A latent decision result based combinatorial clustering scenario reduction method and system

Abstract

The application discloses a kind of combination clustering scene reduction method and system based on potential decision result, the uncertainty factor of each node in power system is expressed as multidimensional curve input domain in the form of time sequence curve;Based on multidimensional curve input domain, the power supply planning problem solving model under different time scale scene is constructed, considering operation checking, multidimensional curve input domain is converted into potential decision result domain for clustering analysis;SOM neural network clustering is carried out to potential decision result, and preliminary clustering result is obtained;SOM and k-medoids combination clustering method is used to finally classify preliminary clustering result, and finally the typical scene used in operation checking is obtained.The application is used to cluster the most representative scene set in power supply planning typical scene operation checking, to reduce the complexity of solving model to the greatest extent under the condition of guaranteeing the reliability of checking result, improve calculation efficiency.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power supply planning, and particularly relates to a combined clustering scenario reduction method and system based on potential decision results. BACKGROUND

[0002] The traditional power supply planning problem meets the power demand of a region in a target year by deciding when, where and how much capacity of generator sets to build. With the large-scale grid connection of renewable energy on the power supply side and the continuous improvement of the demand side load level, decision makers need to conduct operation checking on the proposed power investment decision results in order to cope with power fluctuations from both the source and the load. Whether the operation checking is conducted through a unit commitment model or an economic dispatch model, the time scale involved is usually hourly or even minute-based, and the variables involved include the hourly output of all types of units and the hourly fluctuation of load power. In addition, the existence of numerous nodes in the whole system makes the problem to be solved a large-scale mixed integer linear programming (MILP) problem. At the same time, with the increase of the planning period and the number of nodes in the system, the complexity of the problem to be solved will be further improved.

[0003] Therefore, in order to cope with the solving complexity brought by the annual hourly operation checking, the operation checking of typical scenarios is usually used to replace the high-precision operation checking, which involves scenario reduction technology. The traditional power supply planning scenario reduction method usually adopts a heuristic method, that is, a typical scenario is manually selected according to typical labels such as seasons, holidays, key indicators such as load curves and new energy output curves. However, with the gradual increase of the penetration rate of new energy and the mutual coupling of various energy systems on the demand side, more uncertain scenarios are generated from the uncertainties on both the source and the load sides. The selection by subjective judgment and according to a single standard will no longer be applicable to complex power networks with high penetration of new energy. Therefore, the current power supply planning decision makers use clustering technology to reduce the number of typical scenarios and improve the scenario selection effect.

[0004] The currently widely used clustering methods include spectral clustering, k-means clustering, k-medoids clustering and hierarchical clustering. At the same time, most papers also conclude that there is no absolute best among all different clustering algorithms. The above scenario reduction methods all perform clustering analysis on the input domain of the planning model, such as historical load curves and historical new energy output curves. The advantage is that the operation is simple and the time correlation of the time series can be preserved. However, for the power supply planning model, the relationship between the output results and the input variables is highly nonlinear. In the input domain clustering, two scenarios with the same class may lead to different decision results. Therefore, the clustering results based on the input data of the power supply planning often do not have good practical value and may not be representative in the final scenario reduction effect.

[0005] Therefore, the scenario reduction method based on investment decision cost can more directly distinguish similar scenarios from a large number of scenarios to make more accurate clustering, and in addition, reducing the high-dimensional multiple input data into potential investment results can greatly improve the efficiency of clustering and reduce the difficulty of calculation solution, therefore, it is necessary to summarize a clustering method based on investment decision cost to improve the efficiency of typical scenario analysis. SUMMARY

[0006] The technical problem to be solved by the present application is to provide a combined clustering scenario reduction method and system based on potential decision results to solve the technical problem of scenario number reduction in power source planning operation simulation.

[0007] The present application adopts the following technical solutions:

[0008] A combined clustering scenario reduction method based on potential decision results comprises the following steps:

[0009] S1, representing the uncertainty factors of each node in the power system as a multi-dimensional curve input domain in the form of a time sequence curve;

[0010] S2, constructing a power source planning problem solving model under different time scales considering operation checking based on the multi-dimensional curve input domain obtained in step S1, and converting the multi-dimensional curve input domain into a potential decision result domain for clustering analysis;

[0011] S3, performing SOM neural network clustering on the potential decision results obtained in step S2 to obtain a preliminary clustering result;

[0012] S4, performing final classification on the preliminary clustering result obtained in step S3 by using a SOM and k-medoids combined clustering method to obtain a typical scenario used for operation checking.

[0013] Specifically, in step S1, the multi-dimensional curve input domain Z is represented as:

[0014]

[0015] wherein k / K is the scenario number / scenario set, G RE is the number of new energy power sources, is a multi-dimensional input curve vector in scenario k, N is the total number of nodes in the power system, T is the number of time in a day in scenario k, and R is the real number domain.

[0016] Specifically, in step S2, the scenario-based model is taken as an alternative of the power supply planning solution model based on full-scenario operation checking, the solved scenario-based model is simplified to a power supply planning model containing only a single scenario for solution, the weight of a typical day is set to 1, the power supply commissioning sequence under the scenario k is obtained by solving the single-scenario model, the potential investment decision result is obtained by multiplying the commissioning cost of each type of power supply commissioning sequence, and the potential decision result domain Γ composed of potential decision results is constructed.

[0017] Further, the potential decision result domain Γ is specifically:

[0018]

[0019] Among them, is the potential investment decision result, k / K is the scenario number / scenario set, G is the total number of units, Y is the number of time points in a day in scenario k, and R is the real number domain.

[0020] Further, the expression for solving the power supply planning model containing only a single scenario is:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] Among them, OBJ OD is the objective function, Y is the number of planning years, G is the total number of units, a y,g is the unit capacity cost of the unit, x y,g is the integer variable of the commissioning decision of the unit g in the year y, T is the total number of time points in a year, is the cost coefficient related to the continuous variable, p y,g,t is the continuous variable of the operation decision of the unit g at time t in the year y, is the cost coefficient related to the 0-1 variable, u y,g,t is the 0-1 variable of the operation decision of the unit g at time t in the year y, y is the planning year number, A y-1,g is the constraint coefficient related to the number of units commissioned in the year y-1, x y-1,g and B y,g are the constraint coefficients related to the number of units commissioned in the year y, x y,g is the integer variable of the commissioning decision of the unit g in the year y, d y,g is the upper limit constraint of the commissioning variable, and Cy,g,t is a proportional coefficient related to the number of units in operation at each time, D y,g is a proportional coefficient related to the continuous variable of the operation decision of the unit g at the time t in the year y, p y,g,t is the continuous variable of the operation decision of the unit g at the time t in the year y, E y,g is a proportional coefficient related to the 0-1 variable of the operation decision of the unit g at the time t in the year y, u y,g is the 0-1 variable of the operation decision of the unit g at the time t in the year y, e y,g,t is an upper limit in the operation decision constraint.

[0027] Specifically, in step S3, the input elements of the SOM neural network are read, i.e. the single-scenario potential decision results processed through step S1 and step S2, the input of the SOM neural network is a set containing K M-dimensional vectors , and each neuron in the neural network is defined as a weight vector , where L is the number of neurons and is randomly initialized The following steps are repeated within a given number of iterations:

[0028] An input element is selected The following formula is used to calculate the number of the best matching unit corresponding to each neuron in the neural network .

[0029] The weight vector of the best matching unit and its adjacent neurons is updated using the information of .

[0030] When the learning efficiency is less than a predefined threshold or the number of iterations reaches an upper limit, the learning process ends.

[0031] Further, a quantization error QE is introduced to characterize the average distance between each input element and its best matching unit, and the calculation formula is:

[0032]

[0033] wherein, is the weight vector of the best matching unit of the input element .

[0034] Specifically, step S4 is specifically:

[0035] S401, a dissimilarity matrix is introduced to record the distance between the input elements and each neuron, and the distance is used to represent the similarity between the input elements and each neuron;

[0036] S402, a silhouette coefficient is introduced to determine the optimal cluster number, and the number of the cluster with the highest silhouette coefficient SC is selected as the final cluster number;

[0037] S403, the k-means++ algorithm is used to initialize the cluster center, and the principle of increasing the initial centroid set diffusion degree is followed;

[0038] S404, the elements other than the cluster center are classified, the Euclidean distance of the elements to each cluster center is calculated, and the class to which the nearest cluster center belongs is classified;

[0039] S405, search for the element with the smallest distance sum with other elements in the class from other elements in each class except the cluster center, as the new cluster center;

[0040] S406, repeat steps S402 to S405, further cluster the preliminary clustering results based on SOM, obtain the neuron classification results after merging and reducing, and perform k-medoids clustering on the input elements in each class again to obtain the number of centroids in each class as the result of typical scene reduction.

[0041] Further, in step S402, the contour coefficient sc(i) is:

[0042]

[0043] The total contour coefficient SC of clustering is:

[0044]

[0045] Wherein, a(i) is the average distance of the i-th element to other elements in the same cluster, b(i) is the minimum value of the average distance of the i-th element to all elements in other clusters, and N is the total number of nodes in the power system.

[0046] In a second aspect, an embodiment of the present application provides a combined clustering scene reduction system based on potential decision results, comprising:

[0047] The representation module represents the uncertainty factors of each node in the power system in the form of a time sequence curve as a multi-dimensional curve input domain;

[0048] The transformation module constructs a power supply planning problem solving model considering operation checking under different time scales based on the multi-dimensional curve input domain obtained by the representation module, and transforms the multi-dimensional curve input domain into a potential decision result domain for clustering analysis;

[0049] The clustering module performs SOM neural network clustering on the potential decision results obtained by the transformation module to obtain a preliminary clustering result;

[0050] The output module uses a SOM and k-medoids combined clustering method to perform final classification on the preliminary clustering result obtained by the clustering module to obtain a typical scene used for final operation checking.

[0051] Compared with the prior art, the present application has at least the following beneficial effects:

[0052] The present application is a combination clustering scenario reduction method based on potential decision results, which is used for reducing the number of checking scenarios in power supply planning operation checking, improving the efficiency of scenario selection and reducing the scale and difficulty of operation checking calculation under the premise of ensuring the reliability of operation checking.

[0053] Further, reading the relevant node load data and renewable energy output data from the 365 scenarios throughout the year is the core of the scenario input domain construction and the basis for checking the power supply planning scheme. Ensuring the completeness and reliability of the input data is a necessary condition for improving the reliability of operation checking.

[0054] Further, modeling the power supply investment decision and operation checking under different scenario dimensions, from the most accurate full-scenario model to the typical day operation simulation to the single typical day simulation checking, makes the problem difficulty of the decision model solution gradually reduced, and improves the efficiency of obtaining a single potential decision result. At the same time, the full-scenario operation simulation and the typical scenario checking operation simulation can also play a role in checking the scenario reduction effect from another aspect.

[0055] Further, by solving the power supply planning investment decision of each scenario, the potential result domain Γ of the investment decision is obtained, and the conversion of the multi-dimensional input domain into a lower-dimensional input domain is realized, which provides convenience for further clustering analysis. At the same time, the potential result domain explicitly represents the construction cost problem concerned by the power supply investment decision, thereby providing a key feature in clustering analysis.

[0056] Further, by simplifying the full-scenario power supply planning model to a power supply planning model containing only a single scenario for solving, the operation constraints of each time in each scenario are disassembled and simplified to the operation constraints of each time in a single scenario, which can greatly reduce the complexity of the problem and improve the solving efficiency. Through the solution of the single-scenario power supply planning model, the potential decision result for clustering analysis is obtained.

[0057] Further, the dimensionality reduction of the input domain greatly improves the reliability of the clustering algorithm. Since the similarity measurement result in high-dimensional space is unreliable, mapping high-dimensional data to low-dimensional space is a necessary condition before clustering. Through principal component analysis, the multi-dimensional data in the input domain can be analyzed, and the vectors that have a greater impact on the overall input data are summarized. By retaining the feature vectors corresponding to the principal components, the dimension of the input vector can be significantly reduced without losing the original feature attributes of the data and affecting the results.

[0058] Further, by introducing the quantization error QE to depict the average distance between the input elements and the neurons in the SOM neural network, the best matching unit of each input element can be obtained, and thus the preliminary clustering of the SOM neural network is realized, and the categories to which the input elements belong are obtained.

[0059] Further, the SOM neural network method is used to cluster the potential decision results, the decision results are preliminarily clustered by the interaction mode and the competition mechanism of the self-organizing neural network, and the clustering effect is improved by the artificial intelligence method. Meanwhile, the result of the clustering method can have the self-organizing characteristic by the competition mechanism of the SOM neural network, the unsupervised clustering algorithm k-medoids is used to re-cluster the preliminary result, the distance between the input vector and the neuron vector is recorded by introducing the dissimilarity matrix, and further clustering is performed after the preliminary clustering result is fully captured. The applicability of the clustering method is ensured by using the method of finding the median which is more suitable for the potential decision result. Meanwhile, the k-means++ algorithm is used to initialize the clustering center, and the efficiency of the clustering algorithm is ensured.

[0060] Further, by the combination clustering of the SOM neural network and the k-medoids, the optimal k-modoids clustering number is determined by using the silhouette coefficient, and the preliminary result is further clustered, the number of the centroid in each category is solved, and the final scenario reduction result is obtained.

[0061] It can be understood that the beneficial effects of the above-mentioned second aspect can be referred to the related description in the above-mentioned first aspect, which will not be repeated here.

[0062] In summary, the present application is used for clustering to obtain the most representative scenario set in the power supply planning typical scenario operation checking, the complexity of the solving model is reduced to the greatest extent under the condition of ensuring the credibility of the checking result, and the calculation efficiency is improved.

[0063] The technical solutions of the present application will be further described in detail below with reference to the drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 The flowchart of the present application is shown in the figure;

[0065] Figure 2 The normalized potential decision result is shown in the figure;

[0066] Figure 3 The SOM preliminary clustering result is shown in the figure, wherein (a) is a unified distance matrix, and (b) is an input element distribution diagram;

[0067] Figure 4 The silhouette coefficient diagram of the dissimilarity matrix clustering is shown in the figure;

[0068] Figure 5 to reduce the potential decision result graph of the scenario;

[0069] Figure 6 to reduce the multi-dimensional input graph of the scenario;

[0070] Figure 7 to compare the cumulative installed results of the power supply. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0072] In the description of the present application, it should be understood that the terms "include" and "contain" indicate the existence of described features, whole, steps, operations, elements and / or components, but do not exclude the existence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0073] It should also be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, unless otherwise clear from the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0074] It should be further understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations, for example, A and / or B can represent three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents that the front and rear associated objects are in an "or" relationship.

[0075] It should be understood that although the terms first, second, third, etc. may be used in the embodiments of the present application to describe the preset ranges, etc., these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from each other. For example, the first preset range can also be referred to as the second preset range, and similarly, the second preset range can also be referred to as the first preset range without departing from the scope of the embodiments of the present application.

[0076] Depending on context, the word "if" as used herein can be interpreted to mean "when" or "while" or "in response to determining" or "in response to detecting." Similarly, the phrase "if it is determined" or "if [a stated condition or event] is detected" can be interpreted to mean "when it is determined" or "in response to determining" or "when [a stated condition or event] is detected" or "in response to detecting [a stated condition or event]."

[0077] Various structural diagrams according to the disclosed embodiments of the present application are shown in the drawings. These diagrams are not drawn to scale, in which certain details are exaggerated for clarity and others are omitted. The shapes and relative sizes of the various regions, layers, and the relative positions of these regions / layers shown in the drawings are merely examples and can deviate in actuality due to manufacturing tolerances or technical limitations, and regions / layers with different shapes, sizes, and relative positions can be additionally designed according to actual needs by those skilled in the art.

[0078] The present application provides a combination clustering scene reduction method based on potential decision results, which obtains a most representative scene set of decision results by a combination clustering method of SOM and k-medoids for the potential decision results. A dimension reduction is performed on a full-scene-oriented operation checking model to obtain a typical scene operation checking model with assigned weights, a dimension reduction is performed on the input domain of each typical scene by principal component analysis to reduce the calculation pressure caused by high-dimensional input domain on clustering; potential decision results of all single scenes are obtained by solving a simple model of a single scene, which are taken as input data of SOM clustering analysis, and a simple model is solved one by one instead of a complex model of a full scene, so that the calculation difficulty is greatly reduced; a dissimilarity matrix and a silhouette coefficient are introduced to determine the optimal k-medoids clustering number, and then the k-medoids method is used for clustering of the potential decision results, so that the proximity to the full-scene simulation result is improved under the combination method clustering, thereby reducing the result deviation caused by typical scene simulation.

[0079] Referring to Figure 1 The combination clustering scene reduction method based on potential decision results comprises the following steps:

[0080] S1, the uncertainty factors of each node in the system are taken as input data of the model in the form of a time sequence curve;

[0081] First, the required data is obtained from the relevant departments, including the annual load data of each node in the system and the annual output of the load-side callable resources; the annual output of the renewable energy units in the system.

[0082] The annual load power data of each node in each scene k is defined as:

[0083]

[0084] where T is the number of time points in a day within scenario k; N is the total number of nodes in the system under study.

[0085] Similarly, the wind power resource curve and the photovoltaic resource curve at each time point in each scenario k throughout the year are defined as:

[0086]

[0087]

[0088] where G WT is the number of wind farms; G S is the number of photovoltaic power stations.

[0089] With the increasing availability of resources on the load side, flexibility resources represented by demand response play an increasingly important role in the power system, and their uncertainty should also be considered in each scenario. The upper limit of the demand side flexibility resource response capacity is defined as:

[0090]

[0091] All the data information required in the input domain is obtained, and the four types of data vectors are spliced in the order of the load curve, the wind power resource output curve, the photovoltaic resource output curve, and the upper limit curve of the demand side flexibility resource response in scenario k according to the modeling sequence under the premise of ensuring the time sequence: the spliced vector is the multi-dimensional input curve vector in scenario k:

[0092]

[0093] The multi-dimensional curve input domain of load, new energy, and demand response resources suitable for power source planning and operation checking model is formed as:

[0094]

[0095] where K is the number of scenarios; G RE is the number of new energy power sources.

[0096] S2, based on the defined input domain, a power source planning problem solving model considering operation checking under different time scales is constructed, and the multi-dimensional curve input domain is converted into a potential decision result domain for clustering analysis;

[0097] First, define the power source planning solving model based on full-scenario operation checking (Full model, referred to as FD), and use the input resource curve and load curve in all time periods within the planning period for operation checking:

[0098]

[0099]

[0100]

[0101]

[0102]

[0103] where y is the planning year number; Y is the number of planning years; g / G is the unit number / total number of units; is the time number / total number of times in a year, where x y,g is the integer variable of the construction decision of unit g in year y; p y,g,t is the continuous variable of the operation decision of unit g in year y and time t; u y,g,t is the 0-1 variable of the operation decision of unit g in year y and time t.

[0104] Equation (7) is the objective function of the power source planning problem, aiming to minimize the total construction cost and operation cost, where the operation cost includes the operation cost related to the continuous variable p y,g,t , such as the variable operation cost of the traditional thermal power unit, such as fuel cost, carbon emission cost, and new energy unit penalty cost, and the cost related to the 0-1 operation variable u y,g,t , such as the start-stop cost of the thermal power unit, and the like. Equation (8) is the constraint related to the investment decision of the power source planning, including the construction variable constraint, the new energy installed capacity proportion constraint, the upper limit constraint of the investment budget, the installed capacity adequacy constraint, and the like. Equation (9) is the constraint related to the operation decision in each time in the planning period, including the upper and lower limit constraints of the output of each type of unit, the constraint of the demand response calling time, the power balance constraint of the system in different scenarios, the cross-section constraint, and the like.

[0105] The decision variables involved in the FD problem include the power source investment decision variable of each unit in each year, and the output of different types of units in each scenario in each time of the year, as well as the load size and the demand side response capacity, which has the characteristics of large number of variables, complex time sequence coupling constraints, and high difficulty in solving. In order to reduce the solving time, the unit combination, the thermal power unit simplification, and the like can be used for equivalent and simplification.

[0106] Therefore, in order to reduce the complexity of the overall model solving, the model based on scenario (Model based on scenario, abbreviated as SD) is used as the substitute of the FD problem.

[0107]

[0108]

[0109]

[0110]

[0111]

[0112] where t is the time index within the scenario; T is the set of time indices within the scenario T = 24; k / K is the scenario index / number of scenarios; ωkis the weight of scenario k, the number of scenarios and the weights can be obtained by scenario reduction method. k

[0113] It is worth mentioning that the method of typical scenario simulation cannot guarantee the continuity of the selected typical scenario in terms of unit commitment climbing constraints and energy timing balance constraints of energy storage, so this type of constraints needs to be decoupled between scenarios.

[0114] For the combination clustering method based on potential decision results used in this method, solving the model of multiple typical scenarios is not the optimal method, and then the model is simplified, and the solved SD problem is simplified to a power supply planning model (OD) containing only a single scenario for solving. The expression of the solving model is:

[0115]

[0116]

[0117]

[0118]

[0119]

[0120] Since only a single scenario is used for operation checking simulation, the weight of the typical day is set to 1. By solving the single-scenario model, the power production sequence under scenario k is obtained:

[0121]

[0122] For the planning problem, the cost of the investment decision is often concerned, so the power production sequence of each type of power is multiplied by its construction cost, i.e. the potential investment decision result is obtained:

[0123]

[0124] The potential decision result domain Γ composed of potential decision results is constructed: ​

[0125]

[0126] At this point, the multi-dimensional input domain X is converted into a potential decision result domain Γ that is available for clustering analysis, which has been significantly reduced in input vector dimension and contains information that more obviously reflects the decision effect.

[0127] S3, SOM neural network clustering of the obtained potential decision results;

[0128] First, read the input elements of the SOM neural network, i.e., the single-scene potential decision results processed in steps S1 and S2, and the input of the SOM neural network is a set containing K M-dimensional vectors . At the same time, define the weight vector of each neuron in the neural network as where L is the number of neurons. Randomly initialize Repeat the following steps within a given number of iterations:

[0129] Select an input element Calculate the number of the corresponding best matching unit in the neural network using the following formula:

[0130]

[0131] Update the weight vector of the best matching unit and its adjacent neurons using the information of :

[0132]

[0133] where α is the learning efficiency, which decreases with the increase of the number of iterations; h(s, l, r) is the neighborhood function, usually a Gaussian Mexican hat function, and r is the neighborhood function radius.

[0134] When the learning efficiency is less than a predefined threshold or the upper limit of the number of iterations is reached, the learning process ends. In order to select the best-performing SOM neural network, introduce the quantization error QE to characterize the average distance between each input element and its best matching unit, and the calculation formula is:

[0135]

[0136] where is the weight vector of the best matching unit of the input element .

[0137] S4, using the k-medoids clustering algorithm to perform final classification on the preliminary clustering results.

[0138] S401, in order to capture the preliminary clustering result information, the dissimilarity matrix is introduced to record the distance between input elements and each neuron, and the distance represents the similarity between input elements and each neuron. The smaller the distance between neurons and input elements, the better the representative of the data.

[0139] The specific expression of the similarity matrix is:

[0140]

[0141] S402, select the number of clusters

[0142] In order to determine the number of clusters of k-medoids, the silhouette coefficient is introduced to determine the optimal number of clusters. The silhouette coefficient of a single element can measure the size relationship between the similarity of the element to the elements in its own class and the similarity of the element to the elements in other classes. The specific calculation formula is:

[0143]

[0144] Wherein, a(i) is the average distance of the ith element to other elements in the same cluster, representing the intra-cluster dissimilarity; b(i) is the minimum value of the average distance of the ith element to all elements in other clusters, representing the inter-cluster dissimilarity.

[0145] The total silhouette coefficient SC of clustering is:

[0146]

[0147] The value range of the silhouette coefficient is [-1, 1], the closer to 1, the better the clustering performance, and the closer to -1, the worse the clustering performance. Therefore, the number of clusters with the highest silhouette coefficient is selected as the final number of clusters.

[0148] S403, initialize the cluster center

[0149] After determining the number of clusters, the k-means++ algorithm is used to initialize the cluster center, which follows the principle of increasing the initial centroid set diffusion.

[0150] S404, classify the elements other than the cluster center, calculate the Euclidean distance of the elements to each cluster center, and classify it into the class to which the nearest cluster center belongs.

[0151] S405, search for the element in each class other than the cluster center, which has the smallest distance sum with other elements in the class, and take it as the new cluster center.

[0152] S406, repeatedly performing steps S402 to S405, further clustering the preliminary clustering result based on SOM, and obtaining the neuron classification result after reduction of combination.

[0153] Up to now, the whole process of combining clustering potential decision results to achieve the purpose of scenario reduction is completed.

[0154] In another embodiment of the present application, a system for combining clustering potential decision results is provided, which can be used to implement the above-mentioned method for combining clustering potential decision results, and specifically, the system for combining clustering potential decision results comprises a representation module, a transformation module, a clustering module and an output module.

[0155] The representation module represents the uncertainty factors of each node in the power system as a multi-dimensional curve input domain in the form of a time sequence curve.

[0156] The transformation module constructs a power supply planning problem solving model considering operation checking under different time scale scenarios based on the multi-dimensional curve input domain obtained by the representation module, and transforms the multi-dimensional curve input domain into a potential decision result domain for clustering analysis.

[0157] The clustering module performs SOM neural network clustering on the potential decision result obtained by the transformation module, and obtains a preliminary clustering result.

[0158] The output module adopts a SOM and k-medoids combined clustering method to finally classify the preliminary clustering result obtained by the clustering module, and obtains a typical scenario used for operation checking.

[0159] In still another embodiment of the present application, a terminal device is provided, which comprises a processor and a memory, the memory being configured to store a computer program, the computer program comprising program instructions, and the processor being configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are particularly suitable for loading and executing one or more instructions to implement a corresponding method flow or a corresponding function; the processor in the embodiments of the present application can be used for the operation of the combination clustering scenario reduction method based on potential decision results, including:

[0160] Uncertain factors of each node in the power system are represented in the form of time sequence curves as multi-dimensional curve input domains; a power source planning problem solving model is constructed under different time scale scenarios considering operation checking based on the multi-dimensional curve input domains, the multi-dimensional curve input domains are converted into a potential decision result domain for clustering analysis, SOM neural network clustering is performed on the potential decision results to obtain preliminary clustering results, and a SOM and k-medoids combined clustering method is used to finally classify the preliminary clustering results to obtain typical scenarios used for operation checking.

[0161] In still another embodiment of the present application, the present application further provides a storage medium, specifically a computer readable storage medium (Memory), which is a memory device in the terminal device and is used for storing programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the terminal device, and of course can also include an expansion storage medium supported by the terminal device. The computer readable storage medium provides a storage space, and the storage space stores an operating system of the terminal. Furthermore, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory.

[0162] The one or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the method for reducing the scene combination based on the potential decision result in the above embodiments; the one or more instructions stored in the computer readable storage medium are loaded and executed by the processor to implement the following steps:

[0163] Uncertainty factors of each node in the power system are represented in the form of a time sequence curve as a multi-dimensional curve input domain; a power supply planning problem solving model considering operation checking under different time scales is constructed based on the multi-dimensional curve input domain, the multi-dimensional curve input domain is converted into a potential decision result domain for clustering analysis, the SOM neural network clustering is performed on the potential decision result to obtain a preliminary clustering result, and the SOM and k-medoids combined clustering method is used to finally classify the preliminary clustering result to obtain a typical scene used for operation checking.

[0164] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by a person of ordinary skill in the art without creative work on the basis of the embodiments in the present application belong to the scope of protection of the present application.

[0165] Example analysis

[0166] To verify the effectiveness of the method proposed in the present application, the XJTU-ROTS system is selected for calculation and analysis. To fully reflect the intermittency and volatility of new energy, the new energy curve of a certain actual system in the northwest is used to replace the new energy curve of the test system, and the historical data set contains 365 days. The power sources and wind and light resources in the system are shown in Table 1. The planning is carried out by using a single node model, but the differentiated resource characteristics of the new energy in each region are retained.

[0167] Table 1 System existing power sources and wind and light resource conditions

[0168]

[0169] To control the solving time, the same type of thermal power units are combined to reduce the number of 0-1 variables. The parameters of the selected power sources and energy storage are shown in Tables 2 and 3.

[0170] Table 2 System selected power source parameters

[0171]

[0172] Table 3 System candidate energy storage parameters

[0173]

[0174] According to the method proposed in this invention, the potential decision results for solving the power planning problem in a single-scenario model under 365 scenarios are first obtained, and the results are as follows: Figure 2 As shown.

[0175] Subsequently, using the potential decision results of each scenario as input vectors, a SOM neural network was employed for clustering. After parameter selection through Monte Carlo simulation, the clustering results of the SOM neural network with the best performance and the distribution of input elements were obtained as follows: Figure 3 As shown.

[0176] Based on the preliminary SOM clustering results, the dissimilarity matrix between neurons and input elements was used as the input data for k-medoids clustering. Silhouette coefficients were calculated for different numbers of clusters, and the results are as follows: Figure 4 As shown.

[0177] The classification results of the input elements are obtained based on the best matching relationship in the SOM. For each class of input elements, the k-medoids method is used again to take its centroid as the cluster center. The final 16 potential decision results are as follows: Figure 5 As shown, the multidimensional input domain curve is as follows: Figure 6 As shown.

[0178] To verify the effectiveness of the proposed method, comparative examples are designed:

[0179] Benchmark: Based on the entire input domain, use the FD model to solve the planning problem;

[0180] Case_1: After using the k-medoids method to cluster the input domain to reduce the scenario, the SD model is used for planning and solving;

[0181] Case_2: After using the k-medoids method to cluster the potential decision outcome domain to reduce the scenario, the SD model is used for planning and solving;

[0182] Case_3: After using the SOM+k-medoids combined clustering method to cluster the potential decision outcome domain to reduce the scenario, the SD model is used for planning and solving.

[0183] Introducing benchmark cases as a reference for other cases, the cumulative installed power supply capacity of the four cases under the optimal scenario with reduced number of cases is as follows: Figure 7 As shown.

[0184] Further, the normalized root mean square error (NRMSE) is introduced to measure the deviation of the installed capacity of the geothermal power source in Case_1 to Case_3 from the reference example result, and the calculation formula of NRMSE is:

[0185]

[0186] Wherein, G is the number of power source types; S g is the calculated value of the installed capacity of source g; is the reference value of the installed capacity of power source g.

[0187] The number of reduced scenarios and NRMSE of each example scenario calculated are shown in Table 4:

[0188] Table 4 NRMSE of comparative examples

[0189]

[0190] From the results, from Case_1 to Case_3, the NRMSE value of the power source construction result decreases, the NRMSE of Case_2 compared with Case_1 decreases by 0.0473, which shows that the scenario reduction based on the potential decision result is closer to the actual result than the scenario reduction based on the input domain data. The NRMSE value of Case_3 compared with Case_2 decreases by 0.0294, which shows that the SOM and k-medoids combined clustering scenario reduction method has better performance effect than the k-medoids clustering scenario reduction method.

[0191] In summary, the combination clustering scene reduction method and system based on potential decision results can improve the calculation efficiency of power supply planning operation checking. Compared with traditional panoramic simulation and typical day simulation, the SOM self-organizing neural network and k-medoids clustering algorithm are used to cluster the potential decision results of the whole year multiple scenes, avoiding the high dimension and complex calculation of full scene simulation and the subjectivity and randomness of manual selection of typical days. Through the construction of power supply planning models of different scene scales, a model system for fast solution and accurate checking is summarized. The method can quickly summarize the characteristics of potential decision results under different scenes, and use the advantages of artificial intelligence algorithm to efficiently cluster the characteristics. Finally, by introducing a proper evaluation mechanism, the deviation of the clustering results and the standard simulation results is examined, so that a complete system is formed to reduce the typical scenes of the whole year operation checking. In the current situation of uncertain new energy output prediction and continuous gathering of various flexible resources on the load side, the combination clustering algorithm based on potential decision results can reduce the calculation problem caused by high dimension and multiple scene operation simulation, improve the efficiency of operation simulation in power supply planning, and ensure that the difference between the operation checking results and the panoramic operation checking results is within an acceptable range, which is very suitable for improving the calculation efficiency of the whole model solution of power supply planning.

[0192] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of software functional unit. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the application. The specific working process of the units and modules in the system can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0193] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0194] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0195] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal and method can be implemented in other ways. For example, the apparatus / terminal embodiments described above are merely schematic. The division of the modules or units is merely a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between the units can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.

[0196] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place, or distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0197] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0198] The integrated module / unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the computer-readable medium can include or exclude contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0199] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The flow or multiple flows and / or blocks Figure 1 The apparatus that implements the functions specified in one or more blocks or multiple blocks.

[0200] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction apparatus that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The flow or multiple flows and / or blocks Figure 1 The apparatus that implements the functions specified in one or more blocks or multiple blocks.

[0201] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices to generate a computer implemented process, so that the instructions executed on the computer or other programmable data processing devices provide a process for implementing the functions specified in the flowchart Figure 1 one flow or a plurality of flows and / or the functions specified in the block Figure 1 one block or a plurality of blocks.

[0202] The above is only to illustrate the technical idea of the present application, and cannot limit the protection scope of the present application. Any modification made according to the technical idea of the present application on the basis of the technical scheme falls within the protection scope of the claims of the present application.

Claims

1. A method of reducing scenarios based on a combination of latent decision outcomes, characterized in that, The method comprises the following steps: S1, representing the uncertainty factors of each node in the power system as a multi-dimensional curve input domain in the form of a time sequence curve; S2, based on the multi-dimensional curve input domain obtained in step S1, a power supply planning problem solving model considering operation checking under different time scales is constructed, the multi-dimensional curve input domain is converted into a potential decision result domain for clustering analysis, the scenario-based model is used as a substitute for the power supply planning solving model based on full-scenario operation checking, the solved scenario-based model is simplified into a power supply planning model containing only a single scenario for solving, the weight of the typical day is set to 1, the single-scenario model is solved, and the power supply production sequence under the scenario is obtained, the production sequence of each type of power supply is multiplied by its construction cost to obtain a potential investment decision result, and a potential decision result domain composed of potential decision results is constructed ; Potential decision result domain specifics are: wherein, is a potential investment decision outcome, is a scenario number / scenario set, is the total number of units, is a scenario is the number of time instances within a day, is the real number field; The expression for solving the power source planning model containing only a single scenario is: in, Let be the objective function. To plan the annual quantity, The total number of units. The unit capacity cost of the unit. For the year unit Integer variables for investment and construction decisions, The total number of moments within a year. Cost coefficients related to continuous variables. For the year unit time Continuous variables in operational decisions Cost coefficients related to variables between 0 and 1. For the year unit time The 0-1 variables for running the decision, For the planning year number, To and Constraint coefficients related to the number of generating units put into operation annually and To and Constraint coefficients related to the number of generating units put into operation annually For the year unit Integer variables for investment and construction decisions, To constrain the upper limit of variables for production, This is a proportionality coefficient related to the number of generating units put into operation at each time point. In order to be consistent with the annual unit time The proportional coefficients related to continuous variables in operational decisions. For the year unit time Continuous variables in operational decisions For the year unit time The proportional coefficients related to the 0-1 variables in the decision-making process. For the year unit time The 0-1 variables for running the decision, This represents the upper limit in the operational decision constraints; S3, performing SOM neural network clustering on the potential decision results obtained in step S2 to obtain a preliminary clustering result; S4, performing final classification on the preliminary clustering result obtained in step S3 by using a SOM and k-medoids combined clustering method to obtain a typical scenario used for final operation checking.

2. The method of claim 1, wherein the method is based on a combination of clustering scenarios based on potential decision outcomes. In step S1, a multidimensional curve input domain is represented as: （6） wherein, is a scenario number / scenario set, is a number of new energy power sources, is a scenario is a multi-dimensional input curve vector in the scenario is a total number of nodes in the power system, is a scenario is a number of time instants in a day, is a real number field.

3. The method of claim 1, wherein the method is based on a combination of clustering scenarios based on potential decision outcomes. In step S3, the input elements of the SOM neural network, i.e. the single-scene potential decision results processed in step S1 and step S2, are read, the input of the SOM neural network is a set containing one dimensional vector , and each neuron in the neural network is defined as its weight vector , , where n is the number of neurons, and is randomly initialized , the following steps are repeated within a given number of iterations: Selecting an input element The number of the corresponding best matching cell is calculated in the neural network using the following equation ​ Using the information to update the weight vectors of the best matching unit and its neighboring neurons; When the learning efficiency is less than a predefined threshold or reaches an upper limit of the number of iterations, the learning process ends.

4. The method of claim 3, wherein the method is based on a combination of clustering scenarios based on potential decision outcomes. Introducing quantization error QE The characterization is made by means of an index that represents the average distance between each input element and its best matching cell, calculated as: wherein, is the best matching unit weight vector for the input element .

5. The method of claim 1, wherein, Step S4 specifically comprises: S401, introducing a dissimilarity matrix to record the distance between the input elements and each neuron, and using the distance to represent the similarity between the input elements and each neuron; S402、introducing the contour coefficient to determine the optimal clustering number, selecting the contour coefficient SC The number of the highest cluster is the final clustering number; S403, initializing the clustering centers by using a k-means++ algorithm, and following the principle of increasing the diffusion degree of the initial centroid set; S404, classifying the elements other than the clustering centers, calculating the Euclidean distance of each element to each clustering center, and classifying the elements into the class to which the nearest clustering center belongs; S405, searching for the element with the minimum distance sum to other elements in each class other than the clustering center, and taking the element as a new clustering center; S406, repeating steps S402 to S405, further classifying the preliminary clustering result based on the SOM, obtaining the neuron classification result after reduction, and performing k-medoids clustering on the input elements in each class again to obtain the number of the centroid in each class as the result of typical scenario reduction.

6. The method of claim 5, wherein the clustering is based on a combination of the latent decision results. In step S402, the profile coefficient is: Total profile coefficient of clustering SC is: wherein, is the average distance of the i-th element to other elements in the same cluster, i is the average distance of the i-th element to other elements in the same cluster, is the minimum value of the average distance of the i-th element to all elements in other clusters, i is the minimum value of the average distance of the i-th element to all elements in other clusters, is the total number of nodes in the power system.

7. A latent decision outcome based combined clustering scenario reduction system, characterized in that, The method comprises the following steps: A representation module represents the uncertainty factors of each node in the power system as a multi-dimensional curve input domain in the form of a time sequence curve; The conversion module is configured to construct a power supply planning problem solving model considering operation checking under different time scales based on the multi-dimensional curve input domain obtained by the representation module, convert the multi-dimensional curve input domain into a potential decision result domain for cluster analysis, use the scenario-based model as a substitute for a power supply planning solving model based on full-scenario operation checking, simplify the solved scenario-based model into a power supply planning model containing only a single scenario for solving, set the weight of a typical day to 1, obtain a power supply commissioning sequence under the scenario by solving the single-scenario model, multiply the power supply commissioning sequence of each type of power supply by the construction cost thereof to obtain a potential investment decision result, and construct a potential decision result domain composed of potential decision results . ​ Potential decision result domain specifics are: wherein, is a potential investment decision outcome, is a scenario number / scenario set, is the total number of units, is a scenario is the number of time points within a day, is the real number field; The expression for solving the power source planning model containing only a single scenario is: in, Let be the objective function. To plan the annual quantity, The total number of units. The unit capacity cost of the unit For the year unit Integer variables for investment and construction decisions, The total number of moments within a year. Cost coefficients related to continuous variables. For the year unit time Continuous variables in operational decisions Cost coefficients related to variables between 0 and 1. For the year unit time The 0-1 variables for running the decision, For the planning year number, To and Constraint coefficients related to the number of generating units put into operation annually and To and Constraint coefficients related to the number of generating units put into operation annually For the year unit Integer variables for investment and construction decisions, For production variables upper limit constraints, This is a proportionality coefficient related to the number of generating units put into operation at each time point. In order to be consistent with the annual unit time The proportional coefficients related to continuous variables in operational decisions. For the year unit time Continuous variables in operational decisions For the year unit time The proportional coefficients related to the 0-1 variables in the decision-making process. For the year unit time The 0-1 variables for running the decision, This represents the upper limit in the operational decision constraints; A clustering module performs SOM neural network clustering on the potential decision results obtained by the conversion module to obtain a preliminary clustering result; An output module performs final classification on the preliminary clustering result obtained by the clustering module by using a SOM and k-medoids combined clustering method to obtain a typical scenario used for final operation checking.

Images