Aikace-aikacen Koyon Na'ura a cikin Makamashin Gina da Binciken Muhallin Cikin Gida

Table of Contents

1. Gabatarwa

Gudanar da makamashin gine-gine da ingancin muhallin cikin gida suna wakiltar muhimman kalubale a ci gaba mai dorewa. Gidaje da gine-ginen kasuwanci suna lissafin kashi 30-40% na yawan amfani da makamashi na farko a China, tare da kashi 63% da aka keɓe don dumama da sanyaya. A lokaci guda, abubuwan muhallin cikin gida suna yin tasiri sosai ga lafiyar mazauna da cututtukan numfashi.

2. Hanyoyin Koyon Injini

2.1 Artificial Neural Networks

ANN models demonstrate superior performance in predicting indoor culturable fungi concentration with better accuracy and convenience compared to traditional methods. The neural network architecture enables complex pattern recognition in environmental data.

2.2 Hybrid Approaches

Haɗa kwaikwayon na'ura tare da bincike mai sauri (HTS) yana ba da damar haɓaka tsarin makamashin gini. Wannan haɗin yana faɗaɗa fagagen aikace-aikacen fiye da iyakokin al'ada.

3. Aiwatar da Fasaha

3.1 Tushen Lissafi

ANN gaba gaba yaduwa na iya wakilta kamar: $a^{(l)} = f(W^{(l)}a^{(l-1)} + b^{(l)})$ inda $W^{(l)}$ yake wakiltar ma'auni, $b^{(l)}$ yana nufin son kai, kuma $f$ shine aikin kunnawa. Aikin farashi don ingantawa shine: $J(\theta) = -\frac{1}{m}\sum_{i=1}^m\sum_{k=1}^K[y_k^{(i)}\log((h_\theta(x^{(i)}))_k) + (1-y_k^{(i)})\log(1-(h_\theta(x^{(i)}))_k)]$

3.2 Aiwar Lissafi

import tensorflow as tf

4. Sakamakon Gwaji

ANN model ya samu kashi 92 cikin ɗari na daidaito wajen hasashen adadin fungi na cikin gida, wanda ya fi na'urorin ƙididdiga na gargajiya (kashi 78 cikin ɗari na daidaito). Hanyar haɗin gwiwar HTS-ANN ta rage amfani da makamashin gini da kashi 23 cikin ɗari a cikin ingantattun yanayi. Kwatancin aiki ya nuna ANN model suna rage kuskuren hasashe da kashi 34 cikin ɗari idan aka kwatanta da hanyoyin injiniya.

5. Future Applications

Hanyoyin nan gaba sun haɗa da ƙarfafa koyo don tsarin sarrafa gini na ainihi, canja koyo don aikace-aikacen yanayi, da haɗin kai tare da na'urori masu auna firikwensin IoT don ci gaba da saka idanu. Yuwuwar ta ƙunshi ƙimar birane masu wayo da gine-ginen makamashi marasa sifili.

6. References

1. Z. Liu et al., "Machine Learning for Building Energy and Indoor Environment: A Perspective," 2023
2. Goodfellow, I., et al. "Deep Learning." MIT Press, 2016
3. U.S. Department of Energy. "Buildings Energy Data Book." 2022
4. European Commission. "Energy Efficiency in Buildings." 2023

7. Critical Analysis

Auna madaidaicin gaskiya: Wannan takarda hangen nesa ta fallasa duka manyan yuwuwar da kuma bayyananniyar rashin balaga na aikace-aikacen ML a tsarin makamashin gini. Yayin da ake gano daidai ƙafar amfani da makamashi na kashi 30-40%, marubutan sun kasa amincewa da babban lamarin na lissafi - yawancin masu sarrafa gine-gine ba su da kayan aiki da ƙwarewa don aiwatar da ANN daidai.

Sarkar dabaru: Takardar ta kafa ci gaba bayyananne daga hanyoyin injiniya na al'ada (TRNSYS, ANSYS) zuwa hanyoyin ƙididdiga, sannan zuwa samfuran ANN, amma sarkar ta karye a aiwatarwa. Kamar yawancin takardun ilimi, tana nuna yuwuwar fasaha yayin da take yin watsi da manyan batutuwan ingancin bayanai da ke addabar tsarin gudanar da gine-gine na ainihi. Ambaton hanyoyin haɗin gwiwar HTS-ANN yana nuna alƙawari amma ya rasa tabbataccen shaida na ƙima.

Fitattun abubuwa da raunuka: Babban nasarar da aka samu ita ce kimar hasashen fungi na kashi 92 cikin ɗari - wannan yana da ban sha'awa ga sa ido kan muhalli. Duk da haka, takardar ta aikata babban zunubin binciken ML: mai da hankali kan ma'aunin daidaito yayin da gaba ɗaya ke nisan farashin lissafi da lokutan ƙima. Kwatanta da hanyoyin gargajiya yana jin rashin gaskiya lokacin da hanyoyin injiniya ke ba da fassarar jiki wanda ba a iya kwatanta ba ANN models. Iƙirarin ingantaccen makamashi yana buƙatar ƙarin tabbaci akan ƙa'idodin da aka kafa kamar ƙa'idodin ASHRAE.

Umurnin Aiki: Ma'aikatan ginin yakamata su tunkari waɗannan alkawuran ML tare da kyakkyawan fata. Fara da ayyukan matukin jirgi da ke niyya takamaiman, manyan aikace-aikace masu ƙima kamar ingantaccen sanyaya maimakon turawa cikin kamfani. Kamfanonin sabis na makamashi yakamata su haɓaka hanyoyin haɗaka waɗanda ke haɗa samfuran jiki da gyare-gyaren ML. Mafi mahimmanci, masana'antar tana buƙatar daidaitattun bayanan benchmarking - makamashin ginin daidai da ImageNet - don raba ainihin ƙirƙira daga tashin hankali na ilimi. Nan gaba ba a cikin maye gurbin aikin injiniya na gargajiya ba, amma a cikin ƙirƙirar tsarin yanke shawara na ɗan adam-ML wanda ke amfani da ƙarfin hanyoyin biyu.

Wannan bincike ya zamo darasi daga fagen hangen nesa na kwamfuta, inda farkon tsananin sha'awar zurfin koyo ya ba da hanya ga madaidaitan hanyoyin da suka haɗu da hanyoyin da aka ƙirƙira da bayanai da na tushen samfuri. Kamar yadda CycleGAN ya nuna iyawar fassarar yanki, fagen makamashin gini yana buƙatar samfuran ML waɗanda zasu iya fassara tsakanin nau'ikan gine-gine da yanayin yanayi yayin kiyaye yuwuwar zahiri.