Ann based short term load forecasting for rolling horizon operation of microgrids

Doctoral Dissertation ANN-based Short-term Load Forecasting for Rolling horizon Operation of Microgrids Department of Electrical Engineering Graduate School, Chonnam National University NGO MINH DUC February 2019 Tai ngay!!! Ban co the xoa dong chu nay!!! CONTENTS Pages CONTENTS………………………………………………………………………….i LIST OF FIGURES……………………………………………………………… iv LIST OF TABLES………………………………………………………………….vi NOMENCLARTURE…………………………………………………………….viii ABSTRACT………………………………………………………………………….1 INTRODUCTION……………… BACKGROUND AND LITERATURE REVIEW……………………… 2.1 Introduction to microgrid system………………………………………………4 2.1.1 Each part of microgrid energy management system……………………………5 2.1.2 Application of microgrid in building sector…………………………………….6 2.2 Load forecasting method overview 2.2.1 Long-term load forecasting…………………………………………………… 2.2.2 Medium-term load forecasting………………………………………………….7 2.2.3 Short-term load forecasting…………………………………………………….8 2.3 Literature review of short-term load forecasting system………………… 2.3.1 Conventional approaches…………………………………………………… 10 2.3.2 Machine learning based approaches………………………………………… 13 DATA PREPARERATION AND ANALYSIS……………………………… 16 3.1 Data preparation…………………………………………………………………17 3.2 Characteristic of load profile………………………………………………… 24 i 3.3 Correlation between outdoor factors……………………………………………29 3.4 Conclusion………………………………………………………………………35 DATA FEATURES AND GENERATION…………………………………….37 4.1 Time features……………………………………………………………………37 4.2 Temperature features ……………………………………………………………38 4.3 Previous load features……………………………………………………………39 4.4 Conclusion……………………………………………………………………….39 DESIGN AND IMPLEMENTATION STRATEGY OF LOAD FORECASTING SYSTEM FOR MICROGRID ESS SCHEDULING 40 5.1 Clustering overview…………………………………………………………… 41 5.2 Sparse clustering algorithm…………………………………………………… 44 5.3 Network Architectures………………………………………………………… 46 5.4 Design and implementation strategy of load forecasting system for microgrid ESS scheduling scheme…………………………………………………………… ……49 5.5 Conclusion………………………………………………………………………55 EXPERIMENTAL RESULTS AND DISCUSSION………………………… 56 6.1 Experimental setup………………………………………………………………56 6.2 Evaluation metrics……………………………………………………………….58 6.3 Comparison with other prediction methods…………………………………… 60 6.4 Impact of temperature……………………………………………………………63 6.5 Impact of binary features……………………………………………………… 63 6.6 The impact of different training methods……………………………………… 64 6.7 Evaluate performance of implementation strategies for microgrid ESS scheduling………………………………………………….……………………… 66 ii 6.8 Analysis of forecasting error for rolling horizon operation approach………… 69 6.9 Length evaluation of training data………………………………………………71 6.10 Processing time…………………………………………………………………73 CONCLUSION………………………………………………………………….74 REFERENCES………………………………………………….………………….75 ABSTRACT (in Korean)……………………………………….………………….82 ACKNOWLEDGEMENTS……… ………………………….………………….85 iii LIST OF FIGURES Figure Content Page Fig Microgrid platform Fig Illustration of typical microgrid energy management system platform………………………………………………………… Fig Data preparation process……………………………………… 17 Fig Illustration of impute missing data based on different impute approach………………………………………………………… 19 Fig Identification of outliner position based on Gradient………… 20 Fig Flowchart of proposed preprocessing data 22 Fig Proposed method for removing outliner and missing data…… 23 Fig Daily energy consumption variation for two years…………… 25 Fig Distribution of energy consumption for each month…………… 26 Fig 10 Distribution of energy use for day of the week………………… 27 Fig 11 Characteristic of weekly load pattern………………………… 27 Fig 12 Distribution of hourly energy consumption…………………… 28 Fig 13 Daily average temperature curve……………………………… 30 Fig 14 Correlation between outdoor temperature and energy consumption for a university campus building………………… Fig 15 Scatter plot between outdoor temperature and 30 energy consumption at each hour for a university campus building…… 32 Fig 16 Daily average humidity curve………………………………… 33 Fig 17 Correlation of humidity with energy consumption…………… 34 iv Fig 18 Daily average wind speed curve……………………………… 35 Fig 19 Correlation of wind speed with energy consumption………… 35 Fig 20 Illustration of clustering technique…………………………… 42 Fig 21 Inputs-outputs forecasting model……………………………… 48 Fig 22 Method 1(base case – day ahead approach)…………………… 50 Fig 23 Method 2(weekly update - day ahead approach)……………… 51 Fig 24 Method 3(daily update - day ahead approach)……………… 52 Fig 25 Method 4(Rolling horizon approach)…………………………… 53 Fig 26 Method 5(Rolling horizon approach)…………………………… 54 Fig 27 Method 6(hourly update - Rolling horizon approach)………… 55 Fig 28 The proposed load forecasting scheme… ………………… 57 Fig 29 One week load pattern with different method………………… 60 Fig 30 Correlation between real and predicted value………………… 61 Fig 31 One day load pattern with different methods………………… 62 Fig 32 ANN- Kmeans with different training methods……………… 65 Fig 33 ANN- Kmeans with different training methods……………… 66 Fig 34 MAPE distribution vs methods for building 6………………… 67 Fig 35 MAPE distribution vs methods for buildings ………………… 68 Fig 36 Overestimate MAPE distribution of Building 6………………… 69 Fig 37 Overestimate MAPE distribution of StudenHall……………… 70 Fig 38 Overestimate MAPE distribution of HighSchool……………… 70 Fig 39 Overestimate MAPE distribution of Building3_5……………… 71 v LIST OF TABLES Table Content Page Table Replacing missing data and outliers………………………… 21 Table Statistics of weather variables dataset and relation with energy consumption………………………………………………… 36 Table Summary of selected features in 24 hours………………… 40 Table Basic K-means algorithm…………………………………… 44 Table The algorithm of Sparsified K-means……………………… 45 Table The performance one week of the proposed approach and other methods……………………………………………… Table 60 The performance one day of the proposed approach and other method……………………………………………………… 62 Table Impact of temperature to forecasting accuracy……………… 63 Table Impact of binary features to forecasting accuracy…………… 64 Table 10 MAPE of different training methods………………………… 65 Table 11 Comparison between different implementation strategies in Building 6…………………………………………………… 66 Table 12 Comparison between different implementation strategies in StudentHall…………………………………………………… 67 Table 13 Comparison between different implementation strategies in HighSchool…………………………………………………… 67 Table 14 Comparison between different implementation strategies in Building3_5…………………………………………………… vi 68 Table 15 Length evaluation of training data…………………………… 72 Table 16 The computation cost………………………………………… 73 vii NOMENCLARTURE AI Artificial Intelligence ANN Artificial neural network AR Auto regressive model ARMA Auto regressive moving average model ARIMA Auto regressive integrated moving average model ARIMAX Autoregressive integrated moving average with exogeneous model ADN Active Distribution Network DG Distributed generation DBSCAN Density-based clustering algorithm DNN Deep neural networks DERs Distributed energy resources EMS Energy management system ESS Energy storage system FS Feature selection LR Linear regression LMBP Levenberg-Marquardt back propagation LTLF Long-term load forecasting MA Moving average model MLR Multiple linear regression MG MicroGrid viii Fig 39 Overestimate MAPE distribution of Building3_5 6.9 Length evaluation of training data As stated in the previous subsection, forecasting method based on data-driven, the performance of forecasting model significantly depends on the length of data and the complexity of model There is always exist a trade-off between the range of errors and the length of data In this work, with the availability of two-year datasets of four different buildings, we evaluate the amount of training data with the step size as shown in table 15 The experimental results show that the length of data for training range from 270 to 365 days will result in good performance of forecasting model 71 Table 15 Length evaluation of training data Day Average Day Average Day Average 30 10.74 8.38 8.57 8.5 8.18 11.20 8.66 9.17 30 5.34 8.13 4.11 5.65 7.79 8.56 6.37 6.57 30 14.43 22.03 21.01 12.22 19.96 11.25 17.03 16.85 60 6.03 8.77 7.28 10.1 8.97 14.68 5.38 8.75 60 5.49 7.33 7.28 4.92 7.52 5.42 5.53 6.21 60 17.72 21.37 22.29 10.47 7.18 11.24 9.10 14.20 Building6 90 5.04 7.15 7.47 5.2 4.64 8.39 3.84 5.96 180 5.4 5.70 3.63 3.20 6.55 5.53 7.79 5.40 270 4.0 4.63 4.01 5.14 5.34 4.60 7.05 4.97 365 3.72 3.17 3.14 3.47 3.23 5.13 4.32 3.74 Student_Hall 90 7.71 4.66 4.01 9.24 6.30 9.34 6.10 6.77 180 4.94 8.76 5.87 5.76 6.60 5.77 5.94 6.23 270 3.06 3.39 3.08 3.93 3.63 7.37 7.42 4.27 365 3.45 4.71 2.22 2.92 3.57 4.55 4.71 3.73 Highschool 90 10.83 10.64 14.24 9.90 11.86 6.55 8.69 10.39 180 7.34 8.63 7.70 6.65 6.90 9.42 9.44 8.01 270 6.38 5.57 8.40 6.40 9.27 6.94 7.65 7.23 365 8.92 7.23 5.84 6.32 5.02 4.36 8.61 6.61 72 Day Average 6.10 30 12.77 19.12 24.55 21.15 19.09 27.28 28.57 21.79 60 11.08 9.47 19.66 28.92 34.48 27.11 21.67 21.77 Building3_5 90 28.42 19.75 21.46 22.76 8.63 8.46 15.76 17.89 180 10.26 11.63 9.02 13.22 20.66 11.92 21.74 14.06 270 8.25 7.63 11.86 7.65 8.12 11.02 8.34 8.98 365 8.20 6.10 7.99 6.15 6.31 9.87 16.20 8.69 Processing time The experiments were implemented in Matlab programing language and tested on an Intel Core i7-7700 at 3.60GHz with 16GB of RAM The processing time of training and predicting are presented in Table 16 The proposed method reduces up to 50% processing time by applying clustering approach for all features The number of data for training in each group is smaller and easier for both training and predicting process Consequently, the training process of the proposed method is able to convergence faster than in both cases of using clustering for only load feature and not using any clustering approach The small processing time of proposed forecasting scheme enables real time scheduling operation approach for ESS owner Table 16 The computational cost Training time Prediction time (Second) (Second) ANN without K-means 36.5 0.35 ANN K-means Load only 22.12 0.21 Proposed method 11.94 0.01 Method 73 Conclusion This study described a detailed process of implementing strategies an artificial neural network combining a clustering method with effective input feature selection in order to predict energy consumption of campus building with a highly-nonlinear load profile The analysis performed showed that neural network combined clustering technique is a powerful method for automatic data preprocessing and input binary variable assignment as well as well treating missing data and outliers with an effective algorithm The proposed method is computationally simple and also suitable for analyzing a large set of data whose pattern changes over time The forecasting model is applied to several sets of buildings in order to verify performance of rolling horizon scheduling of energy storage system of microgrid From all the experiment results, as the prediction horizon is decreasing, it can be drawn out that ANN based clustering with hourly update model take more superiority in both robustness and accuracy This enables to apply this method to the real-time forecasting energy consumption applications for microgrid energy management system The proposed forecasting scheme also shows an interesting trade-off between an ability to capture complex relationship in the data and effective estimation processes that not require intensive human intervention In addition, the forecast performance obtained is really good in comparison with other methods and these models can be computed at a low computational cost 74 REFERENCES [1] S Chowdhury, S.P Chowdhury and P Crossley, “Microgrids and Active Distribution Networks” IET Renewable Energy Series 2009 [2] Hai-xiangZhao, and FrédéricMagoulès, “A review on the prediction of building energy consumption” Renewable and Sustainable Energy Reviews, Vol-16(6), pp.3586-3592 2012 [3] Aowabin Rahman, Vivek Srikumar, and Amanda D Smith, “Predicting electricity consumption for commercial and residential buildings using deep recurrent neural networks” Applied Energy, Vol-212, pp.372-385 2018 [4] Korea Energy Economics Institute Energy Consumption Survey 2014; Ministry of Trade, Industry and Energy: Sejong, Korea, 2014 [5] D.J Park, K.H Ju, Y.S Yoon, K.H Kim and S.S Kim, “Analysis of a building energy efficiency certification system in Korea” Sustainability, Vol-7, pp.16086–16107 2015 [6] Martin T Hagan and Suzanne M Behr, “The Time Series Approach to Short term Load Forecasting” IEEE Power Engineering Review, Vol-7(8), pp.56-57 2015 [7] S.A Soliman, “Electrical Load Forecasting Modeling and Model Construction” Butterworth Heinermann Publisher, 1st edition-2010 [8] Nima Amjady and Fashid Keynia, “Mid-term load forecasting of power system by a new prediction method” Energy Conversion and Management Vol49(10), pp.2678-2687 2008 [9] Liu K, “Comparison of Very Short-term Load Forecasting Techniques” IEEE Trans Power Syst, Vol-11(2), pp.877–882 1996 75 [10] S.R Huang, “Short-term Load Forecasting Using Threshold Autoregressive Models” IEEE Proceedings- Generation, Transmission and Distribution Vol144(5), pp477–481 1997 [11] J Nowicka-Zagrajek and R.Weron, “Modeling Electricity Loads in California: ARMA Model with Hyperbolic Noise” Signal Processing Vol82(12), pp.1903-1915 2002 [12] J.F Chen, W.M Wang and C.M Huang, “Analysis of an adaptive timeseries autoregressive moving average (ARMA) model for short-term load forecasting” Electric Power System and Research Vol-34(3), pp.187-196 1995 [13] Juberias G, “A new ARIMA model for hourly load forecasting” In: Proceedings of the 1999 IEEE Transmission and Distribution Conference 1999 [14] Vincenzo Bianco, Oronzio Manca and Sergio Nardini, “The electricity consumption forecasting in Italy using linear regression models” Energy Vol34(9), pp.1413-1421 2009 [15] H Tao, W Pu and H.L Willis, “A naive multiple linear regression benchmark for short term load forecasting” In: Proceedings of the 2011 IEEE power and energy society general meeting 2011 [16] Grzegorz Dudek, “Multivariate Regression Tree for Pattern-Based Forecasting Time Series with Multiple Seasonal Cycles” Information Architecture and Technology-ISAT 2017 Springer International Publishing [17] MA.N Daut, M.Y Hassan, H Abdullah, A.A Rahman, M.P Abdullah and F Hussin, “Building electrical energy consumption forecasting analysis using conventional and artificial intelligence methods: A review” Renewable and Sustainable Energy Reviews Vol-70, pp.1108-1118 2017 76 [18] Wei Ju, Baizhan Li, Yarong Lei and Meng Liu, “Analysis of a residential building energy consumption demand model” Energies Vol-4(3), pp.475-487 2011 [19] Roldán-Blay C, E.E Guillermo, A.B Carlos, R.P Carlos and R.G Javier, “Upgrade of an artificial neural network prediction method for electrical consumption forecasting using an hourly temperature curve model” Energy Build Vol-60, pp.38–46 2013 [20] W Mai, C.Y Chung, W Ting and H Huazhang, “Electric load forecasting for large office building based on radial basis function neural network” In: Proceedings of the 2014 IEEE PES general meeting conference & exposition 2014 [21] K.J Rishee, M.S Kevin, J.C Patricia and E.T John, “Forecasting energy consumption of multi-family residential buildings using support vector regression: Investigating the impact of temporal and spatial monitoring granularity on performance accuracy” Applied Energy Vol-123(15), pp.168178 2014 [22] C Ervin, C Vladimir and B Adrijan, “A Strategy for Short-Term Load Forecasting by Support Vector Regression Machines” IEEE Transactions on Power Systems Vol-28(4), pp.4356-4364 2013 [23] Zhang GP, “Time series forecasting using a hybrid ARIMA and neural network model” Neurocomputing Vol-50, pp.159–175 2003 [24] J.C Lu, D.X Niu and Z.Y Jia, “A Study of Short-term Load Forecasting Based on ARIMA-ANN” In: Proceedings of 2004 International Conference on Machine Learning and Cybernetics 77 [25] H Nie, G Liu, X Liu and Y Wang, “Hybrid of ARIMA and SVMs for Short-term Load Forecasting” Energy Procedia Vol-16, pp.1455-1460 2012 [26] S Vemuri, E.F Hill, and R Balasubramanian, "Load Forecasting using Stochastic Models" In: Proceedings of PICA Conference, pp.565-579 [27] N.D Uri, "System Load Forecasts for an Electric Utility" In: Proceeding of PICA Conference, pp.31-37 [28] A Keyhani and A El-Abiad, "One-Step-Ahead Load Forecasting for OnLine Applications" IEEE Winter Power Meeting, Jan 1975, paper no C75 0278 [29] A Keyhani and T Eliassi Rad, "A Simulation Study for Recursive Prediction of Hourly Load Demands" In: Proceeding of PICA Conference, 1977, pp 228-236 [30] Douglas Montgomery, “Design and Analysis of Experiments” Wiley Publisher 2017 [31] Rajesh Kumar, “Decision Tree for the Weather Forecasting” International Journal of Computer Applications Vol-76(2), pp.31-34 2013 [32] C Liu, Z Hu, Y Li and S Liu, “Forecasting copper prices by decision tree learning” Resources Policy Vol-52, pp.427-434 2017 [33] V.N Vapnik, “The Nature of Statistical Learning” New York: SpringerVerlag 1995 [34] V.N Vapnik, S Golowich, and A Smola, “Support Vector Method for Function Approximation, Regression Estimation and Signal Processing” Advances in Neural Information Processing Systems Vol-9, pp.281–287 1996 78 [35] B Schölkopf and A.J Smola, “Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond” Adaptive Computation and Machine Learning Series Cambridge, MA: MIT Press 2002 [36] E.T Belgin and D Dilara, “Electrical load forecasting using support vector machines” In: Proceedings of th International Conference on Electrical and Electronics Engineering 2011 [37] Zhongyi Hu, Yukun Bao and Tao Xiong, “Electricity Load Forecasting Using Support Vector Regression with Memetic Algorithms” The Scientific World Journal Vol-2013 [38] S Haykin, “Neural networks: A comprehensive foundation” New York: Macmillan Publishing Company 1999 [39] D.E Rumelhart, G.E Hinton and R.J Williams, “Learning representations by back-propagating errors” Nature, Vol-323, pp.533–536 1986 [40] S Kumar, S Mishra and S Gupta, “Short Term Load Forecasting Using ANN and Multiple Linear Regression” Second International Conference on Computational Intelligence and Communication Technology 2016 [41] A.S Ahmad, M.Y Hassan, M.P Abdullah, H.A Rahman, F Hussin, H Abdullah and R Saidur, “A review on applications of ANN and SVM for building electrical energy consumption forecasting” Renewable and Sustainable Energy Reviews Vol-33, pp.102-109 2014 [42] Wan He, “Load Forecasting via Deep Neural Networks” Procedia Computer Science Vol-122, pp.308-314 2017 [43] L.M Daniel, A Kasun and M Milos, “Building Energy Load Forecasting using Deep Neural Networks” Neural and Evolutionary Computing 2016 79 [44] A Kasun, L.M Daniel and M Milos, “Deep Neural Networks for Energy Load Forecasting” IEEE 26th International Symposium on Industrial Electronics 2017 [45] B.S Kwon, R.J Park and K.B Song, “Analysis of Short-Term Load Forecasting Accuracy Based on Various Normalization Methods” Journal of Korean Institute of Illuminating and Electrical Installation Engineers, Vol-32(6), pp.30-33 2018 [46] D Lazos, A.B Sproul and M Kay, “Optimisation of energy management in commercial buildings with weather forecasting inputs: a review” Renewable and Sustainable Energy Reviews Vol-39, pp.587–603 2014 [47] X Jingrui, C Ying, H Tao and D.L Thomas, “Relative Humidity for load forecasting models” IEEE Transaction on Smart Grid Vol-9(1), pp.191-198 2018 [48] X Jingrui and H Tao, “Wind Speed for Load Forecasting Models” Sustainability Vol-9, pp.795-806 2017 [49] L Hernandez, C Balandron, J.M Aguiar, B Carro, S.E Antonio and J Lloret, “Articifial Neural Networks for Short-term Load Forecasting in Microgrids Environment” Energy Vol-75, pp.252-264 2014 [50] https://www.datascience.com/blog/k-means-clustering [51] P.A Fahad and Stephen, “Preconditioned Data Sparsification for Big Data with Applications to PCA and K-means” IEEE Transaction on Information Theory Voi-63, pp.2954-2974 2017 80 [52] A David, S Vassilvitskii, “k-means++: The Advantages of Careful Seeding” In: Proceeding of the 18th Annual ACM-SIAM Symposium on Discrete Algorithms, 2007 [53] https://github.com/stephenbeckr/SparsifiedKMeans [54] D Chirag, S.E Lee, Y Junjing and S Mattheos, “Forecasting diurnal cooling energy load for institutional buildings using Artificial Neural Networks” Energy and Buildings Vol-121(1), pp.284-297 2016 [55] D.E Rumelhart, G.E Hinton and R.J Williams, “Learning representations by back-propagating errors” Nature Vol-323, pp.533–536 1986 [56] M Riedmiller and H Braun, “A direct adaptive method for faster backpropagation learning: The RPROP algorithm” In: Proceedings of IEEE International Conference on Neural Networks, San Francisco, CA, USA, 28 March–1 April 1993, pp 586-591 [57] H.C Gao, J.Ho Choi, S.Y Yun, H.J Lee and S.J Ahn, “Optimal Scheduling and Real-Time Control Schemen of Battery Energy Storage System for Micogrids Considering Contract Demand and Forecast Uncertainty” Energies, Vol-11(6), pp.1-15 2018 [58] J.S Armstrong, “Rule-Based Forecasting: Using Judgment in Time-series Extrapolation Principles of Forecasting-A Handbook for Researchers and Practitioners” Kluwer Academic Publishers, 2001 [59] R.J Hyndman and G Athanasopoulos, “Forecasting: Principles and Practice” 2012 81 ANN-based Short-term Load Forecasting for Rolling horizon Operation of Microgrids NGO MINH DUC 전남대학교 대학원 전기 공학과 (지도교수: 안선주) 국문 초록 소규모 분산 전원은 기존의 대형 중앙집중형 발전기를 대체하는 수단으로써 최근 설치가 지속적으로 증가하고 있다 분산 전원은 온실 가스 배출 감소, 에너지 변환 효율 향상, 계통 운영 비용 감소, 송배전 계통의 손실 감소 등과 같은 다양한 장점을 가지고 있다 이러한 분산 전원의 장점을 극대화하기 위하여 분산 전원과 부하를 하나의 셀처럼 운영하는 개념인 마이크로그리드가 제안되었고 최근 많은 마이크로그리드가 실증 및 상용 운전되고 있다 마이크로그리드 운영 시스템 (EMS)에서 단기 부하 예측은 가장 중요한 입력 데이터 중의 하나이다 전력 부하는 요일, 계절, 온도 등 다양한 요소의 영향을 받기 때문에 단기 부하 예측은 매우 어렵고 복잡한 문제이다 특히, 마이크로그리의 규모가 82 개별 건물 단위와 같이 작은 경우 부하의 변동성이 크고 부하와 다른 요소 사이의 비선형성이 커서 예측이 더욱 어렵다 또한 계측기의 불량, 통신 오류 등으로 인한 부하 데이터의 손실과 비정상적인 데이터는 과거 부하 이력 데이터를 이용해야 하는 부하 예측을 더욱 어렵게 만드는 요인이 된다 본 논문에서는 인공신경망 (artificial neural network, ANN)과 데이터 클러스터링 기법을 이용한 소규모 마이크로그리드 부하의 단기 예측 기법을 다루고 있다 부하와 온도 및 기상 데이터를 분석하여 인공신경망의 입력 특성 변수 (input feature)를 선정하였고, 이에 따라 신경망의 구조 및 최적 뉴런의 수를 결정하였다 이러한 기본적인 절차 외에 ANN 기반 단기 부하예측 기법을 구현하기 위한 데이터 전처리 과정에서 두 가지 방안을 제안하였다 첫 째, 결손 데이터와 이상 데이터를 검출하고 이를 정상데이터로 대체하는 기법을 개발하였다 기존 방법은 모든 데이터의 평균값을 이용하거나 내삽법 (interpolation)을 이용하였다 본 논문에서 제안하는 방법은 gradient 를 이용하여 결손 및 이상 데이터를 검출한 후 이러한 데이터를 포함한 하루의 부하 패턴과 유사한 부하를 갖는 데이터를 선택하여 이상 및 결손 데이터를 대체하였다 둘 째, 이진 특성 변수 (binary feature)의 값을 클러스터링 기반으로 자동으로 결정하는 방법을 제안하였다 기존의 방법은 데이터를 시각화하여 사람의 판단으로 변수의 값을 정하는 방식이었다 본 논문에서 제안하는 방법은 이력 데이터를 클러스터링을 83 이용하여 두 개의 그룹으로 구분하여 이진 변수의 값을 결정하는 것이다 이렇게 함으로써 사람의 개입없이 자동적으로 구현할 수 있으며, 예측의 정확도도 향상 시킬 수 있다 에너지 저장 장치 (ESS)는 마이크로그리드의 에너지 관리에서 가장 중요한 구성 요소이다 ESS 의 운영 스케줄 도출 시 부하예측 오차의 영향을 줄이고 운영 효율을 향상 시키기 위하여 rolling-horizon 기반 운영 기법이 제안되어 적용되고 있다 이 기법은 매 시간 새로운 계통 정보를 이용하여 새로운 운영 스케줄을 도출하는 것이다 이와 같은 스케줄 기법을 적용하기 위해서는 입력 데이터인 단기 부하예측도 매 시간마다 새롭게 업데이트 되어 야 한다 따라서 본 논문에서는 rolling-horizon 기반 ESS 스케줄링에 활용할 수 있도록, 매 시간 업데이트되는 부하를 이용하여 ANN 모델을 다시 도출하여 부하를 예측하는 방법을 구현하였다 이상에서 제안된 ANN 기반 마이크로그리드 단기 부하 예측 기법은 우리 대학의 개 건물에서 계측된 실제 데이터를 이용하여 검증하였다 제안된 기법을 적용하면 기존 방법에 비해 오차의 평균과 표준편차가 모두 작아짐을 시뮬레이션을 통해서 확인할 수 있었다 84 ACKNOWLEDGEMENTS In the first word, I would like to gratefully thank Professor Seon-Ju Ahn for his guidance, warm heart and consideration during my study at Power system Laboratory Without his immense knowledge, this research would not have been finished He encouraged me to grow as an independent thinker and how to become a person who has critical thinking in research environment I learned a lot from him I would like to take this chance to Professor Joon-Ho Choi, Professor Sang-Yun Yun and all members in power system laboratory for sharing their knowledge, friendly as well as making good social environment I considered them as close members in my big family I would like to thank all professors in community defense, especially Professor In-Seon Yeo and Professor Joon-Ho Choi for their valuable comment in completing my thesis I also want to thank the support of my friends, colleagues, staff members and all professors in Department of Electrical Engineering during my study in Chonnam National University I would like to acknowledge the Head of TNUT University, my colleagues at Automation Division and my Government for giving me the chance to study abroad I would like to thank my brother Nguyen Huy Toan and Tran Thai Trung for sharing the life and helping me through tough time I would like to thank my parents, my sisters, brothers for their unconditional love, constant support and dedicated their life for raising my daughter during my studying time Finally, I would like to give my deeply appreciate to my wife Nong Quynh Van, pretty daughter Ngo Bao Chau for their love and understanding The last word goes for my baby boy, who born in my hard time and given me both the challenges and the strength as well as motivation to get the things done 85