**Research Article**

# Load Frequency Control in Interconnected Power System Using Multi-Objective PID Controller

#### Journal of Applied Sciences: Volume 8 (20): 3676-3682, 2008

**
K. Sabahi,
A. Sharifi,
M. Aliyari Sh.,
M. Teshnehlab
and M. Aliasghary
**

#### Abstract

In this study, designing of multi-objective (MO) proportional, integral and derivative (PID) controller for load frequency control (LFC) based on adaptive weighted particle swarm optimization (AWPSO) has been proposed. Unlike single objective optimizations methods, MO optimization can find different solutions in a single run and we can select appropriate and desirable solution based on valuation to the objects. In this study for PID controller design, overshoot/undershoot and settling time are used as objective functions for MO optimization in LFC problem. So that various solutions with different overshoot/undershoot and settling time obtained. From these different PID parameters, one can select a single solution based on valuation to objects and as well as system constraints, reliability etc. The proposed method is used for designing of PID parameters for two area interconnected power system. From the simulation results, efficiency of proposed controller design can be seen.

#### How to cite this article:

K. Sabahi, A. Sharifi, M. Aliyari Sh., M. Teshnehlab and M. Aliasghary, 2008. Load Frequency Control in Interconnected Power System Using Multi-Objective PID Controller.Journal of Applied Sciences, 8: 3676-3682.

DOI:10.3923/jas.2008.3676.3682

URL:https://scialert.net/abstract/?doi=jas.2008.3676.3682

**INTRODUCTION**

One of the principle aspect of Automatic Generation Control (AGC) of
power system is the maintains of frequency and power change over the tie-lines
at their scheduled values. Therefore, it is a simultaneous load frequency
control (LFC) (Saadat, 2002). In LFC problem, each area has its own generator
or generators and it is responsible for its own load and scheduled interchanges
with neighboring areas. The tie-lines are utilities for contracted energy
exchange between areas and provide inter-area support in abnormal conditions.
Area load changes and abnormal conditions lead to mismatches in frequency
and scheduled power interchanges between areas. These mismatches have
to be corrected by LFC, which is defined as the regulation of the power
output of generators within a prescribed area (Kumar, 1997). Therefore,
the LFC task is very important in interconnected power systems. It is
well known that power systems are nonlinear and complex, where the parameters
are a function of the operating point and the loading in power system
is never constant. Over the past decades, many techniques have been developed
for the LFC problem. A number of state feedback controllers based on linear
optimal control theory, robust and conventional adaptive controller have
been proposed to achieve better performance (Aldeen and Trinh, 1994; Pan
and Liaw, 1989; Zribi *et al*., 2005; Bevran and Hiyama, 2008; Shayeghi,
2008a). state adaptive controllers (Kazemi *et al*., 2003) involve
large computational burden and time. Also, Most of proposed techniques
were based on the classical proportional and integral (PI) or proportional,
integral and derivative (PID) controllers. Its use is not only for their
simplicities, but also due to its success in a large number of industrial
applications. In Classical methods, such as Ziegler-Nichols and Cohen-Coon,
these controllers are tuned based on trial-error approaches and, therefore,
have not good performance. To achieve optimal gains for PID controller,
Genetic Algorithm (GA) or Particle Swarm Optimization (PSO) methods were
addressed (Ghoshal and Goswami, 2003; Ghoshal, 2004; Talaq and Al-Basri,
1999; Abdennour, 2002; Shayeghi, 2007, 2008b; Mukherjeea and Ghoshal,
2008). In mentioned study for LFC problem, the area control error (ACE),
which composed of frequency and tie-line error, was defined as fitness
for PSO and GA. Moreover, in these studied, after obtaining PI/PID gains
for some operating point of power system, adaptive fuzzy gain scheduling
technique was proposed. However, the most of real-world control problems
refer to multi-objective control designs that controller must follow several
objectives such as stability, disturbance attenuation and reference tracking
with considering of practical constraints, simultaneously. For this reasons,
in this study Multi-Objective Particle Swarm Optimization (MOPSO) is used
for tuning of PID controller parameters for LFC in interconnected power
system. Unlike classical methods such as Ziegler-Nichols and Cohen-Coon (Astrom
and Wittenmark, 1997) and single objective optimization methods such as
GA (Abdennour, 2002; Shayeghi, 2007; Goldberg, 1989; Demiroren and Zeynelgil,
2007) and PSO (Ghoshal, 2004; Shayeghi, 2008b), the MO optimization can
minimize some important aspect of a system such as overshoot/undershoot
and settling time simultaneously. So that various solutions with different
overshoot/undershoot and settling time obtained. From these different
PID Parameters, one can select a single solution based on valuation to
objects and as well as system constraints, reliability etc. For example,
in such cases overshoot/undershoot has more importance than setting time
and vice versa.

**MATERIALS AND METHODS**

**A two area interconnected power system model:** A two-area system
consists of two single areas connected through a power line called the
tie-line. Each area feeds its user pool and the tie-line allows electric
power to flow between areas. Since both areas are tied together, a load
perturbation in one area affects the output frequencies of both areas
as well as the power flow on the tie-line. The control system of each
area needs information about the transient situation of both areas in
order to brings the local frequency back to its steady state value. Information
about the other areas found in the output frequency fluctuation of that
area and in the tie-line power fluctuation. Therefore, the tie-line power
is sensed and the resulting tie-line power signal is fed back into both
areas (Ertugrul and Kocaarslan, 2005a , b; Yesil and Eksin, 2004). Frequency
control is accomplished by two different control actions in interconnected
two area power systems: primary speed control and supplementary or secondary
control actions. The primary speed control makes the initial coarse readjustment
of the frequency. By its actions, the various generators in the control
area track a load variation and share it in proportion to their capacities.
The speed of the response is limited only by the natural time lags of
the turbine and the system itself. The supplementary speed control takes
over the fine adjustment of the frequency by resetting the frequency error
to zero through an integral action. Schematic of two-area interconnected
power system for the uncontrolled case is shown in Fig.
1. The overall system can be modeled as multi-variable system in the
following from:

(1) |

where, A, B and L are the system matrix, input and disturbance distribution matrices, respectively, x(t), u(t) and d(t) are the state, control and load changes disturbance vectors, respectively and represented as:

(2) |

where, Δ denotes deviation from the nominal values. u_{1}
and u_{2} are the control outputs in Fig. 1.
The system output, which depends on the Area Control Error (ACE) shown
in Fig. 1 and represented as:

(3) |

ACE _{i} = Δp_{yie,
I} + b_{i}Δf_{i} |
_{}(4) |

where, b_{i} is the frequency bias constant, Δf_{i}
is the frequency deviation and ΔP_{tie,i} is the change in
tie-line power for the i-th area and C is the output matrix.

**Basic PSO and AWPSO:** The PSO algorithm is a population based search
algorithm based on the simulation of the social behavior of birds within
a flock. In PSO, individuals referred to as particles, are flown through
hyper dimensional search space. Changes to the position of particles within
the search space are based on the social psychological tendency of individuals
to emulate the success of other individuals. The changes to a particle
within the swarm are therefore influenced by the experience or knowledge,
of its neighbors. The search behavior of a particle is thus affected by
that of other particles within the swarm (PSO is therefore a kind of symbiotic
cooperative algorithm). The consequence of modeling this social behavior
is that the search process is such that particles stochastically return
toward previously successful region in the search space (Engelbrecht,
2006). A swarm consists of a set of particles, where each particle represents
a potential solution. Particles are then flown through the hyperspace,
where the position of each particle is changed according to its own experience
and that of its neighbors. Let
denotes the position of particle P_{i} in hyperspace, at time
step t. The position of P_{i} is then changed by adding a velocity
to the current position as:

(5) |

The velocity vector drives the optimization process and reflects the socially exchange information. Velocity update equation is as follows:

(6) |

where, w is the inertia weight, c_{1} and c_{2} are positive
constants and r_{1} and r_{2} are random numbers obtained
from a uniform random distribution function in the interval [0, 1]. The
parameters
represent the best previous position of the i-th particle and position
of the best particle among all particles in the population respectively
(Engelbrecht, 2006). The inertia weight controls the influence of previous
velocities on the new velocity. Large inertia weights cause larger exploration
of the search space while smaller inertia weights focus the search on
a smaller region. Typically, PSO started with a large inertia weight,
which is decreased over time. Shi and Eberhart (2001) proposed a `fuzzy
adaptation` of the inertia weight due to the fact that a linearly-decreasing
weight would not be adequate to improve the performance of the PSO due
to its non-linear nature. In this study, we use the following formula
to change the inertia weight at each generation:

w = w _{0} + r(1-w_{0}) |
(7) |

where, w_{0 }is the initial positive constant in the interval
[0, 1] and r is random number obtained from a uniform random distribution
function in the interval [0, 1]. The suggest range for w_{0} is
[0, 0.5], which make the weight w randomly varying between w_{0 }and
1.

To improve the performance of the PSO for MO optimization problems, Mahfouf
*et al*. (2004) proposed an Adaptive Weighted PSO (AWPSO) algorithm,
in which the velocity in Eq. 6 is modified as follows:

(8) |

The second term in Eq. 8 can be viewed as an acceleration term, which depends on the distances between the current position the personal best and the global best

The acceleration factor α is defined as follows:

α = α _{0} + t/T |
(9) |

where, t is the current generation, T denotes the number of generations
and the suggest range for α_{0} is [0.5, 1]. As can be seen
from Eq. 8, the acceleration term will increase as the
number of iterations increases, which will enhance the global search ability
at the end of run and help the algorithm to jump out of the local optimum,
especially in the case of multi-modal problems. One of the simplest approaches
to deal with MO problems is to define an aggregate objective function
as a weighted sum of the objectives. Single objective optimization algorithms
can then be applied, without any changes to the algorithm, to find optimum
solutions. We use an* *aggregation* *approach to construct the
evaluation function Eval for MO optimization as follows (Engelbrecht,
2007).

(10) |

where, n is the number of objective functions and k denotes the k-th
particle and the weights w_{i} for each objective are changed
and normalized as follows:

(11) |

where, μ_{i} and μ_{j} are random numbers obtained
from a uniform random distribution function in the interval [0, 1].

**Multi-objective design of PID controller
**

**Outline:**It is well known that the PID controller is the most popular approach for industrial process control, such as LFC problem and several design techniques have been developed. In LFC problem, by taking ACE as the system output, the control vector for a PID controller in continuous form can be given as:

(12) |

where, the K_{P}, K_{D} and K_{I} are the proportional,
derivative and integral gains. In classical methods, there are some approaches
for tuning of PID controller parameters (i.e., Ziegler-Nichols and Cohen-Coon).
In these methods, process, in response to unit step, has been modeled
as a following transfer function:

(13) |

where, k_{p}, L and T are the gain, delay time and constant time
of process, respectively. After this modulation, according to the determined
table, Ziegler-Nichols and Cohen-Coon tables, the PID parameters are achieved.
The application of these methods for PID design have been restricted for
large scale and complicated system due to, lack of accuracy and its cumbrous.
Also, population based techniques (i.e., GA and PSO) have been used for
designing of optimal PID controller parameters. In these approaches the
gains of PID controller, are searched in feasible region of response until
a determined cost function minimized. In design of PID controller parameters,
it is desirable that controlled system include suitable transient and
steady state response. So, some specific feature of system such as overshoot/undershoot,
settling time and rise time must be improved. Therefore, this design can
be mentioned as a MO optimization problem.

**Fitness functions:** For the general control problem, the optimization
of different number of systems performances is desired. The following
simultaneous performance specifications (the objectives) are adopted in
this study:

• | Overshoot/Undershoot minimization: |

(14) |

• | Settling time minimization: |

(15) |

where, OU is the average overshoot or undershoot of areas and T_{N}
is defined as follows:

(16) |

where, T_{settling time} and T_{Total} are the average
settling time of areas and final simulation time, respectively. In fact,
the T_{Total} is used for normalization of settling time object.
Here, aggregation based MO PSO is used to maximize these two objective
functions in order to minimizing overshoot/ undershoot and settling time
simultaneously. That is must be mentioned the ACEs signals in each area
for Overshoot/undershoot and settling time minimization is handled and
the same PID controllers for areas are designed.

**RESULTS AND DISCUSSION**

In this study, the nominal parameters of two area interconnected power system that has been used in the simulation are considered as follow:

T _{g} = 0.08s, T_{12} = 0.545pu,
T_{t} = 0.3s, T_{p} = 20s, k_{p} = 120Hz/pu,
R = 2.4 Hz/pu and b = 0.425Hz/pu |

where, the power system time constant, T_{p}, synchronizing power
coefficient, T_{12} and frequency bias setting b may be changed
according to different operating point of the power system.

The block diagram of controlled system for i-th area is depicted in Fig.
2. For MO optimization of PID parameters we set w_{0} = 0.15,
α_{0} = 0.5, the population size N = 30 and the number of
iteration T = 50. In addition, aggregation-based method is used for MO
PSO. In order to demonstrate the effectiveness of the proposed method,
some simulations were performed. In addition, a nonlinear model of Fig.
3 (with ±0.015 limits) replaces the linear model of a nonreheating
turbine in Fig. 1. This is to take into account the
generating constraint (GRC), i.e., the practical constraint on the response
speed of a turbine.

Fig. 1: | The two-area interconnected power system used in this study |

Fig. 2: | PID controller installed for i-th area |

Fig. 3: | Nonlinear turbine model with GRC |

Fig. 4: | Pareto front for case 1 |

Fig. 5: | Frequency deviation of three samples in first area for case 1 |

**Case 1:** In this case, the system performance with nominal parameters
is tested. The nominal parameters are set as above and apply large load
demands of Δp_{d1}(t) = 0.15 p.u. and Δp_{d2}(t)=
0.10 p.u. MW to first and second area, respectively. The obtained Pareto
front after deleting dominated solutions is shown in Fig.
4. The response of Δf_{1} and Δf_{2}, for
three selected samples (sample No. 1~3) from Pareto front are shown in
Fig. 5 and 6, respectively.

The PID gains for mentioned three samples in case 1 are shown in Table 1. These three samples and clearly other samples in Pareto front, have different PID gains, which causes we have different undershoot and settling time for frequency deviations in areas. In fact, the proposed MO optimization method gives a number of solutions for PID design problem in LFC.

Fig. 6: | Frequency deviation of three samples in second area for case 1 |

Fig. 7: | ACE signal of three samples (Three PID) in first area (case 1) |

Table 1: | PID gains for three samples in Pareto front, case 1 and 2 |

Therefore, we can choose appropriate gains based on valuation to undershoot/overshoot and settling time as well as system constraint, reliability etc. for example, it is clear the PID parameters which remarked by Sample No. 1 have the smallest and largest undershoot and settling time of others, respectively.

lso, the ACE signals for three mentioned samples at two areas are shown in Fig.7 and 8, respectively.

**Case 2:** In this case the other nominal operation conditions parameters
which (T_{p} = 30, T_{12} = 0.545, b = 0.125) are used
for two area and apply load changes of Δp_{d1}(t) = -0. 10
and Δp_{d2}(t) = 0. 15 p.u. MW to first and second areas,
respectively. The obtained Pareto front is shown in Fig.
9. The response of Δf_{1} and Δf_{2}, for
three selected samples from Pareto front are shown in Fig.
10 and 11, respectively.

Fig. 8: | ACE signal of three samples (Three PID) in second area (case 1) |

Fig. 9: | Pareto front for case 2 |

Fig. 10: | Frequency deviation of three samples in first area for case 2 |

In addition, the ACE for three samples at two areas is shown in Fig. 12 and 13, respectively.

The Fig. 12 and 13 show choosing solutions from different parts of Pareto front, cause to different results from the aspect of overshoot/ undershoot and settling time. So one can select a single solution based on system conditions for PID parameters. The obtained PID gains for three samples are given in Table 1.

Fig. 11: | Frequency deviation of three samples in second area for case 2 |

Fig. 12: | ACE signal of three samples (Three PID) in first area (case 2) |

Fig. 13: | ACE signal of three samples (Three PID) in second area (case 2) |

**CONCLUSION**

In this study, designing of PID parameters with multi-objective AWPSO for LFC in interconnected power system has been proposed. Two-area power system is used as a test system to demonstrate the effectiveness of the proposed methods under various operating conditions and area load demand. In this method more than one PID design for each of operating point obtained, so one can select a single solution based on system constraints, overshoot/undershoot and settling time. After selecting appropriate PID gains from Pareto front, we can use an adaptive fuzzy gain-scheduling scheme for tuning off-nominal operating points.

#### References

Abdennour, A., 2002. Adaptive optimal gain scheduling for the load frequency control problem. Elect. Power Components Syst., 30: 45-56.

Direct Link

Aldeen, M. and H. Trinh, 1994. Load-frequency control of interconnected power systems via constrained feedback control schemes. Comput. Elect. Eng., 20: 71-88.

CrossRef

Astrom, K.J. and B. Wittenmark, 1996. Computer Controlled Systems: Theory and Design. 3rd Edn., Prentice Hall, New York, ISBN-10: 0133148998.

Bevrani, H. and T. Hiyama, 2008. Robust decentralised PI based LFC design for time delay power systems. Energy Conversion Manage., 49: 193-204.

Direct Link

Cam, E. and I. Kocaarslan, 2005. Load frequency control in two area power systems using fuzzy logic controller. Energy Conversion Manage., 46: 233-243.

CrossRefDirect Link

Demiroren, A. and H.L. Zeynelgil, 2007. GA application to optimization of AGC in three-area power system after deregulation. Elect. Power Energy Syst., 29: 230-240.

Direct Link

Engelbrecht, A.P., 2006. Foundamentals of Computational Swarm Intelligence. 1st Edn., John Wiley and Sons Ltd., New York, ISBN: 978-0-470-09191-3.

Engelbrecht, A.P., 2007. Computational Intelligence: An Introduction. 2nd Edn., John Wiley and Sons Ltd., New York, USA., ISBN-13: 9780470035610, Pages: 597.

Ertugrul, C. and I. Kocaarslan, 2005. A fuzzy gain scheduling PI controller application for an interconnected electrical-power system. Elect. Power Syst. Res., 73: 267-274.

CrossRefDirect Link

Ghoshal, S.P. and S.K. Goswami, 2003. Application of GA based optimal integral gains in fuzzy based active power-frequency control of non-reheat and reheat thermal generating systems. Elect. Power Syst. Res., 67: 79-88.

CrossRefDirect Link

Ghoshal, S.P., 2004. Optimizations of PID gains by particle swarm optimizations in fuzzy based automatic generation control. Electric Power Syst. Res., 72: 203-212.

CrossRefDirect Link

Goldberg, D.E., 1989. Genetic Algorithms in Search, Optimization and Machine Learning. 1st Edn., Addison-Wesley Publishing Company, New York, USA., ISBN: 0201157675, pp: 36-90.

Kazemi, M.H., M. Karrari and M.B. Menhaj, 2003. Decentralized robust adaptive load frequency control using interactions estimation. Elect. Eng., 85: 219-227.

CrossRef

Kumar, J., N. Kah-Hoe and G. Sheble, 1997. AGC simulator for price-based operation, Part 1: A model. IEEE. Trans. Power Syst., 12: 527-532.

CrossRef

Mahfouf, M., M. Chen and D.A. Linkens, 2004. Adaptive weighted particle swarm optimisation for multi-objective optimal design of alloy steels. Lecture Notes Comput. Sci., 3242: 762-771.

Direct Link

Mukherjeea, V. and S.P. Ghoshal, 2008. Intelligent particle swarm optimized fuzzy PID controller for AVR system. Elect. Power Syst. Res., 77: 1689-1698.

CrossRefDirect Link

Pan, C.T. and C.M. Liaw, 1989. An adaptive controller for power system load-frequency control. IEEE. Trans. Power Syst., 4: 122-128.

CrossRef

Sadat, H., 2002. Power System Analysis. 2nd Edn., McGraw-Hill, New York, USA., ISBN: 0-07-284869-3.

Shayeghi, J.S., 2008. Multi-stage fuzzy load frequency control using PSO. Energ. Conversion Manage.,

Shayeghi, J.S., 2008. A robust mixed H2/H∞ based LFC of a deregulated power system including SMES. Energ. Conversion Manage.,

Shayeghia, H., A. Jalilib and H.A. Shayanfar, 2007. Robust modified GA based multi-stage fuzzy LFC. Energy Conversion Manage., 48: 1656-1670.

CrossRef

Shi, Y. and R. Eberhart, 2001. Fuzzy adaptive particle swarm optimization. Proceedings of the Congress on Evolutionary Computation, May 27-30, 2001, Seoul, South Korea, pp: 101-106.

Talaq, J. and F. Al-Basri, 1999. Adaptive fuzzy gain scheduling for load-frequency control. IEEE. Trans. Power Syst., 14: 145-150.

CrossRefINSPEC

Yesil, E. and M.G. Eksin, 2004. Self tuning fuzzy PID type load and frequency controller. Energ. Conversion Manage., 45: 377-390.

CrossRefDirect Link

Zribi, M., M. Al-Rashed and M. Alrifai, 2005. Adaptive decentralized load frequency control of multi-area power systems. Elect. Power Energy Syst., 27: 575-583.

Direct Link