# Climate Change and Power Security: Power Load Prediction for Rural Electrical Microgrids Using Long Short Term Memory and Artificial Neural Networks

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Methodology: Data Pre-Processing and Models

#### 2.1. Super-Model

#### 2.2. Long Short Term Memory Model

while (step*self.batch_size<self.training_iters) and (testing_loss_value > target_loss): current_learning_rate = self.learning_rate current_learning_rate *= 0.1 ** ((step * self.batch_size) * self.training_iter_step_down_every)

#### 2.3. Artificial Neural Network

## 3. Results and Discussion

## 4. Conclusions

## Author Contributions

## Acknowledgments

## Conflicts of Interest

## References

- Adams, D. Juneau Hydroelectric Power Crisis Lessons-Learned Commission Report; City and Borough of Juneau: Juneau, AK, USA, 2008. [Google Scholar]
- Mauch, L.; Yang, B. A new approach for supervised power disaggregation by using a deep recurrent LSTM network. In Proceedings of the 2015 IEEE Global Conference on IEEE Signal and Information Processing (GlobalSIP), Orlando, FL, USA, 14–16 December 2015; pp. 63–67. [Google Scholar]
- Qu, X.; Kang, X.; Zhang, C.; Jiang, S.; Ma, X. Short-term prediction of wind power based on deep Long Short-Term Memory. In Proceedings of the 2016 IEEE PES Asia-Pacific IEEE, Power and Energy Engineering Conference (APPEEC), Xi’an, China, 25–28 October 2016; pp. 1148–1152. [Google Scholar]
- Zheng, H.; Yuan, J.; Chen, L. Short-term load forecasting using EMD-LSTM neural networks with a Xgboost algorithm for feature importance evaluation. Energies
**2017**, 10, 1168. [Google Scholar] [CrossRef] - Pike, C.; Mueller-Stoffels, M.; Whitney, E. Galena PV Integration Options Guidance on Technical and Regulatory Issues for Distributed Generation in Isolated Grids; Technical Report; Alaska Center for Energy and Power (ACEP), University of Alaska Fairbanks: Fairbanks, AK, USA, 2016. [Google Scholar]
- Meral, M.E.; Dinçer, F. A review of the factors affecting operation and efficiency of photovoltaic based electricity generation systems. Renew. Sustain. Energy Rev.
**2011**, 15, 2176–2184. [Google Scholar] [CrossRef] - Yang, C.T.; Molinas, A. Sediment transport and unit stream power function. J. Hydraul. Div.
**1982**, 108, 774–793. [Google Scholar] - Pryor, S.; Barthelmie, R. Climate change impacts on wind energy: A review. Renew. Sustain. Energy Rev.
**2010**, 14, 430–437. [Google Scholar] [CrossRef] - Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput.
**1997**, 9, 1735–1780. [Google Scholar] [CrossRef] [PubMed] - Zipser, D. Recurrent network model of the neural mechanism of short-term active memory. Neural Comput.
**1991**, 3, 179–193. [Google Scholar] [CrossRef] - Malhotra, P.; Ramakrishnan, A.; Anand, G.; Vig, L.; Agarwal, P.; Shroff, G. LSTM-based encoder-decoder for multi-sensor anomaly detection. arXiv
**2016**, arXiv:1607.00148. [Google Scholar] - Malhotra, P.; Vig, L.; Shroff, G.; Agarwal, P. Long short term memory networks for anomaly detection in time series. In Proceedings of the ESANN 2015: 23rd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Bruges, Belgium, 22–24 April 2015; Presses Universitaires de Louvain: Louvain, Belgium, 2015; p. 89. [Google Scholar]
- Mitchell, T.M. Machine learning and data mining. Commun. ACM
**1999**, 42, 30–36. [Google Scholar] [CrossRef] - Amjady, N.; Keynia, F.; Zareipour, H. Short-term load forecast of microgrids by a new bilevel prediction strategy. IEEE Trans. Smart Grid
**2010**, 1, 286–294. [Google Scholar] [CrossRef] - Sáez, D.; Ávila, F.; Olivares, D.; Cañizares, C.; Marín, L. Fuzzy prediction interval models for forecasting renewable resources and loads in microgrids. IEEE Trans. Smart Grid
**2015**, 6, 548–556. [Google Scholar] [CrossRef] - Kanchev, H.; Lu, D.; Colas, F.; Lazarov, V.; Francois, B. Energy management and operational planning of a microgrid with a PV-based active generator for smart grid applications. IEEE Trans. Ind. Electron.
**2011**, 58, 4583–4592. [Google Scholar] [CrossRef][Green Version] - Iliamna, AK Forecast: Weather Forecast and Reports. Long Range and Local. Available online: https://www.wunderground.com/weather/us/ak/igiugig/99613 (accessed on 11 March 2017).
- Abadi, M.; Barham, P.; Chen, J.; Chen, Z.; Davis, A.; Dean, J.; Devin, M.; Ghemawat, S.; Irving, G.; Isard, M.; et al. TensorFlow: A System for Large-Scale Machine Learning. In Proceedings of the 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI’16), Savannah, GA, USA, 2–4 November 2016; Volume 16, pp. 265–283. [Google Scholar]
- Gers, F.A.; Schmidhuber, J.; Cummins, F. Learning to forget: Continual prediction with LSTM. In Proceedings of the 9th International Conference on Artificial Neural Networks: ICANN ’99, Edinburgh, UK, 7–10 September 1999. [Google Scholar]
- Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv
**2014**, arXiv:1412.6980. [Google Scholar] - LeCun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature
**2015**, 521, 436–444. [Google Scholar] [CrossRef] [PubMed]

**Figure 1.**

**Top**: Annual weather conditions at the Igiugig, AK study site.

**Bottom**: Not an uncommon pattern of weather changes in one day. Notice several sudden temperature changes in access of $10{\phantom{\rule{3.33333pt}{0ex}}}^{\circ}$F occurring over 20 min window.

**Figure 2.**An illustration of the power prediction super-model. The weather conditions, power load and power generation data are showed in the Temporal Data table. The solid arrows indicate the use of data for the model training, the double arrows show the prediction of power load or power generation. While the illustration shows the overall energy microgrid management system (EMMS), we present only the details of the Long Short Term Memory and the Power Load Artificial Neural Networks models. The EMMS simply integrates the power load prediction and the power generation from the renewable sources.

**Figure 4.**LSTM’s training error loss. Eight resulting plot lines represent each LSTM trained on one weather condition. The plots show error loss during the model training. The termination condition for the model training was set if the model generalized the data with a desired error loss of $1.0\times {10}^{-7}$ or $1.0\times {10}^{6}$ training epochs was reached.

**Figure 5.**Illustration of the multilayer perceptron artificial neural network implementation for the power draw predictions. Each neuron uses a rectified linear unit (ReLU) as the activation function. The example network shows six inputs, two hidden layers, and one output.

**Figure 6.**Two sample output of the power load predictions. The blue plot line shows the predicted power load by the ANN that used the LSTM’s outputs of the predicted weather conditions, the red plot line is the ANN’s power load prediction using the measured weather conditions, and the green solid plot line is the actual measured power load. Each major unit on the x-axis is a prediction of 10 min into the future from the start time.

**Figure 7.**The results of the forecast power load. The x-axis has the predicted power load, while the y-axis has the measured power load. (

**Left**) Predicted power load for the current weather conditions with the MSE = $3.61\times {10}^{-3}$. (

**Right**) Predicted power load using the future LSTM weather predictions with the MSE = $3.65\times {10}^{-3}$.

**Figure 8.**The power load predictions results using only the temperature weather conditions. (

**Left**) Predicted power load using the current temperature conditions with the MSE = $1.20\times {10}^{-2}$. (

**Right**) Predicted power load using the future LSTM temperature predictions with the MSE = $1.66\times {10}^{-2}$.

**Figure 9.**The sample outputs of the LSTM models for the events, wind direction, wind gust speed, conditions, temperature, dew point, wind speed and temperature weather features (from the top left to the bottom right in a column-row order). The green plot line is the measured (actual) value of the reported weather feature and the red plot line is the LSTM output predicting the weather conditions for the current and 40 min into the future.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Cenek, M.; Haro, R.; Sayers, B.; Peng, J. Climate Change and Power Security: Power Load Prediction for Rural Electrical Microgrids Using Long Short Term Memory and Artificial Neural Networks. *Appl. Sci.* **2018**, *8*, 749.
https://doi.org/10.3390/app8050749

**AMA Style**

Cenek M, Haro R, Sayers B, Peng J. Climate Change and Power Security: Power Load Prediction for Rural Electrical Microgrids Using Long Short Term Memory and Artificial Neural Networks. *Applied Sciences*. 2018; 8(5):749.
https://doi.org/10.3390/app8050749

**Chicago/Turabian Style**

Cenek, Martin, Rocco Haro, Brandon Sayers, and Jifeng Peng. 2018. "Climate Change and Power Security: Power Load Prediction for Rural Electrical Microgrids Using Long Short Term Memory and Artificial Neural Networks" *Applied Sciences* 8, no. 5: 749.
https://doi.org/10.3390/app8050749