Gina Fasali da Zaɓi don Ƙirar Ƙarfin Hasken Rana na PV: Tsarin Koyon Injin

1. Gabatarwa & Bayyani

Haɗa ƙarfin hasken rana (PV) cikin ayyukan masana'antu wata dabara ce mahimmanci don rage fitar da iskar gas mai gurbata muhalli da haɓaka dorewa. Duk da haka, rashin kwanciyar hankali da sauye-sauyen ƙarfin hasken rana suna haifar da ƙalubale masu mahimmanci ga kwanciyar hankalin cibiyar wutar lantarki da samar da wutar lantarki mai dogaro. Saboda haka, daidaitaccen hasashen ƙarfin samar da wutar lantarki na PV na ɗan gajeren lokaci yana da mahimmanci don ingantaccen sarrafa makamashi, daidaita nauyin lantarki, da tsara ayyukan gudanarwa.

Wannan takarda ta gabatar da sabon tsarin koyon inji don hasashen ƙarfin hasken rana na awa 1 gaba. Babban sabon abu ya ta'allaka ne a hanyarsa ta mataki biyu: na farko, faɗaɗa ainihin fasalin zuwa sararin samaniya mafi girma ta amfani da ƙididdigar Chebyshev da ayyukan trigonometric; na biyu, yin amfani da tsarin zaɓin fasalin da aka keɓance tare da ƙididdigar layi da aka takura don gina ƙirar hasashe na musamman bisa yanayi. Hanyar da aka gabatar tana da nufin ɗaukar rikice-rikice, alaƙar da ba ta layi ba tsakanin sauye-sauyen yanayi da ƙarfin fitarwa yadda ya kamata fiye da daidaitattun ƙira.

2. Hanyar Aiki

2.1 Bayanai & Fasalin Shigarwa

Ƙirar tana amfani da bayanan tarihi na lokaci-lokaci waɗanda suka haɗa da fitarwar tsarin PV da kuma abubuwan muhalli masu dacewa. Manyan fasalin shigarwa sun haɗa da:

Lokaci Mai Girma Kai: Samar da ƙarfin hasken rana daga tazarar mintuna 15 da suka gabata.
Yanayin Yanayi: Bayanan nau'i-nau'i (misali, sarari, girgije, ruwan sama).
Sauye-sauyen Yanayi: Zazzabi, ma'anar raɓa, ɗanɗano, da saurin iska.
Fasalin Lokaci: Ana la'akari da su a ɓoye ta hanyar yanayin bayanan lokaci-lokaci.

2.2 Gina Fasali tare da Ƙididdigar Chebyshev

Don ƙirar yuwuwar abubuwan da ba su layi ba, ana canza ainihin fasalin vector $\mathbf{x}$ zuwa sararin samaniya mafi girma. Ga kowane fasalin shigarwa mai ci gaba $x_i$, ana ƙirƙirar saitin ƙididdigar Chebyshev na nau'in farko $T_k(x_i)$ har zuwa takamaiman digiri $K$. Ƙididdigar Chebyshev na digiri $k$ an ayyana ta ta hanyar maimaitawa:

$T_0(x) = 1$

$T_1(x) = x$

$T_{k+1}(x) = 2xT_k(x) - T_{k-1}(x)$

Hakanan ana ƙara ayyukan trigonometric (sine da cosine) na fasalin don ɗaukar tsarin lokaci-lokaci. Wannan ginin yana haifar da sararin fasali mai wadata, mai bayyanawa $\Phi(\mathbf{x})$ wanda zai iya wakiltar rikice-rikicen alaƙar aiki.

2.3 Zaɓin Fasali & Ƙididdigar da aka Takura

Ba duk fasalin da aka gina ba ne suka dace. Ana amfani da hanyar zaɓin fasali mai tushen lulluɓe don gano mafi yawan hasashe na rukuni don yanayi daban-daban. Daga baya, ana daidaita ƙirar ƙididdigar layi da aka takura:

$\min_{\beta} \| \mathbf{y} - \Phi(\mathbf{X})\beta \|_2^2$

bisa ga ƙuntatawa akan ƙididdiga $\beta$ (misali, ƙuntatawa mara kyau idan alaƙar zahiri ta nuna cewa wasu abubuwan shigarwa yakamata su yi tasiri mai kyau kawai akan fitarwa). Wannan matakin yana tabbatar da ƙirar ƙira da fassarar zahiri yayin kiyaye daidaito.

3. Sakamakon Gwaji & Bincike

3.1 Ma'aunin Aiki

Babban ma'auni don kimantawa shine Matsakaicin Kuskuren Madaidaici (MSE) tsakanin hasashen da ainihin ƙarfin PV na awa 1 gaba. Ƙananan MSE yana nuna mafi girman daidaiton hasashe.

Taƙaitaccen Aiki

Hanyar da aka Gabatar: Ya sami mafi ƙarancin MSE a cikin yanayin gwaji.

Babban Fa'ida: Mafi kyawun aiki a ƙarƙashin yanayi daban-daban, musamman a lokutan canji (misali, girgije masu wucewa).

3.2 Kwatance da Ƙirar Tushe

An yi amfani da tsarin da aka gabatar don kwatanta da wasu ƙirar koyon inji na gargajiya:

Injin Tallafawa Vector (SVM) / Ƙididdigar Tallafawa Vector (SVR)
Gandun Daji Bazuwar (RF)
Bishiyar Yanke Shawara Mai Haɓakawa (GBDT)

Sakamako: Hanyar gina fasalin Chebyshev da zaɓi ta ci gaba da samar da MSE mafi ƙasa fiye da duk ƙirar tushe. Wannan yana nuna ingancin ƙera sararin fasali mai girma da aka keɓance don matsalar hasashen hasken rana, idan aka kwatanta da dogaro kawai akan ikon haɗin fasalin asali na hanyoyin bishiyar gungu ko dabarun kernel a cikin SVM.

4. Cikakkun Bayanai na Fasaha & Tsarin Lissafi

Ana iya taƙaita ƙirar azaman aiki $f$ mai taswira abubuwan shigarwa zuwa hasashen awa 1 gaba $\hat{P}_{t+1}$:

$\hat{P}_{t+1} = f(\mathbf{x}_t) = \beta_0 + \sum_{j \in S} \beta_j \phi_j(\mathbf{x}_t)$

inda:

$\mathbf{x}_t$ shine fasalin vector a lokacin $t$.
$\{\phi_j\}$ sune zaɓaɓɓun ayyukan tushe daga faɗaɗawar Chebyshev/trigonometric.
$S$ shine saitin fihirisa da aka zaɓa ta hanyar algorithm ɗin zaɓin fasali.
$\beta$ sune ƙididdiga da aka ƙididdige ta hanyar madaidaicin ƙididdiga da aka takura.

Ƙuntatawa $\beta_j \geq 0$ ga wasu $j$ ana iya haɗa su don nuna ilimin zahiri (misali, hasken rana yana da alaƙa mai kyau da ƙarfi).

5. Tsarin Bincike: Misali Ba na Lamba ba

Yi la'akari da sauƙaƙan yanayi don hasashen ƙarfi da tsakar rana a rana mai ɗan girgije. Ayyukan tsarin shine:

Shigarwa: Fasali a 11:45 AM: Ƙarfi=150 kW, Zazzabi=25°C, ɗanɗano=60%, Fihirisar Rufe Girgije=0.5 (ɗan girgije).
Gina Fasali: Ƙirƙiri sabbin fasali: $T_2(Zazzabi)=2*(25)^2 -1$, $sin(ɗanɗano)$, $Rufe Girgije * T_1(Zazzabi)$, da sauransu. Wannan na iya haifar da fasalin da aka samo sama da 20.
Zaɓin Fasali (don ƙirar "ɗan Girgije"): Hanyar lulluɓe ta gano cewa 5 kawai daga cikin waɗannan fasalin suna da mahimmanci don hasashe a ƙarƙashin waɗannan yanayi, misali, $Ƙarfi_{t-1}$, $T_2(Zazzabi)$, $Rufe Girgije$, $sin(ɗanɗano)$, da kalmar hulɗa.
Hasashe da aka Takura: Ƙirar ƙididdiga ta musamman ta "ɗan Girgije", ta amfani da fasalin da aka zaɓa 5 kawai da ƙididdigansu da aka koya a baya (tare da ƙuntatawa cewa ƙididdigar rufe girgije ba ta da kyau), ta lissafa hasashen: $\hat{P}_{12:00 PM} = 165 kW$.

6. Aikace-aikacen Gaba & Jagororin Bincike

Ƙirar Haɗin Physics-ML: Haɗa hanyar da aka gabatar ta tushen bayanai tare da ƙirar aikin PV na zahiri (kamar waɗanda ke cikin Tsarin Mai Ba da Shawara na NREL) zai iya haɓaka ƙarfi da ikon fadadawa.
Hasashe Mai Yuwuwa: Faɗaɗa tsarin don fitar da tazarar hasashe (misali, ta hanyar ƙididdiga akan zaɓaɓɓun fasalin) yana da mahimmanci don ayyukan cibiyar wutar lantarki masu sane da haɗari.
Ƙididdiga na Geɓe don PV da aka Rarraba: Tura nau'ikan nau'ikan zaɓin fasali da ƙirar ƙididdiga akan na'urori na geɓe a gonakin hasken rana guda ɗaya don hasashe na gida na ainihin lokaci.
Koyon Canja wuri a cikin Yanayi: Bincika yadda za a iya daidaita ko daidaita saitin fasalin da aka zaɓa don yanki ɗaya na yanki zuwa wani tare da yanayi daban-daban.
Haɗawa tare da Koyo Mai zurfi: Yin amfani da zaɓaɓɓun fasalin Chebyshev azaman abubuwan shigarwa masu bayarwa zuwa cibiyar jijiya mai maimaitawa (RNN) ko ƙirar mai canzawa don ɗaukar dogon lokaci na dogaro fiye da awa ɗaya.

7. Nassoshi

Yang, Y., Mao, J., Nguyen, R., Tohmeh, A., & Yeh, H. G. (Shekara). Gina Fasali da Zaɓi don Ƙirar Ƙarfin Hasken Rana na PV. Sunan Jarida/Taro.
Mellit, A., & Pavan, A. M. (2010). Hasashen hasken rana na awa 24 ta amfani da cibiyar jijiya ta wucin gadi: Aikace-aikace don hasashen aiki na shukar PV da aka haɗa da cibiyar wutar lantarki a Trieste, Italiya. Ƙarfin Rana, 84(5), 807-821.
Laboratorin Makamashi Mai Sabuntawa na Ƙasa (NREL). (2023). Hasashen Rana. https://www.nrel.gov/grid/solar-forecasting.html
Hastie, T., Tibshirani, R., & Friedman, J. (2009). Abubuwan Koyon Ƙididdiga. Springer. (Don tushen faɗaɗawar fasali da daidaitawa).
Isola, P., Zhu, J. Y., Zhou, T., & Efros, A. A. (2017). Fassarar hoto zuwa hoto tare da cibiyoyin adawa na sharadi. Gudanar da taron bitar hangen nesa na kwamfuta da tsarin tsarin (shafi na 1125-1134). (An ambata a matsayin misali na tsarin canzawa a wani yanki na ML, kama da hanyar gina fasali a nan).

8. Ra'ayin Manazarcin: Babban Fahimta & Zargi

Babban Fahimta: Ainihin gudunmawar wannan takarda ba wani ƙirar hasashen rana kawai ba ce; yana da tsari, mataki biyu ka'idar injiniyanci ta fasali wanda ke raba koyon wakilci daga daidaita ƙira. Ta hanyar gina sararin Chebyshev mai girma a fili, yana tilasta ƙirar ta yi la'akari da takamaiman sharuɗɗan da ba su layi ba da hulɗar da ƙirar akwatin baƙi kamar GBDT za su iya ci karo da su cikin rashin inganci ko gaba ɗaya. Matsi ne daga "fatan algorithm ɗin ya same shi" zuwa "ginin sararin da siginar ke rayuwa a ciki." Wannan yana tunawa da falsafar da ke bayan tsarukan nasara a wasu fagage, kamar ƙirar gine-gine na mai samarwa/mai nuna bambanci a cikin CycleGAN waɗanda ke tsara matsalar koyo don fassarar hoto mara biyu.

Kwararar Hankali: Hankali yana da inganci kuma yana da kyau: 1) Amince da rikice-rikicen ilimin lissafi na samar da hasken rana. 2) Kar a jefa bayanan danye kawai a ƙirar da ba ta layi ba; a maimakon haka, faɗaɗa sararin shigarwa da ayyukan tushe masu tabbacin lissafi (ƙididdigar Chebyshev suna da kyau don kusanta). 3) Yi amfani da hanyar lulluɓe don zaɓin fasali—hanyar da ke da tsada amma mai manufa—don datsa wannan sarari zuwa ƙaramin yanki na musamman na yanayi, mai iya fassara. 4) Aiwatar da ƙididdigar da aka takura don shigar da ilimin zahiri na farko (misali, "ƙarin girgije ba zai iya samar da ƙarin ƙarfi ba"). Wannan bututun yana da ƙa'ida fiye da yadda aka saba "binciken grid-sama da hyperparameters" da aka yi amfani da shi ga ƙirar ML da aka cire daga kantin sayar da kayayyaki.

Ƙarfi & Kurakurai:
Ƙarfi: Hanyar ta sami mafi girman MSE, yana tabbatar da ƙimarta ta zahiri. Ƙirar ta musamman ta yanayi tana da hankali. Amfani da ƙuntatawa yana ƙara matakin ƙarfi da fassarar sau da yawa ba su nan a cikin hanyoyin ML masu tsafta. Babban misali ne na "akwatin gilashi" ML don tsarin injiniyanci.
Kurakurai: Farashin lissafi na zaɓin fasali mai tushen lulluɓe ga kowane nau'in yanayi babban cikas ne don daidaitawa na ainihin lokaci ko turawa mai girma. Takardar ba ta da tattaunawa akan kwanciyar hankali na zaɓaɓɓun saitin fasalin—shin suna canzawa da ƙaramin bambancin bayanan horo? Bugu da ƙari, yayin da ya doke SVR, RF, da GBDT yana da kyau, kwatance da ƙirar koyo mai zurfi da aka daidaita (misali, LSTM ko Transformer Fusion na Lokaci) ko ingantaccen aiwatar da haɓakar gradient kamar XGBoost tare da ikon hulɗar fasalin kansa shine wani gibi a cikin bincike na 2023+.

Fahimta Mai Aiki: Ga masu aikin masana'antu, wannan takarda ta zama tsarin gini don ƙirar ƙirar hasashe mai dogaro, na musamman na shafi. Abin da za a iya ɗauka nan take shine saka hannun jari a cikin kayan aikin injiniyanci na fasali kafin tsalle zuwa hadaddun algorithms. Fara da aiwatar da wannan bututun faɗaɗawar Chebyshev akan bayanan tarihinku. Duk da haka, don tsarin aiki, maye gurbin hanyar lulluɓe tare da hanyar tacewa mafi ma'auni (kamar bayanan juna) ko hanyar da aka haɗa (kamar ƙididdigar LASSO) don zaɓin fasali don rage nauyin lissafi. Haɗa kai tare da ƙwararrun yanki don ayyana mafi mahimmancin ƙuntatawa na zahiri don ƙididdiga. Wannan hanyar haɗin gwiwa, mai tunani zai iya haifar da mafi kyawun dawowa fiye da kawai hayar babban shagon gajimare don horar da babbar cibiyar jijiya.