END USE BASED MODEL FOR RESIDENTIAL POWER CONSUMPTION FORECASTING IN NAIROBI REGION, KENYA

Transcription

1 END USE BASED MODEL FOR RESIDENTIAL POWER CONSUMPTION FORECASTING IN NAIROBI REGION, KENYA NAME: REG. NO. MARGARET KANINI NZIA F56/78274/2009 A Repot Submitted in Patial fulfillment of the equiements fo the awad of the Degee of Maste of Science in Enegy Management, Univesity of Naiobi. Depatment of Mechanical and Manufactuing Engineeing School of Engineeing Univesity of Naiobi July 23

2 DECLARATION I heeby declae that this is my oiginal wok and has not been pesented fo a degee in any othe univesity. Signatue.. Date.. Magaet Kanini Nzia This epot has been submitted fo examination with ou appoval as Univesity supevisos. Signatue. Date. D. Alex A. Aganda (Depatment of Mechanical and Manufactuing Engineeing, Univesity of Naiobi) Signatue.. Date.. Pof. Felix M. Luti (Depatment of Mechanical and Manufactuing Engineeing, Univesity of Naiobi) ii

3 ABSTRACT This study set out to geneate baseline data and infomation though the analysis of seconday enegy consumption data available fom the Kenya Powe and Lighting Company Ltd (KPLC) fo domestic households in Naiobi egion, both uban and ual based. The accuacy of the cuent econometic model used fo foecasting esidential enegy demand fo the puposes of planning was assessed. Though inteviews, obsevations and owneship of electical appliances, end use pattens fo the vaious income goups of households wee analysed. It was established that owneship of household appliances and thei standad consumption have an influence on the total enegy demand. The aveage annual consumption pe household in the uban high, medium, and low income goups was 5,767kWh, 1,642kWh and 451kWh espectively, while that of the ual high, middle and low income households was 1,634kWh, 733kWh and 218kWh pe household pe yea. Based on the consumption pe household and consideing demogaphic, socialeconomic and penetation level of appliances, an engineeing (end use based) model was developed fo foecasting esidential electicity demand. The model was used to estimate the esidential consumption in Naiobi egion fom the yea 2009 to 22. Using the model, the pedicted consumption fo the yea 20 was found to compae bette with the actual ecoded consumption in KPLC ecods than the econometic model. The 2009, 21and 22 pedictions, though bette than the econometic model pedictions yielded less accuate esults due to lack of necessay data. The end use model was found to be moe accuate fo foecasting of enegy demand and is ecommended. It is also ecommended that the end use model be developed futhe to include othe foms of enegy and to be used fo enegy demand foecasting fo the County Govenments. Futhe impovements to this model would conside uban ual migation, technological changes (enegy efficiency), migation fom one income goup to the othe and changes in penetation level of appliances. iii

4 ACKNOWLEDGEMENT I wish to expess my pofound gatitude to my supevisos, D. Alex A. Aganda and Pof. Felix M. Luti fo thei encouagement and valuable guidance thoughout the study. To the management of the Kenya Powe & Lighting Company Ltd, I thank you fo the financial suppot which made this study possible. To the staff of KPLC that availed the necessay datafo use in this study, I thank you all. Special mention goes to Eng. Paul Mugo, Eng.Pete Njenga, Eng. Aggey Machasio and M. Julius Mutahi fo poviding logistical suppot whilecaying out the household suvey. I also thank the System Planning Enginees, Eng. Boniface Kinyanjui and M. Amos Nabaala fo poviding vital planning data that assisted this study. The coopeation and assistance of the sampled household heads contibuted immensely to the success of this study and I wish to thank them all. Lastly, special thanks go to my family. Thei encouagement and undestanding contibuted immensely to the success of this study. iv

5 Table of Contents DECLARATION... ii ABSTRACT... iii ACKNOWLEDGEMENT... iv LIST OF FIGURES... vii LIST OF TABLES... vii LIST OF APPENDICES... ix NOTATION... x CHAPTER ONE INTRODUCTION Backgound Electicity Demand in Kenya Planning fo futue electicity demand Poblem Statement Objectives of the study Oganization of the epot LITERATURE REVIEW Electicity Demand and consumption in households Demand foecasting and Powe planning Methods of foecasting enegy demand Backgound infomation on Powe Planning in Kenya METHODOLOGY Geneal Appoach Sampling Selection of samples Detemining the Sample Size Sample design Classification of households Electical Enegy Household suvey Assessment of income level of a given household v

6 3.8 Data Pocessing Data Analysis RESULTS Appliances Data Analysis of the KPLC Bills Distibution of Suveyed Households accoding to aveage annual electicity consumption Powe Consumption Chaacteistics Mean Consumption Sample Standad Deviation Penetation Level of appliances pe income level Penetation level of appliances in Naiobi Region - Uban Penetation level of appliances in Naiobi Region - Rual Mathematical Calculation (Specific Consumption) of consumption pe appliance Adjustment of calculated consumptions to actual consumptions Clusteing of appliances to end uses and thei Specific consumption Electicity Consumption by end use pe income goup High Income uban households consumption pe end use Middle Income Households Consumption by end use Low Income Uban Households Consumption by end use High Income Rual Households Consumption by end use Middle Income Rual Households Consumption by end use Low Income Rual Households Consumption by end use Using the esults of the suvey to pepae a foecasting model Model Veification CHAPTER FIVE DISCUSSIONS Residential Powe Consumption in Naiobi Region Electicity consumption pattens fo households Consumption by end use Foecasting model Residential Powe Consumption in Naiobi Region CONCLUSIONS AND RECOMMENDATIONS vi

7 6.1 Conclusions Recommendations REFERENCES APPENDICES LIST OF FIGURES Figue 1.1 Histoical electicity sales to end uses in Kenya... 3 Figue 2.1 Actual and Pedicted Enegy demand (GWh) in Kenya Figue 4.1 Distibution of annual consumption fo the uban sample Figue 4.2 Distibution of annual consumption fo the ual sample Figue 4.3 Penetation of appliances in Naiobi Region Uban Households Figue 4.4 % Access levels of appliances in Naiobi Rual Figue 4.5 Electicity Consumption by end use fo Naiobi uban High income households Figue 4.6 Electicity Consumption by End Use fo Naiobi Uban Middle income households.. 55 Figue 4.7 % Shae of consumption by end use in low income uban households Figue 4.8 Electicity Consumption by end use in High income ual households Figue 4.9 Electicity Consumption by end use in ual middle income households Figue 4.10 Electicity Consumption by end use fo ual low income households Figue 4.11 Computed Vs actual Residential electicity demand in Naiobi egion Figue Compaison of the of households and the enegy consumption LIST OF TABLES Table 3.1 Categoization of the sample Table 4.1 Categoization of Uban sample Table 4.2 Categoization of Rual Sample Table 4.3 Penetation (% access levels) of Household appliances in Naiobi egion Table 4.4 Mathematical Calculation of Consumption - Uban Sample Table 4.5 Mathematical Calculation of Consumption - Rual sample Table 4.6 Adjustment of Specific Consumptions of appliances Table 4.7 Clusteing of appliances to end uses vii

8 Table 4.8 Adjustment Specific Consumption of end uses pe income level fo Naiobi Uban/ Rual Table 4.9 Electicity consumption by end use in Naiobi High income uban households Table 4.10 Electicity Consumption by end use fo Naiobi Middle Income households Table 4.11 Electicity Consumption by end use fo low income uban households Table 4.12 Electicity Consumption by end use in High income ual Households Table 4.13 Electicity Consumption by end use in Middle income ual households Table 4.14 Electicity Consumption by end use in low income ual households Table 4.15 Residential Electicity Demand in Naiobi egion viii

9 LIST OF APPENDICES Appendix 1 Electicity Demand (GWh) by Secto in Kenya Appendix 2 Map showing KPLC Naiobi Region Appendix 3 Map of the Suveyed Uban Estates Appendix 4 List of Uban estates Suveyed Appendix 5 List of Rual Locations Suveyed Appendix 6 Standad Consumption fo household appliances Appendix 7 Aveage Annual Consumption fo the Sample uban households KPLC ecods. 84 Appendix 8 Aveage annual Consumption fo the sample ual households - KPLC ecods Appendix 9 Calculated consumption of appliances Uban Appendix 10 Calculated Consumption of appliances - Rual Appendix 11 Adjusted Specific consumption of appliances - Uban Appendix 12 Adjusted Specific consumption of appliances - Rual Appendix 13 Adjusted Consumption of appliances - Uban households Appendix 14 Adjusted consumption of appliances - Rual households Appendix 15 Household appliances owneship and usage suvey esults Appendix 16 Questionnaie fo Household appliances owneship ix

10 Units NOTATION % Pecent A GWh h kv kva kw kwh m m/s MVA MW o C V W Ampee Gigawatt-hou hou Kilovolt Kilovolt-ampee Kilowatt Kilowatt-hou mete metes pe second Megavolt-Ampee megawatt Degee Celsius Volt Watt Abbeviations AFD CFL DSM DVD ERC Fig Fench Development Agency Compact Fluoescent Lamp Demand Side Management Digital Video Disk Enegy Regulatoy Commission Figue x

11 GDC GDP GoK Hi HH ICT KEDSPHS KenGen KES KIHBS KIPPRA KNBS KPLC LCPDP Li Mi MoE No. REA UN Video Geothemal Development Company Goss Domestic Poduct Govenment of Kenya High Income Household Infomation Communication Technology Kenya Enegy Demand, Supply and Policy Stategy fo Households, Small Scale Industies and Sevice Establishments Suvey Kenya Electicity Geneating Company Ltd Kenya Shilling Kenya Integated Household Budget Suvey The Kenya Institute fo Public Policy Reseach and Analysis Kenya National Bueau of Statistics The Kenya Powe and Lighting Company Ltd Least Cost Powe Development Plan Low Income Middle Income Ministy of Enegy Numbe Rual Electification Authoity United Nations Video Cassette Playe/ Recode xi

12 CHAPTER ONE 1. INTRODUCTION 1.1 Backgound Electicity penetation ate in Kenya is cuently at about 29% and the of consumes connected to the powe gid has gown apidly in the ecent past. The gowth in demand fo electicity has been significant ove the last 5 yeas diven by a combination of nomal gowth, inceased connections in uban and ual aeas as well as the county s goal to acceleate tansfomation into a apidly industializing middle income county by the yea 2030 as aticulated in the Kenya Vision 2030 [2007]. The of customes connected to the national gid inceased by 19.8% fom 1,463,639 in 20 to 1,753,348 in 21 as detailed in KPLC Annual Repot [21]. This maks the highest pecentage incease in custome connection eve and eflects the acceleated effots in inceasing electicity access duing the peiod. The incease in custome connection is accounted fo lagely by the domestic custome categoy. The sales fo the industial/commecial custome categoy inceased maginally, to 3,4GWh in the yea 20/21 fom 3,153GWh in the yea 2009/20 epesenting a 3.17% incease in demand as in the KPLC Annual Repot [21]. In tems of customes in this categoy, thee was a maginal incease in connections fom 2,728 customes in 2009/10 to 2,844 customes in 20/11, an incease of 4.2%. KPLC Annual epot [21] shows that total domestic consumption in uban aeas inceased fom 1,113GWh in 2007 to 1,424GWh in 21. Duing the same peiod, the of uban customes ose fom 681,060 to 1,444,061 epesenting a dastic 112% incease. Total domestic consumption in ual aeas also evealed the same tend whee annual consumption ose fom 221GWh in 2007 to 307GWh in 21. The shae of ual domestic consumes to uban ones has also inceased since 2008 fom 15% to 18%. 1

13 The acceleated new connections ae due to the delibeate effots by the Govenment to acceleate electification in the county though vaious initiatives. Some of theinitiatives include enhanced funding of the Rual Electification Pogamme, new connections financing evolving fund (Stima loan) by KPLC aimed at availing cedit to potential customes who cannot affod to pay upfont the cost of new connections, tansfome maximization whee customes within a distance of 600m fom a distibution tansfome pay unifom connection chages. Line maximization initiative by KPLC wheeby the Company has ceated a einfocement fund fo the extension of MV lines and installation of tansfomes within a easonable distance to cove existing and potential customes. As pe Sessional Pape No. 4, [2004] on enegy, the Govenment plans to connect at least 1 million customes fom 2008 to the yea 22 and to incease electicity access to 40%of the Kenyan population by The KPLC database of connected customes shows that a majoity of these connections ae in the esidential categoy. Kenya s economic bluepint, Vision 2030 envisages that the county will move fom a low income economy to a middle income economy by the yea 2030 as outlined in Kenya Vision 2030, [2007]. Futhe, the new constitution has put in place a devolved system of Govenment which will take development close to the people though the county system. Each county will equie acceleated electicity connectivity to spu economic gowth and thus enomous enegy esouces will be equied. The powe distibution utility in the county, KPLC has put in place mechanisms to seve the county govenments with electicity, by opening business banches in all the 47 counties. This will ensue connectivity of even moe customes in suppot of the achievement of Vision 2030 and the new constitution. As shown in Figue 1.1, the total enegy consumption inceased fom 3,628GWh in 2002 to 6,123GWh in 21, epesenting a 69% incease ove the peiod. The esidential enegy consumption inceased fom 954GWh in 2002 to 1,769GWh in 21, epesenting 85% incease ove the peiod. The inceased consumption is attibuted to inceased economic activity that saw incease in industies and incease in the commecial secto activities. Inceased connectivity in the esidential secto due to a 2

14 combination of enhanced ual electification, and othe enegy secto policiesto enable the Govenment meet the 40% connectivity by yea 2020 as outlined in Sessional pape No. 4, [2004] on enegy. The tend in electicity consumption by each secto is shown in Appendix 1. Demand (GWh) Total electicity consumption and esidential enegy demand (GWh), ,628 3,802 4,090 4,379 4,580 5,065 5,322 5,432 5,624 6, ,247 1,384 1,569 1,547 1,605 1, Residential Total Sales Figue 1.1 Histoical electicity sales to end uses in Kenya 1.2 Electicity Demand in Kenya The daily load pofile as illustated in Fig. 1.2 shows that the powe system expeiences a peak between 18.30Hs and 22.30hous daily due to inceased demand fom the households duing this peiod.the daily load cuvesindicate that Naiobi egion has the highest demand compaed to all the othe egions, with the egion having a daily demand peak of about half that of the total National peak demand. 3

15 DAILY LOAD CURVES LOAD, MW National Gen. Demand Naiobi Region Coast Region West Region Mt. Kenya Region TIME OF DAY Figue 1.2 Load Cuves- 10th May 21; Souce NCC, KPLC Oveall, thee is a stong electicity demand gowth in Kenyaof about 5.3% pe yea. The demand is being diven by a combination of factos, namely;economic gowth,enhanced ual electificationaising fom the Govenment s commitment to incease connectivity to 40% by the yea 2020 as pe Sessional Pape No.4 on Enegy [2004],and inceased economic activities in a bid to move the county to a middle income economy as envisaged in Vision This gowth is expected to incease esulting fom the implementation of Vision 2030 flagship pojects. Figue 1.3 illustates actual capacity availability and peak demand tend fom 2004 to 21. As illustated, thee has been a sustained annual incease in demand. The peak demand has been met only due to inclusion of emegency powe fom 2006, without which sevee capacity shotages would have esulted. Insufficient eseve magins have esulted in load shedding wheneve powe outages ae expeienced. This clealy illustates the need fo a sound foecasting model and adequate implementation of planned pojects to meet the gowing demand as well as maintain optimum eseve 4

16 magins MW Available Capacity (MW) Emegency Capacity Peak Demand (MW) Figue 1.3 Peak Demand and available capacity Souce: KPLC ecods 1.3 Planning fo futue electicity demand Sound and ealistic foecasting is the key to good planning in any industy. In an electic utility, demand foecasting is the level of electical demand a utility can expect fom a specified of consumes in a specified peiod of time as defined by Islam and Alawi,[1996]. The impotance of electical demand foecasting is thee fold: geneation expansion planning, tansmission expansion planning and financial planning. In addition, foecasting helps utilities to detemine the system s spinning eseve and fuel equiement, and plan thei unit maintenance scheduling. Electical demand foecasting of an electic supply system could eithe be load (MW) o enegy (MWh) fom a specified categoy of customes connected to the system. Electic Powe utilities need to have accuate foecasts if they ae to meet the futue powe needs of thei customes. A bloated electicity demand foecast leads to ove- 5

17 investment, excess capacities and edundancies in the powe system with significant cost implications to consumes and the economy. On the othe hand, a depessed load foecast leads to unde investment in the powe system which causes seious secuity and eliability of supply quality challenges. The accuacy of the foecast is theefoe vey impotant in powe system planning. Kenya s powe geneation, tansmission and distibution system planning is undetaken on the basis of a 20 yea olling Least Cost Powe Development Plan (LCPDP) that is updated evey yea. Electicity Demand Foecasting is a citical input in the LCPDP. A team of pofessionals fom KPLC, KenGen, Ketaco, REA and GDC unde the chaimanship of the Enegy Regulatoy Commission (ERC)pepaes electicity foecasts to be input in the development of the LCPDP. Seveal methodologies have been used ove time in electicity demand foecasting. Initially, long tem electicity demand foecasting was done using tend line analysis wheeby time seies data of past demand fo electicity was analyzed as a function of time using a vaiety of techniques that anged fom simple gaphing to sophisticated mathematical cuve fitting. The esulting tend was extended into the futue, sometimes expessed in tems of compound gowth ates. Because this method implicitly assumed that pesent gowth pattens would emain essentially unchanged in the futue, it was often acceptable in stable socio-economic conditions that encouage electicity demand to incease steadily ove time. Howeve, with time, as moe factos came into play that affect demand fo electicity, othe methodologies wee adopted to povide moe accuate foecasts. 1.4 Poblem Statement Histoically, the county s LCPDPs have been based on electicity demand foecast developed fom a simple econometic analysis. This model uses histoical elationships between electicity demand, Goss Residential Poduct (GDP) and electicity pices to poject electicity demand fo the futue. Howeve, a fundamental assumption of the econometic model is that the elationships between income, pice and demand (coefficients) which existed in the past will continue to hold in the futue. In addition, a 6

18 moe fundamental stuctue of enegy demand, which includes the population, penetation of appliances, technological changes, fuel substitution and othe factos is not analyzed and the model s pedictive capability beaks down if any of the vaiables change.theefoe, demand foecasting using the cuent econometic model does not give accuate esults and a bette methodology is equied which will take in to account othe vaiables that affect electicity demand. The autho is not awae of any study done in Kenya elating todeveloping a demand model that takes to consideation demogaphic factos, social and economic factos, and technological changes. A study done fo the ERC by KIPPRA [20], attempted to eview the electicity consumption pattens fo the vaious consume categoies in the county, but did not captue the uban ual divide and the income class of the households. Technological aspect of consumption was also not consideed. A study on Kenya s Enegy Demand, Supply and Policy Stategy fo Households, Small Scale Industies [KEDSPHS2002] came up with the penetation levels of end uses which included lighting, entetainment, domestic cooking and wate heating among othes. The study howeve did notanalyse the consumption by the vaious end uses to enable estimation of the demand in households by the end uses. This study will theefoe analyse the enegy demand in households acoss the uban ual divide and acoss diffeent income levels and seek to develop a model fo electicity demand foecasting fo the esidential secto that will coelate the enegy demand to the coesponding social, economic, demogaphic and technological factos that affect this demand in Naiobi egion. 1.5 Objectives of the study The geneal objective of the study is to inteogatethe econometic load foecasting model cuently used in Kenya fo powe planning and to develop a suitable model fo foecasting esidential use of electical enegy in Kenya. The specific objectivesae to; 7

19 i) Cay out a household electical enegy suvey in KPLC Naiobiegion. ii) iii) iv) Analysethe data to assess the cuent electicity consumption of households in Naiobi egion. Develop a model fo esidential powe consumption in Naiobi egion. Veify the model developed. 1.6 Oganization of the epot ChapteOne discusses cuent powe situation in the county and outlines the objectives of the study. Chapte Two compises liteatue eview while Chapte Thee discusses the methodology used in this study. The esults and analysis ae pesented in Chapte Fou while discussions ae done in Chapte Five. Conclusions and ecommendations ae pesented in Chapte Six. 8

20 CHAPTER TWO 2.0 LITERATURE REVIEW 2.1 Electicity Demand and consumption in households Electicity demand in Kenya has been gowing apidly ove the last five yeas, with the latest ecoded demand values almost outpacing the pojected demand. Theoetically, electicity demand is a function of electicity pice, of electicity appliances, consume incomes, ambient tempeatue level and consume load pattens, which diffe by egions, consume goups and income levels. A eliable load foecast methodology must coectly estimate the elationships between electicity demand and these factos. Household electicity demand depends on social economic factos, demogaphic factos, geogaphical factos and technological stuctues. Fo accuate foecasts, a foecasting model should take into consideation these vaiables and thei elation to demand. The use/ selection of the pope methodology and compute pogam fo the electic enegy demand foecasting is an extemely impotant task. Many studies wold ove have been caied out to establish the elationship between electicity demand and othe vaiables. Astudyby Magambo [20]on household electicity demand and consumption pattens in a middle income esidential estate in Naiobi by established that weathe and GDP have negligible effect on enegy consumption in the esidential secto in the medium tem. The study futhe established that esidential enegy consumption is best detemined using the end use engineeing models. Howeve, the study was based on a small sample of households in a middle income estate in Naiobi and did not cove esidential uses of electicity acoss all income goups of customes and in diffeent geogaphical locations. The study did not conside the social-economic, demogaphic factos and location impacts on enegy use. In a study of household enegy consumption and pattens in uban Ethiopia, Gamtessa [2002] established that electicity is a substitute fo othe foms of enegy and that its 9

21 adoption and usage inceases with incease in household income and deceases with incease in pices of electicity and substitute fuels. The consumption also inceases with household size.houi and Kofali [2004] in a study of esidential enegy consumption pattens in uban Lebanon showed that seasons and months have a significant impact on enegy consumption. Coelations wee also indicated fo enegy consumption with apatment aea, income and of esidents. McNeil and Letschet [2000] in a study of six counties demonstated that incease in electical appliance owneship is a majo dive fo inceased electicity consumption in developing counties. In the Govenment epot fo Well-Being in Kenya, KNBS [2008] the majoity of poo households have no access to electicity. Among the ual poo, only 0.7% wee connected to electicity compaed to 22.4% uban poo. Among the non-poo, 6.5% of households in the ual aeas ae connected to electicity compaed to 62.0% non-poo in the uban aeas. Fuel based (paaffin) lighting is the most pefeed souce of enegy fo poo and non-poo in ual aeas. Futhe, 0.1% of the poo use electicity as the souce of fuel fo cooking and 1.7% non poo use electicity fo cooking. The epot futhe establishes that the expenditue fo electicity fo poo households was 5.2% of the total expenditue compaed to 20% expenditue on electicity fo the non-poo. A study on Kenya s Enegy Demand, Supply and Policy Stategy fo Households, Small Scale Industies and Sevice Establishments fo the Ministy of Enegy,(KEDSPHS 2002) established that at household level, electicity is used fo lighting by 99%, entetainment by 90%, ioning clothes 69%, efigeation by 35%, wate heating by 26%, domestic cooking by 24%, home businesses by 16% and house heating by 9%. The epot futhe established that the national aveage household pe capita consumption is 694kWh/ yea, with ual aeas using 544kWh/yea and uban aeas using 844kWh/ yea. Highe income uban households consumed 1,354kWh/yea and the low income 606kWh/ yea. The middle income class in uban aeas consumed 921kWh/yea.The KEDSPHS study was aimed at establishing connectivity, end use pattens, quantities and costs. Howeve, the data gatheed only coveed a single month s electicity bill and not the aveage consumption of a household ove a peiod of time. End use pattens, 10

22 the of occupants, owneship of appliances and usage wee not consideed in the study. The Kenya Integated Household Budget Suvey by the GoK [2007] futhe established that 51% of uban households use electicity fo lighting, but only 1.8% use it fo cooking. In ual aeas, 3.9% use electicity fo lighting, but only 0.2% use it fo cooking. The suvey did not categoize the usage of electicity among the vaious income goups in the county based on geogaphical location. A study done fo ERC by KIPPRA [20] on consumption pattens of enegy in Kenya established that the use and connectivity to electicity vaied acoss egions. It futhe established that the demand fo cooking fuels such as wood, chacoal, keosene, LPG and electicity ae diven by cetain key factos and vay depending on whethe the house is located in ual o uban aeas. Fom the study, the key deteminants of choice fo use of electicity wee employment level (income level), pice of wood fuel and education level. The KIPPRA epot did not go futhe to analyze the existing end uses in households and thei contibution to enegy demand. 2.2 Demand foecasting and Powe planning Electical demand foecasting is classified into thee main categoies: Shot tem, Medium tem and long tem. Shot tem foecasts pedict load ove a few hous ahead to a few weeks and coves a time span of 1 yea. They play an impotant ole in the day-to-day opeations of a utility such as unit commitment, economic dispatch, hydo-themal coodination, load management, etc. A shot tem foecast is commonly efeed to as an houly load foecast. Medium tem foecast coves a peiod of 1 to 5 yeas. They ae equied in planning fuel pocuement, scheduling unit maintenance, enegy tading and evenue assessment fo utilities. A medium tem foecast is efeed to as monthly load and enegy foecast. 11

23 A long-tem foecast pedicts load fom 5 25 yeas. This type of foecast is impotant in deciding on the system geneation and tansmission expansion plans. Since the geneation capacity expansion equies huge esouces and a long lead time, a longtem foecast is vey impotant. A long tem foecast is commonly known as an annual peak load and enegy foecast. 2.2 Methods of foecasting enegy demand Seveal models ae used fo electicity demand foecasting. These ae boadly categoized as time seies/ time tend, econometic, engineeing end-use, Hybid and Atificial Neual Netwok (ANN) models. The use of the models depend on the foecast peiod equied, the accuacy and the data available. Some of the foecasting methods ae discussed hee below. Time Seies/ Tend method In this model, a time seies of past demand fo electicity is analyzed as a function of time, using a vaiety of techniques that ange fom simple gaphing to sophisticated mathematical cuve fitting. The esulting tend is extended into the futue, sometimes expessed in tems of compound gowth ates. Because this method implicitly assumes that pesent gowth pattens will emain essentially unchanged in the futue, it is often acceptable when stable socio-economic conditions encouage electicity demand to incease steadily ove time. Equation 2.1 illustates a time seies method, whee the vaiable to be pedicted is expessed as a function of time = + (2.1) Whee, P = Electical Demand F = function C = a Constant 12

24 The main disadvantage with this method is that it does not take into consideation possible inteactions of the vaiable with othe factos. This method is only useful fo shot tem foecasts. Gee [21] indicates that the time seies model was used pio to 1970 in the US (duing the oil cisis) whee the annual gowth in sales of electicity had been aveaging about 6% and many utilities simply foecast sales to gow at this level indefinitely. Howeve, with the shap ise in enegy pices in the 1970 s along with a slowdown in economic activity, the model was not accuate enough and moe sophisticated techniques wee sought. Econometic methods Tintne [1968], defines econometics as the application of mathematical statistics to economic data to lend empiical suppot to the models constucted by mathematical economics and to obtain numeical esults. Econometic methods use aggegated data on GDP, taiffs, etc to pedict load and have a theoetical statistical base. Usually, they ae used fo a whole class of customes and do not take into account the technological stuctue of thei enegy consumption. Damoda [2004] illustates that most common types of econometic models used in enegy studies ae based on the Cobb-Douglas (CD) functional fom as demonstated hee below: Theoetically, electicity is consideed a nomal good whose demand can be expessed in the geneal fom as a function of consume income and electicity pices (taiffs) as shown in Equation 2.2. E d = f( y, p) (2.2) Whee: = Enegy Demand = Income 13

25 = Pice The geneal fom can then be expessed in a specific Cobb-Douglas (CD) functional fom given in Equation 2.3. α β E d = ay p (2.3) Whee: = Constant = Income Elasticity of Demand = Pice Elasticity of Demand The Cobb-Douglas functional fom can then be expessed as a log-linea fomulation and model paamete coefficients αandβ estimated using econometic tools. log E d = a+αlogy βlogp (2.4) The estimated coefficients expess the elationship between electicity demand and the espective vaiable and this elationship is used to foecast electicity demand in the futue. It is expected that a positive elationship exists between electicity demand and consume income and a negative elationship between electicity demand and electicity pices. The econometic model theefoe utilizes histoical time seies data to statistically estimate (by means of egession analysis) the constant and the coefficients α, and β. Howeve, a fundamental assumption of the econometic models is that the elationships between income, pice and demand (coefficients) which existed in the past will continue to hold in the futue. Thee is also inceasing evidence showing that the elationship between enegy, income and pices may vay in the futue when impotant changes in the technological stuctue of enegy demand, consume behavio ae taking place. The model does not conside that the lowe income goup is now able to affod cheap 14

26 appliances now available in the maket, thus leading to moe demand fo electicity. The econometic model theefoe is no longe used by most utilities in the developed wold. In Kenya though, econometic models ae used to poduce long tem enegy foecasts. End-Use methods The end use method captues the impact of owneship and enegy usage pattens of vaious electical appliances and systems. Enegy demand fo each activity is usually a poduct of two factos; namely the level of activity (the enegy sevice) and the enegy intensity (enegy use pe unit of activity). Accoding to Swishe et al [1987] enegy analysis holds the mix of enegy sevices and activities. The level of activity depends on factos such as population, income and economic output. The level of enegy intensity depends on the enegy efficiency, including both technological and opeational aspects. A summation of the poducts of these two factos ove all activities gives the total enegy demand as indicated in equation 2.5. = Q.I (2.5) Whee:Q = Quantity of enegy sevice i, I = Intensity of enegy use fo enegy sevice i The quantity of enegy sevices depends on seveal factos, including the population, the shae of the population using the end sevice, and the extent of use of each sevice as shown in Equation 2.6. = (2.6) Whee, 15

27 = quantity of enegy sevice i, = of customes eligible fo end use i, =penetation (total units/ total customes) of end use i, = magnitude o extent of use of end sevice i, e.g. of hous of use, degee -days, etc End-use models have the difficulties of obtaining accuate data on appliances fom the Households. Besides, most of the time, demogaphic data on households is not egulaly updated and this leads to inaccuacies in the foecasts. Howeve, end-use models ae moe accuate than econometic and time seies models since in end-use models, enegy demand is boken down in to the vaious end-uses, penetation levels of the vaious end uses, consideation of the diffeent socio economic goups is made, demogaphic paametes and well as uban/ ual location of the households ae also taken in to consideation. Bhatachayya [21] explains that since the end use models ae able to captue new technological developments in enegy usage, they have an edge ove othe categoies of models fo enegy demand. The end use model was developed in the late 70 s afte the oil shock due to the need fo accuate foecasts. Many counties, like Fance, Nodic Counties, Thailand and somestates in the United States of Ameica have used the end use model fo analysis of thei electicity demand successfully. Howeve, end use models have difficulties especially in obtaining data. 2.3 Backgound infomation on Powe Planning in Kenya The Wold Bank/ UNDP [1987) unde the Enegy Secto Management Assistance Pogam (ESMAP) pepaed the fist load foecast in the county. This esulted in the pepaation of the Kenya National Powe Development Maste Plan The maste plan coveed among othe things the Load foecast, which was then used as an input in the Least Cost Powe Development Plan fo 20 yeas, fom In foecasting fo powe demand, econometic method was used whee time seies 16

28 analysis models used histoical data fom on Sales and Taiffs obtained fom KPLC and GDP obtained fom Cental Bueau of Statistics (CBS) now Kenya National Bueau of Statistics (KNBS). The foecasts coveed vaious categoies of customes; namely the Residential and small commecial customes, Industial and Commecial Customes and off-peak customes coveing the wate heating loads. The oveall load foecast was obtained by summing up the vaious load foecasts obtained fom the diffeent categoies. An update of the load foecast and consequently the Least Cost Powe Development Plan (LCPDP) was again done in 1991 and the mandate of updating the load foecast annually theeafte was tansfeed to KPLC. The same model has been used fo foecasting enegy demand since then. With the unbundling of the enegy secto, the mandate of pepaing the LCPDP now lies with the ERC with the input of all the enegy secto playes, namely KPLC, KenGen, REA, Ketaco and the Geothemal Development Company. An econometic model has been used to pepae foecasts fo the esidential secto fo input in the cuent long tem Least Cost Powe Development Plan , using Equation 2.7. h =17, (2.7) Whee, h = Pojected enegy demand fo yea t = foecast yea = Non- agicultual Goss Domestic Poduct = Moving Aveage Residential Taiff = Non-Agicultual GDP Gowth elasticity 0.20 = Residential secto moving aveage taiff coefficient of elasticity 17

29 Simila econometic models ae used fo the Commecial/ Industial, off-peak and ual electification enegy demand foecasts. This method of detemining the enegy demand ignoes the individual loads that make up that oveall demand in the esidential secto, the penetation levels of end uses and assumes that the factos that affected powe demand in the past will continue to affect powe demand the same way within the foecasting peiod. The method also ignoes any technological changes that might occu due to policy changes and does not diectly include new load due to inceased connectivity. Fo example, the pedicted demand diffes consideably fom the actual demand fo the yeas using the econometic models as shown in Figue 2.1. Actual and Pedicted Enegy Demand (GWh) in Kenya ( ) Enegy Demand (GWh) 9,000 8,000 7,000 6,000 5,000 4,000 3,000 5,556 4,750 5,809 5,035 Pedicted Demand (E p ) = 326.1t R² = ,185 6,962 6,521 6,064 5,347 5,697 6,169 6,385 7,391 6,489 Actual Demand (E d ) = 302.5t R² = ,734 6,692 8,175 7,303 2,000 1, Pedicted enegy demand (GWh) Linea (Pedicted enegy demand (GWh)) Actual enegy demand (GWh) Linea (Actual enegy demand (GWh)) Figue 2.1 Actual and Pedicted Enegy demand (GWh) in Kenya Souce:LCPDP

30 Pedicted and actual enegy demand is given by Equation 2.8 using a linea model as follows: = ,191 (2.8) and = ,472 (2.9) Whee; = Pedicted demand = Actual demand = yea Figue 2.1 shows that the gowth patten between the actual and the pedicted demand is almost consistent. Based on Equation 2.8 and Equation 2.9, the actual demand is ove foecasted by an aveage of 837GWh. Fom the model, the foecasted demand can be epesented in Equation 2.10 as follows: = + (2.10) Whee; A is a vaiable which is time dependent and C is a constant and t is the foecast yea. The demand foecast deviates fom the actual demand by about -15% using the econometic model. The diffeence between the foecasted demand and the actual demand is quite substantial and this shows that the econometic model used is poo. Futhe, the usual pactice by the planning/ design enginees is that specific esidential demand is usually detemined fom custome applications fo powe supply at the time 19

31 of application fo new supply. Loads ae neve listed in this case; instead the of lighting points and of socket outlets ae listed to estimate the demand. Fo nomal households with odinay loads, the minimum demand is assumed to be 3kVA afte applying a divesity facto efeed to as Afte Divesity Maximum Demand (ADMD) of 80% fo esidential loads. This facto is applied due to assumption that not all the loads contibuting to the maximum demand ae on at the same instant. Afte connection, no suvey/ inspection is caied out to establish the actual loads connected by the custome and this leads to unde/ oveestimates of the demand. In eal life situation, some esidential loads ae ated as high as 7.5kW (e.g. instant showes) and poviding fo a demand of 3kVA is gossly an undeestimate. It is theefoe necessay that a suvey is done to establish the actual specific esidential loads that make up the esidential powe and enegy demand to facilitate accuate foecasts of the enegy demand in the esidential secto. The suvey esults will futhe assist the county to detemine and plan on which Demand Side Pogammes to implement in the esidential secto. 20

32 CHAPTER THREE 3.0 METHODOLOGY In this chapte, the methods and pocedues used in sampling, data gatheing and analysis ae descibed. 3.1 Geneal Appoach The study utilized both pimay and seconday data fo analyzing and developing the model. Pimay data was obtained fom household suvey field wok using pesonal inteviews, billing ecods and obsevations fo household appliances. Seconday data on the othe hand was obtained fom official statistical publications including Statistical Abstacts, Economic Suvey, Census epots, the Kenya Powe and Lighting Company Limited Annual Repots and statistics fom the KPLC custome database, and othe enegy publications. The study was based on KPLCdefined Naiobi egion in both uban and ual geogaphical locations as the domains of estimation. Kenya s administative boundaies wee ovelaid on to the KPLC Regions since the study used the household distibution as given in the KNBS [20] census esults. The disticts coesponding to KPLC Naiobi egion wee sampled fo the study. 3.2 Sampling Selection of samples The study selected household samples within Naiobi egion aleady connected to the electical gid. Convenience sampling method was used to choose samples, whee households wee selected fo the suvey. The study classified vaious espondents in Naiobi Region accoding to thei geogaphical location, uban and ual fo collection of appliances data. 21

33 3.3 Detemining the Sample Size The coect sample size was detemined based on guidelines by the United Nations (UN) Guidelineson household suveys UN [2005]. The sample size equied fo the household suvey was detemined accoding to Equation 3.1. = 1 / (3.1) Whee; = households (sample size) = statistic that defines the level of confidence = estimate of a key indicato to be measued by the suvey = sampling design effect = multiplie to account fo the anticipated ate of non-esponse = popotion of total population accounted fo by the taget population and upon which the paamete,, is based; = aveage household size (no. of pesons pe household) = magin of eo to be attained As pe the UN guidelines, the ecommended values fo some of the paametes to be used in developing countieswhee the actual values ae not available ae as follows: =1.96, defines the level of confidence as 95%; =2; =10%; =5.1 in Kenya based on KIHBS, [2005/6] and the ecommended value of =0.15. Substituting the above gives:- = / (3.2) 22

34 = 0.28 (popotion of powe consumption fo the esidential secto in Naiobi to the total powe consumption in Naiobi= 28% as pe KPLC statistics as detailed in the KPLC Annual epot [21]. = 0.52(the popotion of the connected households in Naiobi egion to the total of households in the egion, based on KNBS, Kenya Population and Housing Census epot Volume II [20]. Substituting the above values to the fomula in equation (2.10) will gives a sample size of 360 households. 3.4 Sample design 360 questionnaies wee sent out to 360 households. Out of those, 351 wee esponsive and 9 wee non- esponsive. Nine (9) households declined to povide infomation to the eseache and wee theefoe left out of the study.kplc Naiobi Region coves both Uban and Rual geogaphical locations. Naiobi egion is futhe divided into 3 sub egions, namely Naiobi Noth, Naiobi South and Naiobi West. The households suveyed wee statified into 2 heteogeneous stata, based on geogaphical location - uban and ual within Naiobi egion. Population and housing census data fom the Kenya National Bueau of Statistics [20] was used to classify all the geogaphic administative units within Naiobi egion into uban o ual, thus allowing the taget population to be statified by this citeion. The study used a combination of suvey (fo pimay data) and non suvey (fo seconday data). A sample size of 360 households has been calculated as shown in Section 3.3, but the suvey esults cove 351 households who esponded to the study. The suvey was caied out in all the thee sub-egions within Naiobi egion to enable pope epesentation of samples i.e. Naiobi West, Naiobi Noth and Naiobi South. Convenience sampling technique was used. The sample size was 236 Households in the uban location and 96Households in the ual locations. 23

35 The suvey coveed Naiobi egion shown on the map as Appendix 2 and 3. Twenty estates within Naiobi uban aeas wee selected and eight sub-locations in the ual aea.the study aeas wee selected to epesent the thee social economic classes of the households, the High Income, Medium Income and Low Income. The selection was done fom acoss Naiobi to epesent the thee sub-egions as pe KPLC administative boundaies.the uban estates sampled fo the study ae attached as Appendix 4 and the ual aeas sampled ae attached as Appendix 5. The study omitted the off gid aeas due to logistical difficulties of caying of the suvey. The off gid aeas fom pat of Naiobi egion and ae supplied by geneation using diesel o wind and sola that is not connected to the gid. 3.5 Classification of households The study classified the 351 sampled espondents in Naiobi Region fo collection of electical appliances data. The espondents (households) wee classified accoding to thei geogaphical location, uban and ual. The classification into eithe ual o uban was done as pe the definition of the 2009 Kenya Population and Housing epot [20]. An uban aea in the epot is defined as an aea with an inceased density of human ceated stuctues in compaison to the aeas suounding it and has a population of 2,000 and above as indicated in KNBS [20]. These include Cities, Municipalities, Town Councils and Uban Councils. A ual aea is defined as lage isolated aeas of open county (in elation to open fields and not foests, etc), often with low population density as indicated KNBS [20]. 3.6 Electical Enegy Household suvey A household electical enegy suvey was undetaken fo the sampled households. The suvey involved collecting data on existing electical appliances in a household. Obsevations as well as pesonal inteviews wee used in the study. A questionnaie was designed to collect the following data among othes: Electical appliances existing in the household; name, type, of each appliance type in the household and powe ating whee possible. 24

36 Numbes of hous each of the appliances ae used in the household, Altenative foms of enegy used in the households if any. KPLC electicity account and mete. The questionnaie used fo the suvey is attached as Appendix Assessment of income level of a given household Afte classification of the samples in to eithe uban and ual, the households wee futhe classified as eithe low income (Li), medium Income (Mi) and High Income (Hi). Thus the study coveed six (6) categoies of electicity uses, i.e. Li_Uban, Mi_Uban,Hi_Ubanand Li_Rual, Mi_Rual, Hi_Rual. The Economic Suvey [20] classifies households in to income goups by expenditue pe month in Naiobi as follows: households spending below KES 23,000 ae classified as low income (LI); households eaning between KES 23,0 and KES 125,000/= as middle income and above KES 125,000/= as high income. To categoize the customes in to the vaious income categoies, the eseache inquied about the income of the householdduing the suvey. The statement of the espondent was then ecoded. This statement was howeve found to be uneliable, and most of the espondents wee not willing to divulge infomation about thei income. This development pesented a daw back to the study and the autho had to find anothe method of classifying the households. Anothe indicato was then chosen, the yealy aveage electicity consumption of the households, accoding to KPLC bills. On aveage, and as many studies show, electicity consumption is popotional to the income level. Fo each household, the mete was ecoded duing the suvey and the last 12 monthly bills fo each account wee obtained fom KPLC. On the basis of this yealy consumption, the household was categoized into one of the 3income goups in the uban and ual aeas.the categoies of income goups wee only identified accoding to the amount of yealy electicity consumption. 25

37 The classification was done accoding to the annual electicity consumption as pe Table 3.1. Table 3.1 Categoization of the sample Uban Sample Consumption Range Categoy (kwh) Rual Sample Consumption Range Categoy (kwh) >3,000 High Income >1,000 High Income 1,000 3,000 Middle Income 500 1,000 Middle Income <1000 Low Income < 500 Lo Income The classification was theefoe not linked to a diect assessment of the monetay income available to the household since this infomation was not eadily available. 3.8 Data Pocessing The data gatheed using the questionnaie fom the households was coded in an excel speadsheet and pogammed to captue the enegy consumption of the vaious appliances and to categoise the households in to the socio economic classes. Appendix gives the data coded in excel woksheets fo analysis. Fo each household, data on electicity consumption fo a yea was obtained fom KPLC database. The aw data fom KPLC was analysed to detemine its validity based on the appliances available in the house and usage. In this epot, the households ae epesented by questionnaie s, fom 1 to 351. The KPLC data was checked fo zeo o nea zeo consumption and uneasonably high consumptionove the peiod. Those houses with an aveage of zeo consumption wee omitted fom the analysis and those with an uneasonably high consumption which donot match the available appliances wee also omitted fom the study. Appliances that could exist in a household wee listed in advance and thei standad consumptions wee used. The infomation collected was the owneship of appliances, the of appliances and usage (whethe a paticula appliance was in use o not). Whee a paticula appliance was not in use, owneship was consideed as zeo. Due to the intusive natue of the suvey, espondents wee eluctant to let the inteviewee 26