1. Introduction
The application of the technology for the Internet of Things (IoT) to monitor and control industrial systems has received a lot of attention around the world. In a very specific way, IoT systems have been gaining space in what is called intelligent electrical systems or Smart Grids [
1]. The integration of IoT technology in intelligent electricity networks has the fundamental purpose of generating control alternatives to guarantee the efficiency and continuity of electrical energy systems. In the electrical industry, a great interest has arisen to determine the patterns at the domestic level, since with them, it is possible the administration of the energy, as well as the identification of failures in installations, identification of defective and/or obsolete appliances, and the generation of mathematical models related to the energy consumption. On the other hand, there is a strong tendency to incorporate this technology to be more efficient in the use of energy by the utilities. However, it is necessary that the end-users of energy at commercial and domestic levels also become aware and generate initiatives in their facilities to optimize the energy consumption [
2].
In this context, household electricity users have always wondered if their energy consumption is that charged by the utility company. Initially, an option to respond to this concern is to place a kWh meter to validate the official meter reading. However, the answer to this first question originates other more complex questions, such as (1) Is the consumption due to energy leaks?; (2) Are the appliances consuming more energy than usual?; (3) How I can reduce the energy consumption?; and finally, (4) Does the consumption can be predicted? Nowadays, the Smart Home Energy Monitor system (SHEM) can response these questions. The main functions of the SHEM are to continuously read the energy consumption, to identify the appliances to be managed intelligently for energy savings, and to provide visual information of the energy consumed by the user [
3]. The hardware needed for acquiring the signals can vary, but the core of the SHEM is composed by the processing and decision algorithms, and the visualization interfaces.
The SHEM can process the data consumption generated from various metering sockets installed in each of the appliances [
4]. This scheme is known as intrusive load monitoring (ILM), and it permits the SHEM system to be more efficient since each appliance is monitored and controlled individually [
5]. The main application appears when a complete home automation is needed, but it is usually composed by complex and expensive systems [
5]. In counterpart, the non-intrusive load monitoring (NILM) scheme becomes the cheapest and the simplest option since only the measurement of total energy is needed [
6]. The identification of habits of consumption becomes a more complex task, since all the appliances are aggregated to the global measurement and disaggregation methods should be used to identify the appliances in operation. This scheme can co-exist with home automation systems too, allowing the connection/disconnection of appliances. Regardless of the scheme used, the purpose of the SHEM is to generate information that allows the user to take decisions for an efficient energy consumption, as well as for diagnosis and prognosis as remedial actions to prevent large consumption in the future [
7].
In the commercial aspect, a large of manufacturers of SHEM exist, which provide different models of different prices. Most of them have user-friendly monitoring interfaces. For example, some devices, such as the presented in
Table 1, are becoming attractive non-intrusive options for home energy monitoring. The development of new algorithms and visualization interfaces is not possible since their architecture is closed.
In counterpart, there exist devices with an open architecture, which are developed by the company Openenergymonitor
® [
8] that is a pioneer in this field. These devices offer various platforms for consumption monitoring such as emonPI
® or emonTx
®. Moreover, they can be linked through the EmonCMS web application to store and process data in the cloud. Although, the devices presented in
Table 1 have a slightly lower cost than
$ 192.72 US, they do not include the web monitoring system, which must be purchased separately; however, they are an attractive option for researching purposes. Moreover, these devices allow implementing new algorithms, but their application to countries such as Mexico is limited due to their high cost and complex scheme of tariffs.
As previously mentioned, diagnosis and prognosis of consumption are inherent attributes in SHEM. These attributes are focus of research around the world. In this direction, several works have been presented with significant contributions. In this context, the prognosis of consumption from real measurements, for improving the use of heavy loads such heating systems, is studied in [
9], where an Artificial Bee’s Colony algorithm (ABC) is proposed with a new metric called habitual average. The authors use a real database for test the ABC algorithm, and their results prove that the use of this metric as an internal variable helps to employ heating systems as an alternative for real implementation in the SHEM. This work is centered only in selecting intelligent decisions for efficient energy consumption, but functionalities to visualize results can be added to it.
The energy consumption behavior can be studied using industrial meters along with low-cost development boards. In this sense, reference [
10] presents the utilization of an Echelon electricity meter for acquiring signals and a Raspberry board for their processing. The main goal in this work is to investigate the behavior of the consumer using data recovery mechanisms and machine learning methods. More recently, energy prediction has been studied in [
11], where the multi-power state idea is introduced for classification and identification of appliances by a supervised learning. The proposed methodology in [
11] shows a better prediction than that obtained with the binary power state model, and it represents an alternative for real-time home energy management. The work could be used with purposes of visualization too. On the other side, the massive application of IoT for smart homes is presented in [
12], where the authors propose a system that uses the IoT and big data analytics tools for managing a large set of users. In this work, a Photon Particle
® board is used, and the authors present interfaces for the final user and the community owner. The interface to visualize the bill is not fully presented, but the results indicate that this board can easily be implemented on large scale smart home with billing purposes.
Recently, web-based processing systems are more common in forecasting due to the use of smart meter systems. The great variety of open source libraries on artificial intelligence allow to generate new ideas, such is the case of the work presented in [
13], where a hybrid combination of SARIMA and metaheuristic firefly algorithm-based least squares support vector regression (MetaFA-LSSVR) is used for forecasting energy consumption. In addition, smart energy metering systems can be developed with low cost boards such as the presented in [
14], where a LoRa-WiFi protocol is employed for reducing the dependency of internet access. This prototype allows detecting theft or abnormal consumptions through a cellular application. The system can be used for billing purposes, but this reference does not present a complete interface. Finally, the manufacturer industry has used development boards for diagnosis of energy consumption in their facilities [
15]. In this case, they collect the measurement signals from industrial meters of the brand Schneider and Elmeasure, and by using a Raspberry board these signals are preprocessed. Interfaces for visualization of daily consumption patterns are generated for purposes of energy savings. As stated by the analysis of the previous works, note that the visualization of energy consumption of SHEM is a key aspect since the information should be displayed timely and easy interpretation. In function of these aspects the user can understand properly and take actions to save energy [
16,
17,
18].
Motivated to this fact, this paper is focused on the development of an interface for real-time billing in the Mexico country, with the main objective of helping the users to avoid large energy consumptions and to save energy. The real-time bill is generated by a proposed IoT system that has the following characteristics: (1) it uses low-cost sensors and the electronic board Particle® Photon; (2) it displays information about the consumption habits of the users; and (3) it activates alarms in case of abnormal or high consumption. The main contributions of this work are (i) the design of an interface that provides a real-time bill with the same characteristics of the bill generated by the company of energy, (ii) the forecast of energy consumption in a period using only one week of measurement, and (iii) the possibility to extend this measurement system for three-phase commercial applications.
This work is organized as follows.
Section 2 describes how the energy in Mexico is charged, emphasizing mainly in the tariffs employed and the bill used by the company of energy. Additionally, the information presented in the bill is analyzed in detail, focusing on its interpretation and how this information can be correctly interpreted to help the user to save energy. On the other hand,
Section 3 discuss the hardware constructed along with the algorithms and its validation. In
Section 4, the interface designed is presented, and finally,
Section 5 describes the application in a real situation.
4. The Proposed Interface
The interface for displaying the screens previously defined in
Section 2, was designed using three programming environments,
html css, and
js.
Figure 10 depicts a schematic diagram of the libraries employed for the interface designing; these libraries are available at [
34].
Using the libraries previously mentioned, an appropriate interface was created.
Figure 11 shows this interface along with the different screens of the previously defined conditions for each variable of interest. The reader can access through the hyperlink given in [
35].
Figure 11a presents a menu where the user can navigate in the different screens. In the screen of the
Figure 11b, the users can provide their personal data (optional) along with their cellphone and email to receive the information of the bill and the different alerts, as previously discussed in
Section 2. Furthermore, in this screen the period of consumption can be selected. In the
Figure 11c, a summary of the bill is displayed in real-time, which is indicated with the horizontal bars of the energy consumption for the different blocks according to the tariff selected. The behavior consumption through the days of the week using gradient colored bars is shown in
Figure 11d. This gradient allows the user to determine if the consumption per day is higher than a limit previously established.
Figure 11e uses the same idea but for the consumption by period. The forecasting of the energy is presented in the screen of
Figure 11f, where consumption is obtained by extrapolation of the mean consumption of the previous days. The user can view announcements in real time as that showed in the screen of the
Figure 11g, where a set of icons that change of color permits determining the level of consumption that has been reached. They also indicate if the user has a correct installation, or if the installation of PV systems needs a revision. Finally, the screen of the
Figure 11h shows the information generated in the official printed bill version.
5. Experimental Results and Discussion
Finally, the developed system was applied in a real situation. It was mounted in a domestic installation in the state of Colima in Mexico during the period of August 19 to October 19 of this year.
Figure 12 shows the image of the conventional bill, and the different screens of the interface show the measurement for this period, as well as the bill forecast obtained in the first week of measurement, and the announcement generated.
Figure 12 informs that the proposed system gives a measurement of 561 kWh (
Figure 12b) vs. 579 kWh reported by the company (
Figure 12a), this represents a minor difference of 18 kWh. The difference observed in the amount is attributed mainly to the block ranges considered in this study. Note that block B is two times higher than that used in the interface. On the other hand, the block IL is lower than used in the interface by 25 kWh, thus the significant difference for block S (129 vs. 211 kWh). According to the position of the pointer on the heat bar color, the householder under study is in the range of moderate high consumption since the Max allowed is 800 kWh.
Daily consumption is shown in
Figure 12c, and it represents an attractive alternative for the users in the evaluation of their consumption day by day. The gradient colored bars help to rapidly determine if the consumption in a determined day if near to the mean allowed. In this case, all days present a similar level of consumption due to the actual pandemic situation of COVID-19, that has caused that people stay at home most of the time.
The forecast of the total energy consumption is shown in
Figure 12d, and it was obtained with the first week of measurement. From this Figure, it is be concluded that the implemented algorithm estimates properly the final reading with a minor difference of 36 kWh (just 0.6 kWh per day), as appreciated in
Figure 12b. Finally, this forecast was used for diagnosis of the installation, and the
Figure 12e indicates in a friendly manner that the facilities are working properly, and changing the technology of lightning or the inverter-based fridges can reduce the energy consumption.