Electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling

Abstract

The invention relates to the technical field of power system demand side management and optimization operation, and discloses an electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling, which comprises the following steps: establishing a directly-heated electric heating indoor heat transfer model to analyze a single user to obtain load characteristics of heating users; a decentralized electric heating load refined aggregation model is constructed, and decentralized heating users are divided into different clusters; an adjustable potential quantification model is constructed, and the adjustable electric quantity and the adjustable potential of each cluster are obtained; and a heating load group optimization operation model based on the master-slave game is constructed, power grid purchase and sale electricity price information, distributed power supply cost information, adjustable electric quantity and adjustable potential are input, and a power utilization behavior adjustment strategy is obtained. According to the method, through refined house type clustering, master-slave game framework design and an improved solution algorithm, precise mining and optimized operation of the heating load group regulation potential are realized, the peak regulation pressure of a power grid is effectively reduced, and the comfort of a user is considered.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system demand side management and optimal operation, in particular to an electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling. BACKGROUND

[0002] Under the drive of the "double carbon" policy, electric heating equipment is promoted on a large scale at the user side due to its cleanliness, high efficiency and strong controllability. However, the disordered access of a large number of electric heating loads contributes to energy low-carbon transformation, but also brings great pressure to power grid peak shaving. The current heating load aggregation model and response potential evaluation model have some problems. The number and characteristics of users are often set as fixed parameters, ignoring the influence of house type difference on the adjustable potential of the cluster, resulting in rough aggregation and inaccurate evaluation. Although the load aggregator (LA) has been introduced to participate in management, the existing strategies mainly focus on the interaction between the grid side and the LA, and the effective game interaction and benefit balance between the LA and the large number of dispersed users are less studied. At the same time, the existing model is insufficient in refined management, and lacks a systematic solution to the fine balance between user comfort and adjustable potential and the optimization of aggregation matching the differentiated dispatching needs of the grid according to the characteristics of the house type.

[0003] Therefore, there is an urgent need for a dispersed heating load group optimization operation strategy that can finely depict user differences, effectively coordinate the interests of the LA and users, and has efficient solving ability. SUMMARY

[0004] The present application provides an electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling, which is aimed at the problems of ignoring the influence of house type difference, lacking effective LA-user interactive game mechanism and model solution being prone to local optimum in the existing dispersed electric heating load aggregation and optimal operation technology. Through fine house type clustering, master-slave game framework design and improved solving algorithm, the method realizes the accurate excavation and optimal operation of the regulation and control potential of the heating load group, effectively reduces the pressure of power grid peak shaving, and at the same time takes into account the LA income and user comfort and economy.

[0005] The present application provides an electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling, which includes:

[0006] S1, establishing a direct heating type electric heating indoor heat transfer model and an adjustable potential basic model, and analyzing a single user by using the direct heating type electric heating indoor heat transfer model to obtain the load characteristics of a single heating user;

[0007] S2, constructing a fine aggregation model of dispersed electric heating load considering house type difference, and dividing the dispersed heating users into different clustering clusters by using an affinity propagation clustering (AP) algorithm;

[0008] S3, constructing an adjustable potential model considering the difference of house types, and obtaining the up-regulated power and down-regulated power of each cluster according to the load characteristics of the single heating user, and quantifying the adjustable potential of the user house group in each cluster;

[0009] S4, constructing a heating load group optimal operation model based on principal-agent game, and adopting a dynamic adjustment strategy of inertia factor parameter in a particle swarm algorithm for optimization;

[0010] S5, inputting the grid buying and selling price information, distributed power cost information, and the up-regulated power and down-regulated power and adjustable potential obtained in step S3 into the optimized heating load group optimal operation model to obtain an optimal selling price strategy and an optimal power consumption behavior adjustment strategy of the user.

[0011] Further, the step S1 specifically comprises:

[0012] S101, establishing an operation model of a single direct electric heating device based on an equivalent thermal parameter ETP model, controlling the start and stop of the electric heating device in real time to depict the heat balance mechanism of the user by comparing the room temperature with the upper and lower limits of the user comfort interval, and establishing an indoor temperature change equation:

[0013] θin(t+1)=θout(t)+(R*Prated*S(t)+R*Psolar-(θin(t)-θout(t))*(Δt / (R*C))

[0014] Wherein, θin(t) is the room temperature at t time; θout(t) is the outdoor temperature at t time; R is the room equivalent thermal resistance; C is the room equivalent thermal capacity; Prated is the rated power of the electric heating device; S(t) is the device start-stop state at t time; Psolar is the solar radiation power; Δt is the time step;

[0015] S102, based on the operation model in step S101, calculating the up-regulated time length Tup and the down-regulated time length Tdown of the single heating device under the constraint of user comfort degree θmin≤θin(t)≤θmax as the adjustable potential basic model:

[0016] Tup=(θmax-θin(t)) / (dθin / dt|D=1)

[0017] Tdown=(θin(t)-θmin / (dθin / dt|S=0)

[0018] Wherein, S is a binary variable, used to represent whether the state or condition is met;

[0019] S103, taking the adjustable potential base model of the single heating device under the user comfort constraint as the load characteristics of the single heating user.

[0020] Further, the step S2 specifically comprises:

[0021] adopting an affinity propagation clustering (AP) algorithm, and performing differential clustering on the distributed heating users according to the rated power P rated and the equivalent thermal parameters of the house; wherein the equivalent thermal parameters of the house include a thermal resistance R and a thermal capacity C; and the differential clustering specifically comprises:

[0022] 1) performing preliminary clustering according to the rated power P rated parameters to obtain preliminary clustering clusters;

[0023] 2) performing fine clustering according to the equivalent thermal parameters R and C of the house type in each preliminary clustering cluster to obtain final secondary clustering clusters j, and adaptively determining the number and members of the clustering clusters according to the data distribution.

[0024] Further, the step S3 specifically comprises:

[0025] S301, calculating the adjustable duration of the user house type i in each clustering cluster j at the time t according to the load characteristics of the single heating user, as the adjustable potential of the user house type group in each clustering cluster, and the calculation formula is:

[0026] Tup j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i)

[0027] Tdown j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i)

[0028] wherein Tup j ,i(t) and Tdown j ,i(t) respectively represent how long the average user in the group can delay heating and how long the average user in the group can turn off heating at the moment, f is a function for calculating the adjustable duration; R j is the equivalent thermal resistance of the house in the jth clustering cluster, i is the user house type in the clustering cluster j, C j is the equivalent thermal capacity of the house in the jth clustering cluster, and Prated_j,i is the rated power of the electric heating device in the jth clustering cluster.

[0029] S302, calculate the up-regulated power E_up_LA(t) and the down-regulated power E_down_LA(t) of each cluster:

[0030]

[0031] wherein M is the total number of aggregated clusters; Lj is the number of house types of the aggregated cluster j; ξ ji is the proportion of the number of users of the house type i of the cluster j to the total number of users, B and B ji,t is the up-regulated power proportion, indicating the proportion of users of the house type i in the jth cluster who choose to up-regulate the power at time t, and is the up-regulated duration and the down-regulated duration of the house type i in the jth cluster at time t, is the rated power of the jth cluster;

[0032] The total up-regulated power and the total down-regulated power are obtained by summing the up-regulated power and the down-regulated power of all the clusters, and then the specific up-regulated power and the specific down-regulated power of each cluster are further calculated according to the proportion of users and the rated power of each cluster.

[0033] Further, the heating load group optimization operation model based on the principal-agent game in the step S4 comprises:

[0034] A principal-agent game framework is established, in which LA is the upper leader and the distributed heating users are the lower followers:

[0035] 1) In the upper LA pricing decision model, the objective function is to maximize the LA net profit Profit_LA:

[0036] MaxProfit_LA = Revenue_sale - Cost_net - Cost_DG - Subsidy_user

[0037] The electricity selling revenue:

[0038] Revenue_sale = ∑ t (ρ_LA(t) * P_sale(t))

[0039] The net electricity purchasing cost:

[0040] Cost_net = ∑ t (ρ_buy_grid(t) * P_buy(t) - ρ_sale_grid(t) * P_sale(t))

[0041] The distributed power supply cost:

[0042] Cost_DG = ∑ t(c_G * p_DG(t))

[0043] User incentive subsidy:

[0044] Subsidy_user =∑ t ∑ j ∑ i (ρ_cut * L_cut_j,i(t))

[0045] Where ρ_LA(t) is the electricity selling price set by LA; P_sale(t) is the electricity selling amount of LA; ρ_buy_grid(t), ρ_sale_grid(t) are the electricity buying / selling price of the grid; P_buy(t), P_sale(t) are the electricity buying / selling amount of LA to the grid; c_G is the distributed power cost coefficient; p_DG(t) is the output of the distributed power at time t; ρ_cut is the electricity reduction subsidy price; L_cut_j,i(t) is the load reduction amount of the cluster j household i at time t.

[0046] 2) In the lower-layer user response decision model, the objective function is to minimize the total electricity cost of the user group Cost_user: Min Cost_user =∑ t ∑ j ∑ i (ρ_LA(t) * P_rated_j,i * S_j,i(t) * Δt); where S_j,i(t) is the absolute value of the electricity consumption state of the i-th user household in the j-th cluster at time t.

[0047] Further, in the upper-layer LA pricing decision model, the constraint conditions include:

[0048] Power balance constraint, electricity buying / selling power constraint (0≤P_buy(t)≤P_buy_max, 0≤P_sale(t)≤P_sale_max); electricity selling price constraint (ρ_LA_min≤ρ_LA(t)≤ρ_LA_max).

[0049] In the lower-layer user response decision model, the constraint conditions include: user room temperature constraint (θmin≤θ_in_j,i(t)≤θmax); indoor temperature change equation constraint based on the ETP model; adjustable capacity constraint of heating device running time 0≤T_run_j,i(t)≤T_run_max_j,i(t), 0≤L_cut_j,i(t)≤L_cut_max_j,i(t);

[0050] Wherein, T_run_j,i(t) is the actual running time of the heating device of the i th user household in the j th cluster at time t, T_run_max_j,i(t) is the maximum possible running time of the heating device of the i th user household in the j th cluster at time t, L_cut_j,i(t) is the actual cut-off (not running) time of the heating device of the i th user household in the j th cluster at time t, and L_cut_max_j,i(t) is the maximum possible cut-off (not running) time of the heating device of the i th user household in the j th cluster at time t.

[0051] The application also provides an electric heating load master-slave game collaborative optimization device based on AP clustering and adjustable potential modeling, comprising:

[0052] An analysis module is configured to establish a direct heating type electric heating indoor heat transfer model and an adjustable potential basic model, and analyze a single user by using the direct heating type electric heating indoor heat transfer model to obtain load characteristics of a single heating user.

[0053] A clustering module is configured to construct a distributed electric heating load refined aggregation model considering household differences, and divide distributed heating users into different clusters by using an affinity propagation clustering (AP) algorithm.

[0054] A quantification module is configured to construct an adjustable potential quantification model considering household differences, and obtain an up-adjustable power and a down-adjustable power of each cluster and quantifies the adjustable potential of a user household group in each cluster according to the load characteristics of the single heating user.

[0055] An optimization module is configured to construct a heating load group optimal operation model based on a master-slave game, and optimize by using a dynamic adjustment strategy of an inertia factor parameter in a particle swarm algorithm.

[0056] A strategy output module is configured to input grid power purchase and sale price information, distributed power supply cost information, and the up-adjustable power and the down-adjustable power and the adjustable potential obtained in step S3 into the optimized heating load group optimal operation model to obtain an optimal power sale price strategy and an optimal power consumption behavior adjustment strategy of a user.

[0057] The application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0058] The application also provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the above method.

[0059] The application has the following beneficial effects:

[0060] 1. The application overcomes the defects of ignoring user differences in the prior art by a fine household differentiation aggregation model based on the AP algorithm, and realizes accurate quantification of the adjustable potential of the decentralized heating load group. Combined with master-slave game optimization operation, it can effectively guide users to adjust electricity consumption behavior, significantly reduce electricity consumption peak, and effectively relieve the pressure of power grid peak regulation.

[0061] 2. In the constructed master-slave game framework, the upper LA maximizes its net profit by optimizing the pricing strategy, and the lower user minimizes its cost by optimizing the electricity consumption plan in response to the price.

[0062] 3. The user room temperature comfort constraint is explicitly considered in the lower user optimization model, ensuring that the user indoor temperature is always within an acceptable comfort range while optimizing electricity consumption behavior and reducing cost.

[0063] 4. The fine aggregation model enables the LA to actively optimize the response characteristics of the entire load group by adjusting the proportion of different household users, thereby more flexibly meeting the diversified dispatching needs of the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 The method flowchart of an embodiment of the application.

[0065] Figure 2 The outdoor temperature and time-of-use price information diagram in the example of the application.

[0066] Figure 3 The price optimization result diagram in the example of the application.

[0067] Figure 4 The demand response result diagram in the example of the application.

[0068] Figure 5 The device structure diagram of an embodiment of the application.

[0069] Figure 6 The computer device internal structure diagram of an embodiment of the application.

[0070] The implementation of the application, functional features and advantages will be further described with reference to the drawings. DETAILED DESCRIPTION

[0071] It should be understood that the specific embodiments described herein are only used to explain the application and not to limit the application.

[0072] The application provides an electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling, researches heating user load characteristics, adopts a proximity propagation clustering algorithm to construct a refined aggregation model, and proposes a load group demand response potential evaluation method to accurately quantify the adjustable potential of the heating load group. An optimization operation strategy based on master-slave game is constructed, the upper layer formulates pricing decisions to maximize the LA net profit, the lower layer optimizes power consumption behavior to minimize user power consumption cost, and the room temperature, heating duration and heat balance constraints are considered. For the master-slave game model, an improved particle swarm algorithm with dynamic adjustment of inertia factor parameters is proposed to improve the global optimality of the algorithm. The application can fully tap the regulation and control potential of distributed electric heating load, reduce the peak regulation pressure of the power grid, balance the LA income and user comfort and economy, and effectively solve the problems of ignoring house type differences, lacking effective LA-user interactive game mechanism and model solution being prone to local optimum in the prior art.

[0073] As Figure 1 shown, the application provides an electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling, which is used for tapping the regulation and control potential of distributed electric heating load and reducing the peak regulation pressure of the power grid. The method comprises the following steps:

[0074] S1, a direct heating type electric heating indoor heat transfer model and an adjustable potential basic model are established, and the direct heating type electric heating indoor heat transfer model is used to analyze a single user to obtain the load characteristics of a single heating user, so as to research the load characteristics of the heating user and analyze the power consumption behavior and heat demand characteristics of different types of heating users. Specifically, it comprises:

[0075] S101, an operation model of a single direct heating type electric heating device is established based on an equivalent thermal parameter (ETP) model, the indoor temperature and the upper and lower limits of the user comfort interval are compared, the electric heating equipment is started and stopped in real time to depict the heat balance mechanism of the user, and the balance between the power consumption and the heat demand is realized under the premise of meeting the user comfort; an indoor temperature change equation is established as follows:

[0076] θin(t+1)=θout(t)+(R*Prated*S(t)+R*Psolar-(θin(t)-θout(t))*(Δt / (R*C))

[0077] Wherein, θin(t) is the room temperature at t time; θout(t) is the outdoor temperature at t time; R is the equivalent thermal resistance of the room; C is the equivalent thermal capacity of the room; Prated is the rated power of the electric heating equipment; S(t) is the equipment start-stop state at t time; Psolar is the solar radiation power; and Δt is the time step.

[0078] The indoor temperature variation power equation considers the influence of outdoor temperature, room equivalent thermal resistance, room equivalent thermal capacity, rated power of electric heating equipment, solar radiation power and equipment start-stop state on the room temperature variation, and can accurately depict the dynamic variation process of indoor temperature.

[0079] S102, based on the operation model in step S101, calculate the up-regulation time T up and the down-regulation time T down of a single heating equipment under the condition that the user comfort constraint θ min ≤ θ in (t) ≤ θ max is met as a basic model of adjustable potential:

[0080] T up = (θ max - θ in (t)) / (dθ in / dt|S=1)

[0081] T down = (θ in (t)-θ min / (dθ in / dt|S=0)

[0082] Wherein, S is a binary variable, used to represent whether the state or condition is met; based on the up-regulation time T up and the down-regulation time T down of each household in the cluster, the overall adjustable potential index of the load group is dynamically aggregated, LA drives the user to adjust the utilization degree of the up-regulation time T up and the down-regulation time T down through the principal and subordinate game pricing strategy, realizes the space-time transfer of the load, and optimizes the peak-valley difference of the power grid.

[0083] S103, taking the adjustable potential basic model of a single heating equipment under the condition that the user comfort constraint is met as the load characteristics of a single heating user.

[0084] In view of the differentiated characteristics of the household parameters, a load refined aggregation model is constructed by using the affinity propagation clustering algorithm, and a load group demand response potential evaluation method is proposed to quantify the adjustable potential of each cluster, specifically as steps S2-S3, which are used as the basic data input for load optimization scheduling and operation of the principal and subordinate game model.

[0085] S2, a distributed electric heating load refined aggregation model considering the difference of household types is constructed, and an affinity propagation clustering (AP) algorithm is used to divide the distributed heating users into different clusters. By using the affinity propagation clustering algorithm, the differentiated characteristics of the household parameters are fully considered, the refined aggregation of the distributed heating load is realized, and the aggregation accuracy and the quantification accuracy of the adjustable potential are improved. The affinity propagation clustering (AP) algorithm can adaptively determine the number and center of the cluster according to the distribution of the equipment power and the household thermal parameters, without pre-specifying the number of clusters, so as to more accurately reflect the differences of the user groups, lay a foundation for accurate quantification and differentiated scheduling of the adjustable potential. Specifically, it includes:

[0086] By using the affinity propagation clustering (Affinity Propagation, AP) algorithm, the number of clusters and the center of each cluster are adaptively determined according to the rated power P ratedand the equivalent thermal parameters (thermal resistance R, thermal capacity C) of houses and the like to differentiate the clustering of distributed heating users; the specific steps of the differentiated clustering are:

[0087] 1) According to the rated power P rated , the preliminary clustering is performed according to the rated power P

[0088] 2) In each preliminary clustering cluster, fine clustering is performed according to the equivalent thermal parameters R and C of the house type, to obtain the final sub-clustering cluster j, and the number and members of the clustering cluster are adaptively determined according to the data distribution.

[0089] The affinity propagation clustering algorithm can automatically determine the number and members of the clustering cluster without prior specification, so that the clustering result is more consistent with the actual house type difference distribution, and the accuracy and adaptability of the load aggregation model are improved.

[0090] S3, a tunable potential quantification model considering the house type difference is constructed, and the up-regulatable power and the down-regulatable power of each clustering cluster are obtained according to the load characteristics of the single heating user, and the tunable potential of the user house type group in each clustering cluster is quantified. The quantification of the tunable potential of the load group specifically includes:

[0091] S301, the tunable duration of the user house type i in each clustering cluster j at time t is calculated according to the load characteristics (tunable potential basic model) of the single heating user, as the tunable potential of the user house type group in each clustering cluster, and the calculation formula is:

[0092] Tup j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_i,i)

[0093] Tdown j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i)

[0094] Wherein, Tup j ,i(t) and Tdown j ,i(t) respectively represent how long this group of users can delay the heating and how long they can turn off the heating at this moment, f is a function for calculating the tunable duration; R j is the equivalent thermal resistance of the house in the jth clustering cluster, i is the user house type in the clustering cluster j, C j is the equivalent thermal capacity of the house in the jth clustering cluster, and Prated_j,i is the rated power of the electric heating device in the jth clustering cluster.

[0095] S302、Calculate the up-regulation power E_up_LA(t) and the down-regulation power E_down_LA(t) of each cluster:

[0096]

[0097] wherein M is the total number of aggregated clusters; Lj is the number of house types of the aggregated cluster j; ξ ji is the proportion of the number of users of the house type i of the cluster j to the total number of users, B and B ji,t is the up-regulation power proportion, indicating the proportion of users of the house type i in the jth cluster who choose to up-regulate the power at time t, and is the up-regulation duration and the down-regulation duration of the house type i in the jth cluster at time t, is the rated power of the jth cluster;

[0098] The total up-regulation power and the total down-regulation power are obtained by summing the up-regulation power and the down-regulation power of all the clusters, and then the specific up-regulation power and the specific down-regulation power of each cluster are further calculated according to the proportion of users and the rated power of each cluster.

[0099] S4, a heating load group optimal operation model based on master-slave game and an LA optimal pricing strategy are constructed, and a dynamic adjustment strategy of inertia factor parameters in a particle swarm algorithm is adopted for optimization to improve the global optimality of the algorithm. Through the interaction of the upper LA pricing decision and the lower user response decision, the load optimal scheduling is realized.

[0100] The constructed upper LA pricing decision model comprehensively considers multiple constraint conditions such as power balance, power purchase and sale, power sale price, and heating load peak value, formulates a reasonable power sale price and a distributed power scheduling plan, and ensures the maximum LA benefit. The improved particle swarm algorithm with linearly decreasing weight through adjustment of inertia factor parameters can effectively overcome the problem that the traditional particle swarm algorithm is easy to fall into a local optimal solution, improve the solving efficiency and global optimality, and provide stronger algorithm support for implementation of the heating load group optimal operation strategy.

[0101] The heating load group optimal operation model based on master-slave game includes:

[0102] A master-slave game framework is established with LA as the leader (upper layer) and distributed heating users as the followers (lower layer):

[0103] 1) Upper layer (LA pricing decision model)

[0104] Objective function: maximize LA net profit Profit_LA:

[0105] MaxProfit_LA=Revenue_sale-Cost_net-Cost_DG-Subsidy_user

[0106] Revenue from electricity sales:

[0107] Revenue_sale=∑ t (ρ_LA(t)*P_sale(t))

[0108] Net electricity purchase cost:

[0109] Cost_net=∑ t (ρ_buy_grid(t)*P_buy(t)-ρ_sale_grid(t)*P_sale(t))

[0110] Distributed power supply cost:

[0111] Cost_DG=∑ t (c_G*p_DG(t))

[0112] User incentive subsidies:

[0113] Subsidy_user=∑ t ∑ j ∑ i (ρ_cut*L_cut_j,i(t))

[0114] Wherein, ρ_LA(t) is the electricity price set by LA; P_sale(t) is the electricity sold by LA; ρ_buy_grid(t) and ρ_sale_grid(t) are the grid purchase / sale electricity prices; P_buy(t) and P_sale(t) are the electricity purchased / sold by LA from the grid; c_G is the distributed generation cost coefficient; p_DG(t) is the output of the distributed generation during time period t; ρ_cut is the subsidy price for reduced electricity; and L_cut_j,i(t) is the load reduction of cluster j and household type i during time period t.

[0115] t represents the amount of electricity load reduced for a specific user type within a specific time period. This directly affects the calculation of the adjustable and reduced electricity loads. These reduced loads can be redistributed to other time periods to optimize overall electricity usage and costs.

[0116] The constraints include: power balance constraints, power purchase and sale constraints (0≤P_buy(t)≤P_buy_max, 0≤P_sale(t)≤P_sale_max); and electricity sales price constraints (ρ_LA_min≤ρ_LA(t)≤ρ_LA_max).

[0117] 2) Lower layer (user response decision model)

[0118] Objective function: Minimize the total electricity cost of users Cost_user: Min Cost_user=∑ t ∑ j ∑ i (ρ_LA(t)*P_rated_j,i*S_j,i(t)*Δt);where S_j,i(t) is the absolute value of the electricity state of the i-th user household in the j-th cluster at time t.

[0119] The constraint conditions include: user room temperature constraint (θmin≤θ_in_j,i(t)≤θmax) ; indoor temperature variation equation constraint based on ETP model; adjustable capacity constraint of heating device running time 0≤T_run_j,i(t)≤T_run_max_j,i(t), 0≤L_cut_j,i(t)≤L_cut_max_j,i(t) ;

[0120] where T_run_j,i(t) is the actual running time of the heating device of the i-th user household in the j-th cluster at time t, T_run_max_j,i(t) is the maximum possible running time of the heating device of the i-th user household in the j-th cluster at time t, L_cut_j,i(t) is the actual cut-off (not running) time of the heating device of the i-th user household in the j-th cluster at time t, and L_cut_max_j,i(t) is the maximum possible cut-off (not running) time of the heating device of the i-th user household in the j-th cluster at time t.

[0121] In the lower user response decision model, the user will comprehensively consider the room temperature comfort requirement, the adjustable capacity of the heating device running time, and the electricity cost and other factors when responding to the electricity price signal formulated by the LA, and realize the minimization of the electricity cost by optimizing the electricity behavior, while ensuring that the comfort requirement of the user is met.

[0122] The master-slave game model can effectively tap the user response potential and realize the "peak load shifting" of the load through the interactive mechanism of the electricity price incentive (upper layer) formulated by the LA and the user optimization of the electricity plan to minimize the cost (lower layer), while ensuring that the room temperature of the user is within the comfort interval.

[0123] S5, input the grid buying and selling electricity price information, distributed power cost information, and the adjustable electricity amount and the adjustable potential obtained in step S3 into the optimized heating load group optimization operation model to obtain the optimal selling electricity price strategy and the optimal electricity behavior adjustment strategy of the user.

[0124] LA formulates dynamic electricity price according to grid buy / sell price information and user adjustable potential to encourage users to increase electricity consumption during low price period, thereby reducing grid peak load. The optimal electricity consumption behavior adjustment strategy of the user is to adjust its electricity consumption plan according to these dynamic price signals to minimize electricity cost while ensuring that the indoor temperature is maintained within the comfortable range.

[0125] To sum up, the output of step S1 is the load characteristics of a single user, which provides basic data for subsequent steps. Steps S2-S3 input the load characteristics and output the user's adjustable power and adjustable amount of each cluster according to equipment power and house type thermal parameters. The users are grouped by clustering algorithm, and the adjustable potential of each group is quantified to provide data support for optimization operation. The input of step S4 is the refined load aggregation model and adjustable potential evaluation result obtained in step S3, as well as grid buy / sell price information and distributed power cost information. The output is the optimal sell price strategy of LA and the optimal electricity consumption behavior adjustment strategy of the user. That is, the user load characteristics model of step S1 provides basic data for clustering of steps S2-S3. The clustering result and adjustable potential evaluation of step S3 are directly used in the master-slave game model of step S4 as the basis for LA pricing and user response.

[0126] Examples:

[0127] In order to verify the effectiveness of the electric heating load master-slave game collaborative optimization method based on AP clustering and adjustable potential modeling proposed in the present application, the present application selects the actual parameters of 168 heating households in a certain district of a certain region to test the practicability of the model proposed in the present application. The outdoor temperature typical day curve and the grid buy / sell price curve to LA are provided, as shown in Figure 2

[0128] Two sets of comparison schemes are set in this example: scheme 1 is that the distributed users do not participate in the master-slave game strategy of the load aggregator (LA); scheme 2 is that the distributed users are uniformly aggregated and controlled by the LA and participate in the market game. Figure 3 and Figure 4 ​The LA electricity selling price curve and the distributed user demand response result in scheme 2 are shown. As can be seen from the figure, the optimization trend of the LA electricity selling price is highly coordinated with the change of the heating load: in the evening peak period (17:00-19:00), the LA electricity selling price is raised to a high level; and in the night valley period, the electricity price is significantly reduced to about 0.15 yuan / kW·h. Under this price incentive, the user electricity consumption curve is significantly improved in smoothness, and the peak-valley difference of the electric load is reduced to 33.75%, and the peak load of the electric heating is reduced by 86.47 kW. This is because the user actively shifts the load in the peak period to the valley period of the electricity price (such as night, 11:00-14:00 and 23:00-24:00) by using the adjustable characteristics of the heating equipment to realize 'peak load shifting and valley filling'. It is worth noting that the average room temperature of the user after optimization is maintained in the comfortable interval of 20.31℃, which verifies the dual effectiveness of the application in optimizing the electricity consumption behavior and ensuring human comfort.

[0129] To quantitatively evaluate the performance of the strategy, two control experiments are set: control group (scheme 1): distributed users independently respond to market electricity price and experimental group (scheme 2): adopt the LA master-slave game aggregation control strategy described in the application. The data comparison is shown in Table 1:

[0130] Table 1 Comparison data table

[0131] Evaluation index Scheme 1 (control group) Scheme 2 (the application) Optimization effect Electric load peak valley difference 51.2％ 33.75％ ↓34.1％ Peak load 298.6 kW 212.13 kW ↓ 86.47 kW Average room temperature fluctuation ±2.8℃ ±1.2℃ Stability is improved by 57%

[0132] The experimental data show that, compared with scheme 1, the strategy described in the application reduces the peak-valley difference by 34.1% and the peak load by 28.9% under the premise of maintaining user comfort, verifying the core advantage of the master-slave game mechanism in load aggregation control. This method of dynamically guiding the spatio-temporal transfer of load through price signals provides effective technical support for high-proportion renewable energy consumption and meets the flexible regulation needs of the new power system.

[0133] As shown in Figure 5 , the application also provides an electric heating load master-slave game cooperative optimization device based on AP clustering and adjustable latent modeling, comprising:

[0134] An analysis module 1 is used to establish a direct heating type electric heating indoor heat transfer model, and analyze a single user by using the direct heating type electric heating indoor heat transfer model to obtain the load characteristics of a single heating user;

[0135] A clustering module 2 is used to construct a fine aggregation model of distributed electric heating load considering house type differences, and uses an affinity propagation clustering (AP) algorithm to divide the distributed heating users into different clusters;

[0136] a quantification module 3 configured to construct an adjustable potential quantification model considering the difference in house types, and to obtain the up-regulatable power and down-regulatable power of each cluster and quantify the adjustable potential of the user house group in each cluster according to the load characteristics of the single heating user;

[0137] an optimization module 4 configured to construct a heating load group optimal operation model based on a principal-agent game, and to perform optimization by using a dynamic adjustment strategy of an inertia factor parameter in a particle swarm algorithm;

[0138] a strategy output module 5 configured to input the grid buying and selling electricity price information, the distributed power cost information, and the up-regulatable power and down-regulatable power and the adjustable potential obtained in step S3 into the optimized heating load group optimal operation model to obtain an optimal selling electricity price strategy and an optimal electricity consumption behavior adjustment strategy of the user.

[0139] In one embodiment, the analysis module 1 specifically comprises:

[0140] a building unit configured to build an operation model of a single direct electric heating device based on an equivalent thermal parameter ETP model, to control the start and stop of the electric heating device in real time to depict the heat balance mechanism of the user by comparing the room temperature with the upper and lower limits of the user comfort interval, and to build an indoor temperature change equation:

[0141] θin(i+1)=θout(t)+(R*Prated*S(t)+R*Psolar-(θin(t)-θout(t))*(Δt / (R*C))

[0142] wherein θin(t) is the room temperature at time t; θout(t) is the outdoor temperature at time t; R is the room equivalent thermal resistance; C is the room equivalent thermal capacity; Prated is the rated power of the electric heating device; S(t) is the device start-stop state at time t; Psolar is the solar radiation power; and Δt is the time step;

[0143] a calculation unit configured to calculate the up-regulatable time length Tup and the down-regulatable time length Tdown of the single heating device under the constraint of user comfort degree θmin≤θin(t)≤θmax based on the operation model in the building unit, as a basic model of adjustable potential:

[0144] Tup=(θmax-θin(t)) / (dθin / dt|S=1)

[0145] Tdown=(θin(t)-θmin / (dθin / dt|S=0)

[0146] wherein S is a binary variable used to represent whether the state or condition is met;

[0147] S103, taking the adjustable potential base model of the single heating device under the user comfort constraint as the load characteristics of the single heating user.

[0148] In one embodiment, the clustering module 2 specifically comprises:

[0149] The AP algorithm is used to cluster the distributed heating users according to the rated power P rated and the equivalent thermal parameters of the house, wherein the equivalent thermal parameters of the house include the thermal resistance R and the thermal capacity C; the specific steps of the differential clustering are as follows:

[0150] 1) Preliminary clustering is performed according to the rated power P rated and the equivalent thermal parameters of the house, and a preliminary clustering cluster is obtained;

[0151] 2) Fine clustering is performed according to the equivalent thermal parameters R and C of the house type in each preliminary clustering cluster, and a final secondary clustering cluster j is obtained, and the number and members of the clustering cluster are adaptively determined according to the data distribution.

[0152] In one embodiment, the quantification module 3 specifically comprises:

[0153] A duration calculation unit is configured to calculate the adjustable duration of the user house type i in each clustering cluster j at time t according to the load characteristics of the single heating user, as the adjustable potential of the user house type group in each clustering cluster, and the calculation formula is as follows:

[0154] Tup j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i)

[0155] Tdown j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i)

[0156] Wherein, Tup j ,i(t) and Tdown j ,i(t) represent how long the average user in this group can delay heating and how long the average user in this group can turn off the heating at this moment, respectively, f is a function for calculating the adjustable duration; R j is the equivalent thermal resistance of the house in the jth clustering cluster, i is the user house type in the clustering cluster j, C j is the equivalent thermal capacity of the house in the jth clustering cluster, and Prated_j,i is the rated power of the electric heating device in the jth clustering cluster.

[0157] An electricity amount calculation unit is configured to calculate an up-regulatable electricity amount E_up_LA(t) and a down-regulatable electricity amount E_down_LA(t) of each cluster:

[0158]

[0159] wherein M is the total number of aggregated clusters; Lj is the number of house types of the aggregated cluster j; ξ ji is the proportion of the number of users of the house type i of the cluster j to the total number of users, B and B ji,t is the up-regulation electricity amount proportion, indicating the proportion of the users of the house type i in the jth cluster who select the up-regulation electricity amount at the time t, and is the up-regulatable time length and the down-regulatable time length of the house type i in the jth cluster at the time t, is the rated power of the jth cluster;

[0160] The total up-regulatable electricity amount and the total down-regulatable electricity amount are obtained by summing the up-regulatable electricity amount and the down-regulatable electricity amount of all the clusters, and then the specific up-regulatable electricity amount and the specific down-regulatable electricity amount of each cluster are further calculated according to the proportion of the users and the rated power of each cluster.

[0161] In an embodiment, in the optimization module 4, the heating load group optimization operation model based on the master-slave game includes:

[0162] A master-slave game framework is established, in which LA is the upper leader and the distributed heating users are the lower followers:

[0163] 1) In the upper LA pricing decision model, the objective function is to maximize the LA net profit Profit_LA:

[0164] Max Profit_LA = Revenue_sale - Cost_net - Cost_DG - Subsidy_user

[0165] The electricity sale revenue is:

[0166] Revenue_sale = ∑ t (ρ_LA(t) * P_sale(t))

[0167] The net electricity purchase cost is:

[0168] Cost_net = ∑ t (ρ_buy_grid(t) * P_buy(t) - ρ_sale_grid(t) * P_sale(t))

[0169] The distributed power supply cost is:

[0170] Cost_DG = ∑t (c_G * p_DG(t))

[0171] User incentive subsidy:

[0172] Subsidy_user =∑ t ∑ j ∑ i (ρ_cut / L_cut_j,i(t))

[0173] Where ρ_LA(t) is the electricity selling price set by LA; P_sale(t) is the electricity selling amount of LA; ρ_buy_grid(t), ρ_sale_grid(t) are the electricity grid buying / selling price; P_buy(t), P_sale(t) are the electricity buying / selling amount of LA to the grid; c_G is the distributed power cost coefficient; p_DG(t) is the distributed power output at time t; ρ_cut is the electricity reduction subsidy price; L_cut_j,i(t) is the load reduction amount of cluster j household i at time t.

[0174] 2) In the lower layer user response decision model, the objective function is to minimize the total electricity cost of the user group Cost_user: Min Cost_user =∑ t ∑ j ∑ i (ρ_LA(t) * P_rated_j,i * S_j,i(t) * Δt); where S_j,i(t) is the absolute value of the electricity consumption state of the i-th user household in the j-th cluster at time t.

[0175] In one embodiment, in the upper layer LA pricing decision model, the constraint conditions include:

[0176] Power balance constraint, electricity buying / selling power constraint (0≤P_buy(t)≤P_buy_max, 0≤P_sale(t)≤P_sale_max); electricity selling price constraint (ρ_LA_min≤ρ_LA(t)≤ρ_LA_max).

[0177] In the lower layer user response decision model, the constraint conditions include: user room temperature constraint (θmin≤θ_in_j,i(t)≤θmax); indoor temperature change equation constraint based on the ETP model; adjustable capacity constraint of heating device running time 0≤T_run_j,i(t)≤T_run_max_j,i(t), 0≤L_cut_j,i(t)≤L_cut_max_j,i(t);

[0178] Wherein, T_rum_j,i(t) is the actual running time length of the heating device of the i th user household in the j th cluster at time t, T_run_max_j,i(t) is the maximum possible running time length of the heating device of the i th user household in the j th cluster at time t, L_cut_j,i(t) is the actual cut-off (not running) time length of the heating device of the i th user household in the j th cluster at time t, and L_cut_max_j,i(t) is the maximum possible cut-off (not running) time length of the heating device of the i th user household in the j th cluster at time t.

[0179] The modules and units are used to perform the steps of the above-mentioned master-slave game collaborative optimization method for electric heating load based on AP clustering and adjustable latent variable modeling, and the specific implementation manners are described in the above-mentioned method embodiments, which will not be described here.

[0180] As shown in Figure 6 The computer device can be a server, and the internal structure thereof can be as shown in Figure 6 The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store all data required by the process of the master-slave game collaborative optimization method for electric heating load based on AP clustering and adjustable latent variable modeling. The network interface of the computer device is used to communicate with the external terminal through network connection. The computer program is executed by the processor to implement the master-slave game collaborative optimization method for electric heating load based on AP clustering and adjustable latent variable modeling.

[0181] Those skilled in the art can understand that Figure 6 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied.

[0182] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program. The computer program is executed by the processor to implement any one of the above-mentioned master-slave game collaborative optimization methods for electric heating load based on AP clustering and adjustable latent variable modeling.

[0183] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiment methods can be included. Any reference to memory, storage, databases, or other media in this application and in examples provided herein, unless specifically stated otherwise, can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0184] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, device, article, or method that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, device, article, or method. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, device, article, or method that includes the element.

[0185] The above description is only the preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, based on the content of the specification and drawings of the present application, are also included in the patent protection scope of the present application.

Claims

1. A collaborative optimization method for electric heating load based on AP clustering and adjustable potential modeling, characterized in that, include: S1. Establish a direct-heating electric heating indoor heat transfer model and an adjustable potential basic model, and use the direct-heating electric heating indoor heat transfer model to analyze a single user and obtain the load characteristics of a single heating user. S2. Construct a refined aggregation model of distributed electric heating load that takes into account the differences in house types, and use the proximity propagation clustering (AP) algorithm to divide distributed heating users into different clusters. S3. Construct a quantitative model of adjustable potential that takes into account the differences in apartment types, and obtain the adjustable power consumption and adjustable power consumption of each cluster based on the load characteristics of the individual heating user, and quantify the adjustable potential of the user apartment type group in each cluster. S4. Construct a heating load group optimization operation model based on master-slave game, and use the dynamic adjustment strategy of the inertia factor parameter in particle swarm optimization algorithm for optimization. S5. Input the grid purchase and sale electricity price information, distributed power source cost information, and the adjustable and de-adjustable power and adjustable potential obtained in step S3 into the optimized heating load group operation model to obtain the optimal electricity sales price strategy and the optimal electricity consumption behavior adjustment strategy for users.

2. The collaborative optimization method for electric heating load based on AP clustering and adjustable potential modeling according to claim 1, characterized in that, Step S1 specifically includes: S101. Based on the Equivalent Thermal Parameter (ETP) model, establish an operation model for a single direct-heating electric heating device. By comparing the room temperature with the upper and lower limits of the user's comfort range, control the start and stop of the electric heating equipment in real time to characterize the user's thermal balance mechanism and establish an indoor temperature change equation: θin(t+1)=θout(t)+(R*Prated*S(t)+R*Psolar-(θin(t)-θout(t))*(Δt / (R*C)) Where θin(t) is the room temperature at time t; θout(t) is the outdoor temperature at time t; R is the equivalent thermal resistance of the room; C is the equivalent heat capacity of the room; Prated is the rated power of the electric heating equipment; S(t) is the start / stop status of the equipment at time t; Psloar is the solar radiation power; Δt is the time step; S102. Based on the operating model in step S101, calculate the adjustable duration Tup and adjustable duration Tdown of a single heating device under the user comfort constraint θmin≤θin(t)≤θmax, as the basic model of adjustable potential: Tup=(θmax-θin(t)) / (dθin / dt|S=1) Tdown=(θin(t)-θmin / (dθin / dt|S=0) Where S is a binary variable used to represent whether a state or condition is satisfied; S103. The adjustable potential basic model of a single heating device under the constraint of user comfort is used as the load characteristics of a single heating user.

3. The collaborative optimization method for electric heating load based on AP clustering and adjustable potential modeling according to claim 1, characterized in that, Step S2 specifically includes: The nearest neighbor propagation clustering algorithm (AP) is used to determine the clustering method based on the rated power P of the heating equipment. rated Differential clustering is performed on decentralized heating users based on the building's equivalent thermal parameters; these parameters include thermal resistance R and heat capacity C; the specific steps of differential clustering are as follows: 1) Based on the rated power P rated The parameters are initially clustered to obtain preliminary clusters. 2) Within each initial cluster, fine clustering is performed based on the equivalent thermal parameters R and C of the apartment type to obtain the final sub-cluster j, and the number and members of the clusters are adaptively determined according to the data distribution.

4. The collaborative optimization method for electric heating load based on AP clustering and adjustable potential modeling according to claim 1, characterized in that, Step S3 specifically includes: S301. Based on the load characteristics of the individual heating user, calculate the adjustable duration of user unit type i within each cluster j at time t, which is used as the adjustable potential of user unit type group in each cluster. The calculation formula is as follows: Tup j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i) Tdown j ,i(t)=f(R j ,i,C j ,i,θin(t),θout(t),Prated_j,i) Among them, Tup j ,i(t) and Tdown j ,i(t) represent the average time that the group of users can delay turning on the heating and the average time they can turn off the heating earlier, respectively, and f is a function used to calculate the adjustable duration; R j Let i represent the equivalent thermal resistance of the houses in the j-th cluster, and let i be the user apartment type within cluster j. j Let Prated_j,i be the equivalent heat capacity of the houses in the j-th cluster, and let Prated_j,i be the rated power of the electric heating equipment in the j-th cluster. S302. Calculate the up-adjustable power E_up_LA(t) and down-adjustable power E_down_LA(t) for each cluster: Where M is the total number of clusters; Lj is the number of apartment types in cluster j; ξ ji Let B and B' represent the proportion of users in cluster j with user type i to the total number of users. ji,t The percentage increase in electricity consumption indicates that at time t, the user of the ith apartment type in the j-th cluster chooses to increase the percentage of their electricity consumption. and Let be the adjustable duration (both upward and downward) of the i-th apartment type in the j-th cluster at time t. Let be the rated power of the j-th cluster; The total adjustable power capacity is obtained by summing the adjustable power capacity of all clusters. Then, based on the user ratio and rated power of each cluster, the specific adjustable power capacity of each cluster is calculated.

5. The collaborative optimization method for electric heating load based on AP clustering and adjustable potential modeling according to claim 1, characterized in that, In step S4, the heating load group optimization operation model based on master-slave game includes: Establish a master-slave game framework with LA as the upper-level leader and distributed heating users as the lower-level followers: 1) In the upper-level LA pricing decision model, the objective function is to maximize the net profit of LA, Profit_LA. Max Profit_LA=Revenue_sale-Cost_net-Cost_DG-Subsidy_user Revenue from electricity sales: Revenue_sale=∑ t (ρ_LA(t)*P_sale(t)) Net electricity purchase cost: Cost_net=∑ t (ρ_buy_grid(t)*P_buy(t)-ρ_sale_grid(t)*P_sale(t)) Distributed power supply cost: Cost_DG=∑ t (c_G*p_DG(t)) User incentive subsidies: Subsidy_user=∑ t ∑ j ∑ i (ρ_cut*L_cut_j,i(t)) Wherein, ρ_LA(t) is the electricity price set by LA; P_sale(t) is the electricity sold by LA; ρ_buy_grid(t) and ρ_sale_grid(t) are the grid purchase / sale electricity prices; P_buy(t) and P_sale(t) are the electricity purchased / sold by LA from the grid; c_G is the distributed generation cost coefficient; p_DG(t) is the output of the distributed generation during time period t; ρ_cut is the subsidy price for reduced electricity; and L_cut_j,i(t) is the load reduction of cluster j and household type i during time period t. 2) In the lower-level user response decision model, the objective function is to minimize the total electricity cost of the user group, Cost_user: MinCost_user=∑ t ∑ j ∑ i (ρ_LA(t)*P_rated_j,i*S_j,i(t)*Δt); where S_j,i(t) is the absolute value of the electricity consumption status of the i-th user unit in the j-th cluster at time t.

6. The collaborative optimization method for electric heating load based on AP clustering and adjustable potential modeling according to claim 5, characterized in that, The constraints in the upper-level LA pricing decision model include: Power balance constraints, power purchase and sale constraints (0≤P_buy(t)≤P_buy_max, 0≤P_sale(t)≤P_sale_max); electricity sales price constraints (ρ_LA_min≤ρ_LA(t)≤ρ_LA_max). The constraints in the lower-level user response decision model include: user room temperature constraints (θmin≤θ-in_j, i(t)≤θmax); indoor temperature change equation constraints based on the ETP model; and adjustable heating equipment runtime constraints 0≤T_run_j, i(t)≤T_run_max_j, i(t), 0≤L_cut_j, i(t)≤L_cut_max_j, i(t). Wherein, T_run_j,i(t) represents the actual running time of the heating equipment for the i-th user unit in the j-th cluster at time t, T_run_max_j,i(t) represents the maximum possible running time of the heating equipment for the i-th user unit in the j-th cluster at time y, L_cut_j,i(t) represents the actual cut-off time of the heating equipment for the i-th user unit in the j-th cluster at time t, and L_cut_max_j,i(t) represents the maximum possible cut-off time of the heating equipment for the i-th user unit in the j-th cluster at time y.

7. A collaborative optimization device for electric heating load based on AP clustering and adjustable potential modeling, characterized in that, include: The analysis module is used to establish an indoor heat transfer model and an adjustable potential basic model for direct-heating electric heating, and to analyze a single user using the indoor heat transfer model of direct-heating electric heating to obtain the load characteristics of a single heating user. The clustering module is used to build a refined aggregation model of distributed electric heating load that takes into account the differences in house types, and uses the proximity propagation clustering (AP) algorithm to divide distributed heating users into different clusters. The quantification module is used to construct an adjustable potential quantification model that takes into account the differences in apartment types, and to obtain the adjustable power consumption and adjustable power consumption of each cluster based on the load characteristics of the individual heating user, as well as to quantify the adjustable potential of user apartment type groups in each cluster. The optimization module is used to construct an optimized operation model for heating load groups based on master-slave game theory, and to optimize it using a dynamic adjustment strategy for the inertia factor parameter in the particle swarm optimization algorithm. The strategy output module is used to input the grid purchase and sale electricity price information, distributed power source cost information, and the adjustable and adjustable power and adjustable potential obtained in step S3 into the optimized heating load group operation model to obtain the optimal electricity sales price strategy and the optimal electricity consumption behavior adjustment strategy for users.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images