論文年份：2019
論文原文：https://eprints.qut.edu.au/128957/
DOI：10.20944/preprints201903.0131.v1
文章目錄
- Energy consumption prediction using machine learning; a review
- Abstract
- 1. Introduction
- 2. Research Methodology
- 3. Machine learning models
- ANFIS（Adaptive neuro fuzzy inference system）
- ANNs
- SVM and SVR
- WNNs（Wavelet neural networks）
- DTs
- ELMs（Extreme learning machine）
- MLPs
- ENSEMBLEs
- HYBRIDs
- DEEP LEARNINGs
- 4. Discussion
Energy consumption prediction using machine learning; a review
Abstract
| orig | trans |
|---|---|
| Machine learning (ML) methods has recently contributed very well in the advancement of the prediction models used for energy consumption. Such models highly improve the accuracy, robustness, and precision and the generalization ability of the conventional time series forecasting tools. This article reviews the state of the art of machine learning models used in the general application of energy consumption. Through a novel search and taxonomy the most relevant literature in the field are classified according to the ML modeling technique, energy type, perdition type, and the application area. A comprehensive review of the literature identifies the major ML methods, their application and a discussion on the evaluation of their effectiveness in energy consumption prediction. This paper further makes a conclusion on the trend and the effectiveness of the ML models. As the result, this research reports an outstanding rise in the accuracy and an ever increasing performance of the prediction technologies using the novel hybrid and ensemble prediction models. | 機器學習（ML）方法最近在用於能源消耗的預測模型的發展中做出了很大的貢獻。這樣的模型極大地提高了常規時間序列預測工具的準確性，魯棒性和精度以及泛化能力。本文回顧了能源消耗的一般應用中使用的機器學習模型的最新狀態。通過新穎的搜索和分類，根據ML建模技術，能量類型，消亡類型和應用領域對本領域中最相關的文獻進行了分類。對文獻的全面回顧指出了主要的機器學習方法，它們的應用以及對它們在能耗預測中的有效性評估的討論。本文還對ML模型的趨勢和有效性進行了總結。結果，這項研究報告了使用新型混合和集成預測模型的預測技術的準確性和不斷提高的性能。 |
| Keywords: energy consumption; prediction; machine learning models; deep learning models; artificial intelligence (AI); computational intelligence (CI); forecasting; soft computing (SC); | 關鍵詞：能耗預測；機器學習模型；深度學習模型；人工智能（AI）；計算智能（CI）；預測；軟計算（SC）； |
1. Introduction
| Energy consumption is among one of the essential topics of energy systems. Energy consumption came under the consideration after the energy crisis in 1970s[1]. Also, It is shown that energy consumption throughout the world is rapidly increasing [2]. Therefore, each country tries to use as less energy as possible in their country in different areas from building to farms, from industrial process to vehicles [3]. As energy comes from three different resources like fossil fuels, renewable and nuclear resources [4], it need so much effort to keep tracking of energy consumption of these types in different area. However by doing so, we can predict the amount of energy, which is consumed in different areas and try to make plans, specialized for a specific usage and area. | 能源消耗是能源系統的基本主題之一。在1970年代的能源危機之後，能源消耗才被考慮在內[1]。同樣，它表明，全世界的能源消耗都在迅速增加[2]。因此，每個國家在從建築到農場，從工業過程到車輛的不同地區，都試圖在本國使用盡可能少的能源[3]。由於能源來自化石燃料，可再生能源和核能等三種不同的資源[4]，因此需要付出巨大的努力來跟蹤不同地區這些類型的能源消耗。但是，通過這樣做，我們可以預測在不同區域中消耗的能量，並嘗試制定針對特定用途和區域的計劃。 |
| For all energy types mentioned above, estimating the usage is useful for decision and policy makers. By knowing how much energy will be used for their process or work, they can be able to think of some changes in them to reduce the amount energy usage. Predicting future energy usage both in Short-term and Long-term manner will help us even to know, that in which type energy is used mostly and try to change the trend, as it is happed in the recent years for fossil fuels and now we have renewable energy. The amount of energy used in different areas is influenced by diffent factors such as water, wind, temperature. Having multiple factors, predicting the energy consumption is a complex problem[5]. | 對於上述所有能源類型，估算使用量對於決策者和決策者都是有用的。通過知道將在他們的過程或工作中使用多少能量，他們可以想到其中的一些變化以減少能量使用量。預測短期和長期的未來能源使用量將幫助我們甚至知道，哪種類型的能源最常使用，並試圖改變趨勢，因爲近年來化石燃料受到限制，現在我們有可再生能源。不同地區使用的能源量受水，風，溫度等不同因素的影響。有多個因素，預測能耗是一個複雜的問題[5]。 |
| Nowadays ML models are being used in different areas because they are useful and the way ML works is like a function which best maps the input data to output. Machine learning models can produce prediction for enery consumption with high accuracy. So they can be used by governments to implement enery-saving policies. For instance, ML models can predict the amount enery used in building [6]. They can also be used to predict the future use of different types of energy like electricity or natural gas[6]. | 如今，ML模型已在不同領域中使用，因爲它們非常有用，並且ML的工作方式就像一個函數，可以最好地將輸入數據映射到輸出。機器學習模型可以準確預測黑膠消耗量。因此，政府可以使用它們來實施節省能源的政策。例如，機器學習模型可以預測建築物中使用的烯醇量[6]。它們還可以用於預測未來將使用不同類型的能源，例如電或天然氣[6]。 |
| This research work has been conducted in the prediction of different enery type usage. Predictions can be done on the usage of energy in a specific procedure in industry [7] or the total amout of energu used in a coutry [8]. This study tries to review the recent studies related to modeling and estimating of energy consumption in different area. | 這項研究工作是在預測不同類型的使用情況下進行的。可以對工業中特定過程中的能源使用情況進行預測[7]，或者對國家中使用的能源總量進行預測[8]。本研究試圖回顧與不同地區能源消耗的建模和估算有關的最新研究。 |
| The organization of this paper is in a way to review different ML models for energy prediction like: ANFIS, ANN, DT, ELM, MLP, SVM/SVR, WNN, ENSEMBLE, HYBRID, and DEEP LEARNING. In each section related to each model we try to review the latest studies which uses ML models for forecasting energy usage in different applications. | 本文的組織方式是回顧用於能量預測的不同ML模型，例如：ANFIS，ANN，DT，ELM，MLP，SVM / SVR，WNN，ENSEMBLE，HYBRID和DEEP LEARNING。在與每個模型有關的每個部分中，我們嘗試回顧使用ML模型預測不同應用中的能源使用情況的最新研究。 |
2. Research Methodology
The database is created using the following search keywords to identify the manuscripts on energy consumption prediction using machine learning models. ISI and Scopus databases had been explored using the essential search keywords, i.e., ( TITLE-ABS-KEY ( “energy consumption” ) AND TITLE-ABS-KEY ( “machine learning” OR “Deep learning” OR “ANN” OR “MLP” OR “ELM” OR “neural network” OR “ANFIS” OR “decision tree” OR wnn ) ).
使用以下搜索關鍵字創建數據庫，以使用機器學習模型識別有關能耗預測的手稿。使用必要的搜索關鍵字對ISI和Scopus數據庫進行了探索，即（TITLE-ABS-KEY（“能源消耗”）和TITLE-ABS-KEY（“機器學習”或“深度學習”或“ ANN”或“ MLP “或” ELM“或”神經網絡“或” ANFIS“或”決策樹“或wnn））。
Study the database shoes a dramatic increase in the number of papers from 2006 to 2018. The database include 4300 papers. The most relevant and original works have been revised in this stateof-the-art.
從2006年到2018年，研究數據庫的論文數量急劇增加。該數據庫包含4300篇論文。在此最新技術中，最相關和最原始的作品已經過修訂。
3. Machine learning models
Here comes the taxonomy chart and one paragraph explanation. The ML models include: ANFIS, ANN, DT, ELM, MLP, SVM/SVR, WNN, ENSEMBLE, HYBRID, and DEEP LEARNING.
ANFIS（Adaptive neuro fuzzy inference system）
ANNs
SVM and SVR
WNNs（Wavelet neural networks）
DTs
ELMs（Extreme learning machine）
MLPs
ENSEMBLEs
HYBRIDs
DEEP LEARNINGs
4. Discussion
The use of ML models have been increased during the last decade. Along with the conventional ML methods, e.g., ANNs and MLP, DTs, the application of hybrids and Ensemble methods has been dramatically increased. Through hybrid methods the researcher aim at higher accuracy and efficiency. The future direction of the research is to develop hybrid models with higher accuracy and higher speeds.
在過去十年中，機器學習模型的使用有所增加。與傳統的ML方法（例如ANN和MLP，DT）一起，混合和集成方法的應用已大大增加。通過混合方法，研究人員致力於更高的準確性和效率。研究的未來方向是開發具有更高準確性和更高速度的混合模型。