Characterization of the Electrical Consumption Pattern of Household Appliances for Home Energy Management Using High-Resolution Measurement Techniques in IoT Environments

Abstract

For future smart cities, smart homes will be required. The key elements are the smart use of energy and smart communication systems that are connected to homes. Along with this, the devices inside the house will need to be monitored and managed efficiently. One of the current proposals is the use of Home Energy Management Systems (HEMS) allowing to solve problems associated with efficient management, the economy of electrical energy, and failures/alarms regarding the operation and safety of appliances. This work proposes a model for the recognition of patterns of energy consumption in household appliances, based on the capture of electrical parameters through Smart Socket, using an intrusive method in the electric charge. The data acquisition system corresponds to an IoT platform that uses automatic meter reading elements, which, connected via Wi-Fi, send data to a cloud service. The results obtained allow a characterization of household appliance consumption profiles, with high levels of reliability and under multiple operating states. Because of the foregoing, the detection, monitoring, and control of household appliances connected to the electrical network allow the reduction of both household billing and CO2 emissions.

1. Introduction

The climate change that the planet is undergoing rapidly has led each country to propose strategies that lead to efficient use of energy resources. The Chilean state has set an energy policy for the year 2050, which establishes energy efficiency as one of its pillars. In this context, the residential consumer has taken an active role in managing their use of electricity and thereby influencing the behavior of their consumption pattern. This recognition facilitates efficient use both for the residential consumer and for the electricity distribution company.

For the distribution company, knowing the behavior profile allows the use of its infrastructure capacity to be maximized, obtaining a reduction in emissions and contributing to sustainable development. On the other hand, for the residential consumer, it allows the responsible use of energy, obtaining a reduction in their monthly billing.

Currently, the development of Management Systems (Energy Management System EMS) to achieve energy efficiency in areas with a high concentration of public, business, and residential loads, is considered a problem of interest in Smart Grids. The use of the Home Energy Management System (HEMS) contributes to managing demand, especially during peak consumption hours, through Demand Response DR programs. Thus, users will be able to participate in the control of demand in real-time, adapting consumption in such a way that demand can be reduced during peak hours and intelligently adapt to times of low activity.

In this article, works related to this research are discussed, mainly on Non-intrusive Automatic Meters (automatic consumption meters) and Intrusive (smart outlets). For this, a low-consumption, high-precision, and low-cost smart socket were designed (15–20 USD). Then, a campaign to measure the electrical consumption of various appliances was carried out in a series of homes for 90 days. With the data obtained, a deep neural network was implemented that allowed the characterization of the device’s consumption profile and its differentiation from other devices simultaneously connected to the electrical network.

The rest of the document is organized as follows: Section II presents the works related to this research; Section III describes the methodology used and development of the proposed system; Section IV exposes the experimental development, the results obtained and their discussion; Section V shows the conclusions of the article; and finally, section VI presents future work.

2. Related work

In this section, some works related to the topic of efficient energy management in a Smart Home – Smart City environment are shown and discussed. The main differences between proposed measurement devices and their relationships, their sampling speed, type of operation, and resolution in the temporal measurement are highlighted.

The state of art shows different alternatives for obtaining the “consumer fingerprint” of electrical devices, an important element that motivates the development of this work, in such a way as to be able to propose an intrusive Smart Socket (SS) device that reaches improvements in the efficiency of device profile recognition and better use of home energy [1]. HEMS use monitoring techniques that are generally classified as Non-Intrusive Load Monitoring (NILM) and Intrusive Load Monitoring (ILM) [2]. The characteristics that differentiate these two techniques is that NILM only uses a monitoring point that is generally located at the entrance of the main power supply of the home electrical system (known as Smart Meter), while ILM uses sensor devices in each of the outlets (Smart Socket; Smart Plug SP) of all or main connected loads [3].

Generally, the data captured from the consumption of the devices is sent over the Internet and stored on a platform available in the cloud. These data mostly correspond to the active power demanded by the artifacts [4], which are processed in a Matlab® environment applying, for example, the Multi-Layer Back-Propagation Neuronal Network (MLBPNN) algorithm that allows detecting the real power profile. Consumed by each electrical device [5], so then the concept of digital footprint of device consumption [6] is proposed. On the other hand, NILM systems have the advantage of not interfering with the home circuit with expensive monitoring devices [7]. On the other hand, ILM systems, by having sensors in different loads of the circuit, provide more precise details about the consumption and alarms referring to each appliance [8], being able to make it possible to monitor very low-power devices and differentiate between variable consumption and consumption among a set of devices [9].

An SS sensor network is designed for use in work and home environments as a device, which is capable of identifying the appliance based on the behavior of the current patterns consumed by the electrical device [10]. The nodes communicate with each other using the environment’s existing Wi-Fi, PLC, or ZigBee wireless architecture. Seamless sensor network integration is important for success in the realm of ubiquitous computing. The hardware and software architectures of the systems that analyze test devices are discussed and the consumption patterns and current profiles of home appliances are explained [11]. In this way, a possible hypothesis of this behavior is presented and the use of the classifier algorithm that does not need training, but only a dynamically updateable database, is explained, thus creating the need for a cloud database connected to all home endpoints. The system implementations and the description of the protocols developed for the device controls are also compared in the works studied [12]. Artifact pattern identification focuses on energy disaggregation and device recognition.

Refs. [13, 14] show the development of a smart system that analyzes the use of devices to extract user behavior patterns in a smart home environment. On the other hand, in [14, 15], it seeks to optimize the cost of electricity, where users can receive a warning of excessive consumption of devices connected to the home electrical system.

In [16, 17] the authors adopt a distributed metering system of smart outlets, designing an algorithm based on ANN (Artificial Neural Network) that exploits low-frequency measurement data with power consumption every two minutes. The works [18, 19, 20] use an algorithm based on fuzzy logic that, by activating the calculation of parameters such as maximum power, average power, and cycle duration, obtains a result for the membership function, achieving the identification of the artifact by degrees of truth. However, the research works related to the automatic recognition of electrical devices described by [21, 22], carry out the sampling in the order of 15 to 10 minutes, since their objective is to determine the energy consumption according to the systems of billing.

We can also mention that the scientific publications on NILM approaches outnumber those of ILM [23], since the NILM approach proves to be much older than Mechanical Gauges that are almost a century of use [24, 25]. On the other hand, since the NILM (Smart Meter with prices of 300 to 400 USD) are based on the use of a single sensor, the installation of the detection system is simple and the data acquisition is of lower quality since we can only monitor general consumption in the home [26, 27]. Device recognition using this type of approach generally suffers from precision due to the inherent problem of summation of consumption signals from different devices [28, 29]. On the other hand, with the ILM meter (SS, SP, price between 20 and 30 USD) the measurement is based on the use of multiple sensors inside the home, having a higher level of detail and resolution in the consumption data, which facilitates the identification of the electrical appliance [30].

In this work, the proposal of the Smart Socket ILM System, to monitor, detect and recognize the main household loads, using automatic measurement (Automatic Meter Reading – AMR) achieves a sampling rate of the order of 1/20 Hz, higher than the works [28, 29, 30] and with a resolution that allows not only the differentiation of each electrical appliance but also the activation of alarms, switching on and off in case of excess current (>16 Amps) or over-voltage (>230 Volts), enabling remote reconnection by the user once the alarm has been reviewed. Finally, the machine learning algorithm manages to discriminate each of the devices with great efficiency with a reduced calculation speed, sending information in real-time to the user and determining the electrical consumption of the home and of each of its devices.

3. Methodology and system development

Figure 1 shows the schematic diagram that briefly describes each of the stages of the methodology implemented in this work.

1. Phase (A):

   A current/voltage measurement module with wireless communication via Wi-Fi using MQTT messaging protocols is designed. This data transmission is carried out over TLS 1.3 (Transport Security Layer) to ensure the privacy of the data on the channel. Maximum value ranges are established for voltage and current. Finally, an overload protection system is configured.

2. Phase (B):

   Measurements of the 22 most common household appliances are recorded, measured in 20 different homes, with a sampling time of 10 seconds and with work cycles that depend on the operation of the appliance.

3. Phase (C):

   It envisions implementing a neural network with a hierarchical framework of synapses in a forward topology, where all nodes in a layer are linked in a unidirectional fashion to identify nonlinear features.

4. Phase (D):

   The data is sent to the cloud to perform the necessary operations to generate representative values for each of the electrical devices to be compared.

5. Phase (E):

   A consumption fingerprint matrix identifier model is created for each of the devices, generating a consumption map of the electrical devices connected in the home. Submatrices are created for testing the errors found and model validation.

Figure 1.

Description of the methodology used to identify consumption patterns of electronic devices in a home.

In the context of the development of the ITCity ELAC T10–0643–Eranet 2017–2020 Conicyt Project, “An ICT platform for sustainable energy ecosystem in Smart cities,” executed by the institutions Federal University of Santa Caterina Brazil, Institute of Energy Physics of Latvia, University of Bucharest, University of Concepción, University of Atacama and Universidad Tecnológica Metropolitana, a system was designed that allows the acquisition of data from a network of intelligent measurement equipment in each outlet, called Smart Sockets (SS). This design corresponds to a Current–Voltage measurement module with wireless communication as an intelligent complement with a maximum capacity of 15 (A) (Figure 2).

Figure 2.

Architecture of a smart socket.

The proposed model has the flexibility of portability of the device to different power outlets within the deployed coverage environment and to be able to change the identity of the device in the user interface from a Smartphone. The SS includes a communication interface, a current-voltage sensor, an MCU (microcontroller unit), and a switching circuit.

The functionality of these SS is the following:

• Measure the instantaneous voltage and current that are present in the connected device.

• Calculate the instantaneous power demanded by the connected device.

• Calculate the power consumed by the connected device.

• Execute communication request algorithm to transmit power consumption and make power on/off decision by SS Device Coordinator.

As shown in Figure 2, consumption measurement is obtained through a current sensor and a voltage sensor, in addition to obtaining the difference between the zero crossings of these two signals and correctly calculating the powers required by the loads (active, reactive, and apparent). Obtaining the current signal is achieved through a non-invasive sensor and the detection of the voltage signal is achieved through a transformer.

Figure 3 describes the architecture of the communication between the devices and the cloud. The devices are connected through a local Wi-Fi network and publish their measurements using the MQTT protocol, identifying each one of the devices through a unique identifier. These messages are sent to a cloud server that acts as an MQTT Broker and forwards the messages to the subscribers. The subscribers store the measurements persistently and process the results, which are exposed through a data analytics platform.

Figure 3.

Smart socket ILM system.

The development of this device allows both the ON/OFF control and the monitoring of the different electrical parameters (voltage, current, active and reactive power, power factor, and frequency, among others) [8]. The SS are connected via Wi-Fi/IEEE 802.11 to the Access Point (AP), which works as a Gateway-IoT that oversees sending the information to the cloud service through communication protocols offered by the Internet Service Provider (Internet Service Provider ISP). The data is stored in a data streaming architecture and analyzed through a web visualization service, which allows its extraction and use for training and testing the neural network.

Given the need for a lightweight and easy-to-implement algorithm for a cheap integrated circuit with low power consumption, it was decided to use a multilayer perception neural network with a recursive propagation algorithm, called MLBPNN [10]. This is a known supervised model, used in this work due to its simplicity and guaranteed convergence. These characteristics allow its future implementation incorporated into a technological package quickly and at a low cost. In this work, the general network of the multilayer perceptron provides a machine-learning algorithm with the device classification task. The analytical model is a hetero-associative network (multilayer perceptron, MLP), where all the inputs connected to the external environment are different from the output that expresses the response of the system. For this work, an active power curve vector of the electrical load is considered as input, and the recognition of the connected appliance is considered as output.

The network elements combine a hierarchical framework of synapses in a forward topology, where all nodes in one layer are unidirectionally linked to those in the next layer, to create the possibility of identifying non-linear features. The ability to detect non-linear relationships, without defining a formal expression that does not require an “a priori” hypothesis about the behavior of the variables, essentially depends on the number of nodes, the number of layers, the transfer function of each node, and the connection weight factors.

Eq. (1) represents the optimization function to find the minimum error between the consumption pattern and the data measured by the SS:

$$\mathit{min}\sum _{p=1}^{\infty }{E}_p\left(x\right)$$E1

where $E_p$ represents a measure of the error related to pattern p (subset) of the training set. This error estimates the gap between the output given in the training set and the output predicted by the network. The recursive propagation algorithm is an iterative method, a heuristic version of the gradient method, commonly applied in multilayer networks due to its high performance in terms of time and precision. The interaction that defines the recursive propagation is given by Eq. 2:

$$w^{k+1}=w^k-a\nabla E_{p\left(k\right)}\left(w^k\right)+\eta \left(w^k-w^{k-1}\right)$$E2

where

• $E_{p\left(k\right)}\left(w^k\right)$ is the gradient of the error function in the current vector $w^k$ of weights.

• $a$ (learning rate), scalar that defines the step along the anti-gradient direction $d^k=\nabla E\left(w^k\right)$, using in each step only the current pattern of inputs and outputs $X^{p\left(k\right)},Y^{p\left(k\right)}$.

• The scalar $\eta$ > 0 (momentum), performs an adaptive step choice or modifies the direction of the investigation to ensure algorithm convergence.

Continuous training of a neural network aims to make it always perform better, but eventually, it reaches a point where forward progress is too slow to be practical. Also, overtraining is harmless, because it can lead to overfitting. This occurs when the mapping function resulting from the training process fits the consumption pattern too well, losing the ability to process new data.

4. Experimental development

Next, the steps executed for the experimental development are highlighted:

4.1 Data set

For this study, a method of multivariate statistics and computational intelligence is designed and implemented, with the objective of identifying electrical appliances connected to the residential electrical network. For this experiment, the proposed system was first installed in 20 homes in Santiago de Chile. Each home has ILM measurement and communication equipment, installed in each electrical outlet in the home. The electrical profiles of household appliances were built and studied based on data from multiple brands and models present in the homes of collaborators in this research.

To obtain electrical consumption data, the operation was measured through an intrusive charging system by means of a low-resolution AMR, which recorded the behavior of the appliance in terms of active power consumed in multiple operating conditions (low, low, medium, and high-performance). Between 50 and 80 samples of each of the variables were obtained for each measured device: kettle, refrigerator, electric oven, microwave, heater, washing machine, and electric grill, among others. The intrusive charging system recorded a sample of electrical consumption measurement in active power, every 10 seconds, and the behavior was measured during 5 minutes of operation, that is, 10 (s), 40 (s), 80 (s), until 300 (s). However, only the first 220 (s) in the ANN training model dataset were used, thus building a data matrix of 11 features, where each one represents an input for the ANN model. It is important to highlight that measurements were made in 20 homes, during 90 days between July 29 and October 30, the spring season. However, a log of 300 (s) is enough to train and recognize artifact profiles. These data were matrix preprocessed but did not require filtering due to the low presence of additive noise.

The measurements are made in 22 devices, of which 9 are shown in the final results of this article, and 6 of these, the most typically used in homes, are used to carry out the ANN training. Figure 4 shows the characteristic behavior of each of the 6 devices based on the active power demanded during 3 minutes of operation.

Figure 4.

Consumption profile of each device.

Although the energy behavior in a global way tends to be complex to visualize, if the curves are separated intuitively, it can be seen that some devices have similar behaviors in relation to the power demanded, as is the case of the kettle, the heater, and the electric oven (Figure 5).

Figure 5.

Active power demanded in an electric kettle, heater, and oven.

These three electrical appliances incorporate resistive elements in their electrical components, which basically describe their main function of generating heat through electrical energy. However, once a maximum temperature is reached, the kettle stops working, unlike other appliances that work under temperature ranges, activating and deactivating its operation for a period.

On the other hand, the consumption curve of a refrigerator, as shown in Figure 6, is very characteristic and is not like any other curve studied.

Figure 6.

Active power profile in a refrigerator.

These elements that are capable of being visualized and differentiated intuitively are the ones that are sought to be measured and parameterized through the use of machine learning (ML) techniques, to detect and predict the electrical artifact measured according to its electrical behavior as a function of time. of use.

Figure 7 shows the behaviors recorded for the microwave and washing machine equipment during the time intervals. For both, a similar work cycle is observed and that generates some complications in their identification.

Figure 7.

Microwave and washing machine active power consumption pattern.

4.2 Security and architecture

Any device and architecture that is exposed to the Internet has a potential risk of intrusion and/or data theft. In the architecture designed for this experiment, the devices send telemetry information to a cloud server, which processes the data, analyzes it, and publishes it on an analytical platform. There are different intrusion scenarios that were considered for the construction of the architecture:

• Someone can communicate and/or have access to the sensor to hack it.

• Someone can impersonate the server and receive the telemetry data from the sensors.

• Someone can impersonate a sensor and send malicious information to a real server.

• There is unauthorized access to the analytics platform.

• Service interruptions and reporting (loss of energy, loss of connection, and unavailability of cloud services).

As a first measure of control and related to the world of IOT, it is necessary that all communications are secure and encrypted at all times, from the Smart Socket report to the publication of analytics. The devices use a Wi-Fi connection and are configured to connect to a LAN using WPA2 encryption. Devices communicate with the cloud through MQTT, this protocol is designed for networks with limited resources and restricted bandwidth. It is also possible to select a QoS (quality of service) level to ensure data transmission. To secure the transmission channel, MQTTS over TLS1.3 is used. The devices are responsible for publishing the telemetry to a previously configured domain.

On the other hand, the cloud contains a firewall service that filters all connections that come from ports other than those defined in the devices (8883). To subscribe to the devices, an MQTT Broker is used, which receives the telemetry and sends it to the persistent storage service. This service stores the data using a time series database, which makes subsequent analysis and consultation more efficient. All this infrastructure is separated from any other service and contains security measures such as Intrusion detection systems (IDS) and intrusion prevention systems (IPS).

4.3 Machine learning process

Once the electrical consumption data of the household appliances studied has been obtained, a data and information matrix is generated that was used to train the model.

This stage is systematized as shown in Figure 8. The first step is to prepare the data to be used. For this, the erroneous data are eliminated, the characteristics that will be used are studied and chosen, and the factors that are not influential in the predictive results are also dismissed.

Figure 8.

Data processing methodology for ANN training and testing.

This matrix is divided into sub-matrices to be used in the training stages of the neural network, in the testing and subsequent validation of the model.

The second stage consists of preparing and parameterizing the model. In the case of the neural network used, it was generated in Matlab® with an improved ANN script, which searches and finds the best neural network based on the prediction results obtained, varying the network parameters such as the number of hidden layers, hidden layer neurons, learning rate, and percentage of data in the training, testing, and validation submatrices [11, 12]. This allows the results obtained from the simulation to be the best and therefore define the structure of the model.

Evaluating the generated model is part of stage three, in which the method used to train the neural network must be tested with a new data set. This evaluation must be made from measurements of new artifacts, not used in the previous training, in such a way that they can be compared, thus carrying out a broader evaluation and, in addition, knowing how accurate the proposed model can be.

4.4 Results

Using the SS, the records of 22 different artifacts are obtained, measured with a resolution of 10 seconds. Of the 9 home appliances, their respective active power consumption patterns are presented (Figures 9–15):

Figure 9.

Active power consumption pattern of a coffee maker. Measurement time in 3-hour intervals.

Figure 10.

Active power consumption pattern of a heater. Measurement time in 48-hour intervals.

Figure 11.

Active power consumption pattern of an electric grill. Measurement time in 12-hour intervals.

Figure 12.

Active power consumption pattern of a microwave. Measurement time in 9-hour intervals.

Figure 13.

Active power consumption pattern of an electric oven. Measurement time in intervals of 30 min.

Figure 14.

Active power consumption pattern of a washing machine. Measurement time in 60 min intervals.

Figure 15.

Active power consumption pattern of a vacuum cleaner. Measurement time in 5-hour intervals.

Models based on Machine Learning techniques were executed with the RapidMiner and Matlab® Software. Using Matlab®, the ANN algorithm was executed, programmed to find the best version and configuration of the model’s hyperparameters. The ANN algorithm is applied to 6 different household appliances with 91.7% efficiency in global terms, as shown in Table 1.

	Kettle	Fridge	Electric oven	Microwave	Heater	Washer
Kettle	86.6%	1.6%	0%	4.2%	8.5%	0%
Fridge	0%	98.4%	1.9%	1.4%	0%	0%
Electric oven	6.7%	0%	94.4%	5.6%	0%	0%
Microwave	0%	0%	0%	83.3%	0%	3.3%
Heater	6.7%	0%	3.7%	1.4%	91.5%	0%
Washer	0%	0%	0%	4.1%	0%	96.7%
Accuracy	86.6%	98.4%	94.4%	83.3%	91.5%	96.7%
Error	13.4%	1.6%	5.6%	16.7%	8.5%	3.3%

Table 1.

Device prediction matrix.

The electric kettle, heater, and oven have resistive components and tend to get confused. The model makes erroneous predictions, as was intuited from the beginning by reviewing the results from Figures 4–7. The operation of these three artifacts is sometimes very similar and the algorithm is not capable of effectively separating the data according to their characteristics. of energy consumption.

In Table 1, each column delivers the recognition results with respect to each artifact defined in each row. For example, the column of the kettle device indicates that the proposed model recognizes its own device with 86.6% accuracy, assimilates it with an electric oven, and with a heater with 6.7%.

The procedure for the rest of the columns is like the one shown. Each column adds a total of 100% of the results obtained. The row sums have no meaning.

The refrigerator, on the other hand, is the artifact with the highest probability of being effectively detected at 98.4%. On the other hand, the microwave and the washing machine are devices that the algorithm tends to confuse since they maintain somewhat similar operating cycles.

To increase the performance of the algorithm, it is proposed to isolate the problem into subcategories, the data set is grouped using clustering techniques, this hypothesis could be validated, since the data from the kettle, heater, and oven measurements were isolated electrical to generate a new dataset with only these targets.

The result of this test is shown in Table 2, where the precision of this algorithm with the previous cluster technique improves the results for the kettle going from 86.6% to 95.1%, in the heater it increases from 91.5% to 95.7% and in the electric oven it changes from 94.4% to 98.1%. These results are better than those reported by [6].

	Kettle	Electric oven	Heater
Kettle	95.1%	1.9%	4.3%
Electric oven	0%	98.1%	0%
Heater	4.9%	0%	95.7%
Accuracy	95.1%	98.1%	95.7%
Error	4.9%	1.9%	4.3%

Table 2.

Comparative prediction matrix between kettle, electric oven, and heater; cluster 1.

The same occurs in the case of cluster 2, which groups microwaves and washing machines, where the performance of the algorithm substantially improves when artifacts are isolated according to similarities, going from 83.3% to 96.7–100% identification, respectively. The results are shown in Table 3.

	Microwave	Washer
Microwave	100%	0%
Washer	0%	100%
Accuracy	100%	100%
Error	0%	0%

Table 3.

Microwave and washing machine; cluster 2.

5. Conclusions

The proposed research makes it possible to recognize the characteristic profiles of a set of household appliances with accuracies ranging from a minimum of 83.3% for the microwave to a maximum of 98.4% for the refrigerator. The incorporation of a cluster stage to the proposed methodology allows for increasing the level of precision, with a minimum of 95.1% for the kettle and 100% for the microwave and washing machine. The difference in the levels of precision between the artifacts is mainly due to the presence of various types of electrical charges: linear, non-linear, constant over time, and variable over time.

From the results obtained, it is concluded that it is possible to predict consumption profiles in household appliances, with high levels of certainty, high sampling speed, and under multiple operating states.

Among the limitations of the research, we can mention the need to validate the proposal on a large scale and the use of a basic neural network that required more than one clustering for early differentiation.

These results encourage the development of solutions that, through the automatic recognition of electrical devices, will allow progress in self-management and the promotion of citizen participation in energy efficiency.

6. Future work

6.1 Appliance recognition

It is expected to develop a recognition model for household appliances where it is possible to compare and/or use different types of machine learning tools such as Support Vector Machines (SVM), k-Nearest Neighbors (kNN), Deep Learning (DL), and Principal Components Analysis (PCA), among others [31, 32, 33] (Figure 16).

Figure 16.

Machine learning techniques for the recognition of consumption patterns.

This work would allow the identification of said household appliances in real-time, depending on the target group being analyzed: homes, neighborhoods, residential buildings, and industrial buildings, among others. However, above all things, it would help to optimize the computing capacity necessary to carry out said work [34].

6.2 Optimization methods for HEMS

Incrementally and in relation to the concept of Smart Grid and Smart Cities, it is expected to inquire about the optimization of a home energy management system (Home Energy Management System – HEMS). On this subject, the state of the art is much more advanced, where it is feasible to find publications related to various types of techniques: Artificial Intelligence, Conventional Methods, and Metaheuristics, among others. With quite ambitious comparisons [35]. Mainly with metaheuristic techniques such as Evolutionary Algorithms or Swarm Intelligence, which are the most used for this type of problem (Figure 17).

Figure 17.

Optimization methods for HEMS.

6.3 Consumption forecast

It is expected to design an AI algorithm that allows predicting the energy consumption of the home/company/grid system as consumption increases over time and complement the consumption control system with this information (Figure 18).

Figure 18.

Consumption prediction techniques.

6.4 Smart socket design

The current work operates under the assumptions already described above. Based on this, it is expected to make the following improvements at the SS design level in order to meet the following objectives (Figure 19):

• That a single SS can work with more than one appliance.

• That the communication of the SS does not depend solely on the presence of Wi-Fi in the home.

• That the SS report their data to a central, which will have communication with the user and the cloud service.

Figure 19.

Future ILM smart socket system.

Acknowledgments

This research has been supported by the project “An ICT platform for sustainable energy ecosystem in smart cities” (ELAC2015/T10-0643) and Red CYTED Ciudades Inteligentes Totalmente Integrables Eficientes y Sostenibles (CITIES-2018).

Thanks to the Programa de Investigación en Radiocomunicación Digital PIRD – Universidad Tecnológica Metropolitana, our alma mater, and our loved ones, for all the support provided during this research.