• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Transactive Demand Response Operation at the Grid Edge using the IEEE 2030.5 Standard

    2020-10-20 08:19:30JavadFattahiMikhakSamadiMelikeErolKantarciHenrySchriemer
    Engineering 2020年7期

    Javad Fattahi, Mikhak Samadi, Melike Erol-Kantarci, Henry Schriemer*

    School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa ON K1N 6N5, Canada

    Keywords:Transactive demand response IEEE 2030.5 Smart grid Multi-agent system Neighborhood coordination

    A B S T R A C T This paper presents a transactive demand response(TDR)scheme for a network of residential customers with generation assets that emphasizes interoperability within a transactive energy architecture.A complete laboratory-based implementation provides the first(to our knowledge)realization of a comprehensive TDR use case that is fully compliant with the Institute of Electrical and Electronics Engineers (IEEE)2030.5 standard, which addresses interoperability within a cybersecure smart energy profile (SEP) context. Verification is provided by a full system integration with commercial hardware using Internet Protocol (IP)-based (local area network (LAN) and Wi-Fi) communication protocols and transport layer security (TLS) 1.2 cryptographic protocol, and validation is provided by emulation using extensive residential smart meter data. The demand response (DR) scheme is designed to accommodate privacy concerns, allows customers to select their DR compliance level, and provides incentives to maximize their participation. The proposed TDR scheme addresses privacy through the implementation of the SEP 2.0 messaging protocol between a transactive agent (TA) and home energy management system (HEMS)agents. Customer response is handled by a multi-input multi-output (MIMO) fuzzy controller that manages negotiation between the customer agent and the TA. We take a multi-agent system approach to neighborhood coordination, with the TA servicing multiple residences on a common transformer, and use a reward mechanism to maximize customer engagement during the event-based optimization.Based on a set of smart meter data acquired over an extended time period, we engage in multiple TDR scenarios, and demonstrate with a fully-functional IEEE 2030.5-compliant implementation that our scheme can reduce network peak power consumption by 22% under realistic conditions.

    1. Introduction

    High penetration of distributed energy resources (DERs) and deferrable loads, such as electric vehicles (EVs), is introducing new challenges to the modern grid,due substantially to the uncertainties that are inherent in their intermittency [1,2]. This is expected to greatly influence the operation and control of the power system [2]. The indeterminacy of load profile and supply from DERs calls for a more flexible approach on the customer side.Demand response (DR), traditionally used by utilities with their larger customers,may now—with advances in two-way communication—be brought to the residential level[3].The fundamentals of DR are noted in Ref.[4],and comprehensively discussed in Ref.[5].For a discussion of DR in the context of demand-side management,see Nadel [6]. For an extensive survey of the DR literature, please see Vardakas et al. [7]. DR addresses consumption modification through particular methodologies. DR approaches range from those without dispatch capability, such as time-of-use (TOU)[8,9], critical peak pricing (CPP) [4,10,11], and real-time pricing(RTP) [12], to those with dispatch capability, such as available demand-side resource capacity control (ADSRCC) [13,14] and available demand-side reserve management (ADSRM) [15–17].

    Existing approaches to DR can be categorized as either centralized or decentralized/hierarchical [18]. In the centralized approach, an optimum solution is sought using extended data across the grid, with the aim of maximizing DR outcomes. Desired grid states are then pursued using direct control strategies in which, based on agreements between the distribution system operator (DSO) and customers, remote regulation of customer assets by the operator is permitted [19]. By contrast, hierarchical/decentralized DR is an indirect control strategy, where controllers are local agents with some degree of intelligence and decisionmaking autonomy. Hierarchical DR schemes can be implemented through a master–slave approach, where a local controller is responsible for balancing local generation and consumption [20].

    Control strategies that enable utilities to engage customer DER can take a number of approaches [21]. Among these, the transactive framework takes an agent-based approach to motivate and manage system behavior [22–24]. Similar to conventional DR programs,transactive platforms may use incentives to secure the necessary resources. Behboodi et al. [25] proposed a transactive control paradigm that enabled thermostatically controlled loads to participate in real-time retail electricity markets. A transactive approach was recently used in commercial building heating,ventilation,and air-conditioning(HVAC)systems for DR[26].Daneshvar et al. [27] proposed transactive energy integration for different rural areas, partially equipped with DERs, to reduce total community energy cost. The transactive approach commonly addresses DER management through some price signal, which could be the market clearing price provided by the independent system operator.Barriers to practical implementation still remain:most prominently, a lack of customer participation in the incentive approaches, and the lack of uniform standards for DR in the electricity market.

    Practical DR implementation is not straightforward. Significant challenges must be addressed before its full potential can be realized within a grid edge context, where DER and nontraditional loads are expected to proliferate. Existing centralized and decentralized approaches assume that necessary customer data are available, which may jeopardize customer privacy [28]. Moreover,there are two significant hurdles to practical implementation:how to address dynamic system state regulation, and how to ensure scalability.Most approaches are therefore deficient in one or more of the following factors: ①accurate control with low latency across a range of time scales;②flexibility in response while maintaining interoperability; ③control law sufficiency while adhering to privacy requirements;④communication reliability while keeping marginal costs low;and ⑤cyber-security.A clustering method was proposed in Refs.[29,30]to address the resiliency of customer lifestyle and responses. Employing distributing controllers in the grid with transactive approaches helps to reduce latency and increase flexibility and scalability, but addressing all needed factors in one platform is still an ongoing research activity.

    We address the above challenges within a transactive energy(TE) framework that exploits interoperability as a reference architecture to realize our DR scheme. The TE framework provides an approach to addressing smart grid operational objectives from the joint perspectives of economics and control systems, using value as a key operational parameter [31]. It is characterized by attributes that describe the characteristics of the particular approach. Chen and Liu [22] reviewed the state of the art of research and industry practice on DR and the new methodology of TE, noting the outstanding barriers that remain to be overcome to advance performance, particularly regarding technology, scalability, system management, and consumer behavior. Nunna and Srinivasan [32] proposed an agent-based TE framework for an inter-microgrid auction-based electricity market that incorporates DR within its energy management strategy. Good et al. [33]recently addressed the TE modeling and assessment framework for DR business cases.Zia et al.[34]very recently presented a multilayer TE architecture along with the distributed ledger approaches needed for the validation of economic transactions.

    Interoperability is the critical TE attribute[35],so in contrast to the top-down approaches noted above, and to address the observed deficiencies, we implement our transactive demand response (TDR) scheme using the Institute of Electrical and Electronics Engineers (IEEE) 2030.5 standard for smart energy profile (SEP) application protocol [36] as central to our reference architecture. This standard uses an Hypertext Transfer Protocol(HTTP) request/response model over Transmission Control Protocol (TCP)/Internet Protocol (IP), and adopts extensible mark-up language(XML)as the data-exchange format.Transport layer security (TLS) 1.2 provides confidential message exchange, to ensure authenticity of the different parties communicating and to ensure message integrity. The information (semantic) and cyber–physical(syntactic)interoperability expressed in our TDR design realization ensure that the pragmatics of economic and business objectives can be met in a secure and scalable fashion if we can functionally demonstrate, through lab-based emulation with real load data on actual hardware, that aggregate customer DR compliance can be driven by price signals through transaction.

    We use an event-driven autonomous distributed architecture with real-time communication between a transactive agent (TA)and multiple home energy management system (HEMS) agents[37,38] compliant with the IEEE 2030.5 standard. Residential customers possess both generation and storage assets, in addition to diverse loads, that are subject to HEMS control. Our TDR scheme takes a multi-agent systems approach [37] to the neighborhood coordination of these agents. Transaction is hierarchical, between an aggregating TA and each HEMS agent.That is,the HEMS agents do not directly interact; they cooperate with the TA, but not with each other. This approach promotes privacy by design, while engaging the cognitive intelligence of the HEMS in finding the best voluntary accommodation of the DR request in coordination with all other homes in the TA neighborhood.The TA seeks to maximize the DR request across this network by negotiation with each HEMS agent, subject to the constraint set by each homeowner regarding the degree to which they may be willing to cooperate with such requests. Enhanced cooperation will be fiscally motivated by the TA through an evolving price signal communicated to each homeowner/customer as a reward for positive participation in the transactive negotiation.

    The paper is structured as follows.In Section 2,we describe the system architecture and its communication hierarchy;we relegate a brief overview of IEEE 2030.5 requirements, and a summary of the function sets and essential elements necessary for TDR, to the Appendix A. In Section 3, HEMS agent operation is presented.Section 4 quantifies customer behavior in both past and present,the former through a description of customer consumption for predictive use, and the later via a reward function that is used to mediate transactive negotiation. In Section 5, our TDR algorithm is presented, and the optimization approach to negotiation is detailed.Emulation results are shown in Section 6 for various scenarios, and concluding remarks are presented in Section 7.

    2. System architecture and SEP 2.0 messaging

    Here, we describe our TE framework that exploits interoperability as a reference architecture to realize our DR scheme. We use an event-driven autonomous distributed architecture with real-time communication between agents. It exploits the IEEE 2030.5 SEP 2.0 standard [36] for interoperability. With SEP 2.0 as the data model,which is based on an International Electrotechnical Commission (IEC) 61968 common information model [39],our application programming interface(API)engages the grid edge resources in a ‘‘RESTful” (where REST stands for representational state transfer) [40] manner. A public key infrastructure (PKI)[41]is used to provide protection against non-repudiation attacks.

    Fig. 1. (a) The TDR structural architecture. (b) Physical implementation of the TDR system in a laboratory test configuration. The top panel shows ①the TA, ②the transformer,③the photovoltaic(PV)inverter,and ④the battery;the bottom panel shows ⑤the user interface and ⑥the HEMS agent;and ⑦the HEMS,smart thermostat,and switches. CCL: customer co~operation level; CSR: customer scheduled return; DRR: dynamic DR request; Pcap: the residential capacity; ΔPava: load shed availability;ΔP:potential load reduction;Δ P:requested load reduction during TDR negotiation;Pk:here,the power consumption of load k for this customer(formally,P(ki ),where the sum over all loads gives the current power consumption for customer i, denoted Pi in the text.

    The structural configuration of the TDR system, shown in Fig. 1(a), consists of a TA and an aggregation of HEMS agents as the primary grid edge agents.The TA is physically co-located with the neighborhood distribution transformer. TA communication with the utility provides centralized grid condition monitoring and control. TA communication with HEMS agents coordinates all DR requests and returns, and monitors the local microgrid on the low-voltage side of the transformer. As illustrated in Fig. 1(a), the TA takes an optimization approach to apportioning the DR request by transactive negotiation with each customer, as further described below.

    The HEMS agent is a residential unit responsible for the control and scheduling of customer assets (here, ‘‘loads” for simplicity). It operates within a SEP 2.0 framework to address TA requests,based on customer choices of consumption mode,comfort level,and load priorities,as further described in Section 3.The HEMS agent translates data from such services to a SEP 2.0-recommended XML file for TDR purposes (see the cyan box in Fig. 1(a)), including local power measurements so that DR action can be verified. It maintains customer privacy by design, acting as a customer agent to firewall any knowledge of load-specific activities and customer behavior from the TA.HEMS agents negotiate with the TA to determine their best response to the DR request,using a fuzzy inference engine to accommodate customer preferences and comfort levels in response to rewards offered by the TA (it may also be directed to comply with a mandatory DR event).This behavior will be more fully described in the next section.

    Fig.1(b)shows a physical implementation of the TDR system in a laboratory test configuration. In the left panel, DC power is brought down from the photovoltaic (PV) panels on the roof to a transformer, and thence to a Tabuchi Electric Eco Intelligent Battery System[42](inverter and battery).The TA is mounted onto the electrical panel.The bottom panel shows the HEMS agent,with which the TA communicates, and the HEMS agent interacts with Energate [43] HoˉlHoˉm smart thermostats and load controllers(the HEMS); also shown is the user interface.

    Fig. 2. Communication hierarchy using SEP 2.0 between the DSO, TA, and HEMS agent for TDR.

    Fig. 2 describes the SEP 2.0 messaging sequence via GET commands for XML files. High-level DR events are initiated by the utility server and are communicated to the TA. Low-level DR events are initiated by the TA in response to local asset monitoring,such as transformer health. The TA communicates with individual HEMS agents,sending XML function sets with their corresponding elements in the sequence shown. HEMS agents send similar messages to the TA as part of the transactive negotiation and monitoring process, and the TA likewise sends messages to the DSO. In our architecture,HEMS agent is a client while the TA is either client or server, as required by the downstream devices. Since the TA is firewalled from engaging customer assets directly, we take a distributed intelligence[44]approach to agent-based energy management [45]. In the following sections, we explain how the information retrieved through SEP is used by HEMS and the fuzzy logic controller(FLC)as a HEMS agent to perform negotiations.The SEP 2.0 function sets used to implement the required TDR activities are described in the Appendix A. Having implemented a complete SEP server, we stress that all agents are fully compliant with the IEEE 2030.5 smart grid interoperability standard,although the present TDR scheme uses but a fraction of its capabilities.

    3. HEMS agent operation

    The HEMS agent is responsible for transactive negotiation with the TA. This agent is implemented using an FLC. It communicates with a HEMS to determine aggregate load and load shed availability, and to pass on the DR request for implementation.Below, we describe how the cognitive intelligence of the FLC is engaged by the TA to find the best voluntary accommodation of the DR request in coordination with all other HEMS agents in the neighborhood. Once the DR negotiation with the HEMS agent is done, it sends this information to the HEMS, which is responsible for accommodating the loads in response.We have used a commercial off-the-shelf HEMS from Energate [43] to do so, ensuring IEEE 2030.5 compatibility with this proprietary HEMS through a separate application programming interface.Given a residential capacity Pcap(i.e., the aggregate nameplate powers of all controllable loads),it does this based on data provided by the HEMS for the load shed availability,ΔPava(i.e., the aggregate consumption of all controllable loads presently in use, which is the maximum DR availability), for the scheduling interval. The HEMS agent determines the most efficacious way to achieve this reduction, whether it be through direct curtailment, the change of load set points and duty cycles,or local DER dispatching(treated here as a negative load).

    The IEEE 2030.5 standard describes the process by which smart grid functionalities may be realized,but it does not define the manner by which they are to be realized. This distinction lies at the heart of the interoperability framework. Our implementation of the standard is within a transactive profile, using the normalized quantities described below to better reveal how the HEMS agent demonstrates cognitive intelligence during the transactive process.Customer constraints are implemented heuristically through a customer cooperation level (CCL) [46] to indicate the degree of willingness to comply with TDR requests. Without loss of generality,it is sufficient to define only two modes: ①good, used when customers will accommodate some behavior modification to comply with TDR requests; and ②poor, used when customers want minimal impact on their usage behavior during TDR events.That is, the CCL is a HEMS agent operation mode, set by the customer, that defines how the FLC will respond to DR requests.

    During transactive negotiation, we need to identify the load reduction ΔPithat the HEMS agent for customer i is willing to accommodate, which is less than or equal to the load shed availability ΔPava. We normalize ΔPiby the residential capacity(i.e., the sum of the nameplate powers of all controllable loads),Pcap, to define the customer scheduled return (CSR) as

    In a complementary fashion, the dynamic DR request (DRR) is defined as the ratio of a potential demand reduction during negotiation,ΔP, to the residential capacity as

    where x~is the normalized output of the FLC.A three-step process is followed to determine this output: fuzzification, fuzzy inference,and defuzzification, with Eqs. (1) and (2) providing the universe of discourse.

    Fuzzification operates on the crisp data of the FLC input, converting these data into linguistic variables, chosen here to be Low–Low (LL), Low (L), Medium (M), High (H), and High–High(HH). Each variable is associated with a membership function defined by a triangular wave function,μ(x), where x is a membership value between zero and one that describes a point in the input space (CSR or DRR). Fig. 3 gives the membership functions for the input(as well as the output)that define how each point in the universe of discourse is mapped to a membership value (or degree of membership) between zero and one.

    The fuzzy inference process combines the membership functions with control rules to present fuzzy outputs in a lookup table,which describes the functional mapping between input and output using the linguistic variables. Using a rule-based FLC allows us to best capture the knowledge variety about the problem, as well as the interactions and relationships between its variables, and best accommodates uncertainties in the system [48]. Tables 1 and 2 give the rule bases for the poor and good CCL modes,respectively.

    Defuzzification is used to convert fuzzy outputs into crisp values. Given the symmetry of the membership functions, the centroid of area (COA) approach is used to determine x~from the evaluated membership functions. The negotiated load reduction is then determined via Eq. (3), and the information is sent to the TA using LoadShedAvailability,DrResponse,and the DER function sets, as explained in Appendix A.

    4. Quantifying customer behavior

    We use real-world data to model customer consumption behavior.Smart meter data within Hydro Ottawa’s[49]service area was collected for ten anonymized residential customers at 1 min intervals over 45 d. The customers were then classified using k-means clustering. This information is used by the TA to initiate the transactive negotiation process with each HEMS agent—see Eq. (7) in Section 5.

    Fig. 3. Membership functions (vertical axis) for the input (DRR, CSR) or output(~x, the defuzzified value) membership values (horizontal axis). Also shown are the linguistic variables,LL,L,M,H,and HH,associated with each membership function.

    Table 1 Fuzzy rule base for poor compliance with TDR requests, where inputs are DRR and CSR, with the resulting output shown in the gray area.

    Table 2 Fuzzy rule base for good compliance with TDR requests, where inputs are DRR and CSR, with the resulting output shown in the gray area.

    To further quantify customer behavior, and to enable efficient negotiation, our TDR algorithm (described in detail in Section 5)incorporates a flexible reward function to accommodate diversity in both power consumption and TDR compliance. It may also be used to monetize participation in TDR requests, and may be adjusted to accommodate market signals. For each negotiation time interval, we define for each customer i a reward function

    Fig. 4. Customer consumption behavior clustered in three groups.

    The reward function is continuously updated during negotiation (through r0) to encourage a trend toward increased DR compliance. It is also incorporated within the MirrorUsagePoint function set in Appendix A to update and record the power usage metrics, and to calculate the present consumption and reward for each residence. We will use these data to make the DR request via an optimization approach.

    5. The TDR algorithm

    Fig.5. Flowchart representation of the TDR algorithm,showing the HEMS agent on the left,and the TA flow on the right.The central block indicates the communication process. PARSE refers to decoding the XML files using the SEP2.0 protocol implemented by the IEEE 2030.5-compliant TDR platform. GET and POST are HTTP methods corresponding to read and create operations, respectively.

    We now describe how our TDR scheme is implemented algorithmically,using a flowchart representation in which the negotiation between the TA and the HEMS agent is operationally accomplished via the TCP/IP exchange of the IEEE 2030.5 SEP 2.0 function sets. DR requests may originate from either the utility side or at the transformer level, and are addressed using EndDeviceControl in Appendix A. We denote the target DRR value by the TA as ΔPDR,for a particular deployment time t across a time horizon th, which is a SEP 2.0-defined time interval.

    A simplified TDR algorithm is given in Fig. 5. On the TA side, if a new DRR value is received, the TA will initiate the TDR negotiation with the HEMS agents. As the HEMS agents confirm compliance, negotiation continues with the remaining HEMS agents, as further detailed below. On the HEMS agent side, if a HEMS agent GETs a DRR value with drProgramMandatory = FALSE,it determines load reduction availability and POSTs it to the TA.Negotiation continues with noncompliant HEMS agents until the evolving DRR value is confirmed (or the iteration limit is reached)and negotiation is terminated. The final DRR value is then allocated to each HEMS agent by setting drProgramMandatory = TRUE,and the HEMS agents then implement their requests. The TDR algorithm optimizes the target DR reduction for each HEMS agent at each stage of the negotiation. Optimization is based on model predictive control (MPC) using a linear programming approach [51]. This event-based DR optimization is a good approach for systems such as ours with complex dynamics and constraints [52].

    Note that, as per Eq. (3), ΔP~i(t ) is the load reduction at time t confirmed by the HEMS agent for customer i during negotiation j using the output of the FLC. Negotiations continue until the target DR request has been met, or until no further progress toward improving individual demand reduction can be made, at which point the control variable is applied to each household by setting drProgramMandatory = TRUE. Note that this does not force the DR request to be obeyed. This is a purely voluntary transaction,although failure to follow through may incur a penalty (see EndDeviceControl in Appendix A). We also define a compliance metric,

    to quantify the efficacy of the transactive process in achieving the target DR request.

    For our 10 customers in three clusters (as described in Section 4), negotiations are initiated as per Eq. (7), with each load prediction determined from the cluster consumption behavior(see Fig. 4) across the given time horizon. Reward diversity in the negotiation process is ensured by differences in the customers’current power consumption and by their TDR compliance. Realistic values of the latter are generated by running multiple different demand cases, initiated with fixed reward r0, to construct a nominal set of compliance histories.Negotiations proceed via Eq.(6)by engaging each HEMS FLC,which defuzzifies for the given CCL mode the fuzzy outputs generated from the CSR ratio (i.e., ΔPi) and the evolving DRR ratio (i.e., ΔP). This introduces the necessary ‘‘give and take” in the dialogue between the TA and the HEMS agent,with ongoing negotiation providing input variation for a quasideterministic response.

    6. Emulation results

    Having developed our TDR scheme and physically realized it using the equipment shown in Fig. 1, we now perform a full operational validation, including negotiation and communication with all agents and loads. We demonstrate the efficacy of our TDR platform by system emulation using real residential energyuse data, as described earlier. These data are used to define the load shed availability (ordinarily provided to the HEMS agent by the HEMS), as further described below. Emulations are performed across a transactive network of 10 customers, with aggregate capacity fixed at 20 kW and aggregate DR availability at 12 kW.All participants are assumed to have the same individual capacities Pcap.As stringent tests of our transactive approach,two specific DR scenarios are investigated across the afternoon/evening demand peak from 5 pm to 8 pm(see Fig.4):①target DR reductions equal to aggregate DR availability (i.e., 12 kW), and ②target DR reductions of about 90% of the aggregate DR availability (i.e., 11 kW).We consider three different CCL mode scenarios: good—all customers; poor—all customers; and mixed—a random distribution of good and poor CCL mode customers(i.e.,the average of 10 such scenarios, evenly distributed about an equal balance of good and poor modes). Ten different load profiles are derived for each customer by applying a random perturbation to the hourly consumption of the source profile while leaving the daily energy unchanged.This creates a dataset of statistically independent but nominally equivalent profiles for each customer that mimics day-to-day variation in consumption.The mean compliance ratio,S—see Eq.(8)—is found by averaging the ten different load assignment simulations.

    Fig. 6 shows the evolution of negotiation for the 11 kW target DR request within a CCL mixed mode network, where triangle symbols denote the increasing aggregate demand reduction ΔPafor all households at each negotiation step, and squares denote the evolving target DR request ΔPdr. Note that after the first negotiation, many customers have already confirmed their requested reduction, for an aggregate reduction of 5 kW. The algorithm then addresses a reduced DR request of 6 kW to the remaining participants,and negotiation continues.After the fifth negotiation,no further progress is seen, so the algorithm reduces the request by 10%and negotiation continues. This results in an immediate marked increase in compliance, but ongoing negotiations affect no further change. Negotiations are terminated after ten iterations, and the final network compliance is determined to be 98%.

    Behavior similar to that of the CCL mixed mode is seen in Fig.7,where the negotiation responses for all residences with CCL good modes and all residences with CCL poor modes are compared.Overall,the responses are similar—they are nearly equal for the initial negotiations—but significant divergence is seen by the fourth negotiation. The subsequent reduction in the DR request after the fifth negotiation moderates, but does not fully eliminate, this divergence.As with all of our negotiations,after the sixth iteration,no significant compliance improvement is found.The end result is a 95% compliance for the network responding with only CCL poor modes, and perfect compliance for the network responding with only CCL good modes.

    Repeating the above for different load demand distributions across the ten home network creates an ensemble of statistically equivalent results, whose average reflects the probability that the target DR request can be accommodated.Evolution of the negotiations for all individual circumstances are found to be similar. The mean compliance ratio,are summarized in Table 3 for all the scenarios introduced above. For even the most stringent of requests,where the DR target equals the network availability, the compliance exceeds 90% regardless of the compliance mode. For target requests of about 90% of aggregate DR availability, compliance exceeds 95%in even the worst case,and is perfect in the best case.CCL mixed mode results are uniformly found to fall midway between good and poor modes,suggesting that there may be a linear relationship between the degree to which customers agree to cooperate and the degree to which the target DR request is filled.All this suggests that our approach may be able to accommodate customers with even less capacity to comply.

    Having seen the DR response in aggregate, let us now consider more generally the individual compliance with ongoing DR requests. For our network of ten customers, the aim is to gain insight into the range of responses under nominally independent DR events, but with an evolving reward function. Specifically, we wish to know,under operationally realistic conditions,how a customer’s historical compliance influences both negotiation and compliance, as well as the reward implications. To that end, for each DR event,we again randomly adapt the measured load profile for each customer to create datasets of statistically independent but nominally equivalent profiles that mimic day-to-day variation in consumption. We then repeat the above analysis for an 11 kW target DR request in CCL mixed mode (i.e., averaging over ten different combinations of CCL poor and CCL good) for 30 successive DR events, updating the reward function after each DR event.

    Fig. 8 shows, for each customer, the incremental compliance with ongoing negotiation. Fig. 8(a) gives the results for the first DR event, and Fig. 8(b) for the 30th. Customer DR compliance is indicated in Watts by the color bar. Note first that negotiation begins with differing requests to each customer, as a consequence of their differing consumption rates and histories—see Eq.(6).Second, note that many customers typically lock-in their compliance early in the negotiation process, which is indicated in the plots by the varying horizontal color bar lengths.Third,we see that there is some variation in customer response during the negotiation processes (both increasing and decreasing responses, and seen in the color changes), as a consequence of the second constraint in Eq.(6). This variation promotes more optimal network response.Finally, comparison between Figs. 8(a) and (b) (DR events #1 and#30,respectively)reveals changes in individual compliances, both in magnitude and in the number of negotiations required. Customers 2, 3, and 10 engaged in one or two more rounds of negotiation, yielding modest increases (about 10%) in their DR response,while other customers were effectively unchanged.Let us consider more specifically how individual behaviors drive—and are driven by—reward evolution.

    Fig. 6. Evolution of negotiation for a target DR request of 11 kW in CCL mixed mode, with specific load demands that differ across the network.

    Fig. 7. Comparison of the evolution of negotiation between CCL good and poor modes for a target DR request of 11 kW, with specific load demands that differ across the network.

    Table 3 Efficacy of transactive energy target DR requests for a 20 kW aggregate capacity and 12 kW aggregate DR availability, assessed over an ensemble of load demand distributions across the ten home networks.

    Fig. 8. Evolution of customer compliance with ongoing negotiation in CCL mixed mode for (a) the first DR event, and (b) the last DR event (#30).

    We consider three representative customers:3,7,and 10.Fig.9 shows how their rewards (black lines) evolve with ongoing DR requests(i.e.,events):Figs.9(a),(c),and(e)show this as a function of their consumption (red lines), while Figs. 9(b), (d), and (f)show this as a function of their compliance(red lines).Also shown are consumption and compliance averages (green lines) for comparison.Note that fluctuations in the consumption between events(and hence compliance variation)arise from the random perturbation of the load profile used to produce the datasets for each DR request; hence, it is only the overall trends that are meaningful.Customer 3 (Figs. 9(a) and (b)) is both high-consumption and high-compliance, with trends well above their means. Its reward is effectively unchanging, fluctuating close to the base reward of r0=1000. Customer 7 (Figs. 9(c) and (d)) has slightly higher consumption (by about 5%), but its compliance is indistinguishable from the mean.Its reward sees a near-factor-of-two decrease.Customer 10 (Figs. 9(e) and (f)) is low-consumption but with initially low compliance that trends toward the mean. Consequently, its reward trends modestly upward. These observations demonstrate the reward bias toward low-consumption customers who maintain meaningful participation, and against high-consumption customers whose participation is proportionally similar(i.e.,average).High-consumption users must participate at better-than-average rates to even maintain reward parity;otherwise,they incur a substantial penalty over time.By contrast,low-consumption users are not negatively impacted, and will preferentially benefit as their compliance increases.These reward trends,which have influenced the negotiation process, are controlled through the consumption and compliance coefficients, c1and c2, respectively, of Eq. (5). We have chosen an equal weighting to demonstrate what may be described as a fair allocation process, without explicit bias toward either consumption or compliance. These coefficients may be adjusted, with the effect that both slope and direction of the reward curves may be altered to promote or penalize specific behaviors.

    Fig.9. Evolution of reward ri (black curve)with successive DR requests,shown against consumption(left column)and compliance(right column)behavior(red curves);also shown are their average behaviors (green curves). (a,b) Customers 3 and (c,d) 7 belong to class III (high-consuming), while (e,f) customer 10 belongs to class I (lowconsuming).

    Network consumption diversity drives the optimization process by interaction with individual customer constraints (i.e., CCL, CSR,and DRR).This interaction is with the FLCs as the local agents that implement the demand reductions. Bounded outputs are ensured by the inherent input bounds on capacity and availability,as incorporated within the functional constraints of Eq. (6). A transparent illustration of algorithm robustness for the bounded response may be obtained by considering the subpopulation statistics (i.e.,the sampling of the load distribution dataset) of Table 3 data,and verifying that the uncertainties decrease with increasing size of the transactive network(10 and 50).We consider,for CCL mixed mode(again averaging over ten different combinations of CCL poor and CCL good), target DR reductions equal to aggregate DR availability.We determine the mean compliance ratio,for three cases,each with its distinct load distribution. The results, presented in Table 4, confirm the behavior expected for bounded outputs. The mean is largely insensitive to the number of participants, but the uncertainties(i.e.,the standard deviations)decrease for increasing network size while remaining nominally unchanged across the three cases (with the possible exception of one outlier).

    Fig.10 shows the impact of transactive negotiation on the total demand trend before and after applying the 11 kW DR request to the ten-home-network in CCL mixed mode across the afternoon/evening demand peak. A peak reduction of 22% (from 45 to 35 kW) is observed, with the demand shifted to later in the evening. This demand shift, enabled by the individual HEMS agents,was accomplished by rescheduling an equal weighting of typical residential interruptible and non-interruptible loads. Fig. 11 provides a complementary view of the impact of the DR event across the network using the peak-to-average ratio(PAR)[53].The PAR is shown for each customer,without and with the DR request,and forthe aggregate network.All customers are seen to contribute to the demand reduction, with the heaviest users seeing the greatest impact. Across the network (all customers), PAR is reduced by 14% (from 3.42 to 2.92). Note that PAR, because it is in reference to the average demand, is a metric that is less sensitive to the immediacy of the DR event, and therefore somewhat undervalues its impact.

    Table 4 Dependence of customer compliance statistics(CCL mixed mode)on network size and load distribution sampling.

    Fig.10. The total demand trend before and after applying the 11 kW DR request for the CCL mixed mode response.

    Fig.11. The peak-to-average ratio(PAR)for all participants before and after the DR event.

    7. Conclusion

    We presented a DR scheme that employed transactive negotiation across a residential network at the transformer level.We took a multi-agent approach that emphasized customer comfort and privacy. A TA negotiated DR requests with customer agents (their HEMS agent), using an event-based optimization technique and a flexible reward mechanism. The agents responded to the DR requests by reducing their consumption based on compliance levels associated with their comfort expectations (expressed via fuzzy rule bases). We showed that our transactive DR scheme was able to reduce the peak demand by 22%, which is equivalent to a 14% PAR reduction. The results were based on anonymized smart meter data from real customers. Notably, our scheme used the IEEE 2030.5 SEP interoperability standard to transfer data between the agents, and thus constitutes the first known TDR use case for this standard. Although we have only used a portion of IEEE 2030.5 functionality within this TDR scheme, a complete SEP server was developed to ensure full agent compliance with the standard.The demonstrated efficacy within this restricted context strongly suggests that its full capabilities—including aggregation for distributed energy resource management systems and virtual power plant control, with market access and all necessary monetization requirements—are within imminent reach of all vendors and solution providers that choose to follow this scalable approach to achieving the smart grid vision.

    Acknowledgements

    This work is supported in part by Natural Sciences and Engineering Council of Canada (CRDPJ 477238-14) and Hydro Ottawa. The Hydro Ottawa led ‘‘GREAT-DR” project is gratefully funded in part by the Ontario Ministry of Energy’s Smart Grid Fund and the LDC Tomorrow Fund, and generously supported by the University of Ottawa, Carleton University, CIMA Canada Inc.,Panasonic Eco-Solutions Canada Inc. (with Tabuchi Electric Company of America Limited), Quadra Power Inc., Energate Inc.,Thorium Technologies Inc.,and the IEEE Standards Association.The authors wish to acknowledge the vision and encouragement provided by Raed Abdullah (Hydro Ottawa Limited), The‘‘GREAT-DR” project’s initiator and leader.

    Compliance with ethics guidelines

    Javad Fattahi, Mikhak Samadi, Melike Erol-Kantarci, and Henry Schriemer declare that they have no conflict of interest or financial conflicts to disclose.

    Appendix A. Supplementary data

    Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2020.06.005.

    又黄又爽又刺激的免费视频.| 免费看美女性在线毛片视频| 18+在线观看网站| 爱豆传媒免费全集在线观看| 精品不卡国产一区二区三区| 国产伦在线观看视频一区| 伦精品一区二区三区| 免费观看精品视频网站| 国产av码专区亚洲av| 亚洲成色77777| 男的添女的下面高潮视频| 男人舔奶头视频| 国产三级中文精品| 美女大奶头视频| av又黄又爽大尺度在线免费看 | 人妻夜夜爽99麻豆av| 三级国产精品欧美在线观看| 26uuu在线亚洲综合色| 国产精品人妻久久久影院| 我要看日韩黄色一级片| 一级二级三级毛片免费看| 青春草亚洲视频在线观看| 岛国在线免费视频观看| 久久韩国三级中文字幕| 久久精品91蜜桃| 99热6这里只有精品| 少妇裸体淫交视频免费看高清| 看非洲黑人一级黄片| 国产午夜精品一二区理论片| 有码 亚洲区| 久久久久久久久久黄片| 国产成年人精品一区二区| 中文精品一卡2卡3卡4更新| 精品人妻偷拍中文字幕| av在线观看视频网站免费| 亚洲国产精品专区欧美| 又粗又硬又长又爽又黄的视频| 少妇的逼好多水| 美女国产视频在线观看| 日本色播在线视频| 色综合亚洲欧美另类图片| 亚洲av.av天堂| 亚洲精品日韩av片在线观看| 国产精品嫩草影院av在线观看| 日本猛色少妇xxxxx猛交久久| 爱豆传媒免费全集在线观看| or卡值多少钱| 美女xxoo啪啪120秒动态图| 亚洲精品日韩在线中文字幕| 禁无遮挡网站| 身体一侧抽搐| 久久精品国产亚洲网站| 亚洲欧美精品综合久久99| 免费观看人在逋| 国产精品人妻久久久影院| 91久久精品国产一区二区三区| 色播亚洲综合网| 91精品伊人久久大香线蕉| 国产视频首页在线观看| 国产成人freesex在线| 成人二区视频| 久99久视频精品免费| 在线免费观看不下载黄p国产| 国产精华一区二区三区| 搡老妇女老女人老熟妇| 男女视频在线观看网站免费| 久久久久网色| 女的被弄到高潮叫床怎么办| 级片在线观看| 国产在线一区二区三区精 | 久久99热这里只频精品6学生 | 国产男人的电影天堂91| 菩萨蛮人人尽说江南好唐韦庄 | 午夜免费激情av| 99久国产av精品国产电影| 午夜精品在线福利| 国产视频内射| 国产探花极品一区二区| 亚洲精品日韩av片在线观看| 一边摸一边抽搐一进一小说| 精品久久久噜噜| 国产成人精品久久久久久| 免费av毛片视频| 日韩欧美精品免费久久| 成人欧美大片| 村上凉子中文字幕在线| 亚洲av熟女| 中文字幕亚洲精品专区| 九九在线视频观看精品| 久久久久久大精品| 国产91av在线免费观看| 亚洲欧美精品自产自拍| 一级二级三级毛片免费看| 国产单亲对白刺激| 一个人看的www免费观看视频| 日韩精品有码人妻一区| 亚洲av免费高清在线观看| 亚洲丝袜综合中文字幕| 亚洲精品久久久久久婷婷小说 | 成人美女网站在线观看视频| 一级毛片aaaaaa免费看小| 日本av手机在线免费观看| 亚洲国产最新在线播放| 一级黄片播放器| 午夜爱爱视频在线播放| 18禁动态无遮挡网站| 18+在线观看网站| 国产乱人视频| 精品久久久久久久久av| 成人av在线播放网站| 欧美成人午夜免费资源| 成人高潮视频无遮挡免费网站| 国产极品天堂在线| 日韩三级伦理在线观看| 亚洲四区av| 午夜福利成人在线免费观看| 国产亚洲精品av在线| 免费观看在线日韩| 色综合亚洲欧美另类图片| av视频在线观看入口| 三级国产精品片| 久久人妻av系列| a级一级毛片免费在线观看| 日本黄大片高清| 亚洲av福利一区| 禁无遮挡网站| 又爽又黄a免费视频| 黄色欧美视频在线观看| 老司机影院毛片| 久久国产乱子免费精品| 日韩强制内射视频| 久久人人爽人人片av| 欧美97在线视频| 少妇的逼水好多| 白带黄色成豆腐渣| 神马国产精品三级电影在线观看| 精品久久久久久久久av| 亚洲国产色片| 老女人水多毛片| 天堂网av新在线| 国产毛片a区久久久久| 波多野结衣高清无吗| 精品久久久久久电影网 | 三级国产精品片| 国产精品蜜桃在线观看| 成年免费大片在线观看| 男女那种视频在线观看| 伦精品一区二区三区| 少妇高潮的动态图| 人妻制服诱惑在线中文字幕| 在线观看av片永久免费下载| 午夜a级毛片| 国产亚洲午夜精品一区二区久久 | 成人欧美大片| 日韩成人伦理影院| 日日摸夜夜添夜夜添av毛片| 日本免费一区二区三区高清不卡| 日本色播在线视频| 国产精品1区2区在线观看.| 亚洲精品日韩在线中文字幕| 91久久精品电影网| 99久久精品热视频| 国国产精品蜜臀av免费| 国产精品熟女久久久久浪| 日韩视频在线欧美| av天堂中文字幕网| 日韩欧美精品v在线| 国产黄a三级三级三级人| 69av精品久久久久久| 水蜜桃什么品种好| 亚洲图色成人| 日本熟妇午夜| av免费观看日本| 最近视频中文字幕2019在线8| 亚洲aⅴ乱码一区二区在线播放| 久99久视频精品免费| 久久久久久久久久久丰满| 久久亚洲国产成人精品v| 日韩av不卡免费在线播放| 国产精品一及| 六月丁香七月| 国产91av在线免费观看| 一级毛片aaaaaa免费看小| 成年女人看的毛片在线观看| 亚洲国产精品sss在线观看| 丰满乱子伦码专区| 成人性生交大片免费视频hd| 又爽又黄a免费视频| 国产精品国产高清国产av| 亚洲精品日韩av片在线观看| 两个人视频免费观看高清| 亚洲欧美清纯卡通| 精品久久久久久成人av| 国产伦在线观看视频一区| 边亲边吃奶的免费视频| 自拍偷自拍亚洲精品老妇| 丰满人妻一区二区三区视频av| 久久久精品94久久精品| 午夜免费激情av| 国产精品蜜桃在线观看| 一个人看的www免费观看视频| 黄片wwwwww| 精品久久久久久久末码| 干丝袜人妻中文字幕| 国产精品.久久久| av免费在线看不卡| 亚洲综合精品二区| 51国产日韩欧美| 色5月婷婷丁香| 欧美日本亚洲视频在线播放| 九九热线精品视视频播放| 亚洲综合精品二区| 成年av动漫网址| 欧美激情国产日韩精品一区| 久久久久久久久久成人| 在线a可以看的网站| 日韩精品有码人妻一区| 淫秽高清视频在线观看| 国产高清国产精品国产三级 | 美女黄网站色视频| 国产视频首页在线观看| 一个人观看的视频www高清免费观看| 蜜桃亚洲精品一区二区三区| 日日摸夜夜添夜夜添av毛片| 亚洲第一区二区三区不卡| 国产一区亚洲一区在线观看| 性插视频无遮挡在线免费观看| 国产v大片淫在线免费观看| 熟妇人妻久久中文字幕3abv| 欧美日韩在线观看h| 18禁在线播放成人免费| 婷婷色av中文字幕| 欧美人与善性xxx| 日韩欧美精品v在线| 日本免费a在线| 极品教师在线视频| 国产伦在线观看视频一区| 亚洲精品亚洲一区二区| 一级毛片电影观看 | 亚洲av成人精品一区久久| 精品一区二区三区人妻视频| 中文精品一卡2卡3卡4更新| 久久午夜福利片| 中文字幕制服av| 日韩国内少妇激情av| 日产精品乱码卡一卡2卡三| 听说在线观看完整版免费高清| 免费在线观看成人毛片| 国产亚洲精品av在线| 亚洲成人精品中文字幕电影| 99九九线精品视频在线观看视频| 桃色一区二区三区在线观看| 亚洲丝袜综合中文字幕| 全区人妻精品视频| 美女黄网站色视频| 好男人在线观看高清免费视频| 夜夜爽夜夜爽视频| 纵有疾风起免费观看全集完整版 | 日本五十路高清| 国产精品精品国产色婷婷| 国产精品久久久久久精品电影小说 | 激情 狠狠 欧美| 桃色一区二区三区在线观看| 岛国毛片在线播放| 国产精品久久久久久精品电影| 麻豆成人午夜福利视频| 亚洲成av人片在线播放无| 国产亚洲5aaaaa淫片| 免费看美女性在线毛片视频| 色视频www国产| 国产精品麻豆人妻色哟哟久久 | 国产极品精品免费视频能看的| 国产高清视频在线观看网站| av女优亚洲男人天堂| 国产精品人妻久久久久久| 欧美日本亚洲视频在线播放| 国产黄片视频在线免费观看| 婷婷色av中文字幕| 中文字幕精品亚洲无线码一区| 国产精品不卡视频一区二区| 亚洲自拍偷在线| 免费搜索国产男女视频| 老师上课跳d突然被开到最大视频| 国产精品久久久久久久电影| 亚州av有码| 在线免费十八禁| 小蜜桃在线观看免费完整版高清| 日产精品乱码卡一卡2卡三| 嫩草影院新地址| 国产精品永久免费网站| 寂寞人妻少妇视频99o| 欧美性猛交黑人性爽| 国产单亲对白刺激| 一个人看的www免费观看视频| 精品午夜福利在线看| 狂野欧美白嫩少妇大欣赏| 内射极品少妇av片p| 男女下面进入的视频免费午夜| 18禁裸乳无遮挡免费网站照片| 又爽又黄a免费视频| 亚洲第一区二区三区不卡| 欧美性猛交黑人性爽| 亚洲高清免费不卡视频| 午夜久久久久精精品| 国产爱豆传媒在线观看| 乱人视频在线观看| 亚洲久久久久久中文字幕| 中文亚洲av片在线观看爽| 麻豆国产97在线/欧美| 啦啦啦啦在线视频资源| 国产精品.久久久| 久久人人爽人人爽人人片va| 亚洲欧美一区二区三区国产| 国产精品国产高清国产av| 亚洲av免费在线观看| 久久99热6这里只有精品| 看十八女毛片水多多多| 日本猛色少妇xxxxx猛交久久| 欧美3d第一页| 熟女电影av网| 亚洲自偷自拍三级| 麻豆成人午夜福利视频| 麻豆成人av视频| 国产欧美另类精品又又久久亚洲欧美| 国产中年淑女户外野战色| 欧美+日韩+精品| 菩萨蛮人人尽说江南好唐韦庄 | 日日摸夜夜添夜夜添av毛片| 精品久久久久久久久亚洲| 天天躁日日操中文字幕| 中文字幕久久专区| av在线蜜桃| 亚洲国产欧洲综合997久久,| 99久国产av精品| 成人毛片60女人毛片免费| 嫩草影院新地址| 蜜桃亚洲精品一区二区三区| 你懂的网址亚洲精品在线观看 | 久久人妻av系列| 日韩欧美 国产精品| 亚洲av电影在线观看一区二区三区 | 在线观看66精品国产| 在线免费观看的www视频| 秋霞在线观看毛片| 亚洲国产精品合色在线| 美女黄网站色视频| 干丝袜人妻中文字幕| 日本黄色视频三级网站网址| 国产高清视频在线观看网站| 亚洲一区高清亚洲精品| 一个人看的www免费观看视频| 久久久精品94久久精品| 久久久国产成人精品二区| 久久精品91蜜桃| 精品免费久久久久久久清纯| 乱码一卡2卡4卡精品| 日日啪夜夜撸| 久久久久久久午夜电影| 精品99又大又爽又粗少妇毛片| 国产黄片美女视频| 男女国产视频网站| 国产精品久久久久久久久免| 午夜福利网站1000一区二区三区| 淫秽高清视频在线观看| 久久精品国产鲁丝片午夜精品| 精品熟女少妇av免费看| 久久久久久大精品| 婷婷色av中文字幕| 最近中文字幕2019免费版| 丰满乱子伦码专区| 久久精品国产亚洲av天美| 我的老师免费观看完整版| 两个人视频免费观看高清| 免费一级毛片在线播放高清视频| 91aial.com中文字幕在线观看| 久久久久网色| 亚洲经典国产精华液单| 青春草亚洲视频在线观看| 国产日韩欧美在线精品| 少妇丰满av| 国产精品国产三级国产专区5o | 亚洲va在线va天堂va国产| 欧美激情国产日韩精品一区| 成年女人永久免费观看视频| 国产乱人偷精品视频| 日韩一本色道免费dvd| 免费搜索国产男女视频| 国产精品野战在线观看| 午夜a级毛片| 最近的中文字幕免费完整| 国产黄片美女视频| 国产亚洲午夜精品一区二区久久 | 国产高清国产精品国产三级 | 永久免费av网站大全| 卡戴珊不雅视频在线播放| 精品国内亚洲2022精品成人| 国产 一区精品| 亚洲欧美一区二区三区国产| 免费看光身美女| 国产黄a三级三级三级人| 免费在线观看成人毛片| 国产av码专区亚洲av| 日本猛色少妇xxxxx猛交久久| 亚洲精品456在线播放app| 久久热精品热| 最后的刺客免费高清国语| 日韩,欧美,国产一区二区三区 | 别揉我奶头 嗯啊视频| 久久久亚洲精品成人影院| 视频中文字幕在线观看| 91精品一卡2卡3卡4卡| 亚洲av免费在线观看| 18禁动态无遮挡网站| 亚洲真实伦在线观看| 精品一区二区三区视频在线| 中文亚洲av片在线观看爽| 免费搜索国产男女视频| 亚洲欧美成人精品一区二区| av.在线天堂| 国产精品女同一区二区软件| 成人午夜精彩视频在线观看| 久久这里只有精品中国| 秋霞在线观看毛片| 日本免费一区二区三区高清不卡| 欧美一区二区亚洲| 国产又色又爽无遮挡免| 亚洲国产欧洲综合997久久,| 国产高清三级在线| 国产精品综合久久久久久久免费| 亚洲四区av| 黄色日韩在线| 国产精品女同一区二区软件| av专区在线播放| 国产成年人精品一区二区| 国产av一区在线观看免费| 五月伊人婷婷丁香| 三级国产精品片| 纵有疾风起免费观看全集完整版 | 国产黄a三级三级三级人| 国产亚洲av片在线观看秒播厂 | 国产精品国产高清国产av| 国产精品女同一区二区软件| 成人午夜精彩视频在线观看| 国产国拍精品亚洲av在线观看| 观看免费一级毛片| 亚洲怡红院男人天堂| 男女那种视频在线观看| 国产精品三级大全| 亚洲性久久影院| 久久韩国三级中文字幕| 精品一区二区三区人妻视频| 18禁在线无遮挡免费观看视频| 国产精品一区二区三区四区免费观看| 午夜福利视频1000在线观看| 日韩一本色道免费dvd| 女人被狂操c到高潮| 亚洲av成人精品一二三区| 一级爰片在线观看| 国产视频首页在线观看| 色综合色国产| 99久久精品国产国产毛片| 日韩一本色道免费dvd| 成人二区视频| 国产精品,欧美在线| 联通29元200g的流量卡| 日韩成人av中文字幕在线观看| av天堂中文字幕网| 精品人妻熟女av久视频| 亚洲精品自拍成人| 国产中年淑女户外野战色| 日韩中字成人| 国产av一区在线观看免费| 亚洲18禁久久av| 国产欧美日韩精品一区二区| 免费观看性生交大片5| 插逼视频在线观看| 久久人人爽人人片av| 国产乱人视频| 极品教师在线视频| av又黄又爽大尺度在线免费看 | 国产黄a三级三级三级人| 伦精品一区二区三区| 午夜福利在线观看吧| 91久久精品电影网| 晚上一个人看的免费电影| 亚洲精品乱久久久久久| 中国美白少妇内射xxxbb| 国产免费视频播放在线视频 | 在线免费观看的www视频| 午夜免费男女啪啪视频观看| 久久6这里有精品| 亚洲中文字幕日韩| 一级黄片播放器| av播播在线观看一区| 在线播放国产精品三级| 欧美一区二区国产精品久久精品| 亚洲欧美日韩东京热| 禁无遮挡网站| 黄片无遮挡物在线观看| 丝袜喷水一区| 我要看日韩黄色一级片| 小说图片视频综合网站| 人妻少妇偷人精品九色| 男人舔女人下体高潮全视频| 夜夜爽夜夜爽视频| 日本爱情动作片www.在线观看| 亚洲欧美日韩东京热| 国产亚洲av嫩草精品影院| 久久99精品国语久久久| 亚洲国产精品国产精品| 综合色av麻豆| h日本视频在线播放| 麻豆成人午夜福利视频| 3wmmmm亚洲av在线观看| 国产免费又黄又爽又色| 在线观看美女被高潮喷水网站| 夫妻性生交免费视频一级片| 非洲黑人性xxxx精品又粗又长| 搡女人真爽免费视频火全软件| 成人欧美大片| 91久久精品国产一区二区成人| 国产女主播在线喷水免费视频网站 | 国产单亲对白刺激| 国产精品人妻久久久影院| 欧美日韩国产亚洲二区| 99久久精品国产国产毛片| 国产午夜精品久久久久久一区二区三区| 欧美高清性xxxxhd video| 国产精品三级大全| 三级经典国产精品| 国产麻豆成人av免费视频| 少妇丰满av| 国产欧美日韩精品一区二区| 免费搜索国产男女视频| 日韩av在线免费看完整版不卡| 日本三级黄在线观看| 日本猛色少妇xxxxx猛交久久| 国产毛片a区久久久久| 国产人妻一区二区三区在| 婷婷色综合大香蕉| 成人欧美大片| 国产成人a∨麻豆精品| 亚洲精品乱码久久久久久按摩| av线在线观看网站| a级毛片免费高清观看在线播放| 黄色日韩在线| 麻豆精品久久久久久蜜桃| 国产成人91sexporn| 久久亚洲国产成人精品v| 亚洲综合精品二区| 国产精品无大码| 天堂网av新在线| 日日干狠狠操夜夜爽| 一区二区三区高清视频在线| 欧美成人午夜免费资源| 狂野欧美激情性xxxx在线观看| .国产精品久久| 三级国产精品片| 在线免费观看的www视频| 欧美人与善性xxx| 欧美色视频一区免费| 男女下面进入的视频免费午夜| 看片在线看免费视频| 卡戴珊不雅视频在线播放| 日韩精品青青久久久久久| 国产精品久久电影中文字幕| 日韩欧美三级三区| 欧美潮喷喷水| 深爱激情五月婷婷| 黄色一级大片看看| 欧美成人一区二区免费高清观看| 亚洲第一区二区三区不卡| 国产老妇伦熟女老妇高清| 日本黄大片高清| 简卡轻食公司| 人人妻人人澡欧美一区二区| 女的被弄到高潮叫床怎么办| 狂野欧美白嫩少妇大欣赏| 亚洲色图av天堂| 一边摸一边抽搐一进一小说| 欧美日韩一区二区视频在线观看视频在线 | 一级毛片我不卡| 亚洲丝袜综合中文字幕| 综合色av麻豆| 黄色配什么色好看| 春色校园在线视频观看| 国产免费一级a男人的天堂| a级毛色黄片| 成人国产麻豆网| av在线老鸭窝| 国产精品精品国产色婷婷| 永久网站在线| 亚洲美女搞黄在线观看| 深爱激情五月婷婷| 国产高潮美女av| 国产精品久久电影中文字幕| 成人漫画全彩无遮挡| 成人毛片60女人毛片免费| 久久精品国产鲁丝片午夜精品| 超碰av人人做人人爽久久| 欧美一区二区精品小视频在线| 欧美成人午夜免费资源| 国产精品女同一区二区软件| 成人毛片a级毛片在线播放| 日本三级黄在线观看| 超碰av人人做人人爽久久| 国产精品综合久久久久久久免费| 精品人妻一区二区三区麻豆| 在线观看美女被高潮喷水网站| 久久久久久九九精品二区国产| videos熟女内射| 亚洲av熟女| 只有这里有精品99| 国产精品电影一区二区三区| 最后的刺客免费高清国语|