Micro-grid master-slave game optimization scheduling method considering user flexible load demand response

Abstract

The invention relates to the technical field of power supply and demand interaction, in order to realize efficient and economical power dispatching, a micro-grid master-slave game optimization dispatching method considering user flexible load demand response comprises the following steps: step 1, establishing a micro-grid day-ahead supply and demand interaction scene in which a single micro-grid operator participates with a plurality of users; 2, constructing an MGO leader optimization scheduling model considering the uncertainty of the user response power; step 3, constructing a user follower demand response model taking maximized participation DR income as a target function; 4, converting the chance constraint into a deterministic linear constraint and a second-order cone constraint which can be solved; and 5, solving the MGO and the master-slave game model of each user by adopting an analytic expression solving method. The method is mainly applied to power dispatching occasions.

Description

Technical Field

[0001] The present invention relates to the technical field of interaction between power supply and demand, and in particular to a microgrid master-slave game optimization scheduling method considering uncertainty in demand response. Background Art

[0002] Incentive demand response (DR) is the main way to achieve supply and demand interaction in a microgrid (MG): the microgrid operator (MGO) sets an appropriate DR compensation price based on the user's electricity parameters and the amount of load that needs to be reduced, and encourages and guides users to reasonably reduce and transfer loads by paying compensation to users participating in DR. The effect of supply and demand interaction depends on the rationality of the DR compensation price setting, but existing studies often set the same DR compensation price for different users, ignoring the differences in the responses of different users to the DR compensation price, making the DR compensation price less targeted and the effect of supply and demand interaction poor.

[0003] In terms of uncertain optimization, in order to cope with the uncertainty of renewable distributed generation (RDG), chance constraints can be used to determine the backup capacity required by MG under the required risk level. Among them, existing studies mostly describe the prediction error of RDG as a specific probability distribution function, such as uniform distribution, Gaussian distribution, Weibull distribution, etc. However, according to statistics, a specific form of probability distribution function cannot accurately describe the random prediction error of RDG. At the same time, in the actual DR execution process, affected by a large number of random factors such as user psychology, consumption behavior, and environmental factors, the user's response power is also uncertain, and as the user's autonomy in electricity management increases, the impact of this uncertainty on MG scheduling is gradually highlighted. However, in the supply and demand interaction model of MG, few studies have taken into account the uncertainty of user response power under the DR mechanism, which will make the MG's scheduling results unable to meet the reliability requirements, thereby affecting the power supply quality of MG.

[0004] On the other hand, in solving the MG optimization scheduling problem with chance constraints, existing research only considers the uncertainty of RDG, and the mean and standard deviation of the random variable are known. The method based on the inverse function of the cumulative probability distribution function can be used to convert the chance constraint into a deterministic constraint. However, after considering the uncertainty of the user's response power, the fluctuation range of the random variable is related to the decision variable (DR compensation price), so the above-mentioned chance constraint solution method is not applicable to the scenario considering the uncertainty of DR response power.

[0005] In terms of solving the model of MG supply and demand interaction, the relationship between MGO and each user can be established as a master-slave game model. Considering that the user's response to DR compensation has segmented characteristics, existing research mainly uses heuristic algorithms such as particle swarm optimization and genetic algorithm to iteratively solve the master-slave game model. However, the heuristic algorithm takes a long time to solve and is prone to fall into the local optimal solution. When the number of users is large, the iterative solution method is often computationally inefficient. Summary of the invention

[0006] In order to overcome the shortcomings of the prior art, the present invention aims to propose a new power dispatching technology to achieve efficient and economical power dispatching. To this end, the technical solution adopted by the present invention is a microgrid master-slave game optimization dispatching method considering the user's flexible load demand response, and the steps are as follows:

[0007] Step 1: Establish a microgrid MG day-ahead supply and demand interaction scenario involving a single microgrid operator MGO and multiple users;

[0008] Step 2: Considering the controllable distributed generation CDG operation constraints and energy storage charging and discharging constraints, with minimization of operation cost and incentive demand response DR compensation cost as the objective function, the opportunity constraint based on Gaussian mixture model GMM is used to deal with the uncertainty of renewable distributed generation RDG output and DR response power, and an MGO leader optimization scheduling model considering the uncertainty of user response power is constructed;

[0009] Step 3: Considering the interruptible load IL reduction constraint and the transferable load SL transfer constraint, a user follower demand response model with maximizing the benefits of participating in DR as the objective function is constructed;

[0010] Step 4: Use ladder linearization and second-order cone relaxation to transform the opportunity constraints into solvable deterministic linear constraints and second-order cone constraints, thereby transforming the nonlinear GMM opportunity constraints in the MGO scheduling model into linear and second-order cone constraints that are easy to solve;

[0011] Step 5: Use analytical solution method to solve the master-slave game model between MGO and each user.

[0012] Steps 2-5 are detailed below:

[0013] Step 2: Considering CDG operation constraints and energy storage charging and discharging constraints, with minimization of operation cost and DR compensation cost as the objective function, the GMM-based opportunity constraint is used to deal with the uncertainty of RDG output and DR response power, and an MGO leader optimization scheduling model considering the uncertainty of user response power is constructed. Among them, the objective function C of the MGO leader optimization scheduling model is MGO for:

[0014]

[0015] Where: and The operating cost and compensation cost of MGO in period t respectively; Including the operating costs of CDG and energy storage and the interaction costs with the upper-level main network:

[0016]

[0017] Where: P t CDG is the output of CDG in period t; Uplink and downlink backup reserved for it; a CDG and b CDG are the quadratic and linear power generation cost coefficients of CDG, c CDG,+ and c CDG,- P is the unit uplink and downlink backup cost of CDG. t ch , P t dch is the charging and discharging power of energy storage; b ESS is the unit charging and discharging cost of energy storage; P t buy and P t sell The power purchased and sold by MG from the upper main grid; and The time-of-use purchase and sale price of electricity of the upper main grid; and The uplink and downlink spare capacity purchased by MGO from the upper main network, c UG,+ and c UG,- It is the unit uplink and downlink standby price of the upper main network;

[0018] MGO sets differentiated DR compensation prices for different users and pays DR compensation to users based on the amount of load reduction of each user. The DR compensation cost Calculate according to the following formula:

[0019]

[0020] Where: is the DR compensation price set by MGO for user i in period t, is the interruptible load reduction of user i in time period t; and The amount of transferable load transferred out and in by user i in time period t, N user It is the user set in MG;

[0021] For MGO, the DR incentive plan is only implemented during the preset period, that is, during the period of DR implementation, Outside of this period, And considering that in order to maximize the benefits of participating in DR, each user will transfer the SL during the DR period to the non-DR period, that is, during the DR period, Therefore, the DR compensation cost is further expressed as:

[0022]

[0023] Where: and is the DR compensation obtained by user i due to IL reduction and SL transfer, expressed as:

[0024]

[0025] Since the RDG output is uncontrollable, considering that the RDG operates according to its predicted output, the controllable devices of MGO include CDG and energy storage, and the following constraints must be met during its operation:

[0026] (1) CDG output, reserve and climbing constraints

[0027]

[0028] Where: is the maximum and minimum output of CDG; is the maximum up and down ramp rate of CDG;

[0029] (2) Energy storage charging and discharging constraints

[0030]

[0031] Where: is the maximum charge and discharge power of energy storage; η ch and η dch The charging and discharging efficiency of energy storage; is the upper limit of the energy state of the energy storage;

[0032] (3) Interaction constraints with the upper-level main network

[0033]

[0034] Where: The upper limit of the interaction power between MG and main network;

[0035] (4) Day-ahead power balance and reserve constraints

[0036] The day-ahead power balance constraint within the MG can be expressed as:

[0037]

[0038] Where: P t RDG It is the day-ahead predicted output of RDG, i.e. the sum of the predicted outputs of wind power WT and photovoltaic PV. is the baseline load of user i in period t, i.e., the load before participating in DR;

[0039] Considering that users are not completely rational and their power environment parameters fluctuate, the DR response power of each user is also uncertain. Since the RDG output and DR response power inside the MG are uncertain, the MG needs to reserve a certain amount of reserve capacity in the day-ahead stage to cope with the fluctuations of RDG and user DR response power. The probability of insufficient reserve does not exceed the given risk level ε. Therefore, the spinning reserve constraint is expressed in the form of an opportunity constraint:

[0040]

[0041] Where: and It is the fluctuation of RDG output and user DR response. If it is greater than 0, it means the actual power is greater than the predicted value. The probability distribution of can be fitted as a Gaussian distribution with a mean of 0, and its probability density function f t DR It is expressed as:

[0042]

[0043] Where: f GPF is the probability density function of the Gaussian distribution, For t DR The standard deviation of the user's total power adjustment is γ DR Times:

[0044]

[0045] Due to the fluctuation of RDG output The probability distribution of does not usually follow a specific probability distribution function. In order to more accurately characterize The probability distribution of The probability density function is fitted. In the GMM model, The probability density function f t RDG Expressed as a weighted sum of the density functions of multiple distributed Gaussians:

[0046]

[0047] Where: N GMM is the total number of Gaussian components. k is the weight of the kth Gaussian component, and is the mean and standard deviation of the corresponding Gaussian component, ω k , and The maximum expectation method can be used to fit the historical data of RDG output fluctuations;

[0048] (5) DR compensation price constraints

[0049]

[0050] Considering that MGO sets different DR compensation prices for different users, To compensate for the price ceiling, T DR A collection of time periods for implementing DR;

[0051] Step 3: Considering the IL reduction constraint and SL transfer constraint, a user follower demand response model with the objective function of maximizing the benefits of participating in DR is constructed:

[0052] User i(i∈N user ) As a follower, according to the DR compensation price set by MGO, adjust its own electricity consumption plan to maximize the benefits of participating in DR The target is the difference between the DR subsidy received by the user and the load adjustment cost:

[0053]

[0054] Where: is the DR compensation obtained by user i from MGO, The load adjustment cost includes IL reduction cost and SL transfer-out and transfer-in cost:

[0055]

[0056] Where: and are the quadratic and linear cost coefficients of IL reduction cost. and The quadratic and linear cost coefficients of SL transfer-out / transfer-in costs;

[0057] Users need to meet the following constraints:

[0058] (1) IL reduction constraints

[0059] Commercial loads, non-emergency lighting loads, and non-important industrial loads are interrupted or reduced in power consumption according to the dispatch requirements of MGO. However, in order to meet the basic power demand of users, the reduction amount of IL is not allowed to exceed its upper limit. Therefore, the reduction constraint of IL is expressed as:

[0060]

[0061] Where: is the proportion of IL in user i, is the maximum reduction ratio of IL;

[0062] (2) SL transfer constraints

[0063] The SL users such as electric vehicle charging load, water heater and washing machine transfer the load in some periods to other periods, and the total power consumption remains unchanged before and after the transfer. Assuming that each user transfers the load in the DR period to the non-DR period in order to maximize the benefits of participating in DR, the reduction constraint of SL is expressed as:

[0064]

[0065] Where: is the proportion of SL among user i;

[0066] Step 4: Use ladder linearization and second-order cone relaxation to transform the chance constraints into solvable deterministic linear constraints and second-order cone constraints;

[0067] Defining random variables Expressed as the power fluctuation of the net load, and considering and Independent of each other. According to the convolution theorem and the operation rules of Gaussian function, Probability density function f t NL It is expressed as:

[0068]

[0069] visible, It also obeys the GMM distribution, and the mean and standard deviation of its kth Gaussian component are and

[0070] Definition F t NL for The cumulative probability distribution function, F t NL It is expressed as:

[0071]

[0072] Where: F GCF is the cumulative probability distribution function of the standard Gaussian distribution;

[0073] Therefore, the upward reserve constraint in the opportunity constraint is transformed into:

[0074]

[0075] Define a new variable x k,t , whose expression is:

[0076]

[0077] Then the upward reserve constraint in the opportunity constraint can be transformed into:

[0078]

[0079] Introduce binary auxiliary variable u k,j and related constraints, j represents the jth step segment, and F is linearized by stepwise stepwise linearization. GCF [(x k,t ) -1 ] is approximately converted into a linear function, then the upward reserve constraint can be converted into:

[0080]

[0081] remember Introduce auxiliary variable z j,t , using the big M method and second-order cone relaxation to k,t Translates to:

[0082]

[0083] Where: M is a sufficiently large constant. t , And the objective function of MGO is about and Monotonically increasing, the second-order cone relaxation error tends to 0 and is ignored. At this time, the uplink reserve constraint in the nonlinear chance constraint has been transformed into a linear and second-order cone constraint without random variables, and the downlink reserve in the GMM chance constraint can also be deterministically transformed in the same way.

[0084] Step 5: Use analytical solution method to solve the master-slave game model between the leader MGO and each user, i.e., the follower:

[0085] The optimization problem of each user in the lower layer is transformed into the following constraint formula by using analytical method and added into the optimization problem of MGO, thus transforming the two-layer optimization problem into a single-layer optimization problem:

[0086]

[0087]

[0088]

[0089] Where: NNDR The number of time periods during which the DR plan is not implemented within the scheduling cycle.

[0090] Therefore, the MGO objective function Item and The items are represented as:

[0091]

[0092]

[0093] After the user's constraint formula is introduced into the optimization problem of MGO, although the two-layer optimization model of the master-slave game can be transformed into a single-layer optimization model, since the above constraints are segmented constraints containing judgment conditions, direct introduction will also lead to the MGO objective function. Item and The term becomes a piecewise function containing judgment conditions, which is difficult to solve directly. Therefore, a binary auxiliary variable is introduced and continuous auxiliary variables The above constraints are transformed into the following linear constraints using the Big M method:

[0094]

[0095]

[0096]

[0097] thus, Item and The term is transformed into a convex function expression as shown below.

[0098]

[0099]

[0100] The characteristics and beneficial effects of the present invention are:

[0101] The invention constructs an MG master-slave game optimization scheduling model that takes into account the uncertainty of DR response power, which improves the response effect of supply and demand interaction within MG under uncertain environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 Overall flow chart of the present invention. DETAILED DESCRIPTION

[0103] In order to solve the problems in the prior art, based on various reasons and advantages and disadvantages, the present invention constructs an MG master-slave game optimization scheduling model considering the uncertainty of DR response power to improve the response effect of supply and demand interaction within MG under uncertain environment.

[0104] The problem to be solved by the present invention is the problem of collaborative optimization scheduling between MG and multiple users under an environment where RDG output and user response power are uncertain.

[0105] The present invention provides a microgrid master-slave game optimization scheduling method considering the uncertainty of demand response, and the method comprises the following steps.

[0106] Step 1: Establish a microgrid MG day-ahead supply and demand interaction scenario involving a single microgrid operator MGO and multiple users;

[0107] Step 2: Considering the controllable distributed generation (CDG) operation constraints and energy storage charging and discharging constraints, with minimizing the operation cost and incentive demand response DR compensation cost as the objective function, the opportunity constraint based on the Gaussian mixture model (GMM) is used to deal with the uncertainty of renewable distributed generation RDG output and DR response power, and an MGO leader optimization scheduling model considering the uncertainty of user response power is constructed;

[0108] Step 3: Considering the interruptible load IL reduction constraint and the transferable load SL transfer constraint, a user follower demand response model with maximizing the benefits of participating in DR as the objective function is constructed;

[0109] Step 4: Use ladder linearization and second-order cone relaxation to transform the chance constraints into solvable deterministic linear constraints and second-order cone constraints;

[0110] Step 5: Use analytical solution method to solve the master-slave game model between MGO and each user.

[0111] Step 2 and step 3 construct the scheduling models of MGO and each user respectively; step 4 transforms the nonlinear GMM opportunity constraints in the MGO scheduling model into linear and second-order cone constraints that are easy to solve; step 5 transforms the two-level optimization problem corresponding to the master-slave game between MGO and multiple users into a single-level optimization problem.

[0112] The user follower demand response model adjusts its own electricity consumption plan according to the DR compensation price set by MGO, and considers the different DR compensation prices set by MGO for different users.

[0113] The present invention is further described in detail below with reference to the accompanying drawings and specific examples.

[0114] An example of the present invention includes the following specific steps:

[0115] Step 1: Figure 1 As shown, a MG day-ahead supply and demand interaction scenario involving a single MGO and multiple users is established.

[0116] Step 2: Considering CDG operation constraints, energy storage charging and discharging constraints, etc., with minimizing operation costs and DR compensation costs as the objective function, the GMM-based opportunity constraint is used to deal with the uncertainty of RDG output and DR response power, and an MGO leader optimization scheduling model considering the uncertainty of user response power is constructed.

[0117] Among them, the objective function C of the MGO leader optimization scheduling model MGO for:

[0118]

[0119] Where: and The operating cost and compensation cost of MGO in period t respectively; Including the operating costs of CDG and energy storage and the interaction costs with the upper-level main network:

[0120]

[0121] Where: P t CDG is the output of CDG in period t; Uplink and downlink backup reserved for it; a CDG and b CDG are the quadratic and linear power generation cost coefficients of CDG, c CDG,+ and c CDG,- P is the unit uplink and downlink backup cost of CDG. t ch , P t dch is the charging and discharging power of energy storage; b ESS P is the unit charging and discharging cost of energy storage. t buy and P t sell The power purchased and sold by MG from the upper main grid; and R is the time-sharing purchase and sale price of the upper main grid; t UG,+ and R t UG,- The uplink and downlink spare capacity purchased by MGO from the upper main network, c UG,+ and c UG,- It is the unit uplink and downlink standby price of the upper main network.

[0122] MGO sets differentiated DR compensation prices for different users and pays DR compensation to users based on the amount of load reduction of each user. The DR compensation cost Calculate according to the following formula:

[0123]

[0124] Where: is the DR compensation price set by MGO for user i in time period t. is the interruptible load reduction of user i in time period t; and The amount of transferable load transferred out and in by user i in time period t. N user It is the user collection in MG.

[0125] The present invention assumes that the MGO implements the DR incentive plan only during a preset period, i.e., during the period of implementing DR, Outside of this period, And considering that in order to maximize the benefits of participating in DR, each user will transfer the SL during the DR period to the non-DR period, that is, during the DR period, Therefore, the DR compensation cost can be further expressed as:

[0126]

[0127] Where: and is the DR compensation obtained by user i due to IL reduction and SL transfer, which can be expressed as:

[0128]

[0129] Since the output of RDG is uncontrollable, the present invention considers that RDG operates according to its predicted output. Therefore, the controllable devices of MGO include CDG and energy storage, and the following constraints must be met during its operation:

[0130] (1) CDG output, reserve and climbing constraints

[0131]

[0132] Where: is the maximum and minimum output of CDG; is the maximum uplink and downlink climbing rate of CDG.

[0133] (2) Energy storage charging and discharging constraints

[0134]

[0135] Where: is the maximum charge and discharge power of energy storage; η ch and η dch The charging and discharging efficiency of energy storage; is the upper limit of the energy storage state.

[0136] (3) Interaction constraints with the upper-level main network

[0137]

[0138] Where: It is the upper limit of the interaction power between MG and main network.

[0139] (4) Day-ahead power balance and reserve constraints

[0140] The day-ahead power balance constraint within the MG can be expressed as:

[0141]

[0142] Where: P t RDG It is the day-ahead predicted output of RDG, i.e. the sum of the predicted outputs of wind turbine (WT) and photovoltaic (PV). is the baseline load of user i in period t, that is, the load before participating in DR.

[0143] Considering that users are not completely rational and their power environment parameters (such as temperature and humidity) fluctuate, the DR response power of each user is also uncertain. Since the RDG output and DR response power inside the MG are uncertain, the MG needs to reserve a certain amount of reserve capacity in the day-ahead stage to cope with the fluctuations of RDG and user DR response power, and the probability of insufficient reserve does not exceed the given risk level ε. Therefore, the spinning reserve constraint can be expressed in the form of an opportunity constraint:

[0144]

[0145] Where: and It is the fluctuation of RDG output and user DR response. A value greater than 0 indicates that the actual power is greater than the predicted value. The probability distribution of can be fitted as a Gaussian distribution with a mean of 0, and its probability density function f t DR It can be expressed as:

[0146]

[0147] Where: f GPF is the probability density function of the Gaussian distribution. for The standard deviation of the user's total power adjustment is γ DR Times:

[0148]

[0149] Due to the fluctuation of RDG output The probability distribution of does not usually follow a specific probability distribution function. In order to more accurately characterize The probability distribution of The probability density function of is fitted. In the GMM model, The probability density function f t RDG It can be expressed as a weighted sum of the density functions of multiple distributed Gaussians:

[0150]

[0151] Where: N GMM is the total number of Gaussian components. k is the weight of the kth Gaussian component, and is the mean and standard deviation of the corresponding Gaussian component. k , and The maximum expectation method can be used to fit the historical data of RDG output fluctuations.

[0152] (5) DR compensation price constraints

[0153]

[0154] Different users have different adjustable load parameters, so their response to DR compensation prices is also different. In order to improve the pertinence of DR compensation prices and promote the supply-demand interaction effect within MG, consider setting different DR compensation prices for different users by MGO. To compensate for the price ceiling, T DR A collection of time periods for implementing DR.

[0155] Step 3: Considering the IL reduction constraint and SL transfer constraint, a user follower demand response model is constructed with the objective function of maximizing the benefits of participating in DR.

[0156] User i(i∈N user ) As a follower, according to the DR compensation price set by MGO, adjust its own electricity consumption plan to maximize the benefits of participating in DR The target is the difference between the DR subsidy received by the user and the load adjustment cost:

[0157]

[0158] Where: is the DR compensation obtained by user i from MGO. The load adjustment cost includes IL reduction cost and SL transfer-out and transfer-in cost:

[0159]

[0160]

[0161] Where: and are the quadratic and linear cost coefficients of IL reduction cost. and It is the quadratic and linear cost coefficients of SL transfer-out / transfer-in costs.

[0162] Users need to meet the following constraints:

[0163] (1) IL reduction constraints

[0164] Commercial loads, non-emergency lighting loads, and non-important industrial loads can interrupt or reduce their power consumption according to the dispatch requirements of MGO. However, in order to meet the basic power demand of users, the reduction amount of IL is not allowed to exceed its upper limit. Therefore, the reduction constraint of IL can be expressed as:

[0165]

[0166] Where: is the proportion of IL in user i, is the maximum reduction ratio of IL.

[0167] (2) SL transfer constraints

[0168] SL users such as electric vehicle charging load, water heaters and washing machines can transfer loads from some periods to other periods, and the total power consumption remains unchanged before and after the transfer. Assuming that each user, in order to maximize the benefits of participating in DR, only transfers the load from the DR period to the non-DR period when transferring loads. Therefore, the reduction constraint of SL can be expressed as:

[0169]

[0170] Where: is the SL proportion of user i.

[0171] Step 4: Use ladder linearization and second-order cone relaxation to transform the chance constraints into solvable deterministic linear constraints and second-order cone constraints.

[0172] Defining random variables Expressed as the power fluctuation of the net load, and considering and Independent of each other. According to the convolution theorem and the operation rules of Gaussian function, Probability density function f tNL It can be expressed as:

[0173]

[0174] visible, It also obeys the GMM distribution, and the mean and standard deviation of its kth Gaussian component are and

[0175] Definition F t NL for The cumulative probability distribution function, F t NL It can be expressed as:

[0176]

[0177] Where: F GCF is the cumulative probability distribution function of the standard Gaussian distribution.

[0178] Therefore, the upward reserve constraint in the opportunity constraint can be transformed into:

[0179]

[0180] Define a new variable x k,t , whose expression is:

[0181]

[0182] Then the upward reserve constraint in the opportunity constraint can be transformed into:

[0183]

[0184] Introduce binary auxiliary variable u k,j and related constraints (j represents the jth step segment), F is linearized by piecewise stepwise linearization. GCF [(x k,t ) -1 ] is approximately converted into a linear function, then the upward reserve constraint can be converted into:

[0185]

[0186] remember Introduce auxiliary variable z j,t , using the big M method and second-order cone relaxation to k,t Translates to:

[0187]

[0188] Where: M is a sufficiently large constant. t , And the objective function of MGO is about and Monotonically increasing, the second-order cone relaxation error approaches 0 and can be ignored. At this point, the uplink reserve constraint in the nonlinear chance constraint has been transformed into a linear and second-order cone constraint without random variables. The downlink reserve in the GMM chance constraint can also be deterministically transformed in the same way.

[0189] Step 5: Use analytical solution method to solve the master-slave game model between MGO (leader) and each user (follower).

[0190] Users are at the bottom of the master-slave game and can only passively accept the subsidy price set by MGO, but the user's response volume will eventually affect MGO's demand response cost. The analytical method can transform the optimization problem of each user at the bottom into the following constraint formula and add it to the optimization problem of MGO, thereby transforming the two-layer optimization problem into a single-layer optimization problem:

[0191]

[0192]

[0193]

[0194] Where: N NDR The number of time periods during which the DR plan is not implemented within the scheduling cycle.

[0195] Therefore, the MGO objective function Item and The items can be expressed as:

[0196]

[0197]

[0198] After the user's constraint formula is introduced into the optimization problem of MGO, although the two-layer optimization model of the master-slave game can be transformed into a single-layer optimization model, since the above constraints are segmented constraints containing judgment conditions, direct introduction will also lead to the MGO objective function. Item and The term becomes a piecewise function containing judgment conditions, which is difficult to solve directly. Therefore, the present invention introduces a binary auxiliary variable and continuous auxiliary variables Using the Big M method, the above constraints can be transformed into the following linear constraints:

[0199]

[0200]

[0201]

[0202] thus, Item and The term can be transformed into a convex function expression as shown below.

[0203]

[0204]

[0205] At this point, the optimization problem of each user in the lower layer can be transformed into the above-mentioned linear constraints and added to the optimization scheduling problem of MGO. After the equivalent transformation, the objective function of MGO is transformed into a quadratic convex function without judgment conditions that is easy to solve. The master-slave game model of MGO and multiple users has been transformed into a mixed integer quadratic programming model, which can be efficiently solved using commercial solvers such as CPLEX and Gurobi.

[0206] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A microgrid master-slave game optimization scheduling method considering user flexible load demand response, characterized in that the steps are: as follows: Step 1: Establish a microgrid MG day-ahead supply and demand interaction scenario involving a single microgrid operator MGO and multiple users; Step 2: Considering the controllable distributed generation CDG operation constraints and energy storage charging and discharging constraints, with minimization of operation cost and incentive demand response DR compensation cost as the objective function, the opportunity constraint based on Gaussian mixture model GMM is used to deal with the uncertainty of renewable distributed generation RDG output and DR response power, and an MGO leader optimization scheduling model considering the uncertainty of user response power is constructed; Step 3: Considering the interruptible load IL reduction constraint and the transferable load SL transfer constraint, a user follower demand response model with maximizing the benefits of participating in DR as the objective function is constructed; Step 4: Use ladder linearization and second-order cone relaxation to transform the opportunity constraints into solvable deterministic linear constraints and second-order cone constraints, thereby transforming the nonlinear GMM opportunity constraints in the MGO scheduling model into linear and second-order cone constraints that are easy to solve; Step 5: Use analytical solution method to solve the master-slave game model between MGO and each user.

2. The microgrid master-slave game optimization scheduling method considering user flexible load demand response as claimed in claim 1 is characterized in that: Steps 2-5 are detailed below: Step 2: Considering CDG operation constraints and energy storage charging and discharging constraints, with minimization of operation cost and DR compensation cost as the objective function, the GMM-based opportunity constraint is used to deal with the uncertainty of RDG output and DR response power, and an MGO leader optimization scheduling model considering the uncertainty of user response power is constructed. Among them, the objective function C of the MGO leader optimization scheduling model is MGO for: Where: and The operating cost and compensation cost of MGO in period t respectively; Including the operating costs of CDG and energy storage and the interaction costs with the upper-level main network: Where: P t CDG is the output of CDG in period t; Uplink and downlink backup reserved for it; a CDG and b CDG are the quadratic and linear generation cost coefficients of CDG, c CDG,+ and c CDG,- P is the unit uplink and downlink backup cost of CDG. t ch , P t dch is the charging and discharging power of energy storage; b ESS is the unit charging and discharging cost of energy storage; P t buy and P t sell The power purchased and sold by MG from the upper main grid; and The time-of-use purchase and sale price of electricity of the upper main grid; and The uplink and downlink spare capacity purchased by MGO from the upper main network, c UG,+ and c UG,- It is the unit uplink and downlink standby price of the upper main network; MGO sets differentiated DR compensation prices for different users and pays DR compensation to users based on the amount of load reduction of each user. The DR compensation cost Calculate according to the following formula: Where: is the DR compensation price set by MGO for user i in period t, is the interruptible load reduction of user i in time period t; and The amount of transferable load transferred out and in by user i in time period t, N user It is the user set in MG; For MGO, the DR incentive plan is only implemented during the preset period, that is, during the period of DR implementation, Outside of this period, And considering that in order to maximize the benefits of participating in DR, each user will transfer the SL during the DR period to the non-DR period, that is, during the DR period, Therefore, the DR compensation cost is further expressed as: Where: and is the DR compensation obtained by user i due to IL reduction and SL transfer, expressed as: Since the RDG output is uncontrollable, considering that the RDG operates according to its predicted output, the controllable devices of MGO include CDG and energy storage, and the following constraints must be met during its operation: (1) CDG output, reserve and climbing constraints Where: is the maximum and minimum output of CDG; is the maximum up and down ramp rate of CDG; (2) Energy storage charging and discharging constraints Where: is the maximum charge and discharge power of energy storage; η ch and η dch The charging and discharging efficiency of energy storage; is the upper limit of the energy state of the energy storage; (3) Interaction constraints with the upper-level main network Where: The upper limit of the interaction power between MG and main network; (4) Day-ahead power balance and reserve constraints The day-ahead power balance constraint within the MG can be expressed as: Where: P t RDG It is the day-ahead predicted output of RDG, i.e. the sum of the predicted outputs of wind power WT and photovoltaic PV. is the baseline load of user i in period t, i.e., the load before participating in DR; Considering that users are not completely rational and their power environment parameters fluctuate, the DR response power of each user is also uncertain. Since the RDG output and DR response power inside the MG are uncertain, the MG needs to reserve a certain amount of reserve capacity in the day-ahead stage to cope with the fluctuations of RDG and user DR response power. The probability of insufficient reserve does not exceed the given risk level ε. Therefore, the spinning reserve constraint is expressed in the form of an opportunity constraint: Where: and It is the fluctuation of RDG output and user DR response. If it is greater than 0, it means the actual power is greater than the predicted value. The probability distribution of can be fitted as a Gaussian distribution with a mean of 0, and its probability density function f t DR It is expressed as: Where: f GPF is the probability density function of the Gaussian distribution, for The standard deviation of the user's total power adjustment is γ DR Times: Due to the fluctuation of RDG output The probability distribution of does not usually follow a specific probability distribution function. In order to more accurately characterize The probability distribution of The probability density function is fitted. In the GMM model, The probability density function f t RDG Expressed as a weighted sum of the density functions of multiple distributed Gaussians: Where: N GMM is the total number of Gaussian components. k is the weight of the kth Gaussian component, and is the mean and standard deviation of the corresponding Gaussian component, ω k , and The maximum expectation method can be used to fit the historical data of RDG output fluctuations; (5) DR compensation price constraints Considering that MGO sets different DR compensation prices for different users, To compensate for the price ceiling, T DR A collection of time periods for implementing DR; Step 3: Considering the IL reduction constraint and SL transfer constraint, a user follower demand response model with the objective function of maximizing the benefits of participating in DR is constructed: user As a follower, adjust your electricity consumption plan according to the DR compensation price set by MGO to maximize the benefits of participating in DR The target is the difference between the DR subsidy received by the user and the load adjustment cost: Where: is the DR compensation obtained by user i from MGO, The load adjustment cost includes IL reduction cost and SL transfer-out and transfer-in cost: Where: and are the quadratic and linear cost coefficients of IL reduction cost. and The quadratic and linear cost coefficients of SL transfer-out / transfer-in costs; Users need to meet the following constraints: (1) IL reduction constraints Commercial loads, non-emergency lighting loads, and non-important industrial loads are interrupted or reduced in power consumption according to the dispatch requirements of MGO. However, in order to meet the basic power demand of users, the reduction amount of IL is not allowed to exceed its upper limit. Therefore, the reduction constraint of IL is expressed as: Where: is the proportion of IL in user i, is the maximum reduction ratio of IL; (2) SL transfer constraints The SL users such as electric vehicle charging load, water heater and washing machine transfer the load in some periods to other periods, and the total power consumption remains unchanged before and after the transfer. Assuming that each user transfers the load in the DR period to the non-DR period in order to maximize the benefits of participating in DR, the reduction constraint of SL is expressed as: Where: is the proportion of SL among user i; Step 4: Use ladder linearization and second-order cone relaxation to transform the chance constraints into solvable deterministic linear constraints and second-order cone constraints; Defining random variables Expressed as the power fluctuation of the net load, and considering and Independent of each other. According to the convolution theorem and the operation rules of Gaussian function, Probability density function f t NL It is expressed as: visible, It also obeys the GMM distribution, and the mean and standard deviation of its kth Gaussian component are and Definition F t NL for The cumulative probability distribution function, F t NL It is expressed as: Where: F GCF is the cumulative probability distribution function of the standard Gaussian distribution; Therefore, the upward reserve constraint in the opportunity constraint is transformed into: Define a new variable x k,t , whose expression is: Then the upward reserve constraint in the opportunity constraint can be transformed into: Introduce binary auxiliary variable u k,j and related constraints, j represents the jth step segment, and F is linearized by stepwise stepwise linearization. GCF [(x k,t ) -1 ] is approximately converted into a linear function, then the upward reserve constraint can be converted into: remember Introduce auxiliary variable z j,t , using the big M method and second-order cone relaxation to k,t Translates to: Where: M is a sufficiently large constant. t , And the objective function of MGO is about and Monotonically increasing, the second-order cone relaxation error tends to 0 and is ignored. At this time, the uplink reserve constraint in the nonlinear chance constraint has been transformed into a linear and second-order cone constraint without random variables, and the downlink reserve in the GMM chance constraint can also be deterministically transformed in the same way. Step 5: Use analytical solution method to solve the master-slave game model between the leader MGO and each user, i.e., the follower: The optimization problem of each user in the lower layer is transformed into the following constraint formula by using analytical method and added into the optimization problem of MGO, thus transforming the two-layer optimization problem into a single-layer optimization problem: Where: N NDR The number of time periods during which the DR plan is not implemented within the scheduling cycle. Therefore, the MGO objective function Item and The items are represented as: After the user's constraint formula is introduced into the optimization problem of MGO, although the two-layer optimization model of the master-slave game can be transformed into a single-layer optimization model, since the above constraints are segmented constraints containing judgment conditions, direct introduction will also lead to the MGO objective function. Item and The term becomes a piecewise function containing judgment conditions, which is difficult to solve directly. Therefore, a binary auxiliary variable is introduced and continuous auxiliary variables The above constraints are transformed into the following linear constraints using the Big M method: thus, Item and The term is transformed into a convex function expression as shown below.

Images