DESAIN MODEL UNTUK PRAKIRAAN BEBAN JANGKA MENENGAH DENGAN REGRESI MULTIPLE DAN JARINGAN SYARAF TIRUAN : Studi Kasus Pada PT.PLN (Persero) P3B Jawa Bali Region IIJawa Barat

• Main Author: Bramana, Andrian (Author)
• Format: Book
• Published: 2014-06-20.
• Subjects: Thesis; NonPeerReviewed
• Online Access: Link Metadata

Summary

Penelitianinimengkajitentangdesain model untukperkiraanbebanlistrikjangkamenengahdenganmetodemultiple regressiondanmetodebackpropagationberbasisjaringansyaraftiruan. Data yang dipakaiadalah data bebanlistrikdari PT. PLN (Persero) P3B Jawa Bali Region II Jawa Barat setiapbulanmulaijanuarisampaidengandesembermulaidaritahun 2007 sampaidengan 2013 yang kemudian data tersebutakandilakukanpembuatan model multiple regressiondenganmelakukanperhitungandenganmicrosoft excel. SedangkanuntukmetodeBackpropagationberbasisjaringansyaraftiruan data tersebutakandibelajarkanpadasistemperangkatlunak yang sudahdirancangdenganalgoritmabackpropagation.Softwarependukunguntukmerancang program tersebutdigunakanMatlab ver. R2009a dariThe MathWork. Inc. melaluiperhitungandarihasildesain model denganmenggunakanmetodemultiple regressionmenunjukkanrata-rata error 0,0114atausebesar 1.14% dandarihasilujiforecastuntuk data digesersatutahunmenunjukan rata-rata error 0.0386 atausebesar 3.86% halinimasihdalamtoleransinilai yang diberikanoleh PT. PLN yaitusebesar 5%. Sedangkandarihasilsimulasidesain model load forecastingdenganmetodebackpropagationberbasisjaringansyaraftiruanmenunjukantingkaterror rata-rata sebesar 0.012% dengannilaiepoch 9000 dannilailearning ratepada 0,5. Dengandemikiandapatdisimpulkanbahwaperamalanbebanlistrikjangkamenengahdenganmenggunakanmetodebackpropagationberbasisjaringansyaraftiruanlebihbaikdibandingkandengandesain model perkiraanbebanlistrikdenganmenggunakanmultiple regression. This studyexamines thedesign ofa modelforthe medium-termelectricityloadforecastby the method ofmultiple regressionandback propagationmethodbased onartificial neuralnetworks. The dataused is theelectricityload datafromPT. PLN(Persero) P3BJawaBaliRegionIIWest Javaeverymonth startingJanuaryuntil Decemberrangingfrom 2007to 2013and thenthe data isperformedmultiple regressionmodelingtoperformcalculationswithMicrosoft Excel. As for themethod ofbackpropagationartificial neuralnetwork-based datawillbe taughtina softwaresystemthat has beendesignedwithback propagationalgorithm. Supporting softwareis usedto designthe programMatlabver. R2009afrom TheMathWork. Inc..throughthe calculationofthe results ofthe designmodelsby usingmultiple regressionmethodshowsan averageerrorof1.14% or0.0114oftest resultsandforecastsforthe datais shiftedoneyearshowedan averageerrorof0.0386or3.86%, this is stillwithinthe tolerancevaluegivenbyPT. PLNis equal to5%. While the results ofthe simulationmodel designloadforecastingmethodbased onback propagationneural networksshowan averageerrorrateof0.012% with a9000epochvalueandthe valueof learning rateat 0.5. It can be concludedthat themid-termelectricload forecastingusing amethodbased onback propagationneural networkis better than thedesignestimatemodelsof electricloadby usingmultiple regression.

Item Description

http://repository.upi.edu/15378/2/S_TE_0902244_Title.pdf
http://repository.upi.edu/15378/1/S_TE_0902244_Abstract.pdf
http://repository.upi.edu/15378/4/S_TE_0902244_Table_of_content.pdf
http://repository.upi.edu/15378/2/S_TE_0902244_Chapter1.pdf
http://repository.upi.edu/15378/2/S_TE_0902244_Chapter2.pdf
http://repository.upi.edu/15378/5/S_TE_0902244_Chapter3.pdf
http://repository.upi.edu/15378/3/S_TE_0902244_Chapter4.pdf
http://repository.upi.edu/15378/3/S_TE_0902244_Chapter5.pdf
http://repository.upi.edu/15378/2/S_TE_0902244_Bibliography.pdf