Load Profiling Approach in Clustering Automated Electrical Power Metering System
Chapter One
AIM AND OBJECTIVE OF THE STUDY
The aim of the research is to develop a load profiling technique in clustering automated electrical power metering system. The specific objectives are:
- To build a category model to support the retail and distribution companies by extraction of knowledge from electricity consumption data.
- Building a data sample for Enugu Electricity Distribution Company EEDC where electricity consumers can be monitored for load management distribution system planning.
- To provide a template that ensures consumers are properly identified, grouped into representative load profiles that uniquely determines their tariff rate.
CHAPTER TWO
LITERATURE REVIEW
INTRODUCTION
A clustering approach to load profiling in monitoring an automated electrical power metering system presents a ground breaking innovation that will revolutionize Nigeria electricity industry. Though smart metering systems are yet to be fully realized in Nigeria, the study offers the advantage of knowing the accurate monthly consumption of different consumers thus enhancing accurate billing through thorough consumer classifications. It also provides load data to assist in the management of energy trading and maximum load demand. It reduces human-meter reading cost and offers other opportunity for further services which include check for real consumer classification based on tariff structure.
What motivated the study are the numerous benefits and large diversity of solution new legislation on metering and load profiling introduces to electricity market of developed economies. It can be seen in [12], that countries that already have new legislation on metering(smart metering systems) and load profiling have different metering time interval variation consumption access accuracy benefits from a quarter of an hour or less to an hour. But for this study circumstance of metering will be considered in Nigeria context where electricity consumption is predominantly measured on monthly basis using either electromechanical meter or less of prepayment scheme (electronic meter). The following section presents overview of Electricity Supply in Nigeria, customer classification, Billing System and Load Profile Using Clustering Technique.
Electricity
Electricity is the science, engineering, technology and physical phenomena associated with the presence and flow of electric charges. Electricity is generated at various kinds of power plant by utilities and independent power producers. In [13] [14], the principle of electricity generation and distribution was extensively described. Electricity is reasons behind industrialization and development of socioeconomic activities. As such, in [16], national demand is presently estimated at well over 5000MW in the year 2020. This implies that actual metering of electricity is paramount for revenue collection. In a survey study in [15], it is established that 48% of Nigeria electricity consumers are over charged or not metered due to faulty metering system and unstructured tariff customer classification. Scientific techniques that will match and measure exactly the load electricity demand of various customer types and generate unambiguous billing to end-users in turn to ensure adequate revenue collection to utility companies has been the trend of research in Liberalized Electric market. Therefore, the study is geared towards creating an alternative method of revenue collection by means of new tariff consumer classification and metering electricity consumers using load profile clustering.
Electricity Billing System
Electricity billing system is a measure whereby electric utilities recover their cost of installation, generation, transmission, distribution, system management, and return on asset on an average basis [17]. It is a machine that produces written detailed statement describing how much of energy is consumed at a defined duration. Electricity billing is categorized into two rate design:
a, Time Independent Rate
b, Time of use
Time Independent Rate: This category of rate design is a charge on customer where price of electricity has been measured and billed on an aggregate basis over the period of a total billing rotation usually a month. In this category, customers pay for electricity based on their entire consumption measured (in Kilowatt) over a billing period regardless of when the electricity is demanded. In Nigeria, an example of time independent rate is the Consumption charge as can be seen in the figure 2.1
Electricity bill showing both time independent rate and Time of use rate
Time of Use Rate: This category of charges on customers considers peak times and off-times. For peak times, the customers are billed at higher rate while lower in off-peak rate. This billing pattern is obtainable in more advanced economies where Time-interval meter is used.
POWER METERING SYSTEM
An electricity power metering system is a power measuring device that measures amount of electric energy consumed by a residence, business or commercial driven machines. These metering systems are as follow: Electromechanical induction meters, or Standard Meters, Variable Rate Electronic meters, Prepayment Electric meters, and Solid State Electric meters.
The birth of National Electric Power Authority (NEPA) in Nigeria on 1st of April 1972 after merging of Electricity Cooperation of Nigeria (ECN) and Nigeria Dam Authority (NDA) repositioned metering systems. Also restructuring NEPA between 1978 and 1983 led to the establishment of the electrification boards whose work is to take power supply to the rural areas and new cities. Part of its achievement by extension is exercising and making sure that electricity consumers use electromechanical induction meter or standard metering system where human interface-meter reader notes the consumed unit of electricity shown on the meter and bill is imposed on the customer along other costs[18].
CHAPTER THREE
RESEARCH METHODOLOGY
INTRODUCTION
The section presents the systematic steps carried out on Load data obtained from Enugu Electricity Distribution Company EEDC operator by ensuring that right clustering technique is adopted; determine the Distinctive Load Profiles (DLPs) of these customers and the assignment of load data into cluster groups using clustering technique. This is realized following the generic flow chart in figure 3.1
Essence of K-means clustering technique
For this work different clustering techniques are tested on electricity load data. K-means proves most effective and suitable clustering technique in this work. This is because it shows best performance indicator which includes the following;
Number of clustering: The selection of cluster number needs an all-encompassing consideration. Clustering performance varies with the number of clusters. Generally, most traditional clustering techniques are run many times to choose the number of clusters that yield the best performance indicators. Follow-the-Leader and fuzzy relation use trial and error method in determining the number of clusters. The optimal number of clusters for k-means is chosen priori (pre-determined before clustering), that is, it is user defined. K-means uses Ratio of within Cluster Sum of Squares to Between Cluster Variation as an adequacy measure to evaluate algorithm effectiveness. Hence, K-means matches the requirements pertaining to the clusters number in the load profiling applications such as tariff design.
Distance metrics. The metric for similarity between two objects is a fundamental issue in clustering. Substantially, the similarity among electrical load data is quantified by Euclidean distance, in [44]. The formula for this distance between load data points A (A1, A2, etc.) and load data points B (B1, B2, etc.) is expressed by the equation 3.1.
(3.1)
Where
Ah = different load values of customer A, for 12-months period
bh = different load values of customer B, for 12-months period
Fuzzy C-means and fuzzy relation do not give clear cut cluster separation in load profile implementation. Unlike Hierarchical clustering, K-means technique uses Euclidean distance measure that does proper data assignments creating boundary between datasets using iterative process while avoiding trial and error approach in its implementation. Equation 3.1 is a true representation measure of the ratio of two distances between load profile data using K-means.
Big data: The scale of electrical consumption data increases as consumer population increases owing to data accruing from population-level consumers. In this study, K-means technique was tested on consumption load values of 850 consumers. K-means is adjudged and known for its capability in clustering large number of data.
CHAPTER FOUR
EXPERIMENT, RESULTS AND DISCUSSION
INTRODUCTION
The aim is to classify the load pattern of different types of customers. Conducting load pattern analysis is an important task in obtaining typical load profiles (TLPs) of customers and grouping them into classes according to their load characteristics. Even if the customer information needed in the classification is correct, some of the customers can simply have such an irregular behaviour pattern that they do not fit in any of the predefined customer class load profiles. The use of clustering algorithm as a data mining technique is an effective means of developing structured electricity tariff (load profile) and electricity load planning. This is because organized customer segmentation will invariably establish and sustain electricity load planning. In this chapter, K-means technique will be applied in clustering monthly electricity consumption load values of Enugu State Electricity Distribution Company, (EEDC) Enugu customers, focusing on the suburb, to be able group customers based on scale attribute: large, medium and small scale customer. Importantly sub groups of customers were also formed from the small scale customers describing different classes of private electricity customers. These latter groups are low electricity customers that are not metered by energy meter and are not regularly visited by utility personnel as a result of their location.
The steps taken using this technique include:
Data pre-processing and sorting was carried out Microsoft excel.
All the analyses were carried out using Python programming language and R script.
The results were all presented in R scripts. The R statistical package returns the clusters, cluster sizes, within sum of squared error for each cluster and for all clusters correspondingly
Three (3) major clusters were developed for electricity consumer types: residential, commercial and industrial with emphasis on private customers. More clusters were identified from residential customer class showing different load groups within private customer’s group.
CHAPTER FIVE
RECOMMENDATION AND CONCLUSION
SUMMARY
At the end of the work, load profiling approach in clustering automated electrical power metering system was developed. K-means R clustering technique was applied to different electricity load data (850 data size) to bring about electricity customer classification. The behaviour of the entire electricity customers was x-rayed in different cluster groups. The earlier three clusters represent the different customer types based on the customer scale attribute: large, medium and small scale customer. Private customer classes were also formed from small scale customer category. The private classes describes category of low electricity customers. The resulting classification confirms the techniques as better alternative to electricity classification used by Nigeria electricity industry. However successive sum of square error was calculated for different clusters for stability and convergence. The best sum of square error value obtained during the iteration by k-means R clustering algorithm is 77.70% for main classification and 95.3% for private customer class classification. This result remains valid and depicts the optimum convergence value
RECOMMENDATIONS FOR FURTHER WORK
This work presents an efficient method for developing a load profiling technique in clustering automated electrical power metering system. The proposed method was implemented as an R analytical program and tested with real data (consumption values). The result showed that the K-means algorithm implemented in R program can classify customers into well separated clusters according to their electricity consumption data, and clearly indicate that the proposed k-means algorithm has significant implications with its efficient and stable nature of the structure in handling the database. Therefore, it is evident that the load profiling technique in clustering electrical power metering system has the potential in efficiently addressing metering challenges and structured tariff. Correspondingly, with little efforts by manipulating the required parameters it is possible to obtain the desired results and reach the goal. In light of this result, it is certainly possible to extend further on these investigations to develop improved load profiling clustering of automated power metering system where in both typical load profiles and the number of clusters with simultaneous fitness function integrated in the investigation of metering. It was proved earlier that, the resulting customer classification is more accurate than the alternative classification methods generated from long non updated data collected over several years Electricity Distribution Companies. The plots particularly that of figure 4.7 clearly indicated that the proposed K-means clusters and load profile data are more accurately a better alternative in metering power metering systems. This technique is recommended to Enugu state Electricity Distribution Company and other electricity distribution utilities for tariff restructuring, consumer management, scheduling maintenance and load planning
CONTRIBUTION
A clustering (data mining) approach to load profiling in monitoring electrical power metering system presents ground breaking innovation that will revolutionize Nigeria electricity market in that it provides a systematic technique that will adequately address
- Load Shedding Formula: Massive load shedding is frequently employed as a way of forcing the demand to be within PHCN’s ability to supply. Therefore selective supplies of those customers considered to be of high priority will be needful which depends on the supply time. The technique proposed in this dissertation will address the matter given that different cluster groups represent customer classification .Consequently different clusters meaning different groups of electricity customers will be selected for supply depending on utility operator’s load shedding formula based on the peak demand. For instance, group (cluster) of customers identified as industrial customers should be supplied during the day to enable these users run production, while residential group will be supplied (disconnecting the industrial group) during night since most industries do not run production at night.
- Scheduling Maintenance: Schedule maintenance is achievable having identified different clusters i.e. electricity customer classification and the transformer/feeder that supplied them. Routine checks could be conducted on those infrastructures (transformers and feeder pillars) supplying these customers. The selective supply should be arranged in such a way that while the first group of customers is disconnected during their off-peak period another group is connected and vice versa. So also is maintenance scheduled interchangeably in ensuring adequate and uninterrupted power supply is achieved and maintained throughout
- Robust Revenue Collection: Robust revenue collection becomes realizable by the formation of clusters and sub cluster using clustering technique. By this procedure, each and every electricity customer will have been metered using representative load profile as the standard.
CONCLUSION
This thesis presented a clustering methodology for creating a representative electricity load profile classes from automated metered load data generated from a particular area covered by electricity network in Enugu, Nigeria. Cluster methods: K-means R tool was used and evaluated against sum of square error for segmenting the load data into disparate patterns of electricity use. Three main profile classes and four sub profile classes were presented as better alternative to electricity customer classification. As a result, it is possible to classify customers based on their load data generated after few months of connection. The result of the proposed technique shows improvement in electricity customer classification for billing consumers and tariff rates.
REFERENCES
- International Energy Agency. 2013 Annual report 2013. Web 13 feburary 2013.
- Omijeh & Ighalo “Design Of A Robust Prepaid Energy Metering And Billing System” JORIND 10 (3), December, 2012. ISSN 1596 – 8308.
- Omijeh & Ighalo “Modelling of GSM-Based Energy Recharge Scheme for Prepaid Meter”IOSR Journal of Electrical and Electronics Engineering ISSN: 2278-1676 Volume 4, Issue 1 (Jan. – Feb. 2013), PP 46-53
- Miyogo N.C, Nyanamba, S.O, Nyangweso G.N “An Assessment of the Effect of Prepaid Service Transition in Electricity Bill Payment on KP Customers, a Survey of Kenya Power, West Kenya Kisumu” American International Journal of Contemporary Research Vol. 3 No. 9; September 2013
- Altir C. D , Bonganay J. C Magno A. G Adrian G. M, John M.E and Morante N.P “ Automated Electric Meter Reading and Monitoring System using ZigBee-Integrated Raspberry Pi Singleb Board Computer via Modbus”
- Islam, M.M, Ahmad, M, Mitul A.F, Malek, M .F and Rasheed, M.A “Electronic Energy Meter with Remote Monitoring and Billing System” 2012 7th International Conference on Electrical and Computer Engineering. Pp(240-244) 20-22 Dec, 2012 Dhaka, Bangladesh.
