Voltage Regulation of Boost Converter using Observer based Sliding Mode Controller

This study dealt with output voltage regulation of boost converter using observer based sliding mode controller comprises of adaptive PI sliding surface. Observer was designed to estimate the inductor current value, such that no sensor was required as a feedback. Adaptive PI sliding surface was constructed from the difference between estimated inductor current and its reference value. The stability of proposed method was ensured by using Lyapunov direct method. To test the system performance, numerical simulation was conducted. The result indicated that the integral absolute error value of proposed method was 0.19, which was 7 times less than sliding mode with PI sliding surface. Consequently, the proposed method was able to estimate accurately the inductor value, track the reference voltage perfectly, and show its robustness against parameter variations.


Introduction
Research focus on power electronics has been growing rapidly due to the more extensively developed renewable energy sources [1].This is the reason why power electronics devices, such as dc-dc converters, have to be better day by day in order to have maximum efficiency and well-controlled.In general, dc-dc converters are employed to alter the voltage level from renewable energy sources to voltage level consumed by either load or grid [2,3].There are many types of dc-dc converter, namely buck [4], boost [5], buck-boost [6], SEPIC [7], Cuk [8], and Zeta [9] converters.Among the others, boost converter is the most widely applied converter, which is able to raise the voltage level.Since boost converter comprises an active component (MOSFET/IGBT) and some passive components (resistor, inductor, and capacitor), regulating the output voltage this converter is quite challenging.In addition, this converter also has nonlinear and non-minimum phase features, in which the controlling of output voltage is not able to be done directly such that inductor current is entailed to regulate it [5].Consequently, the inductor current needs to be measured continuously by using sensors.The problem arises when the additional sensor is installed to measure inductor current.It is very vulnerable to noise and can lead to precision error.Besides, it also can increase the production cost and space requirement.Therefore, the observer is proposed to estimate the inductor current continuously such that no sensor is required for real time measurement.
Several control strategies have been conducted for voltage regulation of boost converters, e.g., PID control [10][11][12][13], adaptive control [14,15], robust control [16,17].There are some drawbacks in using those methods.The procedures for obtaining the control parameters require linearization is some certain operating points, which results narrowness of the operating points.Consequently, the converter is not able to reach all operating range and conditions.To deal with nonlinear system, like boost converter, intelligent control [18][19][20][21][22] strategies have been applied to cope with this problem.However, the stability of this controller is not able to be analyzed.Other approaches are by using Sliding Mode Control [5,[23][24][25][26][27][28][29] strategy to control the output voltage of boost converter.Unlike other controllers, this controller offers good robustness and stability for nonlinear system, especially boost converter [5,[23][24][25][26][27][28][29].In the previous works, the sliding surface is static, whose parameters are fixed.In order to enhance the system ISSN: 1693-6930  performance to be more responsive, fast, and robust; this study applies the dynamic sliding surface, whose parameters are able to change as the system parameters change.The research on dynamic sliding surface has been successfully applied in [30], but there are no literatures discussing and implementing the combination of this method with observer on boost converter.Therefore, the main contribution of this paper is emphasized on how to design the sliding mode controller which has adaptive sliding surface to regulate the output voltage without sensor necessity.

Proposed Method
The mathematical model of boost converter and the procedures in how to design the proposed observer and controller are described in this section.

Observer Design
Boost converter's equivalent circuit is depicted in Figure 1.Both Kirchoff current and voltage laws are employed to obtain the mathematical model of this converter, which results nonlinear features [5].The deeper explanation in how to obtain the dynamic model of this converter is studied in [2,5].This converter is assumed to be an ideal model, in which parasitic effects are not taken into account.In addition, this converter is also regarded in continuous conduction mode (CCM), in which the inductor current never reaches zero or always be positive.As a consequence of those assumptions, the mathematical model of this converter can be obtained as where In this study, the observer of the converter is employed to estimate the proper value of the inductor current and output voltage.Thus, the characteristics of observer has to resemble the real plant.Theferefore, the observer is constructed as where  represent an observer gain; C v ˆ and L i ˆ denote the estimates of C v ˆ and L i ˆ, respectively.By defining estimation errors as . Thus, the estimator error dynamics can be obtained as

Sliding Mode Control with Adaptive PI Sliding Surface
The most important part in designing the sliding mode control is the sliding (switching) surface structure.Sliding surface is determined to force the states from any initial points to the equilibrium.Therefore, this is a crucial factor for control engineers requiring to be considered.A fundamental method in designing observer and controller is focused on Lyapunov stability.The procedure to obtain proposed controller is as follows: Step 1: Determine the observer gain value in sense of Lyapunov stability A positive definite of Lyapunov function is selected as afterwards, the time derivative of ( 4) is obtained by incorporating (3) into (5), it results in order to achive stable in sense of Lyapunov, equation in the bracket should be positive definite.Consequently, it can be determined that Step 2: Determine the sliding surface In this paper, the PI structure is adapted to construct the sliding surface.There are two types of PI sliding surface compared, namely static and dynamic sliding surface.The words "static" and "dynamics" refers to the characteristics of the controller parameters.In static sliding surface, the parameters are determined as constants, while in dynamic, the parameters can be varying regarding to the adaptive mechanism and parameter variations occurence.Therefore, the sliding surface for adaptive PI is designed as where meanwhile, the sliding surface for static PI is simply constructed as where  represents the constant sliding surface parameter.Thus, the difference between ( 8) and ( 10) is only in sliding surface parameter  .
Step 3: Obtaining the control signal In sliding mode control, there are types of control signal, namely equivalent and natural control signal.The equivalent control signal is resulted from the time derivative of sliding surface.By defining control error as and substituting it into (10), it allows consequently, the equivalent control signal for PI sliding surface can be obtained as the natural control signal can be derived from Lyapunov function as by taking time derivative of (12), it can be obtained where sgn( ) denotes signum function.Therefore, the control signal for PI sliding surface becomes (15) on the other hand, by modifying ( 15) into ( 16) and assuming that ( 16) is the control signal for adaptive PI sliding surface as where  ˆ is the switching gain parameters, which varies in time.It can be clearly seen that another probem arises in determining the ˆ and  ˆ Step 4: Obtaining the adaptive mechanism for ˆ and  ˆ The adaptive mechanism for these parameters is obtained by using Lyapunov method.Time derivative of (8) can be obtained by incorporating ( 16) into (17), it results the Lyapunov function is introduced as where  can be assumed as positive definite constant and  ~ is estimation error of  , which is defined as by substituting ( 18) and ( 20) into (21), it is obtained (22) Since (20) has to be negative definite to satisfy Lyapunov stability,   ˆ and  where  is any positive definite constant.As a result, the time derivative of Lyapunov (22) becomes thus, it is theoretically concluded that adaptive sliding surface for sliding mode controller aimed for boost converter is asymptotically stable.

Research Method
Figure 1 depicts the block diagram of whole system.It can be clearly seen that the system does not require inductor current sensor.The input voltage and output voltage are fed to the observer obtained as (2) to estimate the inductor current value.Consequently, the sensor can be replaced and inductor current value can be accurately obtained.The inductor current and output voltage estimation values are considered for controller design.The proposed controller provides gating signal obtained as ( 16) that is able to drive the MOSFET such that the output voltage can follow the desired reference voltage eventhough there are many disturbances occur in this system.

2901
The selection of parameters is considered based on ( 7), (19), and (23), which are derived from Lyapunov stability.Table 1 provides the system parameter used in this research.In Table 1, the parameters include boost converter, sliding mode control with adaptive PI sliding surface, and sliding mode control with PI sliding surface.To verify the proposed method, the numerical simulation is conducted to provide clear and detail descriptions about its properties.In this paper, there are two types of sliding surfaces designed for sliding mode control.As a result, there are further detail discussions about the advantages and drawbacks of those surfaces which are later to be described in Section 4.

Results and Analysis
The performance of both systems are proved through simulation, such that comprehensive results are acquired.There are two types of system testing used in this paper.Firstly, the system is tested by varying input voltage, resistor load, and reference voltage simultaneously in the same time frame.The simulation time is set to be 0.9 seconds.Figure 2 ilustrates the reference voltage, resistor load, and reference voltage changes applied for system testing.The reference voltage is increased by 20V to 70V at t = 0.45 s.In addition, there are many fluctuations on input voltage and resistor load.In the first 0.15 seconds, the input voltage is kept to be constant to 12V, then it is raised to 27V before its dropping by 25% at t = 0.6 s.At the same time, the resistor load doubled from its nominal value in the first 0.3 seconds before its fall to 62.5Ω at t = 0.75 s.Secondly, by using simulation time 0.3 seconds, the resistor load and input voltage are varied at t = 0.15 s.
Figure 3 depicts the comparison of output voltage boost converter between adaptive PI sliding surface and PI sliding surface.Clearly, the performance of sliding mode control with adaptive PI sliding surface is better than PI sliding surface in terms of voltage deviation and time.Although there are any load and input voltage variations, the output voltage of boost converter with adaptive PI sliding surface is able to return to the desired voltage value in less than 6 ms.This can be considered extremely fast compared to PI sliding surface, in which the difference is quite significant almost 19 times faster when the input voltage rises to 2.25 times from its nominal value.The same thing also goes to voltage deviation.Voltage deviation (∆V) is defined as the difference between peak value and reference value.The largest gap of voltage deviation occurs when the resistor load value is added by 25Ω to 75Ω, in which the difference is almost double.The summary of this testing is clearly provided in Table 2. To measure the performance quality, Integral Absolute Error (IAE) is employed.It is clearly seen that the IAE value of adaptive PI sliding surface is about 7 times less than PI sliding surface.Thus, this also indicates that proposed method generates fast, robust, and smooth response for boost converter voltage regulation.The comparison results of load variation and input voltage variation testing are presented in Table 3 and Table 4, respectively.Overall, IAE ratio of both methods experiences gradual increase when the load and input voltage are rised.However, eventhough the adaptive PI sliding surface has shorter recovery time and smaller IAE compared to PI sliding surface, its voltage deviation is quite large for the load added not less than its double.When the load is added more than its double, adaptive PI sliding surface shows its effectiveness.The same thing also happened to input voltage variations, in which the small changes in input voltage lead the voltage deviates quite large.On the other hand, when there is a significant increment of input voltage more than its double, the performance of adaptive PI sliding surface produces the best result regarding to voltage deviation, recovery time, and IAE.

Conclusion
This paper presents a novel technique of controller design using sliding mode control applied for boost control voltage regulation.Based on theoretical proofs and numerical simulation, the proposed method is always able to accurately follow the desired voltage in very short period.In addition, it is more robust in coping the parameter variations compared to another method.Finally, the proposed method consistently produces the least value of IAE and recovery time.

Table 1 .
System Parameters

Table 2 .
Performance Specification of Overall Conditions

Table 3 .
Performance Specification Regarding to Load Variations

Table 4 .
Performance Specification Regarding to Input Voltage Variations