Demand-Side Management for Improvement of the Power Quality in Smart Homes Using Non-Intrusive Identiﬁcation of Appliance Usage Patterns with the True Power Factor

: The proliferation of low-power consumer electronic appliances (LPCEAs) is on the rise in smart homes in order to save energy. On the ﬂip side, the current harmonics induced due to these LPCEAs pollute low-voltage distribution systems’ (LVDSs’) supplies, leading to a poor power factor (PF). Further, the energy meters in an LVDS do not measure both the total harmonic distortion (THD) of the current and the PF, resulting in inaccurate billing for energy consumption. In addition, this impacts the useful lifetime of LPCEAs. A PF that takes the harmonic distortion into account is called the true power factor (TPF). It is imperative to measure it accurately. This article measures the TPF using a four-term minimal sidelobe cosine-windowed enhanced dual-spectrum line interpolated Fast Fourier Transform (FFT). The proposed method was used to measure the TPF with a National Instruments cRIO-9082 real-time (RT) system, and four different LPCEAs in a smart home were considered. The RT results exhibited that the TPF uniquely identiﬁed each usage pattern of the LPCEAs and could use them to improve the TPF by suggesting an alternative usage pattern to the consumer. A positive response behavior on the part of the consumer that is in their interest can improve the power quality in a demand-side management application.


Introduction
Clean and affordable energy-Goal no. 7 of United Nations' Sustainable Development Goals (SDGs)-is targeted for achievement by 2030. Demand-side management (DSM) plays a vital role in accomplishing this goal [1]. Clean energy is normally interpreted as green energy that is generated with renewable non-fossil fuels to reduce climate pollution. Interestingly, in its policies, the Government of India has articulated very well that both green energy and electrical power quality (PQ) are essential and need to be in balance without any ambiguity [2,3]. In this article, the authors consider the latter part of the interpretation of clean electrical energy.
(LED) lamps, have been deployed on a large scale by all nations across the globe, including India. This has led to indiscriminatory usage of consumer electronic appliances (smart TVs, smart phones, smart fridges, etc.) and computers with SMPS. This pervasive usage of low-power consumer electronic appliances (LPCEAs) is defeating the very purpose of SDG 7-clean energy with quality electric power. The traditional approach to DSM is primarily focused on energy conservation, and improvements in the true power factor (TPF) could effectively be achieved by way of a human-in-the-loop DSM [4]. The purpose of this paper is to ensure affordable power and to ensure that LPCEAs do not cause issues of poor PQ in low-voltage distribution system (LVDSs).
The PQ in an LVDS is multi-dimensional and includes both the current harmonics and the power factor (PF). The ill effects of current harmonics have been discussed in detail [5][6][7]. While current harmonics themselves are not healthy, they also result in a poor PF due to the bidirectional exchange of reactive power between the source and the load. The current harmonics and the poor PF negatively affect the accuracy of electricity meters and the PQ of the distribution system's supply [8][9][10]. The existing smart energy meters in our distribution systems are not designed to measure the TPF and reactive power; hence, they are not billed to the consumers. So, the need to bill consumers for poor PQ caused by their behavior is not recognized, and hence, the requirement of compensation of reactive power has gone unnoticed [11][12][13]. Therefore, the authors feel strongly that the DSM for the PQ is the need of the hour, and serious attention is required in order to measure and address issues of poor PQ in LVDSs and to meet SDG 7-affordable, clean, and quality energy for all.

Non-Intrusive Monitoring of the Usage of Appliances
While homes use several electric and electronic appliances, the energy consumption is measured at a single point of common coupling (PCC) of the supply mains from the distribution system. Several techniques have been explored by various researchers since G.W. Hart explored the idea of the disaggregation of electric loads and identified the usage of individual appliances in order to prompt responsible energy consumption and healthy consumer behavior [14]. A cooperative response from consumers to the insights from this critical objective analysis will help in achieving the energy conservation targets set by the United Nations. Hart's work was performed at a time when energy availability was not in abundance. With the advent of energy-efficient smart appliances, the average energy consumption has come down significantly. Yang H. et al. articulated in detail that the DSM in smart homes is focused on the PQ issues introduced by smart appliances, including power savers, in addition to energy conservation [15]. Due to harmonics, the displacement PF is not equal to the TPF. Because household appliances are considered to have low-power consumer electronic loads, their effect on the PQ will be significant when too many of them are in use [10]. The effects of harmonics on the displacement PF and %THD are not additive, so it is not right to ignore the effects of harmonic distortion when metering energy for both consumers and utilities [13].
A typical home uses multiple appliances at different times and for different durations, thus forming different load patterns; their usage is uniform neither over a day nor over a season. As a household might have numerous load points, there are unknown loads that could be connected in open sockets, and several load patterns could possibly reach an unwieldy number (factorial N; N is the number of loads, and some of the loads are unknown), but in practice, they would usually follow a very small number of appliance usage patterns. It is not common for a modern household to use single loads at any point in time [5]. Therefore, it is clear that the focus of DSM should be on appliance usage patterns (a combination of loads), and the usage of individual loads becomes irrelevant. The recommendation to the consumer should be to switch from an appliance usage pattern with a low PQ to an appliance usage pattern with a better PQ, thereby ensuring clean power in the LVDS.

Feature Selection for NILM
The harmonic interactions among several nonlinear appliances change the PQ indexes significantly. For instance, the PF, distortion factor (DF), and current THD are directly impacted; these indexes influence the billing of the electrical energy depending on the appliance usage patterns. Therefore, from the perspective of PQ, a combination of appliances that improve the PQ can determine the amount to be billed. Two different experiences were cited: One suggested the use of fundamental signals (displacement factor) to calculate the surplus of reactive power in electrical grids. Another suggested the use of the TPF. This fact shows that this work is important for the industry in the discussion of the problems caused by nonlinear appliances on electrical energy billing systems in order to rectify PQ issues [12].
R. Gopinath et al. researched the development of robust NILM techniques for the effective management of the energy of appliances in order to support reliable and sustainable DSM [16]. A significant improvement in the system efficiency can be realized if the improvement of the PF and the elimination of harmonics can be applied in the whole network [17].
Therefore, the authors proposed the use of aggregated measures of the PQ, such as the percentage of the total harmonic distortion (THD) and the TPF, in order to understand the appliance usage patterns and for consumers to adapt to PQ-sensitive behaviors. Devarapalli et al. discussed how the percentage THD can be applied in order to disaggregate load patterns and to suggest appropriate load patterns, thus reducing the harmonic pollution produced in the system [5].
This article presents the use of the TPF as a unique feature for identifying appliance usage patterns and for helping consumers by providing them with insights in order to switch from low-PF appliance usage patterns to high-PF appliance usage patterns and to become responsible consumers. The schematic of the non-intrusive identification of appliance usage patterns in smart homes with the TPF is depicted in Figure 1.

Feature Selection for NILM
The harmonic interactions among several nonlinear appliances change the PQ indexes significantly. For instance, the PF, distortion factor (DF), and current THD are directly impacted; these indexes influence the billing of the electrical energy depending on the appliance usage patterns. Therefore, from the perspective of PQ, a combination of appliances that improve the PQ can determine the amount to be billed. Two different experiences were cited: One suggested the use of fundamental signals (displacement factor) to calculate the surplus of reactive power in electrical grids. Another suggested the use of the TPF. This fact shows that this work is important for the industry in the discussion of the problems caused by nonlinear appliances on electrical energy billing systems in order to rectify PQ issues [12].
R. Gopinath et al. researched the development of robust NILM techniques for the effective management of the energy of appliances in order to support reliable and sustainable DSM [16]. A significant improvement in the system efficiency can be realized if the improvement of the PF and the elimination of harmonics can be applied in the whole network [17].
Therefore, the authors proposed the use of aggregated measures of the PQ, such as the percentage of the total harmonic distortion (THD) and the TPF, in order to understand the appliance usage patterns and for consumers to adapt to PQ-sensitive behaviors. Devarapalli et al. discussed how the percentage THD can be applied in order to disaggregate load patterns and to suggest appropriate load patterns, thus reducing the harmonic pollution produced in the system [5].
This article presents the use of the TPF as a unique feature for identifying appliance usage patterns and for helping consumers by providing them with insights in order to switch from low-PF appliance usage patterns to high-PF appliance usage patterns and to become responsible consumers. The schematic of the non-intrusive identification of appliance usage patterns in smart homes with the TPF is depicted in Figure 1.

The Major Contributions and Organization of the Article
• TPF measurement of real-world LPCEAs as per the IEEE 1459-2010 standard [18] by using the four-term minimal sidelobe cosine window (4MSCW)-based enhanced dual-spectrum line interpolated FFT (EDSLIFFT).

•
Development of a virtual instrumentation-based measurement system for TPF measurements.

•
Recommendation of appliance patterns for improvement of the TPF and DSM by using a lookup table in order to improve the utilization indexes of consumers.
This paper is organized into five sections; Section 2 describes the TPF measurement method. Section 3 elaborates on the real-time measurement of the TPF with the 4MCSWbased EDSLIFFT in the NI-cRIO (compact reconfigurable input-output)-based virtual instrumentation environment for various combinations of LPCEAs. Section 4 deliberates on   Recommendation of appliance patterns for improvement of the TPF and DSM by using a lookup table in order to improve the utilization indexes of consumers.
This paper is organized into five sections; Section 2 describes the TPF measurement method. Section 3 elaborates on the real-time measurement of the TPF with the 4MCSWbased EDSLIFFT in the NI-cRIO (compact reconfigurable input-output)-based virtual instrumentation environment for various combinations of LPCEAs. Section 4 deliberates on the results and demonstrates that the TPF is a reliable single feature for identifying appliances' consumption patterns. Section 5 discusses the insights of the results, and Section 6 concludes with a summary of the proposed research work.

Measurement of the TPF Using the Four-Term MSCW-Based EDSLIFFT
In this section, the measurement of the TPF by using the 4-term MSCW-based ED-SLIFFT. The 4-term MSCW-based EDSLIFFT was proposed in [19] for the accurate estimation of the harmonics of LPCEAs. In this article, it is further extended for the computation of the TPF as per the IEEE 1459-2010 standard [18].

Overview of the Four-Term MSCW
Given the better main lobe, sidelobe, and sidelobe roll-off rate of the 4MSCW and the accuracy of the EDSLIFFT algorithm with the RT current harmonic signal analysis [19], they are further extended to measure the TPF in real time. Initially, the 4-term MSCW-based EDSLIFFT was used to measure the spectral amplitude and phase of an LPCEA's harmonic current signal. A brief overview of the 4MSCW is given in the following.
The discrete-time 4-term MSCW is expressed as: where n denotes the sample index, N denotes the total number of samples, m represents the window item index, M is the maximum window item number, and a h denotes the window coefficients. The spectral window corresponding to the 4MSCW is written as:

Processing with the 4MSCW-Based EDSLIFFT Algorithm
Generally, in digital signal processing, the harmonic signal is represented as [20]: where the signal amplitude is denoted as A h , and the signal frequency and phase are represented by f h and ϕ h . The harmonic signal sampling time is denoted as T s . The harmonic order is denoted as h, which starts from 1 and reaches the maximum harmonic order of hmax. The traditional FFT has the issue of spectral leakage due to non-synchronous sampling because of the unstable fundamental frequency. To mitigate the spectral leakage effect, the signal is weighted by window functions. The mathematical representation of the windowed sample signal under non-synchronous sampling is given as: where k = 0, 1, . . . , (N − 1), W indicates the 4-term MSCW function, and k h denotes the division factor of the signal frequency and the frequency resolution, which is expressed as: where f s is the sampling frequency of the harmonic signal, l h is an integer value, and ξ h (0 ≤ ξ h ≤ 1) is the fractional part caused by the non-synchronous sampling. The spectral line corresponding to the h th harmonic lies between the two highest spectral lines-explicitly, either the l eng th and the (l h + 1) th or the l h th and the (l h -1) th . The value of l h is determined by the peak location index search algorithm, and the fractional part ξ h is determined by using the EDSLIFFT algorithm, as shown in [19]. The computation process of the EDSLIFFT algorithm using 4MSCW is illustrated in Figure 2.
Energies 2021, 14, 4837 5 of 19 corresponding to the hth harmonic lies between the two highest spectral lines-explicitly, either the leng th and the (lh + 1) th or the lh th and the (lh -1) th . The value of lh is determined by the peak location index search algorithm, and the fractional part is determined by using the EDSLIFFT algorithm, as shown in [19]. The computation process of the EDSLIFFT algorithm using 4MSCW is illustrated in Figure 2. The frequency spectrum expression used in the EDSLIFFT algorithm is given as follows: The two spectral lines represent the h th harmonic amplitude, and they are represented by lh1 and lh2 (where lh1 = I lh2 = I + 1, lh1 < kh < lh2). The peak locations of the harmonic amplitudes are obtained by using the peak location index search method described in [19].
Consider y1 = |X(I)| and y2 = |X(I + 1)|; then, y1 and y2 are given as follows: The spectral amplitudes are determined with the least-square curve-fitting technique. The symmetrical coefficient α is considered in terms of lh and kh as follows: The even spectral lines in EDSLIFFT, y1 and y2, are represented as: A symmetrical coefficient β in terms of α is considered in order to calculate the harmonic parameters; the expression of β is as follows: From Equations (10) and (11), can be represented as: The frequency spectrum expression used in the EDSLIFFT algorithm is given as follows: The two spectral lines represent the h th harmonic amplitude, and they are represented by l h1 and l h2 (where l h1 = I, l h2 = I + 1, l h1 < k h < l h2 ). The peak locations of the harmonic amplitudes are obtained by using the peak location index search method described in [19].
Consider y 1 = |X(I)| and y 2 = |X(I + 1)|; then, y 1 and y 2 are given as follows: The spectral amplitudes are determined with the least-square curve-fitting technique. The symmetrical coefficient α is considered in terms of l h and k h as follows: The even spectral lines in EDSLIFFT, y 1 and y 2 , are represented as: A symmetrical coefficient β in terms of α is considered in order to calculate the harmonic parameters; the expression of β is as follows: From Equations (10) and (11), β can be represented as: The value of α is determined with two maximum spectral lines and the fitting polynomial g −1 (α). Thereby, the amplitude, frequency, and phase values of the given signal are obtained based upon the following interpolated formulas: An LPCEA's input voltage, appliance current fundamental, and harmonic amplitudes are computed by using Equations (15)- (17). Under nonlinear loading conditions, it is necessary to measure the fundamental and harmonic content accurately; the conventional FFT has issues of spectral leakage and experiences the picket fence effect, and hence, the 4-term MSCW-based EDSLIFFT is adopted in this article. The measurement of the TPF by using the fundamental and harmonic amplitudes is discussed in Section 2.3.

TPF Measurement as per the IEEE 1459-2010 Standard
After measuring the fundamental and harmonic amplitudes of the voltage and current by using Equation (15) from the 4-term MSCW-based EDSLIFFT, the root mean square (RMS) values of the voltage and current are determined as follows: The fundamental to higher-order harmonic voltages are represented by V 1RMS , V 2RMS , V 3RMS , . . . , V hmaxRMS in Equation (18), and the fundamental and harmonic currents are denoted as I 1RMS , I 2RMS , I 3RMS , . . . , I hmaxRMS in Equation (19). As per the IEEE 1459-2010 standard [18], the RMS of the voltage and current can be decomposed into fundamental and harmonic components as follows: One of the most significant parameters for efficient power consumption with a good PQ is the PF. If the PF at the supply mains is at unity or close to unity, then it is designated as having a high PQ. On the contrary, a low PF means that the system is operating with a low efficiency and poor PQ. Under nonlinear appliance conditions, the PF is represented by the TPF, which is computed from the distortion factor and displacement PF. The distortion factor is determined from the current THD (THD I ).
The voltage and current THDs are determined with the following equations: The voltage THD V is negligible compared to the current THD I ; hence, the current THD is considered in the distortion factor computation, as given below: The displacement factor is determined as follows: where P represents the active power consumed by the LPCEAs and S denotes the total apparent power. The computation of the active, reactive, and apparent powers is discussed below.
The active power consumption of the LPCEAs given as: The reactive power consumption of the LPCEAs is represented as: Based on the definitions of the voltage and current RMS values as functions of their fundamental and harmonic components, the total apparent power is expressed as: Using Equation (8), the fundamental and non-fundamental components of the apparent power are written as follows: In Equation (31), S N is the non-fundamental apparent power because of the occurrence of harmonic frequencies in the voltage or current waveforms. The fundamental apparent power S 1 is represented by: The non-fundamental apparent power has three distinct terms, as shown in Equation (34).
The term D 1 denotes the current distortion power and comprises the fundamental voltage (V 1RMS ) and the harmonic current (I HRMS ). It is determined as a function of the fundamental apparent power (S 1 ) and the current total harmonic distortion (THD I ): Similarly, the voltage distortion power, D V , encompasses the harmonic voltage (V HRMS ) and the fundamental current (I 1RMS ): Eventually, the harmonic apparent power, S H , is determined from both the voltage and the current terms, and it contains the total harmonic distortion of the voltage and current, as given below: By substituting the active power given in Equation (26) and the total apparent power represented in (33), the displacement power factor is computed by using Equation (26). Then, the TPF is computed using the following equation: Thus, the TPF is measured by using the 4-term MSCW-based EDLIFFT. The real-time (RT) validation of the proposed method of TPF measurement for identifying appliance usage patterns for DSM is presented in the next section.

Real-Time (RT) Measurement of the TPF
The real-time measurement of the TPF by using the 4-term MSCW-based EDSLIFFT for non-intrusive appliance usage pattern identification for DSM is described in this section. A compact reconfigurable input-output system (cRIO 9082)-based virtual instrumentation testbed from National Instruments (NI) was developed in order to deploy the proposed non-intrusive identification of appliance usage patterns. As per the authors' observation, it is prospective hardware for RT measurement of the TPF as per the requirements of international standards, such as IEEE 1459 [18], 1159 [21], and IEC 61000-4-60 [22]. Table 1. Summarizes the standards compliance of the proposed RT-based measurement method. The NI-cRIO 9082 was equipped with a field-programmable gate array (FPGA) architecture, which was detailed in [23]. The voltage and current signals were acquired by the NI-cRIO 9082 with the NI-9225 (voltage) [24] and NI-9227 (current) modules [25]. Then, they were processed with the 4-term MSCW-based EDSLIFFT, which was deployed in a LabVIEW-configured desktop computer system. The desktop computer was interfaced with the NI-cRIO 9082 through a TCP/IP link, as depicted in Figure 3. CFLs, LEDs, exhaust fans, and SMPSs of personal computers are the LPCEAs that ar most commonly used by consumers in developing countries; they were used here t showcase the severity of the problem for the readers, as every home that uses these appli ances knowingly or unknowingly causes harmonic pollution.
The measurement hardware was coupled at the single-phase 230 V, 50 Hz utility sup ply mains for the TPF measurements. These TPF values were used for non-intrusive iden tification of appliance usage patterns. This non-intrusive appliance usage pattern identi fication approach took advantage of signal processing to reduce the hardware effort asso ciated with systems intended for the identification of intrusive appliances with multipl dedicated sensors.
The TPFs of the different appliance usage patterns were measured by turning th connected appliances on or off, as shown in Figure 2. The detailed flowchart of the non intrusive identification of appliance usage patterns with the TPF is shown in Figure 4.

RT Results
The supply voltage and appliance current waveforms acquired with the NI-cRIO 9082 for various real-world appliance combinations are illustrated in Figures 5-8. The RT values of the average TPF, active, reactive, and apparent power, and percentage THD ove 24 h that were obtained from NI-cRIO at 50 kS/s are tabulated in Tables 2 and 3. The fig ures depict the phase differences between the input voltage and the harmonic current. Th RT data values for P, Q, S, S1, SN, TPF, and %THD were observed and are tabulated i CFLs, LEDs, exhaust fans, and SMPSs of personal computers are the LPCEAs that are most commonly used by consumers in developing countries; they were used here to showcase the severity of the problem for the readers, as every home that uses these appliances knowingly or unknowingly causes harmonic pollution.
The measurement hardware was coupled at the single-phase 230 V, 50 Hz utility supply mains for the TPF measurements. These TPF values were used for non-intrusive identification of appliance usage patterns. This non-intrusive appliance usage pattern identification approach took advantage of signal processing to reduce the hardware effort associated with systems intended for the identification of intrusive appliances with multiple dedicated sensors.
The TPFs of the different appliance usage patterns were measured by turning the connected appliances on or off, as shown in Figure 2. The detailed flowchart of the nonintrusive identification of appliance usage patterns with the TPF is shown in Figure 4.  CFLs, LEDs, exhaust fans, and SMPSs of personal computers are the LPCEAs that are most commonly used by consumers in developing countries; they were used here to showcase the severity of the problem for the readers, as every home that uses these appliances knowingly or unknowingly causes harmonic pollution.
The measurement hardware was coupled at the single-phase 230 V, 50 Hz utility supply mains for the TPF measurements. These TPF values were used for non-intrusive identification of appliance usage patterns. This non-intrusive appliance usage pattern identification approach took advantage of signal processing to reduce the hardware effort associated with systems intended for the identification of intrusive appliances with multiple dedicated sensors.
The TPFs of the different appliance usage patterns were measured by turning the connected appliances on or off, as shown in Figure 2. The detailed flowchart of the nonintrusive identification of appliance usage patterns with the TPF is shown in Figure 4.

RT Results
The supply voltage and appliance current waveforms acquired with the NI-cRIO 9082 for various real-world appliance combinations are illustrated in Figures 5-8. The RT values of the average TPF, active, reactive, and apparent power, and percentage THD over 24 h that were obtained from NI-cRIO at 50 kS/s are tabulated in Tables 2 and 3. The figures depict the phase differences between the input voltage and the harmonic current. The RT data values for P, Q, S, S1, SN, TPF, and %THD were observed and are tabulated in Table 4.

RT Results
The supply voltage and appliance current waveforms acquired with the NI-cRIO 9082 for various real-world appliance combinations are illustrated in Figures 5-8. The RT values of the average TPF, active, reactive, and apparent power, and percentage THD over 24 h that were obtained from NI-cRIO at 50 kS/s are tabulated in Tables 2 and 3. The figures depict the phase differences between the input voltage and the harmonic current. The RT data values for P, Q, S, S1, SN, TPF, and %THD were observed and are tabulated in Table 4.    Table 2. TPFs for all 15 appliance patterns in real time. The combinations of two real-world appliances were observed by turning on the combinations of the different two appliances. The voltage and current waveforms of the combinations of the two real-world appliances acquired with the NI-cRIO 9082 are depicted in Figure 6. The combinations of two real-world appliances were observed by turning on the combinations of the different two appliances. The voltage and current waveforms of the combinations of the two real-world appliances acquired with the NI-cRIO 9082 are depicted in Figure 6. The combinations of three real-world appliances were examined by turning on the combinations of three appliances; the corresponding voltage and current waveforms of the CFL, LED, exhaust fan, and SMPS of a PC are illustrated in Figure 7.  The combinations of three real-world appliances were examined by turning on the combinations of three appliances; the corresponding voltage and current waveforms of the CFL, LED, exhaust fan, and SMPS of a PC are illustrated in Figure 7. The voltage and current waveforms when all of the appliances were turned on are depicted in Figure 8. The voltage and current waveforms when all of the appliances were turned on are depicted in Figure 8.

S.No Combinations of Different Appliances
Pertaining to the TPF values obtained from the operation of a single appliance to the operation of all four appliances, all were unique. Therefore, the TPF can be effectively used to identify appliance usage patterns. The TPFs measured with the four-term MSCW-based EDSLIFFT in the RT NI-cRIO 9082 system environment are depicted in Figure 4 and are tabulated in Table 2.
The performance of the experimentation demonstrated that the TPF could be uniquely identified for all combinations of the appliances. The standard deviation of the TPFs indicated that they were all unique and different. Hence, is the TPF can be used as a key feature of a lookup table in order to discern the appliances being operated.  The TPF variations of the appliance usage patterns according to demand are depicted in Figure 9a. The load patterns 1, 4, 5, 6, 7, 9, 12, 14, and 15 are in the region of poor TPF values. The TPFs of the appliance usage patterns after executing the recommendations given by the actionable insights in Table 3 are illustrated in Figure 9b, which demonstrates the improvement of the TPF. Further, it is necessary to adopt mitigation devices in order to improve the TPF as per the requirements of the IEEE 519-2014 standard [26]. Pertaining to the TPF values obtained from the operation of a single appliance to the operation of all four appliances, all were unique. Therefore, the TPF can be effectively used to identify appliance usage patterns. The TPFs measured with the four-term MSCWbased EDSLIFFT in the RT NI-cRIO 9082 system environment are depicted in Figure 4 and are tabulated in Table 2.
The performance of the experimentation demonstrated that the TPF could be uniquely identified for all combinations of the appliances. The standard deviation of the TPFs indicated that they were all unique and different. Hence, is the TPF can be used as a key feature of a lookup table in order to discern the appliances being operated.  The TPF variations of the appliance usage patterns according to demand are depicted in Figure 9a. The load patterns 1, 4, 5, 6, 7, 9, 12, 14, and 15 are in the region of poor TPF values. The TPFs of the appliance usage patterns after executing the recommendations given by the actionable insights in Table 3 are illustrated in Figure 9b, which demonstrates the improvement of the TPF. Further, it is necessary to adopt mitigation devices in order to improve the TPF as per the requirements of the IEEE 519-2014 standard [26].   The actionable insights for DR management are tabulated in Table 3, which establishes the opportunity to improve the TPFs in smart homes. Utilities can benefit from the reduced malfunctioning of the equipment in their distribution systems, and consumers can benefit from the enhanced lifetimes of their LPCEAs.

Discussion
This section discusses the analysis of the results and the scope of future research.

•
The harmonics are non-additive, and the cumulative TPF takes both the inductive and capacitive factors of various appliance combinations into consideration. • Individual PFs are of no consequence, and the TPF is the value to consider in order to compare the impacts of appliance usage patterns with the impacts of individual appliances, as in our article. Therefore, the TPFs of appliance combinations are nonlinear, making the appliance patterns in use uniquely identifiable.

•
The TPF can be improved by selecting alternative appliance patterns, as suggested in Table 4.

•
Highly nonlinear appliances, such as CFLs, SMPSs, and their combinations, consume more reactive power, which results in a poor TPF.

•
The standard deviation of the TPF from the experimental results was 0.1271 for the demand, and it was reduced to 0.099 according to the recommended consumer responses. Thus, the consumer utilization index improved, as illustrated in Table 4, and the quality of the power was enhanced. • According to the results, it is necessary to adopt current harmonic mitigation devices, such as active power filters and PF correction devices in LVDSs, in order to minimize the malfunctioning of the utility equipment, such as failures of the transformers, false tripping of the circuit breakers, etc., as well as to safeguard LPCEAs from overheating.

•
It is necessary to develop smart energy meters in order to measure the THD and TPF and to recommend PQ-sensitive measures, with a provision in the power tariffs for better consumer behavior.
The quantitative metrics tabulated in Table 5 demonstrate the efficacy of the proposed deterministic method over non-deterministic methods.

Future Research
This paper offers a simple way of identifying appliance usage patterns, but there are a few challenges before taking this concept to market. The authors propose the following as the scope for future research.

•
A smart home may have very sophisticated appliances that could be programmed and operated remotely or controlled through a mobile application. How demand-response systems can take advantage of appropriate communication is of practical interest to explore.

•
The physical/thermal characteristics of electronic appliances may change over time and may, in turn, change the THD and TPF values. The accommodation of these adaptabilities in the demand-side management systems must be considered.

•
The compensation of the reactive power must be studied so that a power factor at unity can be achieved at all times and all nodes in a distribution system.

Conclusions
The goal of this research was to save energy with good PQ in order to meet SDG 7. Due to the huge penetration of LPCEAs and their resultant current THD, the TPFs in LVDSs are poor, and consequently, the energy efficiency and PQ are heavily deteriorated. In LVDSs, poor PFs and LPCEAs are not penalized, which results in losses to the DISCOMs. This article illustrated a deterministic approach that uses the TPF in order to identify the consumption patterns of LPCEAs by using the 4-term MSCW-based EDSLIFFT in a virtual instrumentation environment for various combinations of LPCEAs in real time. It was depicted that the TPF values could be used to effectively identify various LPCEA combinations. This method of the non-intrusive identification of appliance usage patterns is essential for responsible electricity consumption with TPF that are close to unity. It is necessary to streamline the tariff structure based on the PQ. It is also necessary to mandate the correction of the PF to maintain PF that is at unity in an LVDS as per CEA regulations (2010) and to holistically comply with SDG 7. The standard deviation of the TPF is impressive, but is not sufficient to comply with the IEEE 519-2014 standard, which requires the installation of compensation devices in LVDSs when appropriate. The actionable insights recommended in this article highlight the reduction in the active power consumption and kVA requirement; they also demonstrate the improvements in the consumer utilization index. The Republic of India does not specify requirements for domestic requirements at the 230 V level, and hence, no PF penalties are imposed. Hence, the TPF is not even a concern of consumers at this moment. This work intends to highlight the deficiencies in the distribution system that must be fixed.