Decentralized rolling optimization method and system based on resident peak shaving potential expectations

Abstract

The application provides a distributed rolling optimization method and system based on resident peak shaving potential expectation, and peak load information is transmitted downward to a load aggregator by a power distribution network, wherein the peak load information comprises: a total load reduction demand and a corresponding peak period set; the load aggregator issues a next-day peak shaving task to users according to the response potential of each resident user; the net load sequence of each user is collected, aggregated and calculated, and the aggregated load data is transmitted to the power distribution network; the user side load optimization model based on the resident peak shaving potential expectation is constructed by the user through the user home energy management system, considering various costs and benefits, a planned load sequence is generated, and the planned load sequence is fed back to the load aggregator. The application avoids the problem of sharp peak load caused by load transfer between different users, reduces the burden of the power grid, has self-adaptive robustness to user response uncertainty, directly converts the uncertainty problem into a deterministic problem, and significantly reduces the calculation complexity.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart grid, in particular, to a distributed rolling optimization method and system based on resident peak shaving potential expectation. BACKGROUND

[0002] The development of smart grid enables resident users to participate in load peak shaving demand response through home energy management system. However, the following problems exist in the existing research:

[0003] 1. The load transfer scheme based on time-varying price may cause peak load rebound phenomenon;

[0004] 2. The dependence on two-way communication leads to user privacy leakage and reliability reduction;

[0005] 3. Few studies consider the uncertainty of resident response behavior to incentive policy, which significantly affects the actual response potential and effect of user side.

[0006] Therefore, a distributed adaptive robust optimization model is proposed in the art. This distributed optimization framework can minimize the information exchange between users and load aggregators to reduce the dependence on communication, protect privacy, has strong scalability and can be parallel computed to improve efficiency. At the same time, the feedback correction rolling optimization avoids the problem of peak load transfer. However, the existing distributed adaptive robust optimization model still has the following problems:

[0007] 1. Considering that the common random optimization avoids uncertainty in the sense of probability, it is not necessarily feasible in reality;

[0008] 2. Robust optimization is too conservative to meet the requirements in the worst case and sacrifices cost.

[0009] In summary, the current optimization method cannot meet the development needs of smart grid, and no similar technology to the present application has been described or reported, and no domestic or foreign similar data has been collected. SUMMARY

[0010] The present application provides a distributed rolling optimization method and system based on resident peak shaving potential expectation to solve the above problems in the prior art.

[0011] The present application is realized by the following technical solutions.

[0012] According to one aspect of the present application, a distributed rolling optimization method based on resident peak shaving potential expectation is provided, comprising:

[0013] Predicting total load power P of distribution network sch ;

[0014] Initializing distribution network load power and the number of rolling wheels n, so that

[0015] Comparing the distribution network load power with the preset load peak threshold If the maximum value of is greater than , the first round of optimization is entered, at which time the number of rolling wheels is increased by one, n = 1; if the maximum value of is less than or equal to , that is, the load power has met the preset load peak requirement, optimization is ended, and the number of rolling wheels is 0;

[0016] The distribution network transmits load peak information to a load aggregator, wherein the load peak information includes: a total load reduction demand and a corresponding peak period set;

[0017] The load aggregator issues a peak shaving task to users according to the response potential of each residential user:

[0018]

[0019] In the formula: Pj is the flexible load power of the user j; P fl is the total flexible load power of all users,; is the total load reduction demand at period t in the n th round of rolling; is the load reduction demand of the user j at period t in the n th round of rolling; n is the load peak period set in the n th round of rolling;

[0020] A user-side load optimization model based on the expected residential peak shaving potential is constructed to generate a planned load sequence;

[0021] The load aggregator collects the planned load sequences of each user to obtain a planned net load sequence, and performs the following aggregation calculation to transmit the aggregated load data to the distribution network:

[0022]

[0023] In the formula: is the optimized planned net load power of the user j at period t in the n th round of rolling; is the user aggregated net load power at period t in the n th round of rolling; [1, T] indicates that the dispatching period is 1 day, which is divided into T periods; N j is the number of users;

[0024] Return to the step of comparing the distribution network load power with the preset load peak threshold to perform the next rolling optimization process.

[0025] Optionally, the resident user response potential is a proportion of a flexible load of the resident user to a total flexible load of the resident user.

[0026] Optionally, the multiple costs and benefits are considered by the user home energy management system, a user side load optimization model based on resident peak shaving potential expectation is constructed, and a planned load sequence is generated, including:

[0027] According to the obtained multiple costs and benefits, a decision variable of the user side load is established, and the user side load is divided into a shiftable load, an interruptible load and an uncontrollable load according to the decision variable;

[0028] The operation constraint model of the shiftable load, the interruptible load and the uncontrollable load is respectively established;

[0029] According to the operation constraint model, a global constraint model including a power balance constraint model and a peak shaving constraint model based on resident peak shaving potential expectation is established;

[0030] According to the global constraint model, a user side load optimization model based on resident peak shaving potential expectation is constructed;

[0031] The planned load sequence is calculated through the user side load optimization model.

[0032] Optionally, the decision variable includes: a start state, a working state, a running power and a state of charge of the load;

[0033] The shiftable load has a determined working duration, and the start time is controllable and the running time is shiftable;

[0034] The interruptible load can be interrupted multiple times during the running;

[0035] The uncontrollable load is an important load for guaranteeing the basic life of the resident, and has a higher reliability requirement.

[0036] Optionally, the operation constraint model of the shiftable load, the interruptible load and the uncontrollable load is respectively established, including:

[0037] The operation constraint model of the shiftable load is established, including:

[0038]

[0039]

[0040]

[0041]

[0042]

[0043] Where: and are the first The initial state and operating state of a translatable load; is the number of translatable loads of user j; is the operation duration of the i-th translatable load of user j; and are the upper and lower limits of the starting operation period of the i-th translatable load of user j, respectively; is the power of user j’s i-th translatable load in time period t after the n-th round of rolling optimization; is the rated power of the i-th movable load of user j;

[0044] Establishing an operation constraint model of the interruptible load includes:

[0045]

[0046]

[0047]

[0048] Where: is the number of user j in the nth round of rolling An operating state that can interrupt the load; is the number of interruptible loads of user j; and are the upper and lower limits of the operating power of the i-th interruptible load of user j respectively; is the power of the i-th interruptible load of user j in time period t after the n-th round of optimization; and are the upper and lower limits of daily power consumption of the i-th interruptible load of user j respectively; and are the upper and lower limits of the operation period of the i-th interruptible load of user j; σ is the optimization time interval;

[0049] Establishing an operation constraint model of the uncontrollable load includes:

[0050]

[0051] Where: is the first The power of an uncontrollable load in time period t; is the number of uncontrollable loads of user j; is the corresponding day-ahead predicted power of the i-th uncontrollable load of user j.

[0052] Optionally, establishing a global constraint model including a power balance constraint model and a peak shaving constraint model based on residents' expected peak shaving potential according to the operation constraint model includes:

[0053] Establishing the power balance constraint model includes:

[0054]

[0055] Where: is the power generation of the household photovoltaic power generation system of user j in time period t, obtained by day-ahead prediction;

[0056] Establishing the peak shaving constraint model based on residents' expected peak shaving potential includes:

[0057]

[0058] Where: is the benchmark net load power of user j at time t in the nth rolling round, obtained by day-ahead prediction; is the expected response potential of user j at time t in the nth scrolling round; is the upper limit of the load reduction of user j in time period t in the nth rolling round;

[0059]

[0060] Where: n,j is the set of all possible response events of user j in the nth rolling; ε n,j,q is the probability of the qth possible response event occurring; is the net load power of user j at time t under the qth possible response event; the actual response event is divided into two special cases: participating in the peak shaving plan and not participating in the peak shaving plan. is the probability that user j participates in the peak shaving demand response in the nth rolling round. It is approximately expressed by the response frequency obtained from historical data statistics and is the specific parameter expression of residents' expected peak shaving potential.

[0061] Optionally, constructing a user-side load optimization model based on residents' expected peak shaving potential according to the global constraint model to generate a planned load sequence includes:

[0062] Establishing electricity purchase costs Response comfort cost Photovoltaic power generation and electricity sales revenue and incentive compensation benefits The comprehensive objective function minF co :

[0063]

[0064] wherein, is calculated as follows:

[0065]

[0066] wherein: c tou (t) is the time-of-use price function of the user purchasing electricity from the power grid, c g,pv is the price of surplus electricity, c pv is the single price of photovoltaic subsidy; and is the response cost coefficient of the user j; C j,0 is the incentive cost threshold of the power grid paying for the user, is the incentive subsidy coefficient of the user j; is a quadratic function determined by considering the marginal increasing effect of the loss of comfort utility; is a linear function established by considering consumer psychology and having a starting threshold and a saturation threshold; and the nth round of response electricity of the user j is calculated as follows:

[0067]

[0068]

[0069]

[0070] Based on the above steps, a user-side load optimization model based on the expected peak shaving potential of residents is obtained; solving the user-side load optimization model, i.e. obtaining the planned load sequence of each user.

[0071] According to another aspect of the present application, a distributed rolling optimization system based on the expected peak shaving potential of residents is provided, comprising: a power distribution network layer placed at the top layer, a load aggregator layer placed at the middle layer and a user layer placed at the bottom layer; wherein:

[0072] The power distribution network layer is configured to transmit load peak information downward to the load aggregator layer, wherein the load peak information comprises: a total load reduction demand and a corresponding peak period set;

[0073] The load aggregator layer is configured to issue a next-day peak shaving task to the user layer according to the response potential of each residential user; collect the net load sequence of each user, perform aggregation calculation, and transmit the aggregated load data to the power distribution network layer;

[0074] The user layer is used for considering multiple costs and benefits through a user home energy management system, constructing a user-side load optimization model based on resident peak shaving potential expectation, generating a planned load sequence, and feeding back to the load aggregator layer.

[0075] Compared with the prior art, the application has at least one of the following beneficial effects:

[0076] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the application automatically formulate targeted optimization schemes for users with different willingness degrees of participating in incentive demand response by introducing the numerical expectation of user demand response potential (DRP) in the model constraint condition, and have self-adaptive robustness to uncertain factors.

[0077] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the application adopts a multi-layer coordinated system-level distributed optimization framework, avoids the problem that the load transfer strategies between different users are not fully coordinated, which may cause peak load in other periods, and reduces the burden of the power grid.

[0078] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the application does not require users to exchange electricity consumption information, effectively protecting the privacy of users.

[0079] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the application only needs to transmit key information (such as peak period and pre-allocated target reduction amount) of the peak shaving task, and does not need to transmit a system load file in each iteration, thereby reducing the calculation pressure of the distributed home energy management system, further reducing the communication requirements, and improving the reliability and calculation efficiency.

[0080] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the application decomposes the actual response load during the peak shaving event into two components, namely, a reference load and a response load, and when the response probability is smaller, the reference load contributes more, and a smaller optimized load is required to meet the peak shaving requirement. That is, the lower the user's willingness degree, the more conservative the load response plan automatically generated by the algorithm, and the greater the response reliability margin, so that the model has self-adaptive robustness to user response uncertainty.

[0081] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the application directly converts the uncertainty problem into a deterministic problem without converting it into a convex optimization problem with a polynomial complexity, thereby significantly reducing the complexity of calculation. BRIEF DESCRIPTION OF DRAWINGS

[0082] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0083] Figure 1 This is a workflow diagram of a decentralized rolling optimization method based on residents' peak shaving potential expectations in a preferred embodiment of the present invention.

[0084] Figure 2 This is an architectural diagram of a decentralized rolling optimization system based on residents' peak shaving potential expectations in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0085] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention.

[0086] An embodiment of the present invention provides a decentralized rolling optimization method based on residents' peak shaving potential expectations.

[0087] like Figure 1 As shown, the decentralized rolling optimization method based on residents' expected peak shaving potential provided by this embodiment may include the following steps:

[0088] S100, distribution network predicted total load power P sch ;

[0089] S200, initialize distribution network load power and the number of rolling wheels n, so that n=0;

[0090] S300, compare distribution network load power With the preset load peak threshold like The maximum value is greater than Enter S400 and start the first round of optimization. At this time, the number of rolling rounds is automatically increased once, n=1; if The maximum value is less than or equal to Then the optimization ends and the number of rolling rounds n=0;

[0091] S400, the distribution network transmits the load peak information downward to the load aggregator, wherein the load peak information includes: the total load reduction demand and the corresponding peak time period;

[0092] At S500, the load aggregator issues peak shaving tasks to residential users based on their response potential:

[0093]

[0094] wherein: Pj is the flexible load power of user j; P fl P is the total flexible load power of all users, is the total load curtailment demand at time period t in the nth rolling; is the load curtailment demand of user j at time period t in the nth rolling; χ n is the set of load peak time periods of the nth rolling;

[0095] S600, constructing a user-side load optimization model based on the resident peak shaving potential expectation, generating a planned load sequence;

[0096] S700, the load aggregator collects the planned load sequence of each user, obtains a planned net load sequence, and performs the following aggregation calculation, and transmits the aggregated load data to the power distribution network:

[0097]

[0098] wherein: is the optimized planned net load power of user j at time period t in the nth rolling; is the user aggregated net load power at time period t in the nth rolling; [1, T] indicates that the dispatching period is 1 day, which is divided into T time periods; N j is the number of users;

[0099] S800, returning to S300 to perform the next rolling optimization process, and the rolling number n = n + 1.

[0100] In a preferred embodiment of S500, the resident user response potential is: the proportion of the flexible load of the resident user in the total flexible load of the resident user.

[0101] In a preferred embodiment of S600, a user-side load optimization model based on the resident peak shaving potential expectation is constructed by considering various costs and benefits through the user home energy management system, and a planned load sequence is generated, including:

[0102] S601, establishing decision variables for the user-side load according to the obtained various costs and benefits, and dividing the user-side load into translatable load, interruptible load and uncontrollable load according to the decision variables;

[0103] S602, respectively establishing operation constraint models of the translatable load, the interruptible load and the uncontrollable load;

[0104] S603, according to the operation constraint models, establishing a global constraint model including a power balance constraint model and a peak shaving constraint model based on the resident peak shaving potential expectation;

[0105] S604: Based on the global constraint model, a user-side load optimization model based on the expected peak shaving potential of residents is constructed;

[0106] S605: Calculate and obtain a planned load sequence through a user-side load optimization model.

[0107] In a preferred embodiment of S601, the decision variables include: the starting state, working state, operating power and state of charge of the load;

[0108] Shiftable loads with a defined working duration, controllable start-up time, and shiftable operating time;

[0109] Interruptible loads, which can be interrupted multiple times during operation;

[0110] Uncontrollable loads are important loads to ensure the basic lives of residents and have high reliability requirements.

[0111] In a preferred embodiment of S602, operation constraint models for the translatable load, the interruptible load, and the uncontrollable load are established respectively, including:

[0112] Establish an operational constraint model for translatable loads, including:

[0113]

[0114]

[0115]

[0116]

[0117]

[0118] Where: and are the first The initial state and operating state of a translatable load; is the number of translatable loads of user j; is the operation duration of the i-th translatable load of user j; and are the upper and lower limits of the starting operation period of the i-th translatable load of user j, respectively; is the power of user j’s i-th translatable load in time period t after the n-th round of rolling optimization; is the rated power of the i-th movable load of user j;

[0119] Establish an operational constraint model for interruptible loads, including:

[0120]

[0121]

[0122]

[0123] wherein: is the running state of the i-th interruptible load of user j in the n-th rolling; is the number of interruptible loads of user j; is the upper limit of the running power of the i-th interruptible load of user j; is the lower limit of the running power of the i-th interruptible load of user j; is the power of the i-th interruptible load of user j at time period t after the n-th optimization; is the upper limit of the daily power consumption of the i-th interruptible load of user j; is the lower limit of the daily power consumption of the i-th interruptible load of user j; is the upper limit of the running time period of the i-th interruptible load of user j; and is the lower limit of the running time period of the i-th interruptible load of user j; and is the optimization time interval.

[0124] A running constraint model of uncontrollable loads is established, including:

[0125]

[0126] wherein: is the power of the i-th uncontrollable load of user j at time period t; is the number of uncontrollable loads of user j; is the corresponding day-ahead forecast power of the i-th uncontrollable load of user j. In a preferred embodiment of S603, according to the running constraint model, a global constraint model including a power balance constraint model and a peak shaving constraint model based on resident peak shaving potential expectation is established, including:

[0127] The power balance constraint model is established, including:

[0128]

[0129]

[0130] wherein: is the power generation of the household photovoltaic power generation system of user j at time period t, obtained by day-ahead forecast;

[0131] The peak shaving constraint model based on resident peak shaving potential expectation is established, including:

[0132]

[0133] wherein:​ is the benchmark net load power of user j at time t in the nth rolling round, obtained by day-ahead prediction; is the expected response potential of user j at time t in the nth scrolling round; is the upper limit of the load reduction of user j in time period t in the nth rolling round;

[0134]

[0135] Where: n,j is the set of all possible response events of user j in the nth rolling; ε n,j,q is the probability of the qth possible response event occurring; is the net load power of user j at time t under the qth possible response event; the actual response event is divided into two special cases: participating in the peak shaving plan and not participating in the peak shaving plan. is the probability that user j participates in the peak shaving demand response in the nth rolling round. It is approximately expressed by the response frequency obtained from historical data statistics and is the specific parameter expression of residents' expected peak shaving potential.

[0136] In a preferred embodiment of S604, a user-side load optimization model based on residents' expected peak shaving potential is constructed according to the global constraint model to generate a planned load sequence, including:

[0137] Establishing electricity purchase costs Response comfort cost Photovoltaic power generation and electricity sales revenue and incentive compensation benefits The comprehensive objective function minF co :

[0138]

[0139] in, The calculation is as follows:

[0140]

[0141] Where: c tou (t) is the time-of-use electricity price function of the user purchasing electricity from the grid, c g,pv is the on-grid electricity price for surplus electricity, c pv is the unit price of photovoltaic subsidies; and is the response cost coefficient of user j; C j,0 The starting threshold for the incentive cost of grid payment users, is the incentive subsidy coefficient of user j; A quadratic function is determined to take into account the marginal increasing effect of the loss of comfort utility; A linear function of the presence of activation threshold and saturation threshold is established to take into account consumer psychology; and user j's nth round response power The calculation is as follows;

[0142]

[0143]

[0144]

[0145] Based on the above steps, a user-side load optimization model based on the expected peak-shaving potential of residents is obtained; by solving this optimization model, the planned load sequence of each user can be obtained.

[0146] An embodiment of the present invention also provides a decentralized rolling optimization system based on residents' expected peak shaving potential.

[0147] like Figure 2 As shown, the decentralized rolling optimization system based on residents' expected peak shaving potential provided by this embodiment may include: a distribution network layer at the top, a load aggregator layer at the middle, and a user layer at the bottom; wherein:

[0148] The distribution network layer is used to transmit the load peak information downward to the load aggregator layer, where the load peak information includes: the total load reduction demand and the corresponding peak time period;

[0149] The load aggregator layer is used to issue the next day's peak shaving task to the user layer based on the response potential of each residential user; collect the net load sequence of each user, perform aggregate calculations, and transmit the aggregated load data to the distribution network layer;

[0150] The user layer is used to take into account various costs and benefits through the user's home energy management system, build a user-side load optimization model based on the residents' expected peak shaving potential, generate a planned load sequence, and feed it back to the load aggregator layer.

[0151] It should be noted that the steps in the method provided by the present invention can be implemented using the corresponding structural layers in the system. Those skilled in the art can refer to the technical solution of the method to implement the composition of the system. That is, the embodiments in the method can be understood as preferred examples of constructing the system and will not be elaborated here.

[0152] The technical solutions provided by the above embodiments of the present invention are described in further detail below.

[0153] 1. Multi-layer coordinated system-level decentralized optimization framework

[0154] Dynamic electricity price and smart electricity utilization technology can improve the enthusiasm of residential users to participate in the optimization of grid-load interaction. For example, in order to save electricity cost, users are more inclined to transfer flexible load from peak period to off-peak period, and if the load transfer on the user side is effective, the peak shaving pressure on the supply side can be reduced. However, if the load transfer strategies of different users are not fully coordinated, it may lead to sharp peak load in other periods, that is, peak load rebound phenomenon, which increases the burden of the power grid. Therefore, a decentralized rolling optimization system based on the expectation of residential peak shaving potential of a multi-layer coordinated system-level decentralized optimization framework is proposed to avoid this phenomenon.

[0155] The response objects of the three levels in the framework of the system are the distribution network layer, the load aggregator layer and the user layer. The distribution network layer is the top participant, which publishes the total peak shaving task according to the day-ahead load prediction data of the regulation center. The load aggregator layer is the middle participant, which is responsible for processing the upper layer signal and acting as a community agent to allocate sub-tasks to the lower user layer. The bottom layer is executed by decentralized users. The three-party information interaction relationship is as shown in Figure 2 .

[0156] The multi-layer coordinated decentralized rolling optimization system can realize the decentralized rolling optimization method provided by the embodiments of the present application, as shown in Figure 1 . The process is as follows:

[0157] The first step: day-ahead load prediction is performed by the top layer power grid regulation center to obtain the total baseline load sequence as the basis for judging whether to start rolling optimization. Take the rolling wheel number n = 0, at this time the rolling has not started.

[0158] The second step: compare with the peak threshold value set in advance If the maximum value of is greater than , then jump to the third step, the rolling wheel number n = n + 1, and the program enters the main loop of rolling optimization; if does not exceed the threshold value, then jump to the seventh step, and end the optimization.

[0159] The third step: the distribution network transmits the system load peak information (including the total load reduction demand and the corresponding peak period set) to the load aggregator downward;

[0160] The fourth step: the aggregator issues the next day peak shaving task (single user load reduction demand and the corresponding peak period set) to the bottom layer user according to the response potential of each residential user (the proportion of the flexible load of a user to the total flexible load of residents), as formula (1);

[0161]

[0162] In the formula: Pj is the flexible load power of user j; P fl P is the total flexible load power of all users considered; P is the total load curtailment demand at time period t in the nth rolling; P is the load curtailment demand of user j at time period t in the nth rolling; χ n P is the set of load peak time periods in the nth rolling.

[0163] Step 5: The user-side load optimization model based on the resident curtailment potential expectation is calculated by the user home energy management system considering multiple costs and benefits, a planned load sequence is generated, see Section 3 of this paper;

[0164] Step 6: The load aggregator collects the net load sequence of each user, aggregates according to formula (2), and transmits the aggregated load data to the upper layer, and jumps to Step 2;

[0165]

[0166] In the formula: P is the optimized net load power of user j at time period t in the nth rolling; P is the user aggregated net load power at time period t in the nth rolling; the dispatching period is 1 day, divided into T time periods; N j P is the number of users.

[0167] Step 7: Terminate the program.

[0168] As can be seen from the above algorithm, the distributed rolling optimization method solves an independent optimization problem for each user in each rolling, and there is no coupling relationship between the constraints. Although the mutual coordination between end users for weakening the rebound peak phenomenon is ignored in the distributed method, the interaction, comparison and iteration between the upper and lower levels can avoid the rebound peak load phenomenon. Since there is no need to exchange electricity information between users, the privacy protection problem can be solved. In addition, unlike the distributed optimization framework in the prior art, the proposed framework only needs to transmit the key information of the peak shaving task (peak time period and pre-allocated target curtailment amount), without the need to transmit the system load file in each iteration, reducing the calculation pressure of the distributed user energy management system, further reducing the communication requirements, and improving the reliability and calculation efficiency.

[0169] II. User-side flexible load model

[0170] First, the load model is established for different user-side flexibility classifications, including shiftable load, interruptible load and uncontrollable load. The established decision variables include binary variables such as device start state and working state, and continuous variables such as running power and state of charge.

[0171] 1) Shiftable load

[0172] The running time of the dishwasher, microwave oven and other equipment can be shifted, such loads have a certain task duration, and the starting running time is controllable. The running constraint model of the shiftable load is as follows:

[0173]

[0174]

[0175]

[0176]

[0177]

[0178] In the formula: and are the starting state and running state of the jth shiftable load of the user in the nth rolling, taking the value of 0 or 1; is the number of shiftable devices of the user j; is the running duration of the ith shiftable load of the user j; and are the upper and lower limits of the starting running period, respectively; is the power of the jth shiftable load of the user in the nth rolling after optimization; is the rated power.

[0179] 2) Interruptible load

[0180] The water heater, energy storage battery and other equipment can be interrupted many times during operation. Its running constraint model is as follows:

[0181]

[0182]

[0183]

[0184] In the formula: is the running state of the ith interruptible load of the user j in the nth rolling; and are the running power upper and lower limits of the ith interruptible load of the user j, respectively; is the power of the ith interruptible load of the user j in the nth rolling after optimization; and are the daily power consumption upper and lower limits of the ith interruptible load of the user j; and are the upper and lower limits of the running period, respectively; σ is the optimization time interval. ​​

[0185] 3) Uncontrollable load

[0186] Refrigerators, lighting systems, etc. belong to uncontrollable load in demand response scheduling. Because this kind of load is an important load to guarantee the basic life of residents, it should have a higher reliability requirement. The operation constraint model of uncontrollable load is as follows:

[0187]

[0188] In the formula: is the power of the jth uncontrollable load in the tth time period, is the corresponding day-ahead forecast power.

[0189] III. Adaptive robust optimization based on resident peak shaving potential expectation

[0190] 1) Constraint condition

[0191] In addition to the constraint condition established by the user-side flexible load model, global constraints including power balance constraint and peak shaving constraint need to be established. The system should meet the power balance constraint model as follows:

[0192]

[0193] In the formula: is the power generation of the jth user photovoltaic power generation system in the tth time period, which is obtained by day-ahead forecast.

[0194] The load aggregator cannot directly observe the complete information of whether the user participates in demand response, and since there is uncertainty in the willingness of users to participate in demand response, the theoretical response amount of the scheduling plan is usually greater than the actual response amount. Therefore, a certain margin should be left when formulating the scheduling strategy, and the uncertain behavior of user response should be fully considered to ensure the actual effect and reliability of demand response. Based on this, a peak shaving constraint model based on user response potential expectation is proposed, as shown in formula (13).

[0195]

[0196] In the formula: is the jth user's benchmark net load power at t time in the nth round of rolling, which is obtained by day-ahead forecast; is the jth user's response potential expectation at t time in the nth round of rolling, which is obtained by formula (14); is the upper limit of load reduction of the jth user in the tth time period in the nth round of rolling.

[0197]

[0198] In the formula: ψ n,j ​ε is the set of all possible response events of user j in the nth rolling, n,j,q is the probability of the qth case, is the net load power of user j at time t in the qth case; the actual response event is divided into two special cases of participating and not participating in peak shaving plan, is the probability of user j participating in peak shaving demand response in the nth rolling, which can be approximately represented by the response frequency obtained from historical data.

[0199] Equation (14) decomposes the actual response load during the peak shaving event into two components, baseline load and response load. When the response probability is smaller, the baseline load contribution is greater, and the optimization load needs to be smaller to meet the peak shaving requirement of equation (13). That is, the lower the user's willingness, the more conservative the load response plan generated by the algorithm, leaving more response reliability margin, so that the model has adaptive robustness to user response uncertainty. In addition, since the proposed method directly converts the uncertainty problem into a deterministic problem without converting it into a convex optimization problem with polynomial complexity, the computational complexity is significantly reduced.

[0200] 2) Objective function

[0201] The objective function minF considers the purchase cost of electricity , response comfort cost , photovoltaic power generation and electricity sales revenue , and incentive compensation revenue co As shown in equation (15):

[0202]

[0203] Among them, the four cost-benefit functions are calculated as shown in equation (16):

[0204]

[0205] In the equation: c tou (t) is the time-of-use electricity price function of user purchasing electricity from the grid, c g,pv is the price of surplus electricity on the grid, c pv is the photovoltaic subsidy unit price; and are the response cost coefficients of user j; C j,0 is the incentive cost threshold paid by the grid to the user, is the incentive subsidy coefficient of user j; is a quadratic function determined considering the marginal increasing effect of comfort utility loss (see Li Wei, Han Ruidi, Sun Chenjia, Fu Peng, Zhang Jie, Wang Chong. Optimal incentive contract and incentive strategy of demand response with transmissible load based on electricity preference [J / OL]. Proceedings of the CSEE: 1-10 [2021-05-13]. http: / / kns.cnki.net / kcms / detail / 11.2107.TM.20210406.1402.008.html.), and is a linear function with starting threshold and saturation threshold established considering consumer psychology (see Zheng Yingying, Jenkins Bryan M, Kornbluth Kurt, Kendall Alissa, Chresten. Optimization of a biomass-integrated renewable energy microgrid with demand side management under uncertainty. Appl Energy 2018;230: 836-44.). and is determined by formula (17), (18), and the nth round of response electricity of user j is determined by formula (19).

[0206]

[0207]

[0208]

[0209] The distributed rolling optimization method and system based on resident peak shaving potential expectation provided by the above embodiments of the present application have self-adaptive robustness to uncertain factors by introducing the numerical expectation of user demand response potential (DRP) in the model constraint condition to automatically formulate a targeted optimization scheme for users with different willingness degrees of participating in incentive demand response. A multi-layer coordinated system-level distributed optimization framework is adopted to avoid the problem that the load transfer strategy between different users is not fully coordinated, which may cause peak load in other periods, thereby reducing the burden of the power grid. User privacy is effectively protected without exchanging electricity consumption information between users. Only the key information of the peak shaving task (such as the peak period and the pre-allocated target reduction amount) needs to be transmitted, and the system load file does not need to be transmitted in each iteration, thereby reducing the calculation pressure of the distributed home energy management system and further reducing the communication requirements and improving the reliability and calculation efficiency. The actual response load during the peak shaving event is decomposed into two components, namely, the baseline load and the response load. When the response probability is smaller, the baseline load contributes more, and a smaller optimized load is required to meet the peak shaving requirement. That is, the lower the user's willingness degree, the more conservative the load response plan automatically generated by the algorithm, and the greater the response reliability margin, so that the model has self-adaptive robustness to user response uncertainty. The uncertainty problem is directly converted into a deterministic problem without converting it into a convex optimization problem with a polynomial complexity, thereby significantly reducing the complexity of the calculation.

[0210] The details not described in the above embodiments of the present application are known in the art.

[0211] The specific embodiments of the present application are described above. It should be understood that the present application is not limited to the above specific embodiments, and various modifications or changes can be made by those skilled in the art within the scope of the claims, which does not affect the essential content of the present application.

Claims

1. A decentralized rolling optimization method based on resident peak-cut potential expectation, characterized in that, The method comprises the following steps: Predicting total load power P of a distribution network sch ; Initializing power load of power distribution network and a number of rolls n, such that n = 0; Comparing power load of power distribution network with preset load peak threshold If the maximum value of the load power is greater than enter the first round of optimization, at this time the number of rolling rounds is increased by one, n=1; if the maximum value of the load power is less than or equal to that is, the load power has met the preset load peak requirement, the optimization is ended, and the number of rolling rounds is 0; The power distribution network transmits load peak information to the load aggregator, wherein the load peak information comprises a total load reduction demand and a corresponding peak period set; The load aggregator issues a peak shaving task to the user according to the response potential of each residential user: In the formula: Pj is the flexible load power of user j; P fl P is the total flexible load power of all users; Pn is the total load reduction demand of the nth round of rolling at time period t; Pjn is the load reduction demand of user j at time period t in the nth round of rolling; χ n Pn is the load peak period set of the nth round of rolling; A user-side load optimization model based on the expected residential peak shaving potential is constructed by taking into account various costs and benefits through the user home energy management system, and a planned load sequence is generated, comprising: According to the obtained various costs and benefits, a decision variable of the user-side load is established, and the user-side load is divided into a shiftable load, an interruptible load and an uncontrollable load according to the decision variable; Respective operation constraint models of the shiftable load, the interruptible load and the uncontrollable load are established; According to the operation constraint models, a global constraint model comprising a power balance constraint model and a peak shaving constraint model based on the expected residential peak shaving potential is established; According to the global constraint model, a user-side load optimization model based on the expected residential peak shaving potential is constructed; The planned load sequence is calculated through the user-side load optimization model; the load aggregator collects the planned load sequence of each user to obtain a planned net load sequence, performs aggregation calculation, and transmits the aggregated load data to the power distribution network. In the formula: is the optimized net load power of user j in the n th round of rolling in period t; is the user aggregated net load power in period t in the n th round of rolling; [1, T] indicates that the scheduling period is 1 day, which is divided into T periods; N j is the number of users; If the load power of the power distribution network is compared with the preset load peak threshold value, the next rolling optimization process is performed, and the rolling number n is n+1.

2. The resident-based peak-cut potential expectation-based distributed rolling optimization method according to claim 1, characterized in that, The residential user response potential is the proportion of flexible load of the residential user in the total flexible load of the residential user.

3. The resident-based peak-cut potential expectation-based distributed rolling optimization method according to claim 1, wherein, The decision variable comprises a start state, a working state, a running power and a state of charge of the load. The shiftable load has a determined working duration, and the start time is controllable and the running time is shiftable. The interruptible load can be interrupted multiple times during operation. The uncontrollable load is an important load for guaranteeing the basic life of the residential user, and has a reliability requirement.

4. The resident-based peak-cut potential expectation-based distributed rolling optimization method according to claim 1, wherein, The respective operation constraint models of the shiftable load, the interruptible load and the uncontrollable load comprise: The operation constraint model of the shiftable load comprises: wherein: and are the initial state and the running state of the i th shiftable load of user j in the n th rolling; is the number of shiftable loads of user j; is the running duration of the i th shiftable load of user j; and are the upper and lower limits of the initial running period of the i th shiftable load of user j; is the rated power of the i th shiftable load of user j. The operation constraint model of the interruptible load comprises: Where: is the i-th user j in the n-th rolling An operating state that can interrupt the load; is the number of interruptible loads of user j; and are the upper and lower limits of the operating power of the i-th interruptible load of user j respectively; is the power of the i-th interruptible load of user j in time period t after the n-th round of optimization; and are the upper and lower limits of daily power consumption of the i-th interruptible load of user j respectively; and are the upper and lower limits of the operation period of the i-th interruptible load of user j; σ is the optimization time interval; The operation constraint model of the uncontrollable load comprises: In the formula: is the power of the i th uncontrollable load of user j at time period t; is the number of uncontrollable loads of user j; is the corresponding day-ahead forecast power of the i th uncontrollable load of user j.

5. The resident-based peak-cut potential expectation-based distributed rolling optimization method according to claim 4, characterized in that, The global constraint model comprising the power balance constraint model and the peak shaving constraint model based on the expected residential peak shaving potential is established according to the operation constraint models, comprising: The power balance constraint model is established, comprising: In the formula: is the power generation of the household photovoltaic power generation system of user j in time period t, obtained by day-ahead prediction; The peak shaving constraint model based on the expected residential peak shaving potential is established, comprising: In the formula: is the reference net load power of user j at time t in the nth rolling, obtained by day-ahead prediction; is the response potential expectation of user j at time t in the nth rolling; is the upper limit of load reduction amount of user j at time period t in the nth rolling; where: ψ n,j is the set of all possible response events of user j in the nth rolling; ε n,j,q is the probability of the qth possible response event occurring; is the net load power of user j at time t under the qth possible response event; the actual response event is divided into two special cases of participating and not participating in peak shaving plan, is the probability of user j participating in peak shaving demand response in the nth rolling, which is approximately represented by the response frequency obtained from historical data, and is a specific parameter form of the expected resident peak shaving potential.

6. The resident-based peak-cut potential expectation-based distributed rolling optimization method according to claim 5, wherein, The user-side load optimization model based on the expected residential peak shaving potential is constructed according to the global constraint model to generate the planned load sequence, comprising: Establishing the cost of electricity purchase Response comfort cost Photovoltaic power generation and electricity sales revenue And incentive compensation revenue The integrated objective function minF co : wherein The calculation is as follows: wherein: c tou (t) is the time-of-use price function for the user to purchase electricity from the grid, c g,pv is the price of surplus electricity, c pv is the single price of photovoltaic subsidy; and is the response cost coefficient of the user j; C j,0 is the incentive cost starting threshold value paid by the grid to the user, is the incentive subsidy coefficient of the user j; is a quadratic function determined in consideration of the marginal increasing effect of the comfort utility loss; is a linear function established in consideration of consumer psychology and having a starting threshold value and a saturation threshold value; and the nth round of response electricity quantity of the user j is calculated as follows; Based on the above steps, the user-side load optimization model based on the expected residential peak shaving potential is obtained; the user-side load optimization model is solved to obtain the planned load sequence of each user.

7. A distributed rolling optimization system based on resident peak-cut potential expectations, the system configured to perform the optimization method of any one of claims 1-6, wherein, The method comprises the following steps: The power distribution network layer is arranged at the top layer, the load aggregator layer is arranged at the middle layer, and the user layer is arranged at the bottom layer; wherein: The power distribution network layer is configured to transmit load peak information to the load aggregator layer, wherein the load peak information comprises a total load reduction demand and a corresponding peak period set; The power distribution network layer is configured to transmit load peak information to the load aggregator layer, wherein the load peak information comprises a total load reduction demand and a corresponding peak period set; The load aggregator layer is configured to distribute the next-day peak shaving task to the user layer according to the response potential of each resident user, collect the net load sequence of each user, perform aggregation calculation, and transmit the aggregated load data to the power distribution network layer. The user layer is configured to construct a user-side load optimization model based on the resident peak shaving potential expectation by considering various costs and benefits through a user home energy management system, generate a planned load sequence, and feed back to the load aggregator layer.

Images