Artificial Neural Network based FACTS in a Contingency Situation

8 Website: www.ijeer.forexjournal.co.in Artificial Neural Network Based FACTS in a Contingency ░ ABSTRACT - The two biggest issues facing today's energy management systems are the ongoing monitoring of online voltage stability and the improved loadability of the transmission lines for the current electrical power system. As a result, it is highly difficult and time-consuming to assess online voltage stability under diverse loading conditions. This study describes a practical voltage stability monitoring system that automates online voltage monitoring and alerts the operator before voltage drops by computing line voltage stability indices using an ANN. This study compares evaluations of system voltage stability and loadability at the load with FACTS devices. The results suggest increase in system loadability while ensuring the security of power system operation.

The world's power systems are [1][2][3][4][5][6] operating under extraordinarily stressful conditions as a result of a lack of reactive power, which could push the systems closer to their operational limits.It is difficult to keep a power system in a safe operating mode because of the following reasons: • Electricity demand has significantly grown.
• A limited growth of the transmission network is the result of social and environmental constraints All of these factors contribute to stability issues in the power system.When under stress, a power system performs differently than when it is not under stress.The system may lose stability in response to a relatively little perturbation because it is so near to the stability limit.The electricity system is often a connected system, which will have a substantial impact on its functionality and stability.
Discussion is had regarding the basic principles of operating and modeling power systems [1].Problems with voltage stability and rotor angle stability arise simultaneously and are thus related.Despite their connections, one form of instability predominates in many situations [2].Using a modal analysis technique, the voltage stability study of large power systems is done.A reduced Jacobian matrix's allied eigenvectors and explicit range of its minimal Eigen values are computed using a gradual state system model.The Eigen values, each of which is connected to a certain way.The eigenvectors are used to define the mode form and to provide information about the network components and generators involved in each mode [3].Static voltage stability analysis uses a variety of methodologies.
Calculating system load margin is the method used the most frequently to indicate the voltage stability limit.Real and reactive power margins are the two most often utilized indicators [4].The method for measuring the influence of a system on voltage stability through net testing in the outline of an indicator L that fluctuates between zero (no load on the system) and one (voltage collapse) is in [5][6][7].Due to the rapid nature of voltage collapse, quick voltage stability evaluation technologies are required to ensure the safe operation of modern power systems.Thus, a new online method based on ANN that requires the least amount of input to estimate the voltage [8].
Utilizing the method of condensing an influence system network into a single line, voltage stability factors are introduced and regularly used to assess system stability [9], the techniques for using voltage stability indices to identify essential bus/branch and the improvement in voltage stability achieved by FACTS devices [10][11][12][13].

░ 2. PROBLEM FORMULATION AND MATHEMATICAL ANALYSIS 2.1 Line Voltage Stability Index ( 𝐋 𝐦𝐧 ) under Single Line Contingency
The apparent power at the receiving and transmitting ends is given by equations ( 1) and (2) below.
As a result, in order to qualify as stability criteria, the following requirements must be met:

Proper location to install FACTS Device and check the enhancement in stability margin
The Mathematical Analysis of SVC is given below based on the figure 1.The voltage sources are provided for a transmission line of T-type.The capacitance is providing the reactive power which connected at shunt and middle of the transmission line.4), which is shown below, is used to determine the reactive power that the capacitor injects in order to control the transmission line's voltage: Series capacitor banks boost the efficiency of power transmission, enhance system stability, lower system losses, enhance line voltage profiles, and maximize the distribution of current among parallel lines.

Artificial Neural Network (ANN) Technique for Predicting Proper Location for Online Stability Applications
Individually neurons in ANN receives a high number of inputs (Each input has a weight).Bias is added to the summation.Net output is produced with the help of inputs, weights and bias.An activation perform is applied to those inputs which ends up in activation level of neuron.Then the output is obtained.

General Procedure to ANN
Step 1: The inputs like number of neurons, computation weights and activation functions are provided.
Step 3: Check the error with in and out and the desired output.
Step 4: Modify the network weights and the network based on the errors.
Step 5: Update weights and modify the input and output.

Learning process
The flowchart or the procedure of training and testing of the neural network is given in figure 3.

Proposed Algorithm
The proposed algorithm [12] is provided below for determining the line, bus and Lmn respectively.

Read the system's line and bus data; the load (MW & MVAR), generator (MW & MVAR, Qmin & Qmax), and system angle (MW & MVAR) data are assumed to be constants. Perform load flow for potential line outages as the baseline.
For each line outage, determine the Lmn.
Under every line outage with a maximum Lmn, rank the more sensitive line.
Connect SVC to the system and determine Lmn's value.
Compare Lmn. with and without SVC.

Case Study and Results
There are 42 input variables consisting of 14 bus voltages, 6 reactive power limits, 1 reactive power loss, 20 transmission lines and 1 SVC parameters.If SVC is not considered, 41 input variables are considered.To choose the best ANN for the given problem a number of trial-and-error simulations were carried out and an input layer with 41 neurons, 2 hidden layers with 21 and 5 neurons and an output layer with one neuron have given the best performance.The ANN with and without connecting SVC is given in table 1.  Line Ranking for contingency case in figure 4 shows that epoch 39 yields the greatest validation performance of 0.00078228.

░ 4. CONCLUSION
In this study, the interpretation of the maximum load ability limit takes voltage stability into account.The bus and branch in the system are located using voltage stability indices, and voltage instability is felt at various loading margins.These indices are close to 0 when the system is voltage stable and gradually increase closer to 1 as the system approaches the point.The estimated voltage stability indices are compared with various loading scenarios.
Reactive power compensation devices are located where the important bus or branch is predicted by artificial neural networks (ANN).The most powerful devices for controlling voltage and enhancing the stability of the power system are FACTS devices.

Figure 1 :
Figure 1: Single line diagram of Transmission line with SVC Equation (4), which is shown below, is used to determine the reactive power that the capacitor injects in order to control the transmission line's voltage:

Figure 3 :
Figure 3: Training Procedure for Artificial Neural Network with FACTS ░ 3. RESULTS AND DISCUSSION

Figure 4 :
Figure 4: ANN for Lmn in outage scenario

Table 1 : ANN for IEEE-14 Bus System with and without SVC Type of ANN Parameter
3.2.1 Case 1: ANN using prediction of weak buses/branch

.3 Case 3: SVC Compensation with ANNTable 2
shows the Bus Ranking with SVC with ANN.For 100%, 160% is 4, 4 and the value is 0.249571, and 0.480814 respectively.From figure4it is observed that the best validation performance 6.8146e-05 at epoch 8 is obtained.
Figure 5: SVC Compensation with Lmn with ANN From figure 5 it is observed that the best validation performance for Lmn with SVC is 6.8146e-05 at epoch 8 is obtained Lmn.The comparison of various indices is given in table4and it can be observed that the proposed algorithm has given the best performance comparatively.░