Design of a new fast and lossless intelligent charger based on XC164CM

　　According to Maas's law, for fast and lossless charging of batteries, the charging current should be equal to or close to the current that the current battery can accept, so as to ensure the lowest gassing rate and reduce the damage to the battery during fast charging. Recently, advanced intelligent control technology has been introduced into fast charging technology for stop charging control or charging mode selection to improve control accuracy and charging efficiency; however, it does not consider the charging characteristics of the battery itself, lacks adaptive capabilities, and cannot track changes in battery charging characteristics to dynamically adjust the charging current, resulting in a charging current greater than the current that the battery can accept, causing excessive temperature rise and damage to the battery. To this end, it is necessary to design a new type of intelligent charger that can charge the battery safely, losslessly, and quickly.

　　After in-depth research on fast charging theory, starting from the characteristics of nickel-cadmium batteries, we innovatively proposed to introduce adaptive fuzzy neural network (ANFIS) to predict the acceptable current of the battery at different states of charge, so as to adjust the actual charging current; at the same time, negative pulse depolarization is added during charging. On this basis, a new design of fast and lossless intelligent charger is proposed using Infineon's single-chip microcomputer XC164CM and peripheral interface circuits.

　　1 Study on the Charging Process Characteristics of Ni-Cd Batteries

　　The charging curve of a single nickel-cadmium battery is shown in Figure 1. The entire charging process can be roughly divided into four stages.

　　Figure 1 NiCd battery charging characteristic curve

　　When the terminal voltage of the battery is lower than 1.2 V and reaches point A, the discharge should be stopped immediately. Excessive discharge will cause a large temperature rise. During the charging process, the main charging stage is the AB segment. More than 70% of the energy of the entire battery is charged in this stage, and the voltage rise rate is slow. At the same time, the electrochemical reaction in the AB segment generates oxygen at a certain rate, and oxygen recombines with hydrogen at the same rate, so the temperature rise and gas pressure inside the battery are low. During this period, it is suitable to use high-current fast charging, but its charging current must be less than the acceptable current of the battery, otherwise a large amount of gas will be generated, reducing the charging efficiency, and the temperature rise will be too high, causing damage to the battery. In the BC segment, the terminal voltage of the battery rises very quickly, and the internal impedance of the battery increases at this time, so it is appropriate to reduce the charging current. In the CD segment, it enters the stop charging stage. Pay attention to timely stop charging detection and staged. In the OA stage, a small current pre-charging is used; when reaching point A, it enters the fast charging stage, where a large current pulse intelligent charging is used; in the BC segment, a small current supplementary charging is used, and finally to the CD segment to stop charging detection.

　　2 Fast and lossless charging strategy

　　The literature mentions that the battery can be simply regarded as a super-large resistor, and the battery charging process can be regarded as the charging process of an RC circuit. Its time constant τ represents the speed of charging, which is equivalent to the attenuation ratio α in the Maas curve, so τ=1/α. The size of the acceptable current of the battery during charging is only related to the initial current I0. After t=3τ, the acceptable charging current of the battery is about I0/20; when charging to t=5τ, the acceptable current of the battery is already very small.

　　Therefore, it is proposed to use adaptive fuzzy neural network ANFIS to predict the acceptable current of the battery. During the rapid charging process of the battery, the acceptable current is predicted according to the battery's state of charge to ensure that the charging current conforms to the optimal charging curve of Maas, the gas evolution rate is low, and there is no damage to the battery. The basic idea of ​​ANFIS predicting the acceptable current of the battery is: during the charging process, the state parameters of the battery are dynamically detected as the input of the ANFIS prediction model, and the current acceptable current ick is obtained through fuzzy reasoning. When the error between the predicted value ick and the expected value icp does not meet the requirements, the adaptive fuzzy controller generates a control response. Through the self-learning ability of the neural network, the output results of the hidden layer are adaptively corrected, the connection weights between the layers are updated, the fuzzy parameters are optimized, and the output results are recalculated until the error meets the requirements before the prediction result is output, thereby changing the current charging current so that the actual charging current is always close to or equal to the acceptable current. At the same time, negative pulse charging is introduced to eliminate the polarization effect.

　　3 Hardware Design

　　The system hardware circuit mainly includes three parts: power supply circuit, charging/discharging circuit, and current detection and protection control circuit.

　　3.1 Power Circuit

　　In order to reduce the size and improve the power density of the system, the TOPSwitch-Ⅱ series TOP224Y produced by PowerIntegrations was selected to design the power supply circuit. This series of switching power supply chips integrates PWM control circuit, protection circuit and power switch into the same chip. It has the characteristics of high integration, high working efficiency and simple peripheral circuit design, which is very convenient for the design of flyback switching power supply below 150 W. The power supply circuit is shown in Figure 2. [page]

　　Figure 2 24V/40W power supply circuit

　　The performance indicators of the design are as follows:

　　1) Input voltage: Uac = 220 (1 ± 20%) V; 2) Input voltage frequency: f = 50 (1 ± 5%) Hz; 3) Output voltage/maximum output power: 24 V/40 W; 4) Switching power supply efficiency: η ≥ 80%.

　　After the AC input voltage Uac is filtered by the varistor R1 to remove the spike pulse in the AC voltage, the electromagnetic interference (EMI) filter (C1, L1) is used to filter out the differential mode and common mode interference. After full-wave rectification by BR and filtering by C2, a DC high voltage is generated to supply power to the primary winding of the high-frequency transformer. P6KE200 (transient voltage suppressor) and BYV26C (ultra-fast recovery diode) form a clamping circuit to absorb the spike voltage generated by the high-frequency leakage inductance when the TOP224Y is turned off, and can attenuate the ringing voltage to protect the drain. The secondary circuit outputs a 24 V voltage U0 through rectification and filtering by VD3, C3, L2 and C4. The external error amplifier composed of TL431A realizes the dynamic voltage regulation of U0. When the output voltage fluctuates, the sampling voltage is obtained after voltage division by R4 and R5, and then compared with the reference voltage (2.5 V) in TL431A to generate an external control signal, and then the TOP224Y control current is changed through the linear optical coupler PC817A, thereby adjusting the duty cycle to make U0 stable. C7 filters the peak voltage applied to the control end, and also compensates the control loop together with R2 and R5. R3 is the minimum output load, which is used to improve the voltage stability under light load.

　　3.2 Charging Circuit

　　The charging circuit is shown in Figure 3.

　　Figure 3 Schematic diagram of charging circuit

　　The charging circuit adopts a Buck topology structure. C1, L1, and C2 form a π-type filter that can filter out high-frequency components in the DC voltage, where L1 is a differential mode inductor.

　　After filtering and output, PV is the input power of Buck converter, and also the front-stage input power of single-chip control system. L2 is output filter inductor, C5 is output filter capacitor, Q3 is power switch tube, and D3 is freewheeling diode. The input voltage range of charging circuit is Ui=20~28 V, the output voltage range is U0=3~18 V, the load output current is I0=0~3.5 A, the switching frequency is fs=20kHz, the ripple voltage is less than 1%, that is, △U0/U0≤1%. When the load current I0 is 0~0.4 A, the Buck circuit works in the discontinuous inductor current mode; when the load current I0 is 0.4~3.5 A, the circuit works in the continuous inductor current mode.

　　3.3 Negative pulse discharge circuit

　　Nickel-cadmium batteries have a memory effect. Before charging, they should be discharged to eliminate the memory effect. At the same time, in the fast charging process of nickel-cadmium batteries, in order to eliminate the influence of battery polarization, intermittent negative pulse discharge is introduced, and a discharge circuit is designed in the system. The discharge circuit consists of 4 5Ω/3 W power resistors (instant short-time discharge) and 4 control switches.

　　3.4 Current detection and protection circuit

　　The current detection and overcurrent protection circuit is shown in Figure 4. The current sampling input terminal is connected to the negative terminal of the battery pack (BAT-), and between BAT- and the ground are the power switch tube IRF7805 and the constantan wire sampling resistor RS (29 mΩ). When the switch tube is turned on, the on-resistance RDS (on) between the drain and source is 11 mΩ, and the voltage drop at the RS+RDS (on) terminal is used to detect the current. [page]

　　Figure 4 Current detection and overcurrent protection circuit

　　When overcurrent occurs (current exceeds 4 A), comparator U2A outputs a low-level overcurrent signal (FAULT), which is sent to the interrupt trap pin of XC164CM to trigger the microcontroller hardware interrupt. In addition, when FAULT is low, comparator U2B outputs a low-level signal, which also forces the PWM output to be low, forcibly turning off the switch tube Q3 to ensure system safety.

　　4 Software Design

　　The software design of the intelligent charger mainly includes the system main program, nickel-cadmium battery fast charging subroutine, ANFIS prediction current subroutine and fault alarm program, etc., which are mixed programming using C language and assembly language and completed on the Keil C166 software development platform. The system software uses assembly language files in START_V2.A66 to set the special function registers SFR of the XC164 microcontroller, while the control program of the entire charging system uses C language files.

　　After the system is powered on, it enters initialization, reads the parameters in the E2PROM, completes the function settings of each interrupt register and I/O port, and assigns initial values ​​to the corresponding units. After completion, it enters the standby state. When charging starts, it first detects whether there is a battery connection. If a battery is detected, it enters the fast charging process of the battery. The flow chart is shown in Figure 5. The ANFIS predicts the acceptable current subroutine as shown in Figure 6.

　　Figure 5 Single-cell NiCd battery intelligent charging flow chart

　　Figure 6 ANFIS predicts acceptable current subroutine diagram

　　All control procedures are completed through interrupts, including the T12 cycle interrupt to achieve ANFIS prediction current and charging current control, the T3 cycle interrupt to achieve charging control and depolarization subroutine control, and the CCU6 hardware trap interrupt to achieve power supply overcurrent/short circuit protection. [page]

　　For the fast charging stage of nickel-cadmium batteries, a method combining adaptive tracking of the acceptable current change of the battery and negative pulse charging is adopted. In the AC section, the voltage and current information of the battery is detected every 2 minutes as the input data of the ANFIS model. The acceptable current ick of the battery at the next moment is predicted by ANFIS. The ick is output only when the prediction result meets the requirements and sent to the microprocessor as the actual charging current size. The output voltage of the charging circuit is adjusted through the single-chip microcomputer control to provide the battery with the current of ick for constant current charging. In the AB section, intermittent negative pulses are used to eliminate the polarization effect. The charging is stopped for 2 ms to eliminate ohmic polarization, and then the discharge current of about 2.5 times the charging current is used for 3 ms to effectively eliminate concentration difference polarization and electrochemical polarization. The charging current is restarted 5 ms after the discharge is terminated.

　　When the battery is charged to 85%, it is close to being fully charged; after that, the polarization of the battery is serious. At this time, even after adding negative pulse depolarization, the acceptable charging current of the battery is still very small. Therefore, after detecting that the charging current ic≤I0/10, stop calling the negative pulse depolarization subroutine.

　　5 Test results

　　In the intelligent charging process of nickel-cadmium batteries, the maximum charging current of the battery is about 8.75 A. After about 2.3 hours, the voltage rises to 17.6 V in 2.65 hours, and the charging current is reduced to 400 mA (about 0.1 C current). When a voltage drop of 100 mV is detected, the charging is terminated and the full indicator light is on. The total charging time is 2.85 hours. During the entire charging process, the current is relatively large in the early stage of charging. The charging current in the first 50 minutes is greater than 2 A, which can charge the battery quickly; when the charging time is 100 minutes, the charging current is about 1 A (0.22 C), and the amount of electricity charged at this time is 65% C; and the current drops to a very low level in the later stage of charging, which is completely in line with the charging characteristics of the battery itself. During this period, due to the introduction of negative pulses, the influence of polarization is greatly reduced, and the temperature rise and bubble generation caused by polarization are eliminated. Therefore, the highest temperature of the battery during the entire charging process is 38.5°C. The charging current curve is shown in Figure 7. The negative pulse waveform is shown in Figure 8.

　　Figure 7 Charging current curve

　　Figure 8 Negative pulse when charging current is 2.2A

　　6 Conclusion

　　Through in-depth research on the charging characteristics of nickel-cadmium batteries, it is concluded that the charging acceptance rate of batteries under a certain state of charge is certain. This paper innovatively proposes to use the prediction function of neural networks and the decision rules of fuzzy control to predict the acceptable current of the battery, and uses Infineon's single-chip microcomputer to design related hardware circuits. The designed charger introduces fuzzy neural network ANFIS to predict the acceptable current of the battery during the fast charging process, ensuring that the charging current is close to the acceptable current of the battery. The battery is charged under ideal conditions with high charging efficiency, achieving safe, fast and lossless charging, high charging quality, and solving the contradiction between fast charging and battery life.