ABSTRACT

The growth of smart cities is propelled by the rapid integration of advanced technologies, such
as the Internet of Things (IoT), artificial intelligence (AI), and data analytics, into urban
infrastructure and services. Fueled by the challenges of urbanization, governments worldwide
are investing in smart city initiatives to address issues like traffic congestion, resource
optimization, and environmental sustainability. This evolution is marked by the development
of modern infrastructure, citizen engagement through digital platforms, and a data-driven
approach to decision-making. As smart cities strive to enhance efficiency, sustainability, and
quality of life, the sector not only attracts significant investments and fosters economic
opportunities but also reflects a global shift towards technologically integrated urban
environments.

The rapid growth of digitalization in the urban areas results in the high demand of electricity
consumption for smooth operation. Smart home energy conservation plays a pivotal role in the
broader development of smart cities. With the increasing urbanization and focus on sustainable
living, integrating smart home technologies for energy efficiency is essential. Through the
deployment of IoT-enabled devices, machine learning algorithms, and real-time data analytics,
smart homes can optimize energy consumption, reduce wastage, and contribute to overall
resource sustainability. These energy conservation measures not only benefit individual
households by lowering utility bills but also align with the larger goal of creating
environmentally conscious and energy-efficient urban ecosystems. As part of the smart cities'
development, the integration of such smart home energy solutions contributes to a more
intelligent, interconnected, and sustainable urban infrastructure, addressing the growing
challenges of energy demands in densely populated areas.

Smart home energy optimization, driven by artificial intelligence (AI) techniques, emerges as
a critical component in the development of smart cities. Leveraging advanced algorithms and
machine learning, smart homes equipped with AI-enabled systems can intelligently manage and
optimize energy consumption. These technologies analyze real-time data from various sensors
and devices, predicting usage patterns and adapting accordingly. Through personalized
recommendations and automated adjustments, AI enhances energy efficiency, reduces waste,
and contributes to sustainable practices. The integration of such intelligent systems not only

v
empowers residents to make informed energy choices but also aligns with the broader
objectives of smart cities, promoting resource efficiency, environmental sustainability, and a
technologically advanced urban landscape. In this study the data of smart home energy
consumption is being aquired fron the online reposetory, existing online. On this data different
machine learning techniques are being implemented to test the results for the energy
predcitions. To implement this work regression algirthms are being implement namerly;
Desicion Tree Regression, having the prediction error of 0.0439, Random Forest regression,
having the prediction error of 0.0427 K-nearest Neighbour regression, having the prediction
error of 0.0285 Support vector machine, having the prediction error of 0.0.1394 and simple
linear regression having the prediction error of 0.0.9900.

Further, This study performs the smart pattern identification in home energy consumption data
using artificial intelligence techniques. In this work, two different approaches are being used to
achieve the research objectives i.e., the GWO based CNN-LSTM Network and IAP based
SGKRU technique. Grey Wolf Optimization (GWO) with a combination of Convolutional
Neural Network (CNN) and Long Short-Term Memory (LSTM) network predicts the energy
consumption in smart homes. GWO is employed as an optimization algorithm to enhance the
performance of the CNN-LSTM architecture. The Convolutional Neural Network (CNN) is
utilized for spatial feature extraction from input data, focusing on relevant patterns and
relationships. Following this, the Long Short-Term Memory (LSTM) network captures
temporal dependencies, particularly useful for handling sequential data such as time-series
information associated with energy consumption. The Grey Wolf Optimization algorithm is
integrated to optimize the training of the CNN-LSTM network. This involves leveraging the
principles inspired by the social behavior of grey wolves to enhance the convergence and
effectiveness of the optimization process. The combined architecture aims to synergize the
spatial and temporal aspects of data through CNN and LSTM, respectively, while GWO
contributes to refining the model's parameters for better predictive accuracy. The model, thus,
offers an integrated approach to efficiently predict energy consumption in the dynamic context
of smart homes. The utilization of Grey Wolf Optimization enhances the learning process,
making the CNN-LSTM network well-suited for addressing the complexities of energy
consumption prediction in smart home environments. The proposed technique is generating the

vi
energy prediction with the error rate of 0.6213, MSE of 0.3860, which is much lesser than the
error in prediction in the existing techniques.

Another approach is IAP based SGKRU which introduces a novel approach to energy
consumption prediction in smart homes by combining Integration Affinity Propagation
clustering with a Gated Kaiming Recurrent Unit (GRU) enhanced with Swish and softmax
activation functions. Integration Affinity Propagation clustering is employed for data
preprocessing, effectively identifying meaningful clusters in the energy consumption dataset.
These clusters serve as valuable inputs to the subsequent prediction model. The Gated Kaiming
Recurrent Unit is chosen for its ability to model sequential dependencies in the data. The
integration of Swish and softmax activation functions further enhances the GRU's capabilities.
Swish, a smooth and differentiable activation function, is employed for improved gradient flow,
while softmax provides normalized outputs, enhancing interpretability. The model aims to
leverage the strengths of clustering for data grouping and the enhanced capabilities of the GRU
with specific activation functions to capture complex temporal patterns in energy consumption.
This proposed framework seeks to offer a robust solution for accurate and interpretable energy
consumption predictions in the dynamic environment of smart homes. On implementing both
the approaches it is being observed that the best results are obtained using IAP based SGKRU
technique for energy consumption prediction in smart home environment. The predictions of
energy consumptions obtained and discussed in this work for generating the recommendations
for optimized device scheduling in the smart home environment.

Using the above discussed techniques, patterns of energy consumption is being identified in the
smart home environment and provided to implement the recommendation system for optimized
device scheduling. The recommendation system is implemented using two different techniques
i.e., implementing the K-means clustering on the prediction data of energy consumption and
the IAP-based SGKRU approach. GWO based CNN-LSTM network uses the K-Means
Clustring technique for classifying the energy consumption prediction data according to
different levels of consumption identified like; Very high (2 ≤ MAX), High (1≤1.99), medium
(0.26≤0.99), and low (0≤0.25) energy consumptin and the clusters of the weather conditions.
After the energy prediction data and weather conditions data, the high energy consumption
device are disabled and low energy consumption devices are being enabled. The propsoed
approach is able to achieve the energy optimization and generating the recommendations for

vii
eneabling 11 and disabling 2 devices while low energy consumptin, 10 enabled, and 3 disabled,
9 enabling and 4 disabling devices, and 8 enabling and 5 disabling devices in the smart home
environment.The detailed discussion of both techniques reveals that IAP-based SGKRU proves
to be the most effective approach for generating recommendations and optimizing device
scheduling in smart home environments. The recommendations generated using this technique
is achieved using IAP based SGKRU which is having the minimum error in the energy
predictions i.e., 0.2668 RMSE and 0.0712 MSE. The obtained error value is less than the error
values obtained using the existing state-of-art.

The proposed techniques are outperforming the existing work as per the results obtained using
different evolution parameters. Collectively, this comprehensive thesis provides a thorough
examination of smart city concepts, integrates advanced technologies for energy optimization,
and proposes effective recommendation systems for optimal device scheduling. The insights
gained from this work, contributes to the achieve the goal of enhancing the efficiency and
sustainability of urban living. The future work section sets the stage for continued exploration
and advancements in the field, ensuring that the research remains dynamic and relevant.

viii