Power distribution network multi-element load two-stage day-ahead grouping regulation and control method and device based on dynamic clustering

Abstract

The invention discloses a power distribution network multi-element load two-stage day-ahead grouping regulation and control method and device based on dynamic clustering, and the method comprises the steps: constructing a three-dimensional clustering feature vector according to the geographic position of multi-element load virtual energy storage and a load baseline value at a set moment, dividing a cluster through a K-means algorithm, and dynamically adjusting a boundary and a structure; overlapping and constructing an aggregation load model by combining with the single virtual energy storage model; and constructing a day-ahead collaborative optimization regulation and control model by taking the model as a constraint basis, and solving to obtain an optimal day-ahead regulation and control scheme. According to the method, the multi-element load regulation potential can be accurately excavated, the cooperative green and economical operation of the power distribution network and the multi-element load is promoted, and the distributed new energy consumption capability is improved.

Description

Technical Field

[0001] This invention relates to the field of multi-load and distribution network operation and control technology, specifically to a two-stage day-ahead grouping control method and device for multi-load distribution networks based on dynamic clustering. Background Technology

[0002] As a key coupling node between energy infrastructure and regional development, the distribution network not only plays a crucial technical supporting role in promoting my country's energy structure transformation but also shoulders the strategic mission of promoting green and low-carbon development. With the massive integration of distributed renewable energy, the distribution network is evolving from a traditional unidirectional energy supply system into a complex active network with bidirectional energy flow. However, the high proportion of distributed renewable energy integration will lead to systemic problems in distribution network operation, such as difficulties in peak-hour absorption, insufficient supply during off-peak hours, and instability during fluctuations, further triggering operational risks such as voltage exceeding limits. Therefore, it is urgent to establish a collaborative control mechanism for diversified and flexible resources in distribution network operation scenarios to ensure the dynamic balance of system supply and demand and operational safety.

[0003] The power distribution network has a wide coverage area, encompassing diverse load types (such as industrial production, agricultural irrigation, commercial services, residential use, and public utilities). These loads exhibit typical characteristics such as diversity and heterogeneity, large spatial distribution differences, and varying temporal fluctuations. Their operating mechanisms inherently possess dispatchable characteristics and contain abundant demand-side response potential. Different types of loads exhibit differentiated elasticity in their energy consumption time windows, with significant differences in power regulation margins, providing a foundation for flexible regulation. Through scientific load classification and aggregation, and multi-timescale coordinated regulation, the spatiotemporal matching of distributed renewable energy output and load demand can be achieved, thereby significantly enhancing the power distribution network's ability to absorb fluctuating renewable energy.

[0004] Therefore, under the rigid constraint of ensuring the supply of flexible resources in the distribution network, how to deeply explore the spatiotemporal regulation potential of diverse loads and synergistically optimize users' economic costs and energy comfort has become a key scientific issue that urgently needs to be addressed. Summary of the Invention

[0005] To address this, the present invention provides a two-stage day-ahead grouping control method and device for multi-load distribution networks based on dynamic clustering. By using dynamic clustering and two-stage control core technologies, it solves the problems of peak-hour absorption difficulties, valley-hour supply insufficiency, and control instability caused by the high proportion of distributed renewable energy access to the distribution network. At the same time, it solves the problems of low control accuracy and insufficient potential tapping caused by the strong heterogeneity and large spatiotemporal distribution differences of multi-loads, realizes the spatiotemporal matching of loads and renewable energy, improves absorption capacity, and takes into account both economic efficiency and low carbon emissions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering, comprising:

[0007] Based on the geographical location of virtual energy storage for multiple loads under the distribution network and the load baseline value at a set time, a three-dimensional clustering feature vector is constructed.

[0008] Based on the three-dimensional clustering feature vector, the multi-load virtual energy storage is clustered using the K-means clustering algorithm, and an adaptive update mechanism for the clustering structure is introduced to dynamically adjust the boundaries and structure of the clusters to obtain the clustering results.

[0009] By superimposing the individual virtual energy storage corresponding to the clustering results through the individual virtual energy storage model, an aggregated load model under dynamic clustering conditions is constructed.

[0010] Using the aggregated load model under the dynamic clustering conditions as the equivalent constraint basis, a multi-load day-ahead collaborative optimization control model is constructed; based on the set constraints, the multi-load day-ahead collaborative optimization control model is solved to obtain the optimal day-ahead control scheme for multi-loads.

[0011] As a preferred embodiment of the two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering, the expression for the three-dimensional clustering feature vector is:

[0012] a i (t)=(x i ,y i ,z i (t))

[0013] In the formula, a i (t) represents the three-dimensional clustering feature vector; i represents the individual load; x i y i z i (t) are the feature vectors of the three dimensions, (x) i ,y i ) represents the geographical location, z i (t) represents the load baseline value.

[0014] As a preferred embodiment of the two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering, the expression for the single-unit virtual energy storage model is as follows:

[0015]

[0016] In the formula, VSOC i (t) represents the virtual state of charge (SOC) of individual load i at time t; E max Virtual capacity of individual load i Let Δτ be the virtual energy storage of unit load i at time t; Δτ is the virtual energy storage at time z. i (t) is the time it takes for the discharge to be exhausted at the specified power.

[0017] As a preferred embodiment of the two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering, the expression for the aggregated load model under the dynamic clustering condition is as follows:

[0018]

[0019] In the formula, z m (t) represents the load baseline value of the multi-element load cluster m; d m P represents the number of samples contained in cluster m. m (t) represents the actual operating power of the multi-load cluster m at time t; P i (t) represents the actual operating power of unit load i at time t; ΔP i (t) represents the actual capacity of unit load i participating in regulation at time t; P m,c (t) represents the capacity value for regulation involving the multi-load cluster m; These represent the virtual charging and discharging power limits of the multi-load cluster m at time t; VSOC m (t) represents the average value of the VSOC of each multi-element load in the multi-element load cluster m at time t.

[0020] As a preferred option for the two-stage day-ahead clustering control method for multiple loads in a distribution network based on dynamic clustering, the objective function of the day-ahead collaborative optimization control model for multiple loads includes: minimizing the daily operating cost of the multiple load cluster, minimizing the total daily carbon emissions of the multiple load cluster, and maximizing the photovoltaic absorption rate of the distribution network.

[0021] The expression for minimizing the daily operating cost of the multi-load cluster is:

[0022]

[0023] In the formula, σ m,ES P represents the unit operation and maintenance cost coefficient for the multi-load cluster m; K represents the number of multi-load clusters; G (t) represents the electricity price at time t; P m,G (t) represents the power purchased by cluster k at time t;

[0024] The expression for minimizing the total daily carbon emissions of the multi-load cluster is:

[0025]

[0026] In the formula, μ C The carbon emission factor of the power distribution network;

[0027] The expression for maximizing the photovoltaic absorption rate of the distribution network is:

[0028]

[0029] In the formula, P PV (t) represents the photovoltaic output absorbed by the distribution network at time t, P PV,max (t) represents the upper limit of photovoltaic output of the distribution network at time t;

[0030] The overall objective function of the multi-load day-ahead collaborative optimization control model is expressed as follows:

[0031]

[0032] In the formula, μ1, μ2, and μ3 are the weight coefficients of each optimization objective; f 1-0 f 2-0 f 3-0 These are the initial values ​​before optimization for each optimization objective.

[0033] As a preferred option for the two-stage day-ahead grouping control method of multiple loads in the distribution network based on dynamic clustering, in the process of solving the day-ahead collaborative optimization control model of multiple loads based on the set constraints, the set constraints include: power purchase constraints of multiple load clusters, operation constraints of multiple load clusters, power balance constraints of multiple load clusters, and photovoltaic output constraints within the distribution network.

[0034] The expression for the power purchase constraint of the multi-load cluster is:

[0035]

[0036] In the formula, P G,max P G,min These are the upper and lower limits of transmission power for electricity purchase transactions between multi-load clusters and the power grid;

[0037] The expression for the operating constraints of the multi-variable load cluster is:

[0038]

[0039] In the formula, α m,c (t), α m,d (t) are binary parameters that measure the virtual energy storage charging and discharging state of the multi-element load cluster m at time t, respectively. If α m,c If α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased; conversely, if α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased. m,d If (t) = 1, then the virtual energy storage only discharges, meaning the actual operating power of the multi-load cluster is reduced; VSOC m,max (t) and VSOC m,min(t) represents the average of the upper and lower limits of VSOC for each load in the multi-load cluster m at time t;

[0040] The expression for the power balance constraint of the multi-load cluster is:

[0041] P m,G (t)+P m,c (t)+P m,PV (t)=z m (t)

[0042] In the formula, P m,PV (t) represents the photovoltaic output consumed by the multi-load cluster m in the distribution network at time t;

[0043] The expression for the photovoltaic output constraint within the distribution network is:

[0044]

[0045] In the formula, P PV,min (t) represents the lower limit of photovoltaic output of the distribution network at time t.

[0046] This invention also provides a two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, which, based on the above-mentioned two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering, includes:

[0047] The 3D clustering feature vector construction module is used to construct 3D clustering feature vectors based on the geographical location of virtual energy storage of multiple loads under the distribution network and the load baseline value at a set time.

[0048] The clustering result acquisition module is used to divide the multi-load virtual energy storage into clusters based on the three-dimensional clustering feature vector using the K-means clustering algorithm, and introduces an adaptive update mechanism for the clustering structure to dynamically adjust the boundaries and structure of the clusters to obtain the clustering results.

[0049] The aggregated load model construction module is used to superimpose the individual virtual energy storage corresponding to the clustering results through the individual virtual energy storage model to construct an aggregated load model under dynamic clustering conditions.

[0050] The cluster control module is used to construct a multi-load day-ahead collaborative optimization control model based on the aggregated load model under the dynamic clustering conditions as an equivalent constraint basis; and to solve the multi-load day-ahead collaborative optimization control model based on the set constraints to obtain the optimal day-ahead control scheme for the multi-load.

[0051] As a preferred embodiment of a two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, the expression of the three-dimensional clustering feature vector in the three-dimensional clustering feature vector construction module is as follows:

[0052] a i (t)=(x i ,y i ,z i (t))

[0053] In the formula, a i (t) represents the three-dimensional clustering feature vector; i represents the individual load; x i y i z i (t) are the feature vectors of the three dimensions, (x) i ,y i ) represents the geographical location, z i (t) represents the load baseline value.

[0054] As a preferred embodiment of the two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, the expression for the individual virtual energy storage model in the aggregated load model construction module is as follows:

[0055]

[0056] In the formula, VSOC i (t) represents the virtual state of charge (SOC) of individual load i at time t; E max Virtual capacity of individual load i Let Δτ be the virtual energy storage of unit load i at time t; Δτ is the virtual energy storage at time z. i (t) is the time it takes for the discharge to be exhausted at the specified power.

[0057] As a preferred embodiment of the two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, the expression of the aggregated load model under the dynamic clustering condition in the aggregated load model construction module is as follows:

[0058]

[0059] In the formula, z m (t) represents the load baseline value of the multi-element load cluster m; d m P represents the number of samples contained in cluster m. m (t) represents the actual operating power of the multi-load cluster m at time t; P i (t) represents the actual operating power of unit load i at time t; ΔP i (t) represents the actual capacity of unit load i participating in regulation at time t; P m,c (t) represents the capacity value for regulation involving the multi-load cluster m; These represent the virtual charging and discharging power limits of the multi-load cluster m at time t; VSOC m (t) represents the average value of the VSOC of each multi-element load in the multi-element load cluster m at time t.

[0060] As a preferred embodiment of the two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, the objective function of the day-ahead collaborative optimization control model for multiple loads in the grouping control module includes: minimizing the daily operating cost of the multiple load cluster, minimizing the total daily carbon emissions of the multiple load cluster, and maximizing the photovoltaic absorption rate of the distribution network.

[0061] The expression for minimizing the daily operating cost of the multi-load cluster is:

[0062]

[0063] In the formula, σ m,ES P represents the unit operation and maintenance cost coefficient for the multi-load cluster m; K represents the number of multi-load clusters; G (t) represents the electricity price at time t; P m,G (t) represents the power purchased by cluster k at time t;

[0064] The expression for minimizing the total daily carbon emissions of the multi-load cluster is:

[0065]

[0066] In the formula, μ C The carbon emission factor of the power distribution network;

[0067] The expression for maximizing the photovoltaic absorption rate of the distribution network is:

[0068]

[0069] In the formula, P PV (t) represents the photovoltaic output absorbed by the distribution network at time t, P PV,max (t) represents the upper limit of photovoltaic output of the distribution network at time t;

[0070] The overall objective function of the multi-load day-ahead collaborative optimization control model is expressed as follows:

[0071]

[0072] In the formula, μ1, μ2, and μ3 are the weight coefficients of each optimization objective; f 1-0 f 2-0 f 3-0 These are the initial values ​​before optimization for each optimization objective.

[0073] As a preferred embodiment of the two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, in the grouping control module, during the process of solving the day-ahead collaborative optimization control model of the multiple loads based on the set constraints, the set constraints include: power purchase constraints of the multiple load cluster, operation constraints of the multiple load cluster, power balance constraints of the multiple load cluster, and photovoltaic output constraints within the distribution network.

[0074] The expression for the power purchase constraint of the multi-load cluster is:

[0075]

[0076] In the formula, P G,max P G,min These are the upper and lower limits of transmission power for electricity purchase transactions between multi-load clusters and the power grid;

[0077] The expression for the operating constraints of the multi-variable load cluster is:

[0078]

[0079] In the formula, α m,c (t), α m,d (t) are binary parameters that measure the virtual energy storage charging and discharging state of the multi-element load cluster m at time t, respectively. If α m,c If α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased; conversely, if α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased. m,d If (t) = 1, then the virtual energy storage only discharges, meaning the actual operating power of the multi-load cluster is reduced; VSOC m,max (t) and VSOC m,min (t) represents the average of the upper and lower limits of VSOC for each load in the multi-load cluster m at time t;

[0080] The expression for the power balance constraint of the multi-load cluster is:

[0081] P m,G (t)+P m,c (t)+P m,PV (t)=z m (t)

[0082] In the formula, P m,PV (t) represents the photovoltaic output consumed by the multi-load cluster m in the distribution network at time t;

[0083] The expression for the photovoltaic output constraint within the distribution network is:

[0084]

[0085] In the formula, P PV,min(t) represents the lower limit of photovoltaic output of the distribution network at time t.

[0086] This invention has the following advantages: Based on the geographical location of virtual energy storage for multiple loads in a distribution network and the load baseline value at a set time, a three-dimensional clustering feature vector is constructed. Based on this three-dimensional clustering feature vector, the virtual energy storage for multiple loads is clustered using the K-means clustering algorithm, and an adaptive update mechanism for the cluster structure is introduced to dynamically adjust the boundaries and structure of the clusters to obtain the clustering results. Through a single virtual energy storage model, the single virtual energy storage corresponding to the clustering results is superimposed to construct an aggregated load model under dynamic clustering conditions. Using the aggregated load model under dynamic clustering conditions as the basis for equivalent constraints, a day-ahead collaborative optimization control model for multiple loads is constructed. Based on the set constraints, the day-ahead collaborative optimization control model for multiple loads is solved to obtain the optimal day-ahead control scheme for multiple loads. This invention introduces a virtual energy storage model for multiple loads, comprehensively considering the geographical distribution of the loads and their time-varying power consumption characteristics, constructing an accurate cluster aggregation model, and determining the cluster control boundary by quantifying the dynamically adjustable range of the clusters. This invention constructs a day-ahead clustering optimization and control model for multi-load distribution networks based on the equivalent constraint boundary formed by dynamic clustering. It maximizes the utilization of the multi-timescale controllable potential of multi-loads in the optimal scheduling of the distribution network, balancing electricity costs and low-carbon operation. This invention fully considers the spatiotemporal distribution differences of geographical location and load baseline, deeply explores the controllable potential of clusters, achieves dynamic updates of multi-load clustering results, and supports the synergistic optimization of the economic efficiency and low-carbon characteristics of multi-loads, enhancing the distribution network's ability to locally absorb distributed renewable energy. Attached Figure Description

[0087] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0088] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0089] Figure 1This is a flowchart illustrating the two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering provided in Embodiment 1 of the present invention.

[0090] Figure 2 This is a schematic diagram illustrating the specific implementation framework of the two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering provided in Embodiment 1 of the present invention.

[0091] Figure 3 This is a schematic diagram of the dynamic clustering process of multiple loads in the two-stage daytime grouping control method for multiple loads in a distribution network based on dynamic clustering provided in Embodiment 1 of the present invention.

[0092] Figure 4 This is a schematic diagram of the multi-element load distribution in a simulation region in one possible embodiment of Embodiment 1 of the present invention;

[0093] Figure 5 This is a schematic diagram of the dynamic clustering results of massive multivariate loads at certain times in one possible embodiment of Embodiment 1 of the present invention;

[0094] Figure 6 This is a schematic diagram of the operating power of the virtual energy storage for multiple loads in one possible embodiment provided in Embodiment 1 of the present invention;

[0095] Figure 7 This is a schematic diagram illustrating carbon emissions in different scenarios in one possible embodiment of the present invention, provided in Embodiment 1.

[0096] Figure 8 This is a schematic diagram illustrating different photovoltaic power consumption scenarios in one possible embodiment of Embodiment 1 of the present invention;

[0097] Figure 9 This is a schematic diagram of the architecture of the two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, provided in Embodiment 2 of the present invention. Detailed Implementation

[0098] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0099] Example 1

[0100] See Figure 1 and Figure 2Embodiment 1 of the present invention provides a two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering, comprising the following steps:

[0101] S1. Based on the geographical location of virtual energy storage for multiple loads under the distribution network and the load baseline value at a set time, construct a three-dimensional clustering feature vector;

[0102] S2. Based on the three-dimensional clustering feature vector, the multi-load virtual energy storage is divided into clusters using the K-means clustering algorithm, and an adaptive update mechanism for the clustering structure is introduced to dynamically adjust the boundaries and structure of the clusters to obtain the clustering results.

[0103] S3. By superimposing the individual virtual energy storage corresponding to the clustering results through the individual virtual energy storage model, an aggregated load model under dynamic clustering conditions is constructed.

[0104] S4. Using the aggregated load model under the dynamic clustering conditions as the equivalent constraint basis, construct a multi-load day-ahead collaborative optimization control model; based on the set constraint conditions, solve the multi-load day-ahead collaborative optimization control model to obtain the optimal day-ahead control scheme for multi-loads.

[0105] In this embodiment, in step S1, a three-dimensional clustering feature vector is constructed based on the geographical location of the virtual energy storage of multiple loads under the distribution network and the load baseline value at a set time.

[0106] Specifically, the expression for the three-dimensional clustering feature vector is:

[0107] a i (t)=(x i ,y i ,z i (t))

[0108] In the formula, a i (t) represents the three-dimensional clustering feature vector; i represents the individual load; x i y i z i (t) are the feature vectors of the three dimensions, (x) i ,y i ) represents the geographical location, z i (t) represents the load baseline value.

[0109] In this embodiment, in step S2, based on the three-dimensional clustering feature vector, the multi-load virtual energy storage is divided into clusters using the K-means clustering algorithm, and an adaptive update mechanism for the clustering structure is introduced to dynamically adjust the boundaries and structure of the clusters to obtain the clustering results.

[0110] Specifically, this invention employs the classic K-means clustering algorithm, which boasts clear principles, low computational complexity, and fast convergence speed, making it suitable for rapidly partitioning large-scale load samples. Based on this, a dynamic K-means clustering method integrating spatial location, load baseline, and virtual energy storage status is further proposed to achieve dynamic clustering of diverse loads during spatiotemporal evolution, providing support for subsequent clustered collaborative optimization and regulation.

[0111] The basic principle of K-means clustering is as follows: For a given feature set A = {a1, a2, ..., a...} consisting of n clustering features... i ,…,a n}(where a) i Let be the clustering features of the i-th sample (1≤i≤n). The algorithm divides it into K clusters, and the set of each cluster is B={B1,B2,…,B2,N-1}. m ,…,B K}(B m Let m be the m-th cluster, which contains a subset of samples (1 ≤ m ≤ K). The set of centroids of each cluster is c = {c1, c2, ..., c...} m ,…,c K},in:

[0112]

[0113] In the formula, d m Let m be the number of samples contained in cluster m.

[0114] K-means clustering can be viewed as an optimization problem, with the objective function being to optimize the set of features A by maximizing the performance of each sample point a through multiple iterations. i With the corresponding centroid c m The Euclidean distance is minimized, and its expression is:

[0115]

[0116] Based on the above principles, the K-means-based multivariate load dynamic clustering process can be obtained as follows: Figure 3 As shown, the specific steps are as follows:

[0117] T1. Input the feature set A(t) required for clustering = {a1(t), a2(t), ..., a n (t)} and the number of clusters K. Where a i (t) represents a three-dimensional variable, namely the clustering feature composed of geographical location and load baseline value as mentioned above, where a corresponds to individual load i. i (t)=(x i ,y i ,z i (t));

[0118] T2. To ensure accurate clustering results, A(t) needs to be normalized. This invention uses the deviation standardization method, the expression of which is:

[0119]

[0120] In the formula, a i,j (t) is the characteristic quantity a i The j-th physical quantity in (t), namely a i,1 (t), a i,2 (t), a i,3 (t) respectively correspond to x i y i z i (t); a j (t) max With a j (t) min These are the maximum and minimum values ​​of the aforementioned physical quantities in the feature set; This refers to the normalized value of the corresponding physical quantity;

[0121] Calculated Finally, the normalized feature set A can be obtained. * (t)={a1 * (t),a2 * (t),…,a n * (t)}.

[0122] T3. Input a randomly generated initial centroid set c0(t) = {c1(t), c2(t), ..., c K (t)} is used for subsequent clustering processes;

[0123] T4. Based on the K-means principle, calculate a according to the following formula. i * (t) Euclidean distance to the centroid of each multi-element load cluster, select the centroid of the single load i that is closest to it and add it to its corresponding multi-element load cluster;

[0124]

[0125] In the formula, M is a large constant, which serves as a penalty term for cluster B at time t. m (t) Number of multi-element loads d m (t) exceeds its time limit d max Apply a penalty distance to ensure that the number of diverse loads within the cluster is less than or equal to its upper limit;

[0126] T5. Calculate the multi-element load cluster m (whose set is represented by B) using the following formula. m (t) in each ai * The mean of (t) is used as the new centroid, and the updated centroid c' is... m (t) and belong to B m The sum of squared distances between other multi-element loads of (t) is minimized;

[0127]

[0128] T6. Determine whether the centroid has changed. If the following formula is not satisfied for the new centroid, repeat the above steps T4 and T5. Otherwise, output the multivariate load clustering result set B(t) and the corresponding centroid set c(t) at time t.

[0129] c' m (t)=c m (t)T7. Determine if t has reached 24h. If t = 24h, the clustering ends and the final clustering result B of the massive multivariate load is output, and the clustering ends. Otherwise, increase t by 1h and return to step one to continue dynamic clustering.

[0130] In this embodiment, in step S3, the individual virtual energy storage corresponding to the clustering results is superimposed through the individual virtual energy storage model to construct the aggregated load model under dynamic clustering conditions.

[0131] Specifically, to uniformly characterize the controllable capacity of various load types, an equivalent model of multi-load virtual storage (VS) based on power-energy dual-layer constraints is constructed, drawing on the equivalent modeling method of load virtual energy storage. This model simplifies the mapping relationship between load controllability and control variables, achieving a unified mathematical representation and standardized quantification of the controllable characteristics of multi-loads in the distribution network, providing model support for subsequent dynamic clustering and group-based collaborative optimization control. For any type of flexible single load i in the distribution network, the quantitative relationship between its controllable capacity at time t and its load baseline can be expressed as:

[0132]

[0133] In the formula, With Δ P i,a (t) represents the maximum upward and downward adjustable capacity of flexible unit load i at time t; z i (t) represents the baseline value of the daily load of individual unit load i at time t, which is usually obtained through load forecasting; and These are the typical load curve values ​​of individual load i at time t under the highest and lowest power consumption modes, respectively. They can be obtained through analysis of historical load data, with the load reduction direction taken as positive.

[0134] Therefore, the operating power of load i at time t is:

[0135] P i (t)=z i (t)-ΔP i (t),

[0136]

[0137] In the formula, P i (t) represents the actual operating power of unit load i at time t; ΔP i (t) represents the actual capacity of individual load i participating in regulation at time t, which can be obtained from the load cluster regulation results. When ΔP i When (t) < 0, it indicates that the individual load i increases its output power and absorbs electrical energy from the distribution network for "recharging"; similarly, when ΔP i When (t) > 0, it represents that load i reduces its output power and "discharges" into the distribution network. Therefore, in the virtual energy storage model, ΔP i (t) can be defined as the charging / discharging power of the virtual energy storage of a single load i.

[0138] Meanwhile, the virtual state of charge (VSOC) of individual load i in the multi-load VS model is defined as follows:

[0139]

[0140] In the formula, VSOC i (t) represents the virtual state of charge (SOC) of individual load i at time t; E max Virtual capacity of individual load i Energy in z i (t) is the time it takes for the discharge to be exhausted at the specified power.

[0141] In this embodiment, based on the clustering result B = {B1, B2, ..., B} K For simplification, this invention employs a linear superposition method to construct a multi-element load virtual energy storage cluster aggregation model; its expression is:

[0142]

[0143] In the formula, z m (t) represents the load baseline value of the multi-element load cluster m; d m P represents the number of samples contained in cluster m. m (t) represents the actual operating power of the multi-load cluster m at time t; P i (t) represents the actual operating power of unit load i at time t; ΔP i(t) represents the actual capacity of unit load i participating in regulation at time t; P m,c (t) represents the capacity value for regulation involving the multi-load cluster m; These represent the virtual charging and discharging power limits of the multi-load cluster m at time t; VSOC m (t) represents the average value of the VSOC of each multi-element load in the multi-element load cluster m at time t.

[0144] In this embodiment, in step S4, the aggregated load model under the dynamic clustering condition is used as the equivalent constraint basis to construct a multi-load day-ahead collaborative optimization control model; based on the set constraint conditions, the multi-load day-ahead collaborative optimization control model is solved to obtain the optimal day-ahead control scheme for multi-loads.

[0145] Specifically, based on the dynamic clustering results of the multi-load segments and their corresponding aggregation models, and comprehensively considering the economic efficiency and low-carbon characteristics of multi-load operators within the distribution network, as well as the effective absorption of photovoltaic power output during distribution network operation, this invention constructs a day-ahead collaborative optimization model in the cluster control stage, with the optimization of daily operating costs of multi-load segments, total daily carbon emissions, and photovoltaic absorption rate of the distribution network as multiple objectives. Using clusters as the basic control unit, this achieves economical and green collaborative control of large-scale multi-load segments and the power grid.

[0146] The objective functions of the multi-load day-ahead collaborative optimization and control model include: minimizing the daily operating cost of the multi-load cluster, minimizing the total daily carbon emissions of the multi-load cluster, and maximizing the photovoltaic absorption rate of the distribution network.

[0147] The daily operating cost of a multi-load cluster within a distribution network consists of the operation and maintenance costs of the multi-load cluster and the electricity purchase cost; the expression for minimizing the daily operating cost of the multi-load cluster is:

[0148]

[0149] In the formula, σ m,ES P represents the unit operation and maintenance cost coefficient for the multi-load cluster m; K represents the number of multi-load clusters; G (t) represents the electricity price at time t; P m,G (t) represents the power purchased by cluster k at time t;

[0150] The expression for minimizing the total daily carbon emissions of the multi-load cluster is:

[0151]

[0152] In the formula, μ C The carbon emission factor of the power distribution network;

[0153] The expression for maximizing the photovoltaic absorption rate of the distribution network is:

[0154]

[0155] In the formula, P PV (t) represents the photovoltaic output absorbed by the distribution network at time t, P PV,max (t) represents the upper limit of photovoltaic output of the distribution network at time t;

[0156] The overall objective function of the multi-load day-ahead collaborative optimization control model is expressed as follows:

[0157]

[0158] In the formula, μ1, μ2, and μ3 are the weight coefficients of each optimization objective; f 1-0 f 2-0 f 3-0 These are the initial values ​​before optimization for each optimization objective.

[0159] In this embodiment, the corresponding equivalent constraints that the multi-load virtual energy storage cluster needs to meet include: the power purchase constraint of the multi-load cluster, the operation constraint of the multi-load cluster, the power balance constraint of the multi-load cluster, and the photovoltaic output constraint within the distribution network.

[0160] Among these, the purchase of electricity from the distribution network by a large number of diverse loads must meet the upper and lower limits of transmission power constraints within the corresponding area. That is, the expression for the power purchase constraint of the diverse load cluster is:

[0161]

[0162] In the formula, P G,max P G,min These are the upper and lower limits of transmission power for electricity purchase transactions between multi-load clusters and the power grid;

[0163] The operation of a multi-load virtual energy storage cluster must satisfy internal virtual capacity balance constraints, charge / discharge power constraints, charge / discharge state constraints, and state of charge constraints. The expression for the operating constraints of the multi-load cluster is as follows:

[0164]

[0165] In the formula, α m,c (t), α m,d (t) are binary parameters that measure the virtual energy storage charging and discharging state of the multi-element load cluster m at time t, respectively. If α m,c If α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased; conversely, if α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased. m,d If (t) = 1, then the virtual energy storage only discharges, meaning the actual operating power of the multi-load cluster is reduced; VSOC m,max (t) and VSOCm,min (t) represents the average of the upper and lower limits of VSOC for each load in the multi-load cluster m at time t;

[0166] The expression for the power balance constraint of the multi-load cluster is:

[0167] P m,G (t)+P m,c (t)+P m,PV (t)=z m (t)

[0168] In the formula, P m,PV (t) represents the photovoltaic output consumed by the multi-load cluster m in the distribution network at time t;

[0169] The expression for the photovoltaic output constraint within the distribution network is:

[0170]

[0171] In the formula, P PV,min (t) represents the lower limit of photovoltaic output of the distribution network at time t.

[0172] In this embodiment, based on the above objective function and constraints, it can be seen that the multi-load day-ahead collaborative optimization control model is a typical mixed-integer linear programming (MILP) problem. The yalmip and cplex solvers can be used to solve the cluster optimization control model, thereby formulating a multi-load day-ahead cluster optimal control scheme based on dynamic clustering.

[0173] In one possible embodiment, a typical example of rural power distribution network control is provided below:

[0174] I. Example Setup:

[0175] This invention constructs a 2km*2km simulation verification area based on a typical township power distribution network operation scenario. The area is divided into three typical mixed agricultural production and residential areas. The internal load structure consists of two types of small loads and one type of large load: the small loads include 80 electric tobacco curing devices and 160 agricultural wells, while the large loads consist of 6 agricultural machinery charging devices. Based on the characteristics of rural electrification construction, it is assumed that the small loads are evenly distributed within the area, while the large loads are concentrated at the central nodes of each production and residential area, as shown in the distribution diagram. Figure 4 As shown in the figure. To verify the effectiveness and adaptability of the proposed two-stage day-ahead grouping control method for multiple loads in the distribution network based on dynamic clustering, this invention conducts simulation experiments using the aforementioned region as a typical scenario. The typical operating parameters of the multiple loads involved and the agricultural time-of-use electricity price are shown in Tables 1 and 2.

[0176] Parameter name Parameter value Rated power of electric tobacco curing equipment (kW) 4 Rated power of well (kW) 7.5 Rated charging power for agricultural machinery (kW) 30

[0177] Table 1 Typical operating parameters of multi-load systems

[0178]

[0179]

[0180] Table 2 Agricultural Time-of-Use Electricity Prices

[0181] Since agricultural machinery charging loads differ significantly from the other two types of small loads in terms of capacity and regulation characteristics, this invention treats them as independent clusters to directly participate in the group regulation and optimization process; the dynamic clustering method is only applicable to small multi-element loads with large scale and similar response characteristics.

[0182] II. Simulation Result Analysis:

[0183] 1. Clustering-based computational speed analysis:

[0184] The total computation time T of the two-stage method (i.e., clustering and regulation) proposed in this invention corresponding to each cluster number K. cal And T as a function of K cal The corresponding year-on-year growth rates are shown in Table 3:

[0185] K <![CDATA[Calculate duration T cal / s]]> Year-on-year growth rate / % 1 16.23 / 2 19.50 20.15 4 23.23 19.13 6 29.10 25.27 8 42.44 45.84 10 54.92 29.41 240 383.54 /

[0186] Table 3. Computation time for each cluster number K

[0187] As shown in Table 3, as the value of K increases, the corresponding computation time gradually increases while the computation speed decreases. When the value of K increases from 6 to 8, the year-on-year growth rate of computation time is the highest, meaning that the computation speed decreases most significantly at this point. Therefore, this invention focuses on optimizing control efficiency and selects the number of clusters K=6 for subsequent control and analysis. When K=240, under the traditional centralized optimization control method, the computation time increases by 1218.01% compared to K=6. Therefore, the two-stage day-ahead cluster control method for multivariate loads based on dynamic clustering proposed in this invention can significantly improve the solution efficiency of massive multivariate load optimization control.

[0188] 2. Analysis of multivariate load clustering results:

[0189] Based on the dynamic clustering method proposed in this invention, Figure 4 The massive multi-variable loads within the simulation area were clustered, with K=6 as the number of clusters. For ease of observation and analysis, the three-dimensional distribution of the clustering results at certain time points and their corresponding two-dimensional plane mappings are obtained as follows: Figure 5 As shown.

[0190] Depend on Figure 5It can be seen that each multi-dimensional load cluster contains different types of loads. This means that under this method, multi-dimensional load clustering strictly follows the principles of geographical proximity and similar load baseline values, rather than aggregating based on inherent load types. Therefore, it can ensure that loads with similar controllable potential are clustered into clusters at each time point. Furthermore, the three-dimensional sample set composed of the baseline values ​​and geographical locations of the massive multi-dimensional loads changes dynamically at different times, leading to differences in clustering results at each time point.

[0191] At t=6h, the baseline values ​​of various multivariate loads were all low, geographical location played a dominant role in clustering, and the clusters were nearly evenly distributed.

[0192] As time progressed, the spatiotemporal differences of the multivariate load baseline values ​​at t=15h gradually became significant, and the influence of geographical location on clustering was relatively weakened. Among them, the load baseline values ​​of clusters 1, 4, and 6 were much higher than those of clusters 2, 3, and 5. Furthermore, under the dominance of the load baseline, clusters 2, 3, and 4 all exhibited cross-regional aggregation.

[0193] At t=24h, the multivariate load clustering results change dynamically again with the changing load baseline. Taking cluster 4 as an example, there are still loads with relatively high baseline values ​​in cluster 4. Therefore, cross-regional aggregation is carried out. The same applies to clusters 1 and 5. The load baseline values ​​of clusters 2, 3 and 6 are closer, so their distribution is more uniform.

[0194] In summary, the clustering results for each time period change dynamically with the time-varying three-dimensional dynamic clustering index. This can aggregate different types of mixed-distribution multi-dimensional loads that are geographically close and have similar load baseline values ​​into a unified cluster and continuously update its composition. This can provide equivalent constraint support for the cluster optimization and control of massive multi-dimensional loads.

[0195] 3. Analysis of the results of virtual energy storage optimization and control for multiple loads:

[0196] 3.1 Power Analysis of Virtual Energy Storage Operation for Multiple Loads

[0197] Depend on Figure 6It can be observed that: from 01:00 to 04:00, due to the high virtual energy storage capacity of the diversified loads, a certain amount of discharge can be carried out, thereby reducing the amount of electricity purchased and carbon emissions; from 05:00 to 07:00, during the off-peak electricity price period, the virtual energy storage of the diversified loads charges to reserve capacity for subsequent discharge, maintaining supply and demand balance; from 08:00 to 20:00, during the normal and peak electricity price periods, photovoltaics becomes one of the important power supply forms for diversified loads. When photovoltaic output is insufficient, electricity is purchased from the grid or supplemented by the discharge of virtual energy storage of the diversified loads. When photovoltaic output is excessive, it charges the virtual energy storage of the diversified loads, which is the reserve capacity for virtual energy storage discharge at night. After the energy storage capacity reaches its limit, the remaining photovoltaic output is difficult to absorb, so there is still curtailment. At this time, the purchase of electricity from the grid is reduced, thus reducing operating costs and carbon emissions; from 08:00 to 09:00 and from 18:00 to 24:00, during the normal and peak electricity price periods, the discharge of virtual energy storage of the diversified loads reduces the amount of electricity purchased, reducing operating costs and carbon emissions.

[0198] 3.2 Comparison of Optimized Operation Results of Virtual Energy Storage for Multiple Loads

[0199] To visually compare and analyze the superior effects of the proposed method in improving the economy, environmental protection, and renewable energy absorption of diversified loads in distribution networks, three scenarios are proposed for verification, as shown in Table 4:

[0200] Scene Consider configuring photovoltaics in the power distribution network Considering virtual energy storage characteristics Ⅰ √ × Ⅱ × √ Ⅲ √ √

[0201] Table 4 Comparison of Scene Settings

[0202] In Table 4, Scenario III refers to the multi-load day-ahead optimization and control under the method proposed in this invention, while Scenario I and II are set up as comparison scenarios. Thus, the comparison of the optimization operation results of each part under various scenarios is shown in Table 5.

[0203] Scene Ⅰ Ⅱ Ⅲ Electricity purchased / kW·h 3420.5 13655.0 1982.2 Electricity purchase cost / yuan 1950.5 8229.7 892.7 Multi-load operation and maintenance cost / yuan / 483.0 347.4 Photovoltaic power generation operation and maintenance cost / yuan 411.2 / 466.8 Cost of penalty for abandoning light / yuan 379.9 / 32.8 Photovoltaic grid integration rate / % 87.1 / 98.9 Total operating cost / yuan 1950.5 8712.7 1240.1 Carbon emissions / kg 2000.6 7986.8 1159.4

[0204] Table 5 Optimization Results for Various Scenarios

[0205] As can be seen from Table 5, compared with Scenario I, Scenario III introduces the virtual energy storage theory in the operation of multiple loads, making multiple loads quantifiable, flexible and controllable resources. Therefore, the photovoltaic absorption capacity of the distribution network is enhanced, and the corresponding photovoltaic power generation operation and maintenance costs increase, curtailment penalties decrease, electricity purchases decrease, and carbon emissions decrease.

[0206] Compared to Scenario II, Scenario III introduces photovoltaic output into the operation of virtual energy storage for multiple loads, reducing the reliance of multiple loads on traditional power generation output. As a result, the amount of electricity purchased, the cost of electricity purchase, and carbon emissions are significantly reduced. Meanwhile, in Scenario II, which relies solely on virtual energy storage for multiple loads to maintain the system's supply and demand balance, more charging and discharging phenomena occur, thus slightly increasing the operation and maintenance costs of multiple loads.

[0207] In summary, compared to scenario I, the method proposed in this invention increases the photovoltaic absorption rate of the distribution network by 11.8%, and reduces the total operating cost and carbon emissions of multiple loads by 36.42% and 42.05%, respectively. Compared to scenario II, the method of this invention reduces the total operating cost and carbon emissions of the loads by 85.77% and 85.48%, respectively. Therefore, the multi-load cluster optimization and control method that considers the potential of virtual energy storage proposed in this invention effectively saves the operating cost of multiple loads, promotes green electricity use by loads, and facilitates the efficient utilization of new energy sources within the distribution network.

[0208] 4. Comparative analysis of carbon emissions from multiple loads:

[0209] Depend on Figure 7 It can be known that:

[0210] Considering only the operation of virtual energy storage for multiple loads (i.e., Scenario II), the carbon emissions are high at all times, which has a significant impact on the environment; while in Scenario I and Scenario III, the carbon emissions are effectively reduced by using photovoltaics configured in the distribution network to supply power to multiple loads.

[0211] Scenario III, compared to Scenario I, further considers the optimized operation of virtual energy storage for diverse loads, fully utilizing the potential for multi-load regulation. However, its carbon emissions fluctuate significantly under the influence of virtual energy storage charging and discharging. Combined with... Figure 6 It can be seen that during the period from 05:00 to 07:00, the virtual energy storage charges and stores electricity, which leads to an increase in the amount of electricity purchased at this time and exacerbates carbon emissions. However, during periods such as 01:00 to 04:00, 08:00 to 09:00, and 18:00 to 24:00, the virtual energy storage discharges to meet the system's supply and demand balance, and the corresponding purchase of electricity from the grid decreases, resulting in a decrease in carbon emissions that are even close to zero.

[0212] In summary, the optimized control method proposed in this invention, which considers the virtual energy storage characteristics of multiple loads, can fully leverage the potential for flexible control of multiple loads and achieve time-shifted photovoltaic output. As shown in Table 5, the total carbon emissions of this method are effectively reduced compared to Scenario I, further promoting the green operation of multiple loads in the distribution network.

[0213] 5. Comparative analysis of photovoltaic power grid absorption:

[0214] Depend on Figure 8 It can be seen that, compared with Scenario I which only considers photovoltaic output, the optimized operation based on the virtual energy storage characteristics of multi-load can significantly absorb excess photovoltaic power through virtual energy storage charging during peak photovoltaic output periods such as 08:00-18:00, thereby promoting the effective utilization of photovoltaic power and reducing curtailment.

[0215] In summary, this invention is applicable to distribution network scenarios with a high proportion of distributed renewable energy access, and is particularly suitable for areas with diverse loads including industrial production, agricultural irrigation, commercial services, residential life, and public utilities. For example, in rural distribution networks, there are diverse loads such as electric tobacco curing equipment, agricultural wells, and agricultural machinery charging equipment, and when a large number of new energy sources such as photovoltaics are connected, this method can be used to achieve dynamic clustering and grouping control of diverse loads, solve problems such as peak-hour absorption difficulties and insufficient supply during off-peak hours, improve the renewable energy absorption rate, and ensure the green and economical operation of the distribution network.

[0216] This invention constructs a day-ahead clustering optimization and control model for multi-load distribution networks based on the equivalent constraint boundary formed by dynamic clustering. It maximizes the utilization of the multi-timescale controllable potential of multi-loads in the optimal scheduling of the distribution network, balancing electricity costs and low-carbon operation. This invention fully considers the spatiotemporal distribution differences of geographical location and load baseline, deeply explores the controllable potential of clusters, achieves dynamic updates of multi-load clustering results, and supports the synergistic optimization of the economic efficiency and low-carbon characteristics of multi-loads, enhancing the distribution network's ability to locally absorb distributed renewable energy.

[0217] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0218] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0219] Example 2

[0220] See Figure 9 Embodiment 2 of the present invention also provides a two-stage day-ahead grouping control device for multiple loads in a distribution network based on dynamic clustering, comprising:

[0221] The 3D clustering feature vector construction module 001 is used to construct a 3D clustering feature vector based on the geographical location of virtual energy storage of multiple loads under the distribution network and the load baseline value at a set time.

[0222] The clustering result acquisition module 002 is used to divide the multi-load virtual energy storage into clusters based on the three-dimensional clustering feature vector using the K-means clustering algorithm, and introduces an adaptive update mechanism for the clustering structure to dynamically adjust the boundaries and structure of the clusters to obtain the clustering results.

[0223] The aggregated load model construction module 003 is used to superimpose the individual virtual energy storage corresponding to the clustering results through the individual virtual energy storage model to construct an aggregated load model under dynamic clustering conditions.

[0224] The cluster control module 004 is used to construct a multi-load day-ahead collaborative optimization control model based on the aggregated load model under the dynamic clustering conditions as an equivalent constraint basis; and to solve the multi-load day-ahead collaborative optimization control model based on the set constraints to obtain the optimal day-ahead control scheme for the multi-load.

[0225] In this embodiment, the expression for the three-dimensional clustering feature vector in the three-dimensional clustering feature vector construction module 001 is:

[0226] a i (t)=(x i ,y i ,z i (t))

[0227] In the formula, a i (t) represents the three-dimensional clustering feature vector; i represents the individual load; x i y i z i (t) are the feature vectors of the three dimensions, (x) i ,y i ) represents the geographical location, z i (t) represents the load baseline value.

[0228] In this embodiment, the expression for the single-unit virtual energy storage model in the aggregated load model construction module 003 is:

[0229]

[0230] In the formula, VSOC i (t) represents the virtual state of charge (SOC) of individual load i at time t; E max Virtual capacity of individual load i Let Δτ be the virtual energy storage of unit load i at time t; Δτ is the virtual energy storage at time z. i (t) is the time it takes for the discharge to be exhausted at the specified power.

[0231] In this embodiment, the expression for the aggregated load model under dynamic clustering conditions in the aggregated load model construction module 003 is as follows:

[0232]

[0233] In the formula, z m (t) represents the load baseline value of the multi-element load cluster m; d m P represents the number of samples contained in cluster m. m (t) represents the actual operating power of the multi-load cluster m at time t; P i (t) represents the actual operating power of unit load i at time t; ΔP i (t) represents the actual capacity of unit load i participating in regulation at time t; P m,c (t) represents the capacity value for regulation involving the multi-load cluster m; These represent the virtual charging and discharging power limits of the multi-load cluster m at time t; VSOC m (t) represents the average value of the VSOC of each multi-element load in the multi-element load cluster m at time t.

[0234] In this embodiment, the objective function of the multi-load day-ahead collaborative optimization control model in the group control module 004 includes: minimizing the daily operating cost of the multi-load cluster, minimizing the total daily carbon emissions of the multi-load cluster, and maximizing the photovoltaic absorption rate of the distribution network.

[0235] The expression for minimizing the daily operating cost of the multi-load cluster is:

[0236]

[0237] In the formula, σ m,ES P represents the unit operation and maintenance cost coefficient for the multi-load cluster m; K represents the number of multi-load clusters; G (t) represents the electricity price at time t; P m,G (t) represents the power purchased by cluster k at time t;

[0238] The expression for minimizing the total daily carbon emissions of the multi-load cluster is:

[0239]

[0240] In the formula, μ C The carbon emission factor of the power distribution network;

[0241] The expression for maximizing the photovoltaic absorption rate of the distribution network is:

[0242]

[0243] In the formula, P PV (t) represents the photovoltaic output absorbed by the distribution network at time t, P PV,max (t) represents the upper limit of photovoltaic output of the distribution network at time t;

[0244] The overall objective function of the multi-load day-ahead collaborative optimization control model is expressed as follows:

[0245]

[0246] In the formula, μ1, μ2, and μ3 are the weight coefficients of each optimization objective; f 1-0 f 2-0 f 3-0 These are the initial values ​​before optimization for each optimization objective.

[0247] In this embodiment, in the process of solving the multi-load day-ahead collaborative optimization control model based on the set constraints in the group control module 004, the set constraints include: multi-load cluster power purchase constraints, multi-load cluster operation constraints, multi-load cluster power balance constraints, and photovoltaic output constraints within the distribution network.

[0248] The expression for the power purchase constraint of the multi-load cluster is:

[0249]

[0250] In the formula, P G,max P G,min These are the upper and lower limits of transmission power for electricity purchase transactions between multi-load clusters and the power grid;

[0251] The expression for the operating constraints of the multi-variable load cluster is:

[0252]

[0253] In the formula, α m,c (t), α m,d (t) are binary parameters that measure the virtual energy storage charging and discharging state of the multi-element load cluster m at time t, respectively. If α m,c If α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased; conversely, if α = 1, then the virtual energy storage only charges, meaning the actual operating power of the multi-load cluster is increased. m,d If (t) = 1, then the virtual energy storage only discharges, meaning the actual operating power of the multi-load cluster is reduced; VSOC m,max (t) and VSOC m,min (t) represents the average of the upper and lower limits of VSOC for each load in the multi-load cluster m at time t;

[0254] The expression for the power balance constraint of the multi-load cluster is:

[0255] P m,G (t)+P m,c (t)+P m,PV (t)=zm (t)

[0256] In the formula, P m,PV (t) represents the photovoltaic output consumed by the multi-load cluster m in the distribution network at time t;

[0257] The expression for the photovoltaic output constraint within the distribution network is:

[0258]

[0259] In the formula, P PV,min (t) represents the lower limit of photovoltaic output of the distribution network at time t.

[0260] It should be noted that the information interaction and execution process between the modules of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0261] Example 3

[0262] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, wherein the computer-readable storage medium stores program code for a two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering. The program code includes instructions for executing the two-stage day-ahead grouping control method for multiple loads in a distribution network based on dynamic clustering as described in Embodiment 1 or any possible implementation thereof.

[0263] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives, SSDs).

[0264] Example 4

[0265] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0266] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can call the program instructions to execute the dynamic clustering-based two-stage day-ahead grouping control method for multiple loads in the distribution network according to Embodiment 1 or any possible implementation thereof.

[0267] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0268] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0269] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using program code executable by a computing system, thereby storing them in a storage system for execution by the computing system. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0270] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A two-stage day-ahead clustering control method for multi-element load of power distribution network based on dynamic clustering, characterized in that, The method comprises the following steps: constructing a three-dimensional clustering feature vector based on the geographical position of the multi-element load virtual energy storage under the power distribution network and the load baseline value at a set time; based on the three-dimensional clustering feature vector, the multi-element load virtual energy storage is clustered and divided by a K-means clustering algorithm, and a clustering structure adaptive updating mechanism is introduced to dynamically adjust the boundary and structure of the cluster to obtain a clustering result; by means of a single virtual energy storage model, the single virtual energy storage corresponding to the clustering result is superimposed to construct an aggregated load model under dynamic clustering conditions; taking the aggregated load model under the dynamic clustering conditions as an equivalent constraint basis, a multi-element load day-ahead collaborative optimization regulation model is constructed; based on the set constraint condition, the multi-element load day-ahead collaborative optimization regulation model is solved to obtain an optimal multi-element load day-ahead regulation scheme.

2. The dynamic clustering-based multi-energy load two-stage day-ahead grouping control method for power distribution networks according to claim 1, characterized in that, The expression of the three-dimensional clustering feature vector is: a i (t) = (x i ,y i ,z i (t)) wherein a i (t) is a three-dimensional clustering feature vector; i For single load; x i , y i , z i (t) are eigenvectors of three dimensions respectively, (x i , y i ) are geographical positions, z i (t) are load baseline values.

3. The dynamic clustering-based multi-energy load two-stage day-ahead grouping control method for power distribution networks according to claim 2, characterized in that, The expression of the single virtual energy storage model is: wherein VSOC i (t) is the virtual state of charge of the monobloc i at time t; E max is the virtual capacity of the monobloc i is the virtual energy stored in the monobloc i at the current time t; Δτ is the time during which the monobloc i is discharged at the power P i (t) is the time during which the monobloc i is discharged at the power P 4. The dynamic clustering-based multi-energy load two-stage day-ahead grouping control method for power distribution networks according to claim 3, characterized in that, The expression of the aggregated load model under the dynamic clustering conditions is: In the formula, z m (t) is the load baseline value of the multi-element load cluster m; d m is the number of samples contained in the cluster m; P m (t) is the actual running power of the multi-element load cluster m at time t; P i (t) is the actual running power of the single-element load i at time t; ΔP i (t) is the actual capacity of monomer load i participating in regulation at time t; P m,c (t) is the capacity value of the multi-load cluster m participating in the regulation; VSOC (t) is the virtual charging and discharging power upper limit of the multi-load cluster m at time t; VSOC m (t) is the average value of the VSOC of each multi-load in the multi-load cluster m at time t.

5. The dynamic clustering-based multi-energy load two-stage day-ahead grouping control method for power distribution networks according to claim 4, characterized in that, The objective function of the multi-element load day-ahead collaborative optimization regulation model comprises: minimization of multi-element load cluster daily operation cost, minimization of multi-element load cluster daily carbon emission total amount, and maximization of power distribution network photovoltaic consumption rate; The expression of the minimization of multi-element load cluster daily operation cost is: In the formula, σ m,ES is the unit operation cost coefficient of the multi-element load cluster m; K is the number of multi-element load clusters; P G (t) is the electricity price at time t; P m,G (t) is the power purchased by cluster k at time t; The expression of the minimization of multi-element load cluster daily carbon emission total amount is: wherein μ C is the carbon emission factor for the distribution grid; The expression of the maximization of power distribution network photovoltaic consumption rate is: In the formula, P PV (t) is the photovoltaic output of the distribution network at time t PV,max (t) is the upper limit of the photovoltaic output of the distribution network at time t The expression of the total objective function of the multi-element load day-ahead collaborative optimization regulation model is: In the formula, μ1, μ2, μ3 are weight coefficients of respective optimization objectives; f 1-0 , f 2-0 , f 3-0 are initial values of respective optimization objectives before participation in optimization.

6. The dynamic clustering-based multi-energy load two-stage day-ahead grouping control method for power distribution networks according to claim 5, characterized in that, In the process of solving the multi-element load day-ahead collaborative optimization regulation model based on the set constraint condition, the set constraint condition comprises: multi-element load cluster power purchase power constraint, multi-element load cluster operation constraint, multi-element load cluster power balance constraint, and power distribution network photovoltaic output constraint; The expression of the multi-element load cluster power purchase power constraint is: In the formula, P G,max , P G,min are the upper and lower limits of transmission power for the multi-element load cluster to purchase electricity from the power grid, respectively. The expression of the multi-element load cluster operation constraint is: wherein α m,c (t) and α m,d (t) are binary parameters respectively measuring the virtual energy storage charging and discharging state of the multi-load cluster m at time t, if α m,c (t) = 1, the corresponding virtual energy storage only charges, i.e. the actual multi-load cluster operating power is up-regulated, otherwise if α m,d (t) = 1, the corresponding virtual energy storage only discharges, i.e. the actual multi-load cluster operating power is down-regulated; VSOC m,max (t) and VSOC m,min (t) are the average values of the upper and lower limits of the VSOC of each load in the multi-load cluster m at time t, respectively. The expression of the multi-element load cluster power balance constraint is: P m,G (t)+P m,c (t)+P m,PV (t)=z m (t) In the formula, P m,PV (t) is the photovoltaic output consumed by the multi-element load cluster m in the power distribution network at time t; The expression of the power distribution network photovoltaic output constraint is: In the formula, P PV,min (t) is the lower limit of photovoltaic output of the power distribution network at time t.

7. The power distribution network multi-element load two-stage day-ahead grouping regulation device based on dynamic clustering, adopting the power distribution network multi-element load two-stage day-ahead grouping regulation method based on dynamic clustering in any one of claims 1-6, characterized in that, The method comprises the following steps: a three-dimensional clustering feature vector construction module is used to construct a three-dimensional clustering feature vector based on the geographical position of the multi-element load virtual energy storage under the power distribution network and the load baseline value at a set time; a clustering result acquisition module is used to cluster and divide the multi-element load virtual energy storage based on the three-dimensional clustering feature vector by means of a K-means clustering algorithm, and a clustering structure adaptive updating mechanism is introduced to dynamically adjust the boundary and structure of the cluster to obtain a clustering result; an aggregated load model construction module is used to superimpose the single virtual energy storage corresponding to the clustering result by means of a single virtual energy storage model to construct an aggregated load model under dynamic clustering conditions; a cluster regulation module is used to take the aggregated load model under the dynamic clustering conditions as an equivalent constraint basis to construct a multi-element load day-ahead collaborative optimization regulation model; based on the set constraint condition, the multi-element load day-ahead collaborative optimization regulation model is solved to obtain an optimal multi-element load day-ahead regulation scheme. 8.The dynamic clustering based power distribution network multi-element load two-stage day-ahead grouping regulation device according to claim 7, characterized in that, In the three-dimensional clustering feature vector construction module, the expression of the three-dimensional clustering feature vector is: a i (t) = (x i ,y i ,z i (t)) In the formula, a i (t) is a three-dimensional clustering feature vector; i for single load; x i , y i , z i (t) are eigenvectors of three dimensions, respectively, (x i , y i ) are geographical positions, z i (t) are load baseline values. 9.The dynamic clustering based power distribution network multi-element load two-stage day-ahead grouping regulation device according to claim 8, characterized in that, In the aggregated load model construction module, the expression of the monomer virtual energy storage model is: wherein VSOC i (t) is the virtual state of charge of the monobloc i at time t; E max is the virtual capacity of the monobloc i is the virtual energy storage of the monobloc i at the current time t; Δτ is the time for the monobloc i to discharge the virtual energy storage at the current time t by the power P i (t) is the time for the monobloc i to discharge the virtual energy storage at the current time t by the power P 10. The dynamic clustering based power distribution network multi-energy load two-stage day-ahead grouping regulation device according to claim 9, characterized in that, In the aggregated load model construction module, the expression of the aggregated load model under the dynamic clustering condition is: In the formula, z m (t) is the load baseline value of the multi-element load cluster m; d m is the number of samples contained in the cluster m; P m (t) is the actual running power of the multi-element load cluster m at time t; P i (t) is the actual running power of the single-element load i at time t; ΔP i (t) is the actual capacity of monomer load i participating in regulation at time t; P m,c (t) is the capacity value of the multi-load cluster m participating in the regulation; respectively, the virtual charging and discharging power upper limit of the multi-load cluster m at time t; VSOC m (t) is the average value of each multi-load VSOC in the multi-load cluster m at time t.

Images