Jung-Hyun Kim, “태양광 배터리 충전기를 위한 적응형 신경회로망-퍼지로직기반의 센서리스 MPPT,” 아주대학교 공학석사 학위 논문, 2014. > Paper

Recently, there is a rising problem of energy source depletion along with an environmental problem due to the rising use of energy

such as oil or coal. To resolve this problem, various renewable energies has been researched, and, among them, the photovoltaic

(PV) generation system is one of the most widely accepted renewable energy [1]-[3]. The maximum power of solar cell module varies

with external factors such as the solar radiation, load, external temperature and so forth; hence, the maximum power point tracking

(MPPT) control is required. In the case of MPPT control, the most prominent methods are the perturbation and observation (P&O)

method and incremental conductance (IncCond) method[4]-[7]. However, these conventional methods track maximum power point

(MPP) by detecting voltages and currents of solar cell outputs, thus, requiring a sensor for solar cell outputs. These sensors raise the

cost of systems; in the case of breakdown, the halt of the system becomes inevitable. To reduce the number of sensors used for the

MPPT operation, the sensorless MPPT method that do not use voltages and currents of the solar cell is proposed[8]-[9]. In the

sensorless MPPT method, the response characteristic and stability are very important for tracking MPP. The conventional MPPT

method performs the MPPT control by fixing the duty ratio. Such fixed duty MPPT control considers transient-state response

characteristic and the vibration is generated at MPP when a large changing rate of duty is configured. The vibration from MPP causes

ripples of battery input current leading to the raise of battery temperature and reducing the battery life cycle. On the other hand, if

the amount of change of duty is set small prioritizing the steady-state stability, the MPPT tracking speed slows down. Also, in the

abrupt solar radiation changing condition, the MPPT control could fail [10]-[12]. If such phenomenon lasts for a long period time,

the power loss accumulates; to improve such a situation, the fuzzy controller can be applied[13]. In the fuzzy controller, depending

on the fuzzy inference method and de-fuzzification method, the membership function width causes influence on the response

performance of the controller. Since such a width of membership function is determined at the beginning of the system design, due

to the solar cell characteristics (like solar radiation, temperatures, and etc.), the adaptive control to a variety of changing parameters

are still insufficient. To resolve this problem, the optimized fuzzy membership function width is needed for the duty and the output

current. Therefore, by using the neural-network that can learn and has an excellent information distribution process ability, the

resulting value for the membership function width is taught while determining the membership width leading to the improvement

of the operation of the fuzzy controller[14].