Power oscillation damping control using PI-neuron network controller for distributed generator grid connection with MVDC

Distributed generator (DG) connection to the system with the DC grid called medium voltage direct current (MVDC) grid connection has received attention and gradually integrated into the distribution grid. The linear controller, such as the PI controller, usually uses the MVDC grid control for power oscillation damping. The PI controller is limited and does not show satisfactory results when the load and parameters of the system itself are changed. This paper proposes the PI with neuron network (NN) as a feed-forward controller (PINNF) to improve the control of the DG grid connection with the MVDC. The proposed PINNF controller is applied to control the power oscillation damping. Since the NN can estimate the proper feed-forward control signals in each situation to the control system, the proposed PINNF controller performs better than the conventional PI controller. The effectiveness of the proposed PINNF controller is validated using nonlinear dynamic simulations on the MATLAB/Simulink program. Four case tests are presented and discussed in this paper. Results indicate the improvement of power oscillation damping stability and performance in the MVDC grid connection with the proposed PINNF controller.


INTRODUCTION
Distributed power generation technology has been developed to support distributed energy sources in many areas.The advantage of distributed power generation in terms of power management and distribution is that it distributes power generation that allows the power system to have a two-way flow [1].The DGs are usually small sets of synchronous generators ranging from a few kilowatts to tens of megawatts designed, installed, and operated in a distribution network [2].However, connecting the DG to the system has more stringent user conditions and requirements for power quality that traditional AC distribution networks may not be able to support [3], [4].The power oscillation problem due to the inertia of the synchronous generator is another quality issue that has been a concern for DG connections [5]− [7].The oscillation is observed while the synchronous DG is connected to the system.Therefore, the oscillation problem is essential and may lead to not continuing operations to connect the synchronization DG in the power system [8].In the transmission system, the high voltage direct current (HVDC) transmission line and the flexible AC transmission systems (FACTS) family with power oscillation damping mod has been popularly applied to power grids to improve the power oscillation [9]− [11].Since the HVDC and FACTS include the main part with the Voltage Source Converter (VSC) that has the advantage of independent and fast control of active and reactive powers, the power oscillation can be controlled.Concurrently, a similar smaller-scale technology called medium voltage direct current (MVDC) has been interested and gradually integrated into the distribution grid.The integration of DG on many different levels of the system offers potential benefits relying on MVDC technology compared to traditional AC systems [12].Figure 1 shows the MVDC grid connection with a small 5MW DG synchronous generator connected to the DC bus.The MVDC grids have a better transfer capacity, flexibility, controllability, and power supply reliability.The MVDC is gaining more attention for distribution in power systems, mainly used successfully between distribution grids [13]− [15], where it can be used for various renewable power sources.

Figure 1. Schematic diagram grid connection with MVDC system
There are several applications of the MVDC in the distribution system [16], [17].Joseph et al., [16] proposed MVDC link between the mainland in North Wales and the island of Anglesey, while the MVDC and Grid Interties applications are presented in [17].The MVDC in DC microgrid application is illustrated in [18], covering AC interfaces, architectures, possible grounding schemes, power quality issues, and communication systems.Furthermore, the MVDC is introduced to improve the power oscillation damping of the DG grid connection, as presented in [12] and [19].Therefore, to improve the power oscillation of the DG connection, the MVDC grid connection is proposed and investigated in this paper.
However, the DG grid connection with the MVDC is a nonlinear system, leading to being hard to determine adequate parameters and design the control.The linear control theory is proposed based on the linear approximation technique [20] for analyzing and designing the control system.The PI controller is the linear controller popularly used with the rotating reference frames using Park transformation, a widespread technique popularly applied to control various VSCs, as illustrated in [21]− [22].Mehmood et al., [23] proposes a doubleclosed-loop control topology based on the rotating reference frames with the PI controller to control the MVDC.Therefore, the PI controller based on the rotating reference frames is used in this paper.
Although PI controller is generally applied to MVDC control, it is linear and symmetric.So, it has limitations for highly nonlinear systems.The PI controller is designed using a linear model, often obtained by linearization at specific operating conditions.It is limited when a defined boundary is unavailable or may not perform well when it is used outside a particular boundary.In addition, the PI controller has the overshoots and undershoots in the control system's output, and its parameters must be adjusted if the load and the system itself are changed.Hence, tuning and optimizing this controller is challenging and complex, particularly under varying load conditions, parameter changes, and abnormal modes of operation [24].The performance of the VSC with the PI controller to parameter variations and uncertainties can be significantly increased with the help of feed-forward control, as shown in [25], [26].However, the feed-forward control signals of those studies are obtained by measurement and calculation with the system's parameters that variations and uncertainties so that it does not give satisfactory results when the parameters of the system itself are changed.This problem can ISSN: 2088-8694  Power oscillation damping control using PI-neuron network controller for distributed … (Kittaya Somsai) 2543 be mitigated by using the NN to estimate the proper feed-forward control signals in each situation to the control system.
The NN of living things inspires many technologies.Researchers from many disciplines have designed NN to solve pattern recognition, prediction, optimization, and control systems [27]- [30].The NN can operate beyond a defined boundary and become an alternative solution for control systems.Controllers based on NN principles benefiting from the learning capabilities of NN are suitable for adaptive controls where the controller needs to adapt to environmental changes.The application of the NN to damp power oscillation is shown in [31], and the results show that the NN can emulate nonlinear dynamics and show a promising robust nonlinear performance.
This paper presents the study of the DG with MVDC grid connection on the distribution system to solve traditional AC connection problems.This study focuses on the MVDC control to solve power quality in the event of power oscillation at the connection point of the DG.The conventional PI controller and the proposed PINNF controller are tested and compared.The four case tests are implemented by MATLAB/Simulink program simulation.

MODELING OF DISTRIBUTION GENERATOR GRID CONNECTION WITH MVDC
This paper presents the DG grid connection with the MVDC.The schematic diagram is shown in Figure 2. It consists of three parts: 1) the DG and local load 2, 2) the MVDC, and 3) the Utility Electrical Power Station (EPS) and local load 1.In part 1, the DG is connected to the MVDC through the step-up transformer.At this point, the local load 2 has been installed.This part is called DG ZONE.The MVDC in part 2 consists of two connected back-to-back VSCs with a common DC link voltage.Two VSCs have to be connected to the AC system with the interface impedances so that the interface impedances can function as the AC filter for VSC.The last one, part 3, consists of the EPS and local load 1.This part is directly connected to the MVDC.The equivalent circuit of distributed generator grid connection with MVDC is shown in Figure 3.In this figure, the load is modeled by a series of resistance and inductance.Simplification assumes that the source, load, and MVDC are balanced.By applying the abc-to-dq transformation, the mathematical model of the system can be obtained, as shown in ( 1)-( 16).

Mathematical of DG ZONE
In this paper, the three-phase salient-pole synchronous generator is used.The dynamic model of the generator referring to the rotor reference frame can be expressed in [32].The (1)-( 2) are the stator voltage equations, and the (3)-( 6) are the rotor voltage equations, while the relationship between the torque and rotor speed can be written as shown in (7).
Where and r kq2 ' , and the resistance of field winding is r fd ' .The v kd 'r , v kq1 'r and v kq2 'r are the voltage of three damper windings, and the v fd 'r is the field winding voltage.The ω r is rotor speed used as the reference frame speed or the reference frame fixed in the rotor.The input and electromagnetic torque in the rotor reference frame are T m and T e , respectively.Where J is the rotor's inertia, while P is the number of poles in the machine.
The currents (i qs r , i ds r , i kq1 'r , i kq2 'r , i kd 'r and i fd 'r ) in the (1)-( 6) can be obtained by using (9). [ The i qs r and i ds r are the stator currents in dq axes, the i kq1 'r , i kq2 'r and i kd 'r are the rotor currents of three damper windings, and the i fd 'r is the current of the field winding.The L md and L mq are the magnetizing inductances in dq axes, and the L ls is the stator leakage inductance, while the leakage inductances of three damper windings are represented byL lkq1 , L lkq2 and L lkd , and the field winding is represented by L lfd .
The DG is connected to the VSC at BUS 2 via a step-up transformer with the ratio a.However, the transformers are assumed to be ideal for more accessible consideration.Therefore, the impedance (R T and L T ) of the transformer can be ignored.It means , the a is a step-up transformer ratio.The local load 2 resistance and inductance are represented by R l2 and L l2 , respectively, while the time constant Power oscillation damping control using PI-neuron network controller for distributed … (Kittaya Somsai) 2545

Mathematical of MVDC
Although MVDC consists of two VSCs, this paper highlights the details of the VSC2.The paper's main objective is to implement power oscillation damping control at the DG, which can be achieved with the control at the VSC2.Therefore, in this paper, the details of the mathematical model of the VSC1 are not presented.The VSC2 has connected the DG ZONE at BUS V2 with interface impedance (R con2 , L con2 and C con2 ).In addition, the R con2 , L con2 and C con2 are the RLC filter of the VSC2.The mathematical model of the VSC2 can be expressed as follows: Where the i gd =i ds r and i gq =i qs r .The i con2d and i con2q are the VSC2 currents on dq axes.The output voltages on dq axes of VSC2 are represented by v con2d and v con2q , respectively.The VSC2 output voltages can be expanded by v con2d =k p u 2d v dc and v con2q =k p u 2q v dc .Whereas the v dc is the DC voltage of the MVDC, and the k p is a constant value of VSC.The control signal u 2d and u 2q can control the VSC2 output voltages.According to [22], showing the relation between AC power and DC power of the VSC, the power flow at the VSC2 can be derived as, where C dc2 is the DC link capacitor, and the resistor R dc2 represents the loss in the VSC2.The i dc is the DC current from the VSC1.

CONTROL SYSTEM AND DESIGN
The detail of distributed generator grid connection with MVDC and its controller is shown in Figure 4.As can be seen in this figure, the control system consists of two controllers: a VSC1 and VSC2 control.The functional control of the VSC1 regulates the DC voltage at DC BUS and controls the AC voltage at BUS V1.For the VSC2 control, the primary functional control is the AC voltage control at the load bus.Besides, the DG is connected to the system at this point, and the VSC2 is also used to control the power oscillation.Therefore, the VSC2 control consists of the AC voltage and the power oscillation controls.The detail of VSC1 and VSC2 controls can be expressed in the following sections.

Control strategy of the VSC1 and VSC2
The DC voltage is regulated by controlling the current on the d axis, while the AC voltage can be controlled by controlling the current on the q axis of the VSC1.The PI controllers are used to achieve those control functions, while the current control loop is successful using the hysteresis control.The DC and AC ).Inverse Park transformation transform these currents to the abc axes [21].The reference angle (  ) obtained from the phase-lock-loop (PLL) is applied in this process.The hysteresis current control compares the reference currents (i con1,abc * ) with the feedback currents (i con1,abc ) to generate the switching six pulses for the VSC1.The detail of the VSC1 control diagram is shown in Figure 5(a).
In this paper, the local load 2 is connected to the system at the VSC2 zone.In addition, the DG has to connect to the system at this point.The AC load voltage control is necessary for VSC2.Moreover, installing the power oscillation damping control is added for VSC2 to enhance the power quality of the DG.The AC voltage control is achieved by controlling the current on the q axis, while the power oscillation damping control can be successful by controlling the current on the d axis.This paper proposes PINNF to control the AC voltage and damp the power oscillation.The outputs of the NN are the feed-forward reference currents on dq axes (i NNd * and i NNq * ) .These currents are then combined with the PI controller's reference currents on dq axes (i d * and i d * ).Similar to the VSC1, the total reference currents (i con2d * and i con2q * ) are transformed to the abc axes using inverse Park transformation.The reference angle (θ r ) obtained from the rotor speed of DG will be applied in this process.The hysteresis current control compares the reference currents (i con2,abc * ) with the feedback currents (i con2,abc ) to generate the switching six pulses for the VSC2.The detail of the VSC2 control diagram is shown in Figure 5(b).

Control design 3.2.1. VSC1 controller design
Considering the connection network of VSC1 to the power system, as can be seen in Figure 4, it is found that the connection characteristics are similar to the connection of D-STATCOM devices to the power system.However, there is a difference that the VSC1 can exchange real power with the system.Therefore, the controller design process for D-STATCOM devices can be effectively applied to VSC1.This paper proposes the classical method called the loop shaping method [33] for tuning PI controllers of VSC1.By selecting the gain and phase margin of the control system, around 6.00 dB and 60.00 degrees,

VSC2 controller design
The dynamic equations of the DG grid connection with the MVDC are shown in ( 1)-( 7), ( 10)- (11), and ( 12)-( 16).This paper uses the linear approximation technique [20] to study the system's dynamic behavior.A set of linear equations can be derived using the partial derivatives of the dynamic equations around an operating point (17).
The ( 17) can be rewritten as, where Δx(t) and Δu(t) are the state of the system and the control input, respectively.The state (or system) matrix A 0 and input matrix B 0 can be expressed in (19) and (20).
Then, the system's output Δy(t) can be obtained, as shown in (21).
Where C is the output matrix, and D is the feedthrough (or feed-forward) matrix.The multi-input-multi-output transfer function of the system can be obtained, as shown in (22).
Where G(s) is the multi-input-multi-output transfer function, and I is an identity matrix.The system's dynamic performance can be obtained by using the transfer function in (22).The system's behavior depends the operating point, and when the operating point is changed, the system's behavior will also be changed.Two variables: generator power and current of local load 2, can affect the system's behavior, as can be seen in Figures 6 and 7.An increase in the generator power under an improper excitation voltage condition will eventually lead to an unstable system corresponding to the roots and zeros of the system, as shown in Figure 6.In addition, the next point of concern is that excessive load currents can cause the system to become unstable.In this case, the roots and zeros of the system are shown in Figure 7.
For designing the VSC2 control system, the transfer function in ( 22) is used to determine the parameters of the PI controller.Similar to VSC1, the loop shaping method is also used for VSC2.The parameters of the PI controller are obtained by selecting the control system's gain and phase margin at nearly 6.00 dB and 60.00 degrees.The parameters of the PI controller of the AC voltage and the power oscillation damping controls are K P,V =-0.3,K I,V =30.0 and K P,ω =20,000, K I,ω =900,000, respectively.Where K P,V and K I,V are the P and I controller parameters of the AC voltage control, and K P,ω and K I,ω are the P and I controller parameters of the power oscillation damping control.
Although the proposed PI controller can be used to control AC voltage and power oscillations with good stability, the PI controller is a linear controller with good response and stability only around the design working point.Therefore, a sudden change of working point in a wide area can eventually lead to instability.In order to provide the VSC2 control with good response and stability in the broader area, this paper proposed a NN as a feed-forward controller in combination with the PI controller.The NN estimates a reasonable reference current for operating conditions and feeds it to the control system while the remaining control errors are eliminated with the PI controller.The next section will show the architecture and training process of neuron networks.

Pole-Zero Map
Real Axis Real Axis dominant pole Figure 7. Roots and zeros of the system when increasing local load 2

Architecture and training process of neuron networks
The NN is usually composed of multiple neurons connected in parallel with a single or multiple layers.This paper used a two-layers (one-hidden layer and output layer) NN for model predictive to represent the plant's forward dynamics.The structure of the NN is shown in Figure 8.As shown in Figure 8, there are four neurons for the hidden layer and two for the output layer.The output in each layer can be written as in ( 23)- (24).
Where j and i are the j th input and the i th neuron of the hidden layer, respectively, and k is the k th neuron of the output layer.The y i h and p i are the output and input of the hidden layer, respectively.Meanwhile, the W i,j in and b i h are weight matrix and bias matrix of the hidden layer, respectively.This paper uses the tansigmoid transfer function (f h =tansig(n i h )) as a hidden layer transfer function when the n i h is the net input of the hidden layer.For the output layer, the output of the hidden layer y i h is used as the input, while the output is y k o .
Meanwhile, the weight matrix and bias matrix are

RESULTS AND DISCUSSIONS
This section presents and discusses comparing the proposed PINNF and the conventional PI controllers.The main objective of control is to solve power quality in power oscillation by connecting the DG to the distribution system with the MVDC.The test system and parameters are shown in Figure 10 and Table 1.The study is validated by simulation on the MATLAB/Simulink program.Among the four cases, case 1) the DG power (P G ) is increased from 0.5 MW to 2.8 MW of local load 2 power (P L2 ) at a constant of 0.6 MW, case 2) the P L2 is increased from 0.6MW to 1.6 MW of P G at a constant of 2.8 MW, case 3) the P G is increased from 0.6 MW to 4.0 MW of P L2 at a constant of 1.6 MW, and case 4) the P L2 is increased from 0.6 MW to 1.6 MW of P G at a constant of 4.0 MW.In case 1, the P G increases from 0.5 MW to 2.8 MW at t = 1sec, while the P L2 is constant at 0.6 MW, as shown in Figure 11(a).As can be seen in this figure, the P G is immediately increased and has the overshoots caused by the acceleration of the DG, while the P L2 is slightly affected.The response of P L2 is related to local load 2 bus voltage (V L2 ).The V L2 is regulated at 22.00 kV, as shown in Figure 11(b).In this figure, the responses of the proposed PINNF and the conventional PI controller are similar in performance.The disturbance response of V L2 is increased due to an immediate increase in the P G with the peak voltage of about 30.00kV, which the P L2 is slightly increased as well.And then, after about 40 msec, the PINNF and PI controllers have returned the V L2 to nearly 22.00 kV.Similarly, the P L2 is turning to the normal value.The peak voltage of the excursion is bounded by the P controller, while the I controller will eliminate the error until it is less than 100V (0.45%) in the steady-state.Generally, the rotor speed deviation of DG (Δω r ) can be affected by the P G changes and may cause oscillation.Fortunately, the DG grid connection oscillation with MVDC and its control is rapidly settled, as shown in Figure 11(c).The disturbance response of Δω r is increased due to an immediate increase in the P G with the peak of about 0.0005 p.u. and 0.0006 p.u. for the PINNF and PI controllers, respectively.Then, after about 70 msec, the PINNF and PI controllers have returned the Δω r to settle with about zero error in the steady-state.Furthermore, both the PINNF and PI control systems are stable.Although the proposed PINNF and the conventional PI control give similar responses, in this case, it is noticed that the PINNF gives slightly better results.It can be seen that the feed-forward signals of the PINNF controller show a slight improvement in the peak value and response time of the disturbance response of V L2 and Δω r .

2551
In case 2, the P L2 suddenly changed from 0.6 MW to 1.6 MW at t = 2 sec, while the P G remains constant at 2.8 MW, as shown in Figure 12(a).The increased P L2 affects the load voltage V L2 to sag.However, the V L2 is regulated at the desired value of 22.00 kV by the MVDC and its control.The disturbance responses of V L2 are shown in Figure 12(b).As can be seen in this figure, the V L2 can reach the desired value within 10 msec for the proposed PINNF controller, while the conventional PI controller gives a slower response time at about 50 msec.In addition, the proposed PINNF controller shows a better sag voltage improvement than the conventional PI controller.The PINNF controller has a maximum sag voltage of about 2.00 kV (9.09%), while it is about 7.00 kV (31.82%) of the PI controller.Similarly, to the V L2 , the increased P L2 also affects Δω r as shown in Figure 12(c).In this case, it seems that the DG has been interrupted and is trying to return to the normal operating point by accelerating and decelerating the speed of the DG; this causes the DG's power to oscillate, or the Δω r oscillation.If the DG connects to the system with traditional AC distribution networks, the oscillation can take several seconds or even minutes, adversely affecting the power system.However, this oscillation can be better improved by using MVDC, especially under the proposed PINNF controller.The maximum oscillation is limited at 0.0001 p.u. and 0.0005 p.u. for the PINNF and PI controllers, respectively.In addition, the PINNF controller gives a fast time response to the settling Δω r of about 10 msec, the PI controller has a time response of about 100 msec.However, the PINNF and PI controllers can also eliminate the steady-state error to zero with stable systems.In this case, the simulation results showed that the proposed PINNF controller provided significantly better performance than the PI controller, which means the PINNF controller shows greater performance in the disturbance response of V L2 and Δω r and nonlinear systems handling.In case 3, the P G increases from 0.5 MW to 4.0 MW, while the P L2 constant at 0.6MW is tested at time t = 1sec, as seen in Figure 13(a).In this case, a relatively higher increase in power results in a high nonlinear system.This substantial increase in power (about 8 times) brings the system closer to the region of instability, as seen from the analysis of the roots path of the system in Figure 6.The DG has highly increased the power that highly affects the load voltage V L2 and rotor speed deviation Δω r .disturbance responses of V L2 in this case, are shown in Figure 13(b).The PINNF and PI controllers can regulate the V L2 at the desired value within 40 msec under a steady-state error of less than 100V (0.45%) with the stable system.However, the PI controller's response shows two waves of the peak voltage at about 35.00 kV, whereas the PINNF shows only one peak voltage.It means that the PINNF controller can eliminate the disturbance response of V L2 better than the PI controller.Figure 13(c) presents the responses of the Δω r .As seen in this figure, the simulation results show that the PINNF controller reveals a better response than that of the PI controller, with the peak of the Δω r of 0.0015 p.u. and 0.0016 p.u for the PINNF and PI controllers, respectively.And the Δω r are settled within about 20 msec for the PINNF, while the PI controller is about 40 msec.However, in this case, the PINNF and PI controllers can also eliminate the steady-state error to zero with stable systems.
In the last case, case 4, the P L2 is increased from 0.6MW to 1.6MW at time t = 2sec, while the P G is constant at 4.0MW, as presented in Figure 14(a).In this case, the system has operated near the critical point, which is highly sensitive to changes.Although the increased power P L2 in this case is the same as in case 1, the system operates near the critical point, so it is susceptible to power changes.This results in a sharp sag voltage V L2 .So, the PI controller cannot control this voltage (unstable system), as seen in Figure 14(b).However, with the PINNF controller, the NN estimates a reasonable reference current for operating conditions and feeds it to the control system, which can regulate the V L2 at the desired value within 10 msec with a small sag voltage of about 2.00 kV.Moreover, the system with the PINNF controller shows a less steady-state error than 100V (0.45%) with stability.In this case, the responses of the Δω r are presented in Figure 14(c).As can be noticed in this figure, the simulation results show that the PINNF controller can improve the oscillation of the Δω r to settle within about 10 msec with the steady-state error to zero, while the PI controller cannot control the Δω r with an unstable system.
The graphics analysis has revealed that the proposed PINNF controller approach has yielded far more favorable results than the conventional PI controller.It has been observed that the PI controller is unsuitable for controlling such that the system operates near the critical point.On the other hand, the system remains stable when using the PINNF controller, and the oscillations are quickly settled.It can be concluded that the proposed PINNF controller provides better performance, mainly when the system operates near the critical point and high nonlinear system.

CONCLUSION
This paper presented the connection of the DG to the distribution system with MVDC.The advantages of MVDC are obtained.Examples are the possibility of quickly controlling the constant AC voltage on both sides of the DG and distribution system and improving power quality in significantly mitigating power oscillations.In this paper, the main objective of the MVDC is to improve power oscillation at the connection point of the DG to remain in synchronism.It can continuously supply power to the system in every possible change, such as the increased DG power and local load power.The results of two controllers, the proposed PINNF and the conventional PI controllers, are compared and discussed.The results of four cases confirm that the MVDC could improve the power oscillation.It is also found that the MVDC with the proposed PINNF controller gives a better response than the conventional PI controller, especially in cases where the system is The proposed PINNF control performs well and can handle high nonlinear systems because the NN as a feed-forward path can estimate the proper reference currents in each situation to the control system.In addition, the feed-forward path is outside the loop and thus does not respond to disturbances leading to a better performance of the PINNF controller.

Figure 2 .Figure 3 .
Figure 2. Schematic diagram of distributed generator grid connection with MVDC

2 ,
which the v g is the output voltage of DG at BUS G.The load currents at BUS 2 (i l2d and i l2q ) can be derived.Where v 2d =

Figure 4 .
Figure 4. Detail of distributed generator grid connection with MVDC and its controller

Figure 5 .
Figure 5. Diagram of VSC1 and VSC2 control (a) VSCI control and (b) VCS2 control Int J Pow Elec & Dri Syst ISSN: 2088-8694  Power oscillation damping control using PI-neuron network controller for distributed … (Kittaya Somsai) 2547 respectively, K P,Vdc =-0.01,K I,Vdc =-40.0 and K P,Vac =-0.1, K I,Vac =-2.0 are used.Where K P,Vdc and K I,Vdc are the P and I controller parameters of the DC voltage control, and K P,Vac and K I,Vac are the P and I controller parameters of the AC voltage control.

Figure 6 .
Figure 6.Roots and zeros of the system when increasing generator power

Figure 8 .Figure 9 .
Figure 8. Architecture of the two-lawyers neural network for model predictive

Figure 10 .
Figure 10.DG grid connection with MVDC and control on MATLAB/Simulink program Int J Pow Elec & Dri Syst ISSN: 2088-8694  Power oscillation damping control using PI-neuron network controller for distributed … (Kittaya Somsai)

Figure 11 .Figure 12 .
Figure 11.The P G increases from 0.5MW to 2.8MW of the P L2 at a constant of 0.6MW for (a) active powers at the DG and local load, (b) the load bus voltage, and (c) the rotor speed deviation Figure 12.The P L2 increases from 0.6MW to 1.6MW of the P G remains constant at 2.8MW for (a) active powers at the DG and local load, (b) the load bus voltage, and (c) the rotor speed deviation

Figure 13 .Figure 14 .
Figure 13.The P G increases from 0.5MW to 4.0 MW, while the P L2 is constant with 0.6MW for (a) active powers at the DG and local load, (b) the load bus voltage and (c) the rotor speed deviation Figure 14.The P L2 is increased from 0.6MW to 1.6 MW, while the P G is constant with 4.0MW for (a) active powers at the DG and local load, (b) the load bus voltage and (c) the rotor speed deviation Int J Pow Elec & Dri Syst ISSN: 2088-8694  Power oscillation damping control using PI-neuron network controller for distributed … (Kittaya Somsai) 2553 susceptible to working point change.