Gina Siffofi da Zaɓe 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 hayaki mai gurbata yanayi da haɓaka dorewa. Duk da haka, rashin kwanciyar hankali da sauye-sauyen ƙarfin hasken rana na haifar da ƙalubale masu mahimmanci ga kwanciyar hankalin cibiyar wutar lantarki da samar da makamashi mai dogaro. Saboda haka, daidaitaccen hasashen samar da ƙarfin hasken rana na ɗan gajeren lokaci yana da mahimmanci don ingantaccen sarrafa makamashi, daidaita nauyin lantarki, da tsara aiki.

Wannan takarda ta gabatar da sabon tsarin koyon inji don hasashen ƙarfin hasken rana na sa'a 1 gaba. Babban sabon abu ya ta'allaka ne a tsarinsa na ƙirar siffofi. Maimakon dogaro kawai akan bayanan tarihi da sauye-sauyen yanayi, hanyar tana gina sararin siffofi mai girma ta amfani da ƙididdiga na Chebyshev da ayyukan trigonometric. Sai kuma a yi amfani da tsarin zaɓen siffofi tare da ƙayyadaddun lissafi na layi don gina ingantaccen samfurin hasashe mai fassara wanda aka keɓance don nau'ikan yanayi daban-daban.

2. Hanyar Aiki

2.1 Bayanai da Siffofin Shigarwa

Samfurin yana amfani da haɗin shigarwar lokaci, yanayi, da na kai-da-kai:

• Sauye-sauyen Yanayi: Haske, zafin jiki, ma'aunin raɓa, ɗanɗano, saurin iska.
• Rarraba Nau'in Yanayi: Ana rarraba abubuwan shigarwa bisa yanayin yanayi da ke faruwa (misali, sarari, girgije, ruwan sama).
• Kalmar Kai-da-Kai: Samar da ƙarfin hasken rana daga lokacin da ya gabata (misali, mintuna 15 da suka gabata) an haɗa shi don ɗaukar dogaro na lokaci.

2.2 Gina Siffofi tare da Ƙididdiga na Chebyshev

Ana canza ainihin siffofin shigarwa zuwa sarari mai wadata, mafi girma. Don wani sauyi na shigarwa $x$, ana amfani da ƙididdiga na Chebyshev na nau'in farko, $T_n(x)$. Ana bayyana waɗannan ƙididdiga ta hanyar dangantakar maimaitawa:

$T_0(x) = 1$

$T_1(x) = x$

$T_{n+1}(x) = 2xT_n(x) - T_{n-1}(x)$

Ana gina siffofi kamar $T_n(x)$ don $n$ har zuwa wani tsari da aka ƙayyade, kuma yana iya haɗawa da sharuɗɗan haɗin kai (misali, $T_i(x) \cdot T_j(y)$) da ayyukan trigonometric (misali, $\sin(\omega t)$, $\cos(\omega t)$) don ɗaukar alamu na lokaci-lokaci.

2.3 Tsarin Zaɓen Siffofi

Ana amfani da hanyar nade don zaɓar mafi dacewar siffofi daga cikin faɗaɗaɗɗen saiti. Ana yin wannan tsari daban-daban ga kowane nau'in yanayi don yin la'akari da bambancin tasirin abubuwa a ƙarƙashin yanayi daban-daban. Zaɓen yana nufin daidaita rikitaccen samfuri da ƙarfin hasashe, guje wa yin wuce gona da iri.

2.4 Ƙayyadaddun Samfurin Lissafi na Layi

Bayan zaɓen siffofi, ana gina samfurin lissafi na layi: $\hat{y} = \mathbf{w}^T \mathbf{x} + b$, inda $\mathbf{x}$ shine juzu'in zaɓaɓɓun siffofi. Don haɓaka dacewar zahiri da kwanciyar hankali, an tsara lissafin a matsayin matsala mafi ƙanƙanta murabba'ai masu ƙayyadaddu. Ƙayyadaddun na iya haɗawa da rashin tabbataccen ƙima akan wasu ma'auni (misali, haske ya kamata ya sami tasiri mara kyau ga fitar da wutar lantarki) ko iyakoki akan girman ma'auni.

3. Sakamakon Gwaji & Aiki

3.1 Saitin Gwaji

An gwada tsarin da aka gabatar akan bayanan tsohon shukar PV. An raba bayanan zuwa saitin horo da gwaji, tare da kimanta aiki ta amfani da Matsakaicin Kuskuren Murabba'ai (MSE) da yuwuwar wasu ma'auni kamar Matsakaicin Kuskure na Cikakke (MAE).

3.2 Kwatance da Samfuran Tushe

Takardar ta kwatanta hanyarta da wasu ƙa'idodin koyon inji da aka kafa:

• Na'urar Tallafawa Vector (SVM)/Lissafin Tallafawa Vector (SVR)
• Daji bazuwar (RF)
• Bishiyar yanke shawara mai Haɓakawa (GBDT)

Babban Bincike: Samfurin lissafin da aka gabata na tushen ƙididdiga na Chebyshev tare da zaɓen siffofi ya sami MSE mafi ƙasa fiye da duk hanyoyin gargajiya da aka kwatanta.

3.3 Aiki A Ko'ina cikin Yanayin Yanayi

Hanyar ƙirar samfurin ta musamman ga nau'in yanayi ta nuna daidaitawar da ta fi dacewa. Misali, a ƙarƙashin yanayi masu sauye-sauye na girgije, zaɓaɓɓun siffofin samfurin (watakila sharuɗɗan ƙididdiga mafi girma waɗanda ke ɗaukar tasirin haske mara layi) za su bambanta da waɗanda aka zaɓa don yanayin sararin sama mai kwanciyar hankali, wanda zai haifar da ƙarin daidaitattun hasashe a ko'ina.

4. Cikakkun Bayanai na Fasaha & Tsarin Lissafi

Ana iya taƙaita babbar matsalar ingantawa kamar haka:

1. Faɗaɗawar Siffofi: Ƙirƙiri juzu'in siffofi mai faɗaɗawa $\mathbf{\Phi}(\mathbf{z}) = [T_0(z_1), T_1(z_1), ..., T_n(z_m), \text{ sharuɗɗan haɗin kai}, \text{ sharuɗɗan trig}]$ daga ainihin juzu'in shigarwa $\mathbf{z}$.
2. Zaɓen Siffofi: Nemo wani yanki $\mathbf{x} \subset \mathbf{\Phi}(\mathbf{z})$ wanda ke rage kuskuren hasashe ga takamaiman nau'in yanayi $k$.
3. Ƙayyadaddun Lissafi: Warware ma'auni $\mathbf{w}$:
   $\min_{\mathbf{w}} ||\mathbf{y} - \mathbf{X}\mathbf{w}||^2_2$
   bisa la'akari da: $\mathbf{A}\mathbf{w} \leq \mathbf{b}$ (ƙayyadaddun rashin daidaito na layi, misali, $w_i \geq 0$).

5. Tsarin Nazari: Misali Ba na Lamba ba

Yi la'akari da wani sauƙaƙan yanayi don hasashen wutar lantarki da tsakar rana a rana mai ɗan girgije. Abubuwan shigarwa na danye sune: Haske ($I=600 W/m^2$), Zafin jiki ($T=25^\circ C$), da ƙarfin da ya gabata ($P_{t-1}=300 kW$).

1. Gina Siffofi: Don haske $I$, samar da sharuɗɗan Chebyshev har zuwa oda 2: $T_0(I)=1$, $T_1(I)=600$, $T_2(I)=2*600*600 - 1 = 719,999$. Ana yin irin wannan faɗaɗawa ga $T$ da $P_{t-1}$. Hakanan ana ƙirƙirar sharuɗɗan haɗin kai kamar $T_1(I)*T_1(T)$.
2. Zaɓen Siffofi (don samfurin "ɗan Girgije"): Algorithm ɗin zaɓi na iya riƙe $T_1(I)$ (haske na layi), $T_2(I)$ (ɗaukar tasirin jikewa mara layi), $T_1(T)$, da $P_{t-1}$, yayin da yake watsi da yawancin sauran siffofin da aka gina a matsayin marasa amfani ga wannan nau'in yanayi.
3. Hasashe: Hasashen ƙarshe shine haɗin layi: $\hat{P} = w_1*600 + w_2*719,999 + w_3*25 + w_4*300 + b$, inda $w_1, w_2 \geq 0$ saboda ƙayyadaddu.

6. Babban Fahimta & Ra'ayin Manazarcin

Babban Fahimta: Ainihin nasarar wannan takarda ba sabon algorithm baƙar fata ba ne, amma ingantaccen, tsarin aikin ƙirar siffofi mai sanin kimiyyar lissafi. Yana gane cewa dangantakar tsakanin yanayi da fitar da PV ba kawai layi ba ne ko kuma ana iya ɗauka cikin sauƙi ta hanyar bishiyoyin yanke shawara na yau da kullun. Ta hanyar gina sarari na tushe (ƙididdiga na Chebyshev) da aka sani da kyawawan kaddarorin kiyasin aiki sannan kuma a yi amfani da zaɓi mai haifar da yalwa, hanyar tana gina samfuran da za a iya fassara su, masu inganci waɗanda aka keɓance don takamaiman tsarin aiki (nau'ikan yanayi). Wannan shine amfani da ML mai hikima fiye da aikace-aikacen ƙarfi na koyon zurfi, musamman a cikin saitunan masana'antu masu iyakacin bayanai.

Kwararar Ma'ana: Ma'ana tana da inganci: 1) Amince da rikitaccen matsalar (mara layi, dogaro da yanayi). 2) Faɗaɗa sararin shigarwa bisa tsari don wakiltar yuwuwar rikitattun alaƙa. 3) Yanke baya da ƙarfi tare da zaɓi mai sanin yanki (nau'in yanayi) don guje wa wuce gona da iri. 4) Aiwatar da sauƙaƙan, ƙayyadaddun samfuran layi akan ingantattun siffofi don kwanciyar hankali da fahimta. Wannan bututun yayi daidai da mafi kyawun ayyuka a cikin ML na zamani, mai kama da falsafar da ke bayan faɗaɗawar tushe a cikin ƙirar ƙari na gabaɗaya ko koyon siffofi a cikin yankuna masu tsari.

Ƙarfi & Kurakurai:
Ƙarfi: Hanyar tana da fassara—kuna iya ganin waɗanne sharuɗɗan ƙididdiga suke da mahimmanci ga wane yanayi. Yana da haske a lissafi fiye da horar da manyan ƙungiyoyi ko hanyoyin jijiyoyi ga kowane nau'in yanayi. Ƙayyadaddun sun tilasta gaskiyar zahiri, wani mataki da yawa ya ɓace a cikin samfuran da aka ƙirƙira ta hanyar bayanai kawai. Yin fiye da RF da GBDT akan bayanansa yana da ƙarfi, saboda waɗannan ƙa'idodi ne masu ƙarfi.
Kurakurai: Babban iyakancewa shine dogaro da daidaitaccen nau'in yanayi na ainihin lokaci, wanda shi kansa matsala ce ta hasashe. Hanyar na iya fuskantar wahala tare da sauye-sauyen yanayi cikin sauri ko gauraye waɗanda ba a kama su da kyau a cikin rukunin horo. Bugu da ƙari, duk da yake ya fi kyau fiye da ma'auni a nan, iyakar aikin ƙarshe na samfurin layi akan zaɓaɓɓun siffofi na iya zama ƙasa fiye da cikakken daidaitaccen, rikitaccen samfuri don manyan bayanai, kamar yadda aka gani a cikin yankuna kamar hangen nesa inda samfura kamar CycleGAN (Zhu et al., 2017) suka bunƙasa akan bayanan pixel ɗin danye ba tare da gina siffofi da hannu ba.

Fahimta Mai Aiki: Ga masu aikin masana'antu, abin da za a ɗauka a bayyane yake: Saka hannun jari a cikin aikin ƙirar siffofi kafin rikitaccen samfuri. Kafin tura hanyar jijiyoyi, gwada faɗaɗa tsarin abubuwan shigar ku tare da ƙididdiga masu kusurwa biyu ko sharuɗɗan Fourier. Aiwatar da samfuran na musamman na yanayi ko tsarin. Koyaushe yi la'akari da ƙara sauƙaƙan ƙayyadaddu don daidaita samfura tare da ilimin yanki. Ga masu bincike, mataki na gaba shine haɗa wannan hanyar: yi amfani da gina siffofi/zaɓe ta atomatik a matsayin mai sarrafa shigarwa zuwa ƙarin ci-gaban samfura (misali, zaɓaɓɓun siffofin sun zama abubuwan shigarwa ga hanyar jijiyoyi mai maimaitawa don ƙirar jeri), ko haɗa matakin rarraba yanayi kai tsaye cikin tsarin koyo na ƙarshe-zuwa-ƙarshe.

7. Aikace-aikacen Gaba & Jagororin Bincike

• Haɗawa da Koyon Zurfi: Za a iya haɗa Layer ɗin gina siffofi a matsayin Layer na al'ada a cikin hanyar jijiyoyi, yana ba da damar samfurin ya koyi mafi kyawun haɗin ayyukan tushe.
• Hasashen Yiwuwa: Faɗaɗa tsarin ƙayyadaddun lissafi don samar da tazara na hasashe, mahimmanci ga sarrafa cibiyar wutar lantarki mai sanin haɗari. Ana iya bincika dabarun kamar lissafin Tsarin Gaussian tare da maɓuɓɓuka na al'ada waɗanda aka yi wahayi ta hanyar ƙididdiga na Chebyshev.
• Canja Koyo A Tsakanin Wurare: Bincika ko tsarin zaɓen siffofi (waɗanne ƙididdiga suke da mahimmanci ga yanayin "girgije") ana iya canjawa tsakanin wurare daban-daban na yanayin yanayi, yana rage buƙatun bayanai don sabbin shigarwar PV.
• Zaɓi na Daidaitawa na Ainihin Lokaci: Haɗa sigogin koyo na kan layi na algorithm wanda zai iya daidaita saitin siffofi a hankali yayin da yanayin yanayi ke canzawa, ya wuce buckets na nau'in yanayi na tsaye.
• Aikace-aikacen Makamashi Mai Faɗi: Aiwatar da falsafar gina siffofi/zaɓe ɗaya zuwa wasu hasashen sabuntawar da ke katsewa, kamar ƙarfin iska, ko kuma ga matsalolin da ke da alaƙa kamar hasashen nauyin makamashin gini.

8. Nassoshi

1. Yang, Y., Mao, J., Nguyen, R., Tohmeh, A., & Yeh, H. (Shekara). Gina Siffofi da Zaɓe don Ƙirar Ƙarfin Hasken Rana na PV. Sunan Jarida/Taron.
2. Zhu, J., Park, T., Isola, P., & Efros, A. A. (2017). Fassarar Hotuna-zuwa-Hotuna mara Haɗin gwiwa ta amfani da Cibiyoyin Adawa na Da'ira-Mai daidaituwa. Gudanar da Taron Ƙasa da Ƙasa na IEEE akan Hangen Nesa na Kwamfuta (ICCV).
3. Hukumar Makamashi ta Duniya (IEA). (2023). Sabuntawa 2023: Nazari da hasashe zuwa 2028. Littattafan IEA. [Tushen waje akan ci gaban makamashi mai sabuntawa]
4. Mason, K., & Ghanem, R. (2021). Koyon Ƙididdiga don Hasashen Makamashi Mai Sabuntawa. Wiley.
5. Laboratorin Makamashi Mai Sabuntawa na Ƙasa (NREL). (b.t.k.). Hasashen Rana. An samo daga https://www.nrel.gov/grid/solar-forecasting.html [Mai inganci tushen waje akan binciken hasashen rana]