltage, power ,[V. sub. [theta]]and ie. [V. sub. [theta]]
Voltage drop and [angle]i. sub. [theta]]
Is the angle of the current drop when discharging the battery.
These values are determined by calculating the difference in values over 400 seconds.
This time value is determined by trying and being considered a unit of time. [[DELTA]. sub. V]
Is the difference in voltage value, and [[DELTA]. sub. i]
Is the current value difference after 400 seconds. ie is arctan(Ai)and [V. sub. [theta]]is arctan([[DELTA]. sub. V]).
CCNN\'s input values are normalized between 0 and 1, dividing the input values by the absolute value of the maximum value of the input vector.
In this equation, x is the normalized value and x is the value to be normalized ,[
Absolute value of X]
The maximum value of the input vector.
NN has five outputs and gives a result between 0 and 1 for each neuron.
The maximum value of these values represents the type of battery.
Determining the architecture of the CCNN of the battery SoC is given in the figure4.
The architecture has four inputs, one hidden layer and one output.
The input values are current, voltage, power, and time (t).
The output of NN is between 0 and 1.
0 represents a fully discharged battery, and 1 represents a fully charged battery.
The t value is from (6).
It is calculated according to the change of voltage value.
For each architecture, determine the number of neurons in the hidden layer by trying.
The figures that gave the best results were accepted in October 21, 2017. Caption: Fig. 1.
Measurement settings. Caption: Fig. 2.
Cascading architecture. Caption: Fig. 3.
CCNN structure for Determining battery type. Caption: Fig. 4.
CCNN structure used to determine the battery SoC. Caption: Fig. 5.
Battery analysis.