Energy consumption prediction in buildings using LSTM and SVR modified by developed Henry gas solubility optimization – Nature

Executive Summary

Accurate prediction of building energy consumption is fundamental to achieving key Sustainable Development Goals (SDGs), particularly those related to energy, infrastructure, and climate action. This report details a novel hybrid forecasting model designed to enhance energy management and sustainability. The model integrates wavelet decomposition for feature extraction, Long Short-Term Memory (LSTM) networks for temporal analysis, and Support Vector Regression (SVR) for refined estimation. All model parameters are optimized using a Developed Henry Gas Solubility Optimization (DHGSO) algorithm. Validated on a robust dataset of two years of hourly energy data from seven campuses, the proposed model demonstrates superior performance, achieving a 20% reduction in Root Mean Square Error (RMSE) and a 15% reduction in Mean Absolute Percentage Error (MAPE) compared to standalone models. These results underscore the model’s potential to support proactive energy management strategies, thereby advancing SDG 7 (Affordable and Clean Energy), SDG 11 (Sustainable Cities and Communities), and SDG 13 (Climate Action) through improved demand response, renewable energy integration, and operational efficiency.

Introduction: Aligning Energy Management with Sustainable Development Goals

The imperative to manage building energy consumption effectively is directly linked to the global agenda for sustainable development. Accurate energy forecasting is a critical enabler for strategies that support several SDGs. By providing precise insights into energy usage patterns, such forecasts empower facility managers to optimize resource allocation, a core tenet of SDG 12 (Responsible Consumption and Production). This report outlines a novel predictive model that enhances these capabilities, contributing to a more sustainable and resilient built environment.

The Role of Forecasting in Sustainable Operations

Effective energy forecasting is instrumental in achieving multiple sustainability objectives:

• Advancing SDG 7 (Affordable and Clean Energy): Accurate predictions are vital for integrating intermittent renewable energy sources. By forecasting demand, building managers can optimize the use of solar or wind power, reducing reliance on fossil fuels and supporting the transition to clean energy.
• Supporting SDG 11 (Sustainable Cities and Communities): Buildings are major energy consumers in urban areas. Efficient energy management, enabled by precise forecasting, reduces the overall environmental footprint of cities, making them more sustainable and resilient.
• Promoting SDG 13 (Climate Action): By minimizing energy waste and facilitating the use of renewables, advanced forecasting models directly contribute to the reduction of greenhouse gas emissions, a primary driver of climate change.
• Enhancing SDG 9 (Industry, Innovation, and Infrastructure): The development and deployment of sophisticated forecasting models represent an innovation in sustainable infrastructure management. This fosters resilient infrastructure and promotes inclusive and sustainable industrialization.

This study addresses the limitations of existing models by proposing a hybrid framework that combines multi-resolution feature engineering, deep learning, and advanced metaheuristic optimization to deliver a robust tool for sustainable energy management.

Methodology for Sustainable Energy Forecasting

The proposed methodology is designed to capture the complex, non-linear, and temporal dynamics of building energy consumption, providing a robust foundation for decisions that align with sustainability targets.

Data Acquisition and Preprocessing

The model was developed using a comprehensive dataset collected over two years from seven distinct campuses, ensuring a diverse representation of building types and operational contexts. This aligns with the need for scalable solutions applicable to diverse urban infrastructures under SDG 11.

1. Data Sources: Hourly energy consumption data was collected from building energy management systems, supplemented by meteorological data (temperature, humidity) and temporal indicators (day of week, time of day).
2. Data Imputation: A multi-stage process was used to handle missing data, preserving temporal trends and ensuring data integrity without introducing synthetic bias.
3. Normalization: All predictive variables were scaled to a [0, 1] range using Min-Max normalization to ensure stable model convergence and comparability across features.

Hybrid Forecasting Model Architecture

The core of the proposed solution is a hybrid model that leverages the complementary strengths of three distinct techniques to achieve superior predictive accuracy.

• Wavelet Decomposition: This technique is used for multi-resolution feature extraction. It decomposes the energy consumption time series into high-frequency and low-frequency components, allowing the model to capture both short-term fluctuations and long-term trends. The Haar wavelet was selected for its computational efficiency and suitability for non-stationary signals.
• Long Short-Term Memory (LSTM): An advanced type of recurrent neural network, LSTM is employed to model the long-term temporal dependencies identified in the decomposed data. This is crucial for understanding seasonal and weekly patterns that drive energy use.
• Support Vector Regression (SVR): SVR is used to refine the predictions from the LSTM network. It excels at modeling non-linear relationships and acts as a smoothing mechanism, improving the final forecast’s stability and accuracy by focusing on linear trends and reducing noise.

Developed Henry Gas Solubility Optimization (DHGSO) Algorithm

To ensure the hybrid model operates at peak performance, a novel metaheuristic algorithm, the Developed Henry Gas Solubility Optimization (DHGSO), was used to tune the hyperparameters of both the LSTM and SVR components. This optimization is critical for balancing the exploration of the solution space with the exploitation of promising regions, thereby finding the optimal model configuration. By automating and enhancing this process, the DHGSO algorithm ensures that the forecasting tool is both highly accurate and robust, making it a reliable instrument for advancing energy efficiency goals under SDG 7 and SDG 12.

Results and Performance Analysis

The proposed hybrid model, optimized with the DHGSO algorithm, was rigorously evaluated against standalone LSTM and SVR models, as well as other state-of-the-art methods. The results confirm its superior performance in predicting building energy consumption, providing a more reliable tool for sustainable management.

Key Performance Metrics

The model’s accuracy was assessed using standard statistical metrics across various building types, including university dormitories, laboratories, and offices.

• Root Mean Square Error (RMSE): The proposed model achieved an average reduction of 20% in RMSE compared to baseline models. This significant decrease in error translates to more reliable energy procurement and budget planning.
• Mean Absolute Percentage Error (MAPE): An average reduction of 15% in MAPE was observed, indicating a substantial improvement in the relative accuracy of the forecasts.

These quantitative improvements demonstrate the model’s capacity to provide actionable intelligence for energy managers. Such precision is essential for implementing effective demand-response programs and optimizing the integration of renewable energy systems, directly contributing to the objectives of SDG 7 and SDG 11.

Comparative Analysis

When compared to seven other advanced prediction algorithms, including Gradient Boosting Regression Tree (GBRT) and Deep Reinforcement Learning (DRL), the proposed framework consistently demonstrated the lowest prediction error. This validates the synergistic benefits of combining wavelet decomposition, deep learning, and metaheuristic optimization for this application.

Discussion: Implications for Sustainable Development

The superior performance of the proposed forecasting model has significant practical implications for advancing the Sustainable Development Goals within the context of building and urban energy management.

Contribution to SDG 7: Affordable and Clean Energy

By delivering highly accurate energy demand forecasts, the model serves as a critical tool for managing the transition to cleaner energy systems. It enables building operators to:

1. Optimize Renewable Energy Integration: Forecasts allow for better alignment of energy consumption with periods of high renewable energy generation, maximizing the use of clean power and reducing reliance on the grid.
2. Enhance Grid Stability: Participation in demand-response programs becomes more effective, as buildings can reliably reduce their load during peak periods, contributing to a more stable and efficient energy grid for the entire community.

Contribution to SDG 11 and SDG 13: Sustainable Cities and Climate Action

Efficient energy use in buildings is a cornerstone of sustainable urban development and a key strategy for climate change mitigation. The model supports these goals by:

• Reducing Urban Energy Footprint: Minimized energy waste through optimized operations leads to lower overall consumption in cities, reducing the strain on resources and infrastructure.
• Lowering Carbon Emissions: A 20% reduction in forecasting error can prevent significant over-procurement of energy, which in turn avoids unnecessary generation and the associated CO2 emissions, directly supporting climate action.

Contribution to SDG 9 and SDG 12: Innovation and Responsible Consumption

The model itself is an innovation in sustainable infrastructure management (SDG 9), while its application promotes more responsible resource use (SDG 12).

• Proactive Maintenance Scheduling: Forecasts can predict periods of high equipment stress, allowing for proactive maintenance that extends asset lifespan, reduces waste from equipment failure, and lowers operational costs.
• Informed Budgeting and Resource Allocation: Accurate cost projections enable better financial planning, ensuring that resources are allocated efficiently to support sustainability initiatives.

Conclusion and Future Directions

This report has presented a novel hybrid model for building energy consumption forecasting that demonstrates significant improvements in accuracy and reliability. By integrating wavelet decomposition, LSTM, SVR, and the DHGSO optimization algorithm, the framework provides a powerful tool for building managers to enhance operational efficiency and advance sustainability objectives. The validated 20% reduction in RMSE and 15% reduction in MAPE confirm its superiority over existing methods.

The practical applications of this model directly support the achievement of multiple Sustainable Development Goals, including SDG 7 (Affordable and Clean Energy), SDG 9 (Industry, Innovation, and Infrastructure), SDG 11 (Sustainable Cities and Communities), SDG 12 (Responsible Consumption and Production), and SDG 13 (Climate Action). It empowers stakeholders to make data-driven decisions that reduce energy waste, facilitate renewable energy integration, and lower carbon emissions.

Future work will focus on testing the model’s scalability on larger, more diverse datasets and exploring its adaptability to different climatic regions and building typologies. Further research into integrating additional variables, such as occupancy patterns and real-time weather data, could enhance its predictive power and broaden its applicability in dynamic urban environments.

Analysis of Sustainable Development Goals in the Article

1. Which SDGs are addressed or connected to the issues highlighted in the article?

The article on predicting building energy consumption connects to several Sustainable Development Goals (SDGs) by focusing on energy efficiency, technological innovation, sustainable infrastructure, and climate action. The primary SDGs addressed are:

• SDG 7: Affordable and Clean Energy: The core theme of the article is the efficient management of energy consumption in buildings. By developing a model to accurately predict energy needs, the study directly contributes to optimizing energy use, which is a cornerstone of energy efficiency. The article also explicitly mentions that accurate forecasts are “vital for integrating renewable energy,” which aligns with the goal of increasing the share of clean energy.
• SDG 9: Industry, Innovation, and Infrastructure: The study is fundamentally about innovation. It proposes a “novel hybrid forecasting model” that integrates advanced technologies like Long Short-Term Memory (LSTM) networks, Support Vector Regression (SVR), and a “Developed Henry Gas Solubility Optimization (DHGSO) algorithm.” This represents a technological advancement aimed at improving the operational efficiency and sustainability of infrastructure (buildings).
• SDG 11: Sustainable Cities and Communities: The research is set in the context of “sustainable building management” across multiple university campuses. Buildings are a critical component of urban infrastructure, and their energy consumption is a major factor in the environmental impact of cities. Improving energy efficiency in large-scale building systems contributes directly to making communities more sustainable and resource-efficient.
• SDG 12: Responsible Consumption and Production: The article highlights that accurate energy forecasts help managers “allocate resources efficiently, identify savings opportunities,” and “reduce waste from faulty equipment.” This promotes more sustainable consumption patterns by ensuring that energy resources are used more effectively and waste is minimized, aligning with the principles of responsible resource management.
• SDG 13: Climate Action: By improving energy efficiency and facilitating the “integration of renewable energy sources,” the proposed model helps in “reducing reliance on conventional sources.” The discussion section quantifies this impact, stating that improved accuracy “equates to cutting W tons of CO2 emissions per annum.” This directly addresses the need to take urgent action to combat climate change and its impacts.

2. What specific targets under those SDGs can be identified based on the article’s content?

Based on the article’s focus, the following specific SDG targets can be identified:

1. Target 7.2: Increase the share of renewable energy. The article states that accurate forecasts are essential for “renewable energy integration,” as they help in “properly sizing, managing, and optimizing clean energy systems.” This directly supports the goal of increasing the proportion of renewable energy in the energy mix.
2. Target 7.3: Double the global rate of improvement in energy efficiency. The entire study is centered on “optimizing energy management” and improving “operational efficiency.” The proposed model, which achieves significant reductions in prediction errors, is a tool designed to enhance energy efficiency in buildings, contributing to this global target.
3. Target 9.4: Upgrade infrastructure to make it sustainable and resource-efficient. The research provides an advanced technological tool for managing building energy systems, which are a key part of infrastructure. By enabling more efficient energy use, the model helps upgrade the management of this infrastructure to be more sustainable and less resource-intensive.
4. Target 11.6: Reduce the adverse per capita environmental impact of cities. Efficiently managing energy consumption in large buildings, as demonstrated in the study across seven campuses, directly reduces the overall energy footprint and associated emissions of urban and community areas, thus lowering their environmental impact.
5. Target 12.2: Achieve the sustainable management and efficient use of natural resources. The article emphasizes that its forecasting model allows managers to “allocate resources efficiently” and supports “proactive demand response.” This promotes the efficient use of energy, a critical natural resource.
6. Target 13.2: Integrate climate change measures into policies and planning. The model serves as a practical tool for planning and decision-making in energy management. As noted in the discussion, its application leads to quantifiable reductions in CO2 emissions, demonstrating how technological solutions can be integrated into operational planning to achieve climate goals.

3. Are there any indicators mentioned or implied in the article that can be used to measure progress towards the identified targets?

Yes, the article mentions and implies several indicators that can be used to measure progress towards the identified targets:

• Directly Mentioned Performance Metrics: The article quantifies the improvement of its model using specific error metrics. These serve as direct indicators of increased efficiency in energy forecasting, which is a proxy for better energy management.
  • Reduction in Root Mean Square Error (RMSE): The study reports a “20% reduction in RMSE,” indicating a higher accuracy in predicting energy consumption.
  • Reduction in Mean Absolute Percentage Error (MAPE): A “15% reduction in MAPE” is also cited, further validating the model’s improved predictive performance.
• Implied Quantitative Indicators: The article discusses the practical outcomes of improved forecasting, which can be measured as indicators of progress towards sustainability goals.
  • Energy Intensity (Indicator 7.3.1): While not explicitly named, the concept of managing “hourly energy consumption data” for buildings with known floor areas directly relates to measuring and improving energy intensity (energy consumed per square meter).
  • CO2 Emission Reductions: The discussion section explicitly mentions that avoiding unnecessary energy generation “equates to cutting W tons of CO2 emissions per annum,” which is a key indicator for climate action (SDG 13).
  • Cost Savings: The article notes that improved accuracy translates into “cost saving of roughly USD Z annually,” serving as an economic indicator of resource efficiency (SDG 12).
  • Share of Renewable Energy (Indicator 7.2.1): The model’s role in facilitating “renewable energy integration” implies that its adoption could be measured by an increase in the share of renewables used in the managed buildings.

4. Summary Table of SDGs, Targets, and Indicators

Source: nature.com