Summary Of Engineering Doctoral Thesis: Building artificial intelligence algorithm for reconfiguration distribution network problem

The thesis aims to propose the Heuristic method algorithm applied to the problem of power grid restructuring with the objective of reducing power loss in case of having / without distributed power source connecting to the distribution grid. Study the effect of distributed power source when connecting to the distribution grid, affecting the power grid restructuring problem. Proposing algorithms according to the new method of Meta Heuristic for grid reconstruction problem with the objective of reducing power loss in the cases with / without and in case of considering the position and capacity of the power source scattered when connecting distribution grid.

MINISTRY OF EDUCATION AND TRAINING VIETNAM ACADEMY OF SCIENCE

AND TECHNOLOGY

GRADUATE UNIVERSITY SCIENCE AND TECHNOLOGY

………***………

Ng

NGUYEN TUNG LINH

BUILDING ARTIFICIAL INTELLIGENCE ALGORITHM FOR RECONFIGURATION DISTRIBUTION NETWORK PROBLEM

Code major: Control theory and Optimization control

(Automation and Control Engineering Technology)

SUMMARY OF ENGINEERING DOCTORAL THESIS

Ha Noi - 2018

INTRODUCTION

1 Sep up the problem

According to the statistics of Vietnam Electricity Corporation, total power loss in recent years is about 9-15% of electricity production volume, in which distribution of the electrical network holds 5-7% Then, researching methods for power loss reduction in the distribution network is very urgent demand Reconfiguaration distribution network is one of methods to minimize power loss which is researched most

Currently, solution of reconfiguaration distribution network problem is optimal one under NP-hard class and then in order to solve this problem, there are some following methods to approach and solve the problem:

- Seek by optimal mathematical method

- Seek heuristic to seek enough good solution

- Seek near correct solution by natural emulation algorithm such as: Simulated Annealingalgorithm, genetic algorithm, herd optimum, etc

2 Objectives and tasks of the thesis

 Propose algorithm by Heuristic method to apply for the problem of reconfiguaration distribution network with objective function for capacity loss reduction considered in the case that there is/is not dispersal power connected into electrical distribution network

 Research the effect of dispersal power when connect to the electrical distribution network, affecting the network reconstruction problem

 Propose algorithm by new Meta Heuristic method for reconfiguration distribution

network problem with a goal of power losses reudction in the the case that there is/is not as well

as consideration to the location and capacity of dispersal power source when it is connected to electrical distribution network

5 New points of the thesis: the thesis achieves some the following researching contents:

 Propose method for the problem of reconfiguaration distribution network on “Heuristic” experience rules with objective function of capacity loss reduction under the consideration of two cases: there is not connection to dispersal power source and there is connection to dispersal power source

 Propose the improvement for algorithm of metallurgical emulation for the problem of reconfiguaration distribution network with subjective function of power loss reduction

 Propose using genetic algorithm for the reconfiguarationin the consideration of the location and capacity of dispersal power when connect to distribution network with objective function of capacity loss reduction

6 Practical value of the thesis: Researching results of the thesis achieved some practical value in the problem of re-configuration solution and actual applications

 The method which proposes by the way of researching Heuristic, again can affirm that application of experience rules and optimal method for optimum problems are still used well in

some cases

 The researching method according to MetaHeuristic for the problem of electrical

network re-configuration is proposed by the author to use Simulated Annealing algorithm and genetic algorithm for the problem of reconfiguaration distribution network with objective function of capacity loss reduction in the case there is dispersal power and there is not dispersal power as well as the case considered to the location, dispersal power capacity connected to distribution network

 This is supporting tool for deciding design and operation of distribution network when

participate into competitive electrical market

7 Thesis layout: The thesis is divided into 4 chapters

Chapter 1: General view of distribution network and the problem of reconfiguaration

distribution network

Chapter 2: Heuristic method for the problem of electrical reconfiguaration distribution

network

Chapter 3: MetaHeuristic method for the problem of electrical distribution problem

Chapter 4: Genetic algorithm for the problem of Reconfiguaration distribution network in the

consideration of dispersal power planning

CHAPTER 1: OVERVIEW OF ELECTRICAL DISTRIBUTION NETWORK AND THE PROBLEM OF ELECTRICAL DISTRIBUTION NETWORK RE-

CONFIGURATION 1.1 Introduction of electrical distribution network

1.1.1 Characteristics of electrical distribution network

Electrical distribution network is an important component in the supply of electricity from the production site to the electricity consumers, spreading across the whole territory of the country The distribution network can be designed with a loop structure or a beam structure, but for technical reasons and operating conditions, it is operated in a beam structure Thanks to

the open operating structure, the relay protection system only uses over current relay

Distribution network operating conditions must always meet the following conditions:

- Open operating structure

- All loads are provided with electricity, pressure drop within the allowed range

- The relay protection system must be changed suitably

- Lines, transformers and other equipment are not overloaded

1.1.2 Introduction of electrical reconfiguration distribution network

a Introduction of electrical reconfiguration distribution network:

Electrical reconfiguaration distribution network problem is the status control of switching equipments in the distribution network, in some operation cases to ensure for some objectives

b Classification of electrical reconfiguaration distribution network

*Classification by objective function:

Problem 1: Determine electrical network by load diagram in certain duration for the operation

Problem 4: Re-configure distribution network to balance load (among lines, transformers at

stations) to improve loading capacity of electrical network

Problem 5: Recover electrical network after incident or cutting off power for repairing

Problem 6: Determine structure of electrical network by many objectives such as: capacity loss

is minimum, highest loading balance, minimum loading transmission, minimum pressure drop

at the end of network concurrently happens out, affects of dispersal power source to

re-configuration electrical distribution network, etc

Problem 7: Determine electrical network to ensure for objectives of power reduction and

stopping supply or improving reliability of power supply

*Classification by researching methods:

Figure 1.3 Classification of reconfiguration distribution network by researching method

* Some researching results of reconfiguration distribution network

 Problem 3 – Determination of reduction network structure P is the most important problem

The problem determines the structure of effective capacity loss reduction network – problem is an important one, it is considered as one module to solve other problems in the system of reconfiguration distribution network It is proved through alogorithms of last researches

 Problem 1 – Minimal function of operation cost

This objective function is very function with distribution network with flexible and low load transmission cost in the operation; network structure can be changed in many times in the day This function is very suitable with distribution network with flexible and low load transmission cost, network structure can be changed in many time in the day

Classification of methods for electrical distribution network

reconfiguration solutions

Analytic method Heuristic method

(Experience law)

Meta Heuristic (Use AI algorithm)

 Problem 2 – Minimal function of power loss

In the fact, evenly in developed industrial countries, load transmission cost affects greatly the decision on network structure changes because sometime these costs are bigger than gained benefits

Thus, for problem 2 – Determine non-changeable electrical network structure in surveying time for power loss to be minimal

 Problem 4 – Balance capacity among lines and transformer stations

This algorithm can be applied suitably for areas which are usually affected by overload or incertain loads, In [91], Tim Taylor, etc

 Problem 5 – Re-structure distribution network after incident

This is objective which is mentioned by almost scientists in their researches

 Problem 6 – Re-structure network under objective function

In the operation of distribution network, there are many operation objectives which the controlid must select suitably with characteristics of local electrical networks

1.1.3 Current status of electrical distribution network in Vietnam

Current status of Vietnam’s electrical network

- By history of development and in each country, there many grades of distribution voltages and these grades in regions are also different each other (6.6, 10, 15, 22, 35 kV)

- Recloser and loaded cutter (LBS) are not controlled remotely and quantity is not considerable then switching cost is big and time for loading transmission is long

Problems during the process of Vietnam’s electrical grip operation are presentated in problem 1 to problem 7

1.1.4 Model of electrical reconfiguration distribution network

a Mathematical model of reconfiguaration distribution network

For mathematics, reconfiguration network is the problem for planning discrete curvilinear

by the capacity line running on branches, at [78] the model as presented below

ij ij n

i n j

ij ij

ij I R C L C

max n

b Some assumptions to simplize reconfiguaration distribution network

 Reactive power compensation when consider reconfiguration distribution network

Ross Baldick [49]: “It is possible to pass away reactive power compensator in the network

after solving the problem of electrical distribution network structure determination.”

 Some other assumptions for the problem of electrical distribution network

re-configuration

- Switching operation to transmiss load does not make electrical system to be incertain

- Voltage at loading buttons does not change and has value nearly equal to Uđm

- When solve capacity distribution problem in the beam network, pass away power loss

- Reliability of power supply in the distribution network seems not to change when network structure changes

1.2 Overview of researches on re-configuration solutions with objective functions of power loss reduction

1.2.1 Heuristics combination and optimization

a Merlin and Back algorithms – close loop technique

b Other algorithms

1.2.2 Pure algorithms on Heuristics

a Algorithms of Civanlar and partners – Technique of branch change

b Some other algorithms

1.2.3 Algorithms based on artificial intelligence

1.3 Comments and evaluation

1.3.1 Electrical reconfiguration distribution network with the objective of electrical network

control

- Almost reconfiguration problems approach different objectives but all of them use

problem 3 – to determine structure of electrical distribution network to reduce power loss as

main modules during the re-solution problem

- When solve problem 3, algorithms baseds on the researching method by the technique

of branch change of Civanlar [17] or close loop technique of Merlin and Back [65] then they usually fall into local minimal and use algorithms of artificial intelligence and evolution in which the most effective one is Gen algorithm and algorithm of metallurgical emulation

- Algorithms in problem 3 find out solutions to reduce directly power loss function value for whole network and then it wastes a lot of time because the problem of power distribution must be solved in many times during the repeating process

1.3.2 Technique for solving electrical network re-structure problem

- When approach electrical reconfiguration distribution network, scientists determine that mathematical analystics is not as effective as search algorithm

- Search algorithms are used in the electrical reconfiguration distribution network can

be divided into three main directions such as: heuristic search algorithm combines with optimal algorithm; algorithm only uses heuristic rule in expert system; use artificial intelligence including expert system, genetic algorithm, noron network, etc

- Almost algorithms of network re-structure do not show out that network structure has minimal power loss, not prove the point to find out minimal point in overall

Under above reasons, in the own research, the author propses the search of electrical

reconfiguration distribution network by 2 directions: research by the way of using Heuristic

and MetaHeuristic methods with objective function on power loss reduction

CHAPTER 2: HEURISTIC METHOD FOR ELECTRICAL RECONFIGURATION

DISTRIBUTION NETWORK 2.1 Heuristic method for electrical reconfiguration distribution network

2.1.1 Introduction

Heuristic algorithm is an expansion to algorithm definition It presents problem solution with the following characteristics:

- Usually seek good solution (but not surely for the best solution)

- Solve problem by Heuristic algorithm easily and quickly to give out results rather than optimal algorithm

- Heuristic algorithm is regularly presented naturally, nearly with the thinking way and action of human

In the electrical reconfiguration distribution network, Heuristic algorithm was used since long time Since 1975 until now, there are nearly 80 researches on this problem by using Heuristic as published on famous magazines

2.1.2 Emulation of electrical reconfiguration distribution network

Emulation of electrical network and conventions:

Convention: - IPi, IQi are action and reactive current of the branch i; Ri electrical resistance of branch i,

- Call k as electrical switch number to ensure for open electrical network operation On the branch, there is open electrical switch at the jth with the sign: MNj and j = 1 k

Convention choose a set of independent loops in order for each independent loop to cross unique one open electrical switch MNj; the positive direction is counter to the switchwise in the figure 2.1 as hereafter:

- Vjh is the set of crossing branches between loop j and loop h;

- Vjj is the set under the loop j;

- Rj loop is resistance of loop j;

- MNj is the branch with openning switch of the jth loop

Indicator Aij states the correlation between the jth loop and natural distribution direction at the the ith branch in the open network

Aij = 1: when direction of j loop is the same to directions of IPi and IQi;

Aij = -1: When direction of j loop is counter to directions of IPi and IQi;

Aij = 0: when the ith branch does not belong to j loop

b Mathematical description of load re-distribution operation Following with the figure 2.3

Consider a simple distribution network consisting of: a single source and a closed loop at the MN branch, as shown in Figure 2.3 (or at the MN branch with IPMN = 0 and IQMN = 0)

We need to determine the key to open (on the circuit) to minimize power loss (Open key may include on-openning key) Assuming that the result is a switch on the MN branch and un switch

on the branch AB, then the change in load distribution can be similar to the in / out of the two poles on the branch MN at the current until the current on the AB line is zero

c Conditions for power loss to be minimal after re-distributing P branch load on branches of electrical networks after re-distributing load:

i n

sau

R I R

I

R I

A I B I

R I

A I B I

P

1 2 1

2.1.3 Comments and evaluation

- With any distribution network, deriven from a certain (non-optimal) open configuration, if inject/ draw out at the opening switch of an electric current by the formula (2.11), (2.12) to create loop currents through branches, the objective function P will be smallest

- The switch dependent ring number on the loop to inject in / draw out an electric current Theoretically, if switch is selected for the loop current to inject in or draw out equally to 0, then

it is the optimal point

- The expression (2.13) is the sum of the voltage drops on the branches of the j independent circuit if the circuit is pure (or homogeneous circuit) This shows that the optimum value of obtained current under (2.11) and (2.12) is the branch line of the closed distribution network When close all electrical switches, the losses P in the distribution network are minimal

- - When there are DG resources involved in the distribution network, the optimal expression for the received current will be added to the second component (in expressions (2.11) and (2.12)

2.2 Propose Heuristic algorithm for electrical reconfiguration distribution network 2.2.1 Objective function of the problem Objective funciton of the problem:

vong j opt MN Qj j

vong j opt MN

( 2

) ( 2

) (

min

P increasing less than loop network

(2.5)

G function has just considered as the value of loss P, just considering the resistance factor

of the distribution network (R loops), thus it is possible to consider the interactions between the electric switches and the DG to whole distribution network This is difference in comparison with previous studies If we ignore the resistance factor (the "distance" of the electrical switch comparing to the power supply), we usually select the branch with the smallest lines in the distribution network to open first This leads to open electric switches which are far from the power source, in the factthat it is not normally open (because if they are open, the rear loads will not have electricity) Therefore, finding optimum P can avoid the local minimum and does not take the time to re-examine whether all loads are being supplied with electricity It also makes sense to compare the value of the function G when there are many competitive pairs in the distribution network

Figure 2.7: Diagram of reconfiguaration distribution network algorithm

Solve capacity distribution problem in the distribution network with DG and

compensator, pass away line resistance sensor

Determine independent rings by definitions of distribution network

Calculate reduction of G function in each independent ring

Calculate reduction of G function in each independent ring

Choose closing/opening switch couple in independent ring with most

 G reduction

Correct

Wrong

Closeopenning electrical switch and solve capacity distribution problem

in distribution network with one closed loop (1 independent loop)

Open electrical switch with the smallest current in closed loop

Distribution capacity in the beam distribution network

Process 1

Open switch with the next smallest current in closed loop

G Function reduce?

Open remaining switches with the smallest current

Check pressure drop, overload ?

Current through openning switch is the smallest one?

Begin

Implement with independent loops

End

has DG, seek the smallest P

2.3 Emulation and evaluation of researching results

2.3.1 Emulation of researching results

Considering the distribution network of 16 nodes with 21 branches; there are 6 open switches; there are 2 DGs proposed by G.Celli in [39] described in Figure 2.7

a Describes the process of finding the network configuration without DG

b Describe the process of finding the network configuration when there are 2 DGs at node 9 and node 13

c Describe the process of finding the network configuration when there is a DG generator at node 9

d Describe the process of finding the network configuration when there is a DG generator at node 13

e Evaluation of simulation results:

After performing simulations on the sample network and comparing with some other methodologies summarized in Table 2.7

Table 2.7 Results of surveying summary on the distribution network with 16 nodes

DG1 – node 9 (kW)

DG2 – node 13 (kW)

Table 2.9 Comparison before and upon performing electrical network re-structure

Repe ated loop The system does no thave DG

Consider distribution network with 69 nodes including 73 branches, 5 switches usually open and total load capacity is 3.802 + j 3.696 MW

Table 2.10 Comparison before and opon performing electrical network re-configuration

Repe ated loop The system does no thave DG

2.3.2 Evaluation of emulation results:

Configuration of distribution network in 2 cases without DG and with DG as proposed by algorithm has the same results to TOPO module results in PSS / ADEPT 5.0, but Pdecrease is better than proposed method by G Celli or R Srinivasa [4] about (11.5 - 18.8)% The power loss gain function G fully describes the relationship between the electric switches and the branch currents in the distribution network, so all independent loops are considered at once, so the number of operations is few and straight to configuration with minimal P

The above results are calculated with a fixed DG load and capacity When the DG capacity and load change over time, the process is repeated similarly, so there will be many combinations

of key pairs on the network This time it is necessary to determine the combination of the key pair for the smallest P, or to operate with a constant change configuration As such, accurate information on load forecasting, DG capacity at future times, and the problem of operating the electricity distribution network should be taken into account in order to reduce damage caused

by power failure when re-distributing load

Conclusion: In this research, the author has developed the explicit expression of the influence

of DG on the problem of reconfiguration of the distribution network in order to increase the least P The objective function G is proposed For this purpose, the configuration of a distribution network is guaranteed to have P to be as small as possible, demonstrating the effect of DG on the distribution of current across distribution network function G serves as the basis for the reconfiguration of the distribution network with DG to reduce P and is very good for application in online operation Research results show that:

• The method of experience rule proposal for the reconfiguration distribution network

Rule 1: Close all switches forming a closed net, where the loss going through closed loop

will be smallest

Rule 2: Calculate the function G following with (2.14) and G according to (2.16), make

a list with descending G and choose to open a pair of electrical switch in the independent loop, starting with the largest G loop

Rule 3: When close /unswitch a independent switch, in order to reduce the P, it will unswitch the smallest electric current and close switch with the maximum current

The calculation of G function is not complicated and faster than the direct calculation of P Applying the G function allows for showing out the last configuration to have an increase of

P at least compared to the closed network, as there is a comparison of G functions at all independent loops across the distribution network

Due to the characteristics on step by step value degradation of the G function, it is possible

to develop this function to solve the problems of reconfiguration of the distribution network with DGs such as configuring the distribution network to operate online, power supply restoration, load balancing, and reload redistribution

The proposed algorithm is suitable for the online operation of the distribution network with

DG on the basis of the comparison of the deviation c.A with the switching cost When there

is information on the load forecast, DG as well as the percentage of additional industrial load and load redistribution time will help the controller of the Electricity decide to change the distribution network configuration or not

The proposed method was tested on the IEEE network system and compared with other methods with the same data set that showed reliable and accurate research results

CHAPTER 3: META HEURISTIC METHOD FOR ELECTRICAL DISTRIBUTION RECONFIGURATION NETWORK 3.1 Mehthod uses stimulated annealing algorithm for electrical distribution network re- struturing problem

3 1.1 Introduction to stimulated annealing algorithm

Application of SA algorithm for reconfiguaration distribution network: SA algorithm was

proposed for the first time in the 1990s by the researching team of Chiang and Jumeau [46 Chang and Kuo [47], Jiang and Baldick [30] and Su and Lee [89], Jeon et al [56], Chen et al Jeon and Kim [57], Augugliaro et al [4] and Su et al [88], Jeon and Kim [57], Augugliaro et al [4], Su et al [88] for electrical reconfiguration distribution network It is noted that the research

of the application of the SA algorithm for the network reconfiguration problem is applied to a variety of objective functions and that the algorithm improvement is also focused on the process

of speeding up the calculation, aim to meet the real-time, these processes focus on two phases

of the algorithm that is the process of reducing the temperature for metropolis process and the stop condition of the algorithm

3.1.2 Proposal for improvement of SA algorithm for electrical reconfiguration distribution network

a Apply SA algorithm in the case of not connecting to dispersal power

* Objective function of the problem

Objective function of loss reduction is determined by capacity current as follows:

2 2 NL

at the temperature under the investigation In contrast, the Metropolis process is carried out according to the formula (3.9)

Lj j

S SW

b Apply SA algorithm with loss reduction objective in consideration to connection of dispersal power (DG) into the network Objective function of the problem