隨著全球能源情勢日益嚴峻，我國政府也逐漸開發多項能源開發，其中，太陽能是最被受注目的綠色能源之一。為了加速太陽能產業的發展，政府在民國103年時推動了第二期能源國家型科技計畫，其中包含了智慧電網中心。在智慧電網的計畫中，有許多子項目需要太陽能發電預測才能進行，例如用戶端需量交易市場、輸配電網智慧化以及智慧輸電技術開發等項目。因此，設計一個太陽能發電模型成為重要的課題。
在這篇論文中，我們採用3種不同的神經網路架構來進行太陽能發電模型。第一個架構為透過遞迴神經網路輸出帶有時間特徵的向量後，透過全連接層來進行發電預測，並且使用卷神經網路取代全連接層做為第二種架構。最後，Temporal Convolutional Network也可以運用在時間序列的預測上，因此將此架構做為第三種模型架構。為了驗證架構的精確度有提升，我們比較三種架構之間的準確度，並且以過往不同的模型作為基準，皆可證明本模型對於太陽能發電有一定程度的提升。

As the global energy situation becomes increasingly serious, the government has gradually developed a number of energy developments. Among them, solar energy is one of the most high-profile green energy sources. In order to promote the development of the solar industry, the government proposes the second phase of the national energy science and technology program in 103 years. In the smart grid, there are many sub-projects that require solar power generation forecastings such as the customer demand trading market, smart transmission and distribution network, and smart transmission technology development projects. Therefore, designing a solar power generation model has become an important issue.
In this paper, w we use three different neural network architectures to design the solar power generation model. Using the recurrent neural network combined with a fully connected layer model as our first model architecture, and replacing a fully connected layer by a convolutional neural network as the second model architecture. Temporal Convolutional Network can also be used in time series predictions, so we choose this neural network as the last model. In order to verify whether the accuracy has improved, we compare the accuracy between the three model architectures and choose the model published in IET journal in 2018 as a baseline to prove that the model has a certain degree of improvement for solar power generation.

[1] J.L. Torres, A. Garcia, M. De Blas, A. De Francisco Forecast of hourly average wind speed with ARMA models in Navarre Solar Energy, 79 (1) (2005), pp. 65-77
[2] Y. Li, Y. Su, L. ShuAn ARMAX model for forecasting the power output of a grid connected photovoltaic system Renew. Energy, 66 (2014), pp. 78-89
[3] G. Reikard, Predicting Solar Radiation at High Resolutions: A Comparison of Time Series Forecasts, Solar Energy, vol. 83, no. 3, pp. 342349, 2009
[4] J. Shi, W.-J. Lee, Y. Liu, Y. Yang, and P. Wang, Forecasting Power Output of Photovoltaic Systems Based on Weather Classification and Support Vector Machines, IEEE Trans. Ind. Appl., vol. 48, no. 3, pp. 10641069, May/Jun. 2012.
[5] M. Abuella, B. Chowdhury, Random Forest Ensemble of Support Vector Regression Models for Solar Power Forecasting, 2017. ArXiv170500033 Cs, http://arxiv.org/abs/1705.00033 (accessed February 21, 2018)
[6] A. Saberian, H. Hizam, M.A.M. Radzi, M.Z.A. Ab Kadir, M. Mirzaei Modelling and Prediction of Photovoltaic Power Output Using Artificial Neural Networks Int J Photo Energy (2014), pp. 1-10
[7] P. Ramsami, V. Oree A hybrid method for forecasting the energy output of photovoltaic systems Energy Convers. Manage., 95 (2015), pp. 406-413
[8] A. Alzahrani, P. Shamsi, M. Ferdowsi, C. Dagli, Solar Irradiance Forecasting Using Deep Neural Networks, Procedia Computer Science, 2017, ISSN 1877-0509.
[9] A. Gensler, J. Henze, B. Sick, and N. Raabe. Deep Learning for solar power forecastingAn approach using AutoEncoder and LSTM Neural Networks. In Systems, Man, and Cybernetics (SMC), 2016 IEEE International Conference on (pp. 002858-002865), Oct. 2016.
[10] Bengio, Y., Simard, P., Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE transactions on neural networks, 5(2), 157-166.
[11] Hochreiter, S., Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
[12] Junyoung Chung, Caglar Gulcehre, Kyunghyun Cho, and Yoshua Bengio. Empirical evaluation of gated recurrent neural networks on sequence modeling. CoRR, abs/1412.3555, 2014.
[13] Shaojie Bai, J Zico Kolter, and Vladlen Koltun. 2018. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv preprint arXiv:1803.01271.
[14] J. Long, E. Shelhamer, and T. Darrell, “Fully convolutional networks for semantic segmentation,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2015, pp. 3431–3440.
[15] R. Girshick. Fast R-CNN. arXiv:1504.08083, 2015
[16] Zang H., Cheng L., Ding T., Cheung K.W., Liang Z., Wei Z., Sun G.Hybrid method for short-term photovoltaic power forecasting based on deep convolutional neural networkIET Generation, Transmission Distrib., 12 (2019), pp. 4557-4567