# Short-Term Load Forecasting on Individual Consumers

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

- Propose the use of the mutual information technique as one of the criteria to select the most important features for load forecasting;
- Propose a methodology for choosing significant features based on the joint application of mutual information and correlation techniques;
- Compare STLF results with different machine learning techniques;
- Improved short-term load forecasting in individual consumers by pre-processing and feature selection.

- Support Vector Regression (SVR);
- Feedforward Multilayer Perceptron (MLP);
- Long short-term memory (LSTM);
- ARTMAP-Fuzzy.

## 2. Related Work

## 3. Methodology

#### 3.1. Data Acquisition and Pre-Processing

#### 3.2. Feature Selection

#### 3.3. Model Creation

- Support vector regression (SVR);
- Feedforward multilayer perceptron (MLP);
- Long short-term memory (LSTM);
- ARTMAP-Fuzzy.

#### 3.4. Optimization

#### 3.5. Tests

^{2}), which indicates how fitted the model is in relation to the real values.

^{2}), in turn, is a metric for determining the fit of a model to real values. Its result ranges from 0 to 1 and indicates that the model is more accurate the higher its result. Its calculation is performed as shown in (4).

^{2}coefficient. The objective is to determine the model that presents the best results in relation to the forecast accuracy.

## 4. Results

#### 4.1. Feature Selection

#### 4.2. Otimization

#### 4.3. Tests

^{2}and their average results are presented in Table 11 and Table 12.

#### 4.4. Discussion of Results

^{2}metric, which indicates curve fitting quality, the SVR also obtained the best results, around 0.7, a very high value considering the randomness of the analyzed dataset.

^{2}), reaching 19.8% and 36.31% improvement, respectively, for MAPE and R

^{2}of the SVR model.

## 5. Conclusions

^{2}) increased by 36.31% in the best case (SVR) when compared to the baseline approach.

## Author Contributions

## Funding

## Institutional Review Board Statement

## Data Availability Statement

## Conflicts of Interest

## References

- Ziekow, H.; Goebel, C.; Struker, J.; Jacobsen, H. The potential of smart home sensors in forecasting household electricity demand. In Proceedings of the 2013 IEEE International Conference on Smart Grid Communications (SmartGridComm), Vancouver, BC, Canada, 21–24 October 2013. [Google Scholar] [CrossRef]
- Gielen, D.; Boshell, F.; Saygin, D.; Bazilian, M.D.; Wagner, N.; Gorini, R. The role of renewable energy in the global energy transformation. Energy Strategy Rev.
**2019**, 24, 38–50. [Google Scholar] [CrossRef] - Avancini, B.; Rodrigues, C.; Martins, B.; Rabêlo, L.; Al-Muhtadi, J.; Solic, P. Energy meters evolution in smart grids: A review. J. Clean. Prod.
**2019**, 217, 702–715. [Google Scholar] [CrossRef] - Quilumba, L.; Lee, W.; Huang, H.; Wang, Y.; Szabados, L. Using smart meter data to improve the accuracy of intraday load forecasting considering customer behavior similarities. IEEE Trans. Smart Grid
**2014**, 6, 911–918. [Google Scholar] [CrossRef] - Kipping, A.; Trømborg, E. Modeling aggregate hourly energy consumption in a regional building stock. Energies
**2018**, 11, 78. [Google Scholar] [CrossRef] - Shi, H.; Xu, M.; Li, R. Deep learning for household load forecasting—A novel pooling deep RNN. IEEE Trans. Smart Grid
**2017**, 9, 5271–5280. [Google Scholar] [CrossRef] - Edwards, E.; New, J.; Parker, E. Predicting future hourly residential electrical consumption: A machine learning case study. Energy Build.
**2012**, 49, 591–603. [Google Scholar] [CrossRef] - Arvanitidis, A.I.; Bargiotas, D.; Daskalopulu, A.; Kontogiannis, D.; Panapakidis, I.P.; Tsoukalas, L.H. Clustering Informed MLP Models for Fast and Accurate Short-Term Load Forecasting. Energies
**2022**, 15, 1295. [Google Scholar] [CrossRef] - Ghofrani, M.; Hassanzadeh, M.; Etezadi-Amoli, M.; Fadali, S. Smart meter based short-term load forecasting for residential customers. In Proceedings of the North American Power Symposium, Boston, MA, EUA, 4–6 August 2011. [Google Scholar] [CrossRef]
- Park, S.; Jung, S.; Jung, S.; Rho, S.; Hwang, E. Sliding window-based LightGBM model for electric load forecasting using anomaly repair. J. Supercomput.
**2021**, 77, 27–30. [Google Scholar] [CrossRef] - Kong, W.; Dong, Z.Y.; Hill, D.J.; Luo, F.; Xu, Y. Short-term residential load forecasting based on resident behaviour learning. IEEE Trans. Power Syst.
**2018**, 33, 1087–1088. [Google Scholar] [CrossRef] - Kong, W.; Dong, Z.Y.; Jia, Y.; Hill, D.J.; Xu, Y.; Zhang, Y. Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural Network. IEEE Trans. Smart Grid
**2019**, 10, 841–851. [Google Scholar] [CrossRef] - Alves, M.F. Previsão de Cargas Não Residenciais Mistas Por Redes Neurais Artmap Fuzzy. Ph.D. Thesis, Universidade Estadual Paulista, Ilha Solteira, Brazil, 2019. [Google Scholar]
- Haq, E.U.; Lyu, X.; Jia, Y.; Hua, M.; Ahmad, F. Forecasting household electric appliances consumption and peak demand based on hybrid machine learning approach. Energy Rep.
**2020**, 6, 1099–1105. [Google Scholar] [CrossRef] - Moon, J.; Jung, S.; Rew, J.; Rho, S.; Hwang, E. Combination of short-term load forecasting models based on a stacking ensemble approach. Energy Build
**2020**, 216, 109921. [Google Scholar] [CrossRef] - Ayub, N.; Irfan, M.; Awais, M.; Ali, U.; Ali, T.; Hamdi, M.; Alghamdi, A.; Mahammad, F. Big Data Analytics for Short and Medium-Term Electricity Load Forecasting Using an AI Techniques Ensembler. Energies
**2020**, 13, 5193. [Google Scholar] [CrossRef] - Roth, J.; Chadalawada, J.; Jain, R.K.; Miller, C. Uncertainty Matters: Bayesian Probabilistic Forecasting for Residential Smart Meter Prediction, Segmentation, and Behavioral Measurement and Verification. Energies
**2021**, 14, 1481. [Google Scholar] [CrossRef] - Jung, S.; Moon, J.; Park, S.; Hwang, E. An Attention-Based Multilayer GRU Model for Multistep-Ahead Short-Term Load Forecasting. Sensors
**2021**, 21, 1639. [Google Scholar] [CrossRef] - Tkachenko, R. An Integral Software Solution of the SGTM Neural-like Structures Implementation for Solving Different Data Mining Tasks. Lect. Notes Data Eng. Commun. Technol.
**2022**, 77, 696–713. [Google Scholar] [CrossRef] - Izonin, I.; Tkachenko, R.; Kryvinska, N.; Tkachenko, P.; Greguš ml, M. Multiple Linear Regression Based on Coefficients Identification Using Non-iterative SGTM Neural-like Structure. Lect. Notes Comput. Sci.
**2019**, 11506, 467–479. [Google Scholar] [CrossRef] - Prakash, K.; Sydulu, M. Non iterative-state estimation based neural network for short term load forecasting of distribution systems. In Proceedings of the 2009 IEEE Power & Energy Society General Meeting, Calgary, AB, Canada, 26–30 July 2009; pp. 1–8. [Google Scholar] [CrossRef]
- Ishwaran, H.; Rao, J.S. Spike and slab variable selection: Frequentist and bayesian strategies. Ann. Stat.
**2005**, 33, 730–773. [Google Scholar] [CrossRef] - Tkachenko, R.; Izonin, I. Model and Principles for the Implementation of Neural-like Structures Based on Geometric Data Transformations. Adv. Intell. Syst. Comput.
**2019**, 754, 578–587. [Google Scholar] [CrossRef] - Cover, M.; Thomas, A. Elements of Information Theory, 2nd ed.; John Wiley & Sons: Hoboken, NJ, USA, 2006. [Google Scholar]
- Kim, S.H.; Geem, Z.W.; Han, G.T. Hyperparameter Optimization Method Based on Harmony Search Algorithm to Improve Performance of 1D CNN Human Respiration Pattern Recognition System. Sensors
**2020**, 20, 3697. [Google Scholar] [CrossRef] - Bergstra, J.; Bengio, Y. Random search for hyper-parameter optimization. J. Mach. Learn. Res.
**2012**, 13, 281–305. [Google Scholar] - Jain, R.K.; Smith, K.M.; Culligan, P.J.; Taylor, J.E. Forecasting energy consumption of multi-family residential buildings using support vector regression: Investigating the impact of temporal and spatial monitoring granularity on performance accuracy. Appl. Energy
**2014**, 123, 168–178. [Google Scholar] [CrossRef] - Gajowniczek, K.; Zabkowski, T. Short term electricity forecasting using individual smart meter data. Procedia Comput. Sci.
**2014**, 35, 589–597. [Google Scholar] [CrossRef]

**Figure 3.**Predictive models for SVR and ARTMAP-Fuzzy network with a horizon of (

**a**) 30 min and (

**b**) 6 h.

Reference | Techniques | Granularity |
---|---|---|

[11] | LSTM | 30 min |

[12] | LSTM | 30 min |

[13] | ARTMAP-Fuzzy | 15 min |

[14] | K-medoids and ANN | Daily |

[10] | Sliding window-based LightGBM | 15 min |

[15] | Stacking ensemble MLP | 15 min |

[16] | RF, GB and SVM, GRU and CNN | Daily |

[17] | BSTS | 1 h |

[18] | GRU | 1 h |

[19] | Non-Iterative Ito Decomposition and SGTM | - |

[20] | Non-iterative SGTM | - |

[21] | State Estimation NN | 30 min |

Type | Feature |
---|---|

Predictors | Energy Consumption (t−25 to t−1) ^{1} |

Day (t) | |

Hour (t) | |

Rush hour (t) | |

Objective | Energy Consumption (t) |

^{1}“t−n” is the number of measurements prior to the current value “t”.

Type of Feature | Feature |
---|---|

Position | Maximum |

Minimum | |

Average | |

Median | |

Quantile (15%) ^{1} | |

Dispersion | Range |

Variance | |

Standard deviation | |

Distribution | Obliquity |

Kurtosis |

^{1}Value that is not exceeded in 15% of measurements.

Forecast Model | Hyperparameter | Bottom Limit | Upper Limit | N° |
---|---|---|---|---|

SVR | Epsilon | 0.001 | 0.01 | 50 |

C | 0 | 100 | ||

MLP | N° Layers | 1 | 5 | 60 |

N° Neurons | 5 | 20 | ||

LSTM | N° Layers | 1 | 5 | 72 |

N° Memory Loops | 7 | 25 | ||

ARTMAP-Fuzzy | Surveillance | 0.8 | 0.98 | 200 |

Alpha | 0.0001 | 0.01 |

Forecast Model | Hyperparameter | Value |
---|---|---|

SVR | Kernel | RBF |

MLP | Activation | Sigmoid |

Training | Backpropagation | |

LSTM | Activation | Hyperbolic Tangent |

Recurring Activation | Sigmoid | |

ARTMAP-Fuzzy | Training Rate | 1 |

Tracking | 0.001 |

Feature |
---|

Energy Consumption (t−25 to t−1) |

Day (t) |

Hour (t) |

Rush hour (t) |

Maximum |

Minimum |

Average |

Median |

Quantile (15%) |

Range |

Variance |

Standard deviation |

Obliquity |

Kurtosis |

Feature 1 | Feature 2 | MI Normalized |
---|---|---|

STD | Variance | 0.9724 |

Rush hour | Hour | 0.7849 |

Day | Maximum | 0.7475 |

Maximum | Range | 0.6303 |

Feature | MI Normalized |
---|---|

Hour | 0.5480 |

Variance | 0.1343 |

STD | 0.1329 |

Maximum | 0.1164 |

Range | 0.1046 |

Rush | 0.0286 |

Feature |
---|

Energy Consumption (t−25 to t−15) |

Energy Consumption (t−11 to t−1) |

Day (t) |

Hour (t) |

Maximum |

Minimum |

Average |

Variance |

Obliquity |

Forecast Model | Hyperparameter | Value |
---|---|---|

SVR | Epsilon | 0.0012 |

C | 0.1633 | |

MLP | N° Layers | 2 |

N° Neurons | 15 | |

LSTM | N° Layers | 3 |

N° Memory Loops | 25 | |

ARTMAP-Fuzzy | Surveillance | 0.8 |

Alpha | 0.0071 |

Forecast Model | MAPE (%) | R^{2} |
---|---|---|

SVR | 8.88 | 0.72 |

MLP | 10.62 | 0.62 |

LSTM | 9.62 | 0.68 |

ARTMAP-Fuzzy | 12.23 | 0.45 |

Forecast Model | MAPE (%) | R^{2} |
---|---|---|

SVR | 9.31 | 0.67 |

MLP | 11.02 | 0.56 |

LSTM | 10.58 | 0.61 |

ARTMAP-Fuzzy | 12.74 | 0.31 |

Time Horizon | MAPE Reducing (%) | R^{2} Increasing (%) |
---|---|---|

30 min | 1.79 | 0.65 |

6 h | 2.37 | 0.71 |

**Table 14.**MAPE and R

^{2}with a 30-min horizon using the implemented models compared to a persistent model.

Forecast Model | MAPE Reducing to Baseline (%) | R^{2} Increasing to Baseline (%) |
---|---|---|

SVR | 19.80 | 36.31 |

MLP | 4.08 | 17.38 |

LSTM | 13.11 | 28.74 |

ARTMAP-Fuzzy | −1.05 | −13.42 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2022 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 (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Melo, J.V.J.; Lira, G.R.S.; Costa, E.G.; Leite Neto, A.F.; Oliveira, I.B.
Short-Term Load Forecasting on Individual Consumers. *Energies* **2022**, *15*, 5856.
https://doi.org/10.3390/en15165856

**AMA Style**

Melo JVJ, Lira GRS, Costa EG, Leite Neto AF, Oliveira IB.
Short-Term Load Forecasting on Individual Consumers. *Energies*. 2022; 15(16):5856.
https://doi.org/10.3390/en15165856

**Chicago/Turabian Style**

Melo, João Victor Jales, George Rossany Soares Lira, Edson Guedes Costa, Antonio F. Leite Neto, and Iago B. Oliveira.
2022. "Short-Term Load Forecasting on Individual Consumers" *Energies* 15, no. 16: 5856.
https://doi.org/10.3390/en15165856