Data-driven method and device for supporting adjustable resource scheduling domain aggregation evaluation

Abstract

The present invention relates to the field of electric power control technology, and in particular to a data-driven method and device for supporting aggregation evaluation of adjustable resource scheduling domains, wherein the method comprises: generating basic flexible resource parameters and the number of aggregators in a target power grid in an offline stage to calculate each aggregation feasible domain and the feasible domain of each basic flexible resource; training a feedforward neural network using each aggregation feasible domain and the feasible domain of each basic flexible resource to obtain a trusted evaluation model; dynamically accessing new flexible resource parameters in an online stage, and dynamically dividing the new flexible resources to obtain the corresponding optimal aggregation domain; inputting the new flexible resource parameters and the optimal aggregation domain into a feedforward neural network for trusted evaluation to obtain a trusted regulation domain of the flexible resources. Thus, the problem that the existing aggregation technology is difficult to quickly calculate the feasible domain of flexible resources after dynamic evaluation and division in practical applications is solved.

Description

Technical Field

[0001] The present invention relates to the field of power control technology, and in particular to a data-driven method and device for supporting adjustable resource scheduling domain aggregation evaluation. Background Art

[0002] As the scale of new energy grid connection gradually increases, the addition of a large number of small-capacity flexible resources poses a huge challenge to the regulation of the power system. In order to ensure the reasonable dispatch and safe and reliable operation of the power grid, flexible resources are usually aggregated to achieve efficient management and control of a large number of flexible resources. However, due to the mutual coupling of the power input / output, adjustment range and energy state of flexible resources, their feasible domains also affect each other. In addition, due to the large number of flexible resources, although the existing aggregation technology can accurately calculate the maximum aggregation feasible domain of flexible resources, its calculation scale is large and the calculation time is long. It is difficult to quickly calculate the feasible domain of flexible resources after dynamic evaluation and division in practical applications. Summary of the invention

[0003] The present invention provides a data-driven method and device for supporting aggregation evaluation of adjustable resource scheduling domains, so as to solve the problem that the existing aggregation technology is difficult to quickly calculate the feasible domain of flexible resources after dynamic evaluation and division in practical applications.

[0004] The first aspect of the present invention provides a data-driven method for supporting aggregation evaluation of adjustable resource scheduling domains, comprising the following steps: generating basic flexible resource parameters and the number of aggregators in a target power grid in an offline stage; calculating each aggregation feasible domain and the feasible domain of each basic flexible resource according to the basic flexible resource parameters and the number of aggregators; training a feedforward neural network using each aggregation feasible domain and the feasible domain of each basic flexible resource to obtain a trusted evaluation model; dynamically accessing new flexible resource parameters in an online stage, and dynamically classifying the new flexible resources to obtain their corresponding optimal aggregation domains; inputting the new flexible resource parameters and the optimal aggregation domain into the trusted evaluation model for trusted evaluation to obtain a trusted regulation domain of flexible resources.

[0005] Optionally, the calculating of each aggregation feasible domain and the feasible domain of each basic flexible resource according to the basic flexible resource parameters and the number of aggregators includes: establishing models of the aggregation feasible domain and the flexible resource feasible domain, wherein the models of the aggregation feasible domain and the flexible resource feasible domain include an operation model of flexible resources and an aggregation model of flexible resources; inputting the basic flexible resource parameters and the number of aggregators into the models of the aggregation feasible domain and the flexible resource feasible domain to calculate each aggregation feasible domain and the feasible domain of each basic flexible resource.

[0006] Optionally, the loss function expression of the feedforward neural network is:

[0007]

[0008]

[0009] Where m is the number of new flexible resource parameters, Predict the output value for the neural network, is the true value of the predicted parameter of the flexible resource, s is the control switch, r is the adjustable hyperparameter, and φ is the penalty function.

[0010] Optionally, dynamically classifying the new flexible resources to obtain their corresponding optimal aggregation domains includes: calculating proximity based on each aggregation feasible domain and the feasible domain of each flexible resource; and dynamically classifying the new flexible resource parameters using a K-nearest neighbor algorithm and the proximity to obtain the optimal aggregation domain.

[0011] The second aspect of the present invention provides a data-driven device that supports aggregation evaluation of adjustable resource scheduling domains, including: an offline generation module, which is used to generate basic flexible resource parameters and the number of aggregators in a target power grid in an offline stage; an offline calculation module, which is used to calculate each aggregation feasible domain and the feasible domain of each basic flexible resource according to the basic flexible resource parameters and the number of aggregators; an offline training module, which is used to train a feedforward neural network using each aggregation feasible domain and the feasible domain of each basic flexible resource to obtain a trusted evaluation model; an online classification module, which is used to dynamically access new flexible resource parameters in an online stage, and dynamically classify the new flexible resources to obtain their corresponding optimal aggregation domains; an online trusted evaluation module, which is used to input the new flexible resource parameters and the optimal aggregation domain into the trusted evaluation model for trusted evaluation to obtain a trusted regulation domain of flexible resources.

[0012] Optionally, the offline calculation module includes: an offline construction unit, used to establish models of the aggregation feasible domain and the flexible resource feasible domain, wherein the models of the aggregation feasible domain and the flexible resource feasible domain include an operation model of flexible resources and an aggregation model of flexible resources; an offline calculation unit, used to input the basic flexible resource parameters and the number of aggregators into the models of the aggregation feasible domain and the flexible resource feasible domain to calculate the feasible domain of each aggregation feasible domain and each basic flexible resource.

[0013] Optionally, the loss function expression of the feedforward neural network is:

[0014]

[0015]

[0016] Where m is the number of new flexible resource parameters, Predict the output value for the neural network, is the true value of the predicted parameter of the flexible resource, s is the control switch, r is the adjustable hyperparameter, and φ is the penalty function.

[0017] Optionally, the online classification module includes: an online calculation unit, used to calculate the proximity according to the feasible domain of each aggregation and the feasible domain of each flexible resource; an online dynamic classification unit, used to dynamically classify the new flexible resource parameters using the K nearest neighbor algorithm and the proximity to obtain the optimal aggregation domain.

[0018] An embodiment of the third aspect of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a data-driven method for supporting adjustable resource scheduling domain aggregation evaluation as described in the above embodiments.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the data-driven method for supporting adjustable resource scheduling domain aggregation evaluation as described above.

[0020] The data-driven method and device for supporting aggregation evaluation of adjustable resource scheduling domains proposed in the embodiments of the present invention obtain various parameters of flexible resources and construct a rapid evaluation framework combining offline and online according to the two-stage concept; based on the framework, in the offline stage, a multi-aggregator operation model is constructed according to a large number of flexible resources, and the neural network is trained to complete time-consuming work; in the online stage, the dynamically accessed flexible resources are quickly classified into aggregation domains and their feasible domains are evaluated, forming an optimal strategy for aggregation and dimensionality reduction of a large number of flexible resource adjustable spaces, so that the flexible resources can be classified and the feasible domain can be quickly evaluated, which can provide important support for the real-time scheduling and rapid access of flexible resources to the power grid.

[0021] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0023] Figure 1 A flow chart of a data-driven method for supporting adjustable resource scheduling domain aggregation evaluation provided by an embodiment of the present invention;

[0024] Figure 2A schematic diagram of the execution of the data-driven flexible resource trusted scheduling domain rapid evaluation provided by an embodiment of the present invention;

[0025] Figure 3 A schematic diagram of the KNN algorithm provided by an embodiment of the present invention;

[0026] Figure 4 A schematic diagram of a feedforward neural network structure provided by an embodiment of the present invention;

[0027] Figure 5 A schematic diagram of a brief working process of a neural network provided by an embodiment of the present invention;

[0028] Figure 6 A block diagram of a data-driven device for supporting adjustable resource scheduling domain aggregation evaluation provided by an embodiment of the present invention;

[0029] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0030] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0031] The following describes a data-driven method and apparatus for supporting adjustable resource scheduling domain aggregation evaluation according to an embodiment of the present invention with reference to the accompanying drawings.

[0032] Specifically, Figure 1 A flow chart of a data-driven method for supporting adjustable resource scheduling domain aggregation evaluation provided by an embodiment of the present invention.

[0033] like Figure 1 As shown, the data-driven method for supporting adjustable resource scheduling domain aggregation evaluation includes the following steps:

[0034] In step S101, basic flexible resource parameters and the number of aggregators in the target power grid are generated in an offline phase, and each aggregation feasible region and the feasible region of each basic flexible resource are calculated according to the basic flexible resource parameters and the number of aggregators.

[0035] In the actual implementation process, Figure 2 As shown, the number of parameters required for generating basic flexible resource parameters and aggregation in the offline stage, wherein the basic flexible resource parameters include the maximum / minimum power input / output of the flexible resource, the adjustment range, the maximum / minimum boundary of the residual energy, the energy state, the dissipation rate, the conversion coefficient between power and energy, etc.

[0036] It should be noted that, considering different types of flexible resources, a large amount of flexible resource offline training data is generated respectively. Therefore, according to the probability distribution of each type of resources, the Monte Carlo sampling method is used to randomly generate flexibility resource parameters.

[0037] In step S102, each aggregation feasible region and each basic flexible resource feasible region are calculated according to the basic flexible resource parameters and the number of aggregators.

[0038] In some embodiments, calculating each aggregation feasible domain and each basic flexible resource feasible domain according to the basic flexible resource parameters and the number of aggregators includes:

[0039] Establishing models of the aggregation feasible domain and the flexible resource feasible domain, wherein the models of the aggregation feasible domain and the flexible resource feasible domain include an operation model of the flexible resources and an aggregation model of the flexible resources;

[0040] The basic flexible resource parameters and the number of aggregators are input into the models of the aggregation feasible domain and the flexible resource feasible domain to calculate the feasible domain of each aggregation feasible domain and each basic flexible resource.

[0041] It should be noted that the models of the aggregation feasible domain and the flexible resource feasible domain include the operation model of the flexible resources and the aggregation model of the flexible resources, wherein the general operation model of the flexible resources is as follows:

[0042] p i,min (t)≤p i (t)≤p i,max (t) (1)

[0043] r i,min (t)≤r i (t)≤r i,max (t) (2)

[0044] e i,min (t)≤e i (t)≤e i,max (t) (3)

[0045] e i (t) = λ i e i (t-1)+ΔTη i p i (t) (4)

[0046]

[0047] Wherein, formulas (1)-(3) represent the constraint ranges of flexible resource active power, regulation service and residual energy respectively, formula (4) represents the relationship between residual energy and input / output, and formula (5) represents the constraint of flexible resource regulation range. i,min (t) is the minimum active power of flexible resource i at time t, p i,max (t) is the maximum active power of flexible resource i at time t, e i (t) is the residual energy of flexible resource i at time t, r i (t) is the adjustment capacity of flexible resource i at time t; e i,min (t) / e i,max (t) is the lower and upper limits of residual energy, p i,min (t) / p i,max (t) is the lower and upper limits of the input and output power of flexible resource i at time t, r i,max (t) is the upper limit of the power regulation capacity of flexible resource i, λ i and e i (t-1) are the dissipation rate of flexible resource i and the initial energy of flexible resource, η i It is the conversion coefficient between active power and energy of flexible resources. is the maximum power adjustment range of the equipment, ΔT is the planned time interval, and γ is the inverse of the product of the energy reserve coefficient and the planned time interval.

[0048] Different flexible resources have different parameters, so their feasible domains are also different. Taking active power input / output and regulation range as independent variables, formulas (1)-(5) can be simplified to the following form:

[0049] Ω i ={X i ∈R 2T |J i X i ≤K i} (6)

[0050] Among them, Ω i is the exact feasible range of flexible resource i, vector X i =[P i ,R i ] T The independent variable P i =[p i (t)] and R i =[r i (t)] composition, J i and K i is the matrix composed of input and output in formulas (1)-(5), and T is the total number of time points.

[0051] J i and K i The structure and coefficients of the matrix are as follows:

[0052]

[0053]

[0054]

[0055] Among them, I e is the T-dimensional unit matrix, A i , B i is a T-dimensional square matrix; C i , P i,min , P i,max , R i,min , R i,max 、E i,min and E i,max is a T-dimensional column vector:

[0056]

[0057] B i =diag[ΔTη i ] T×T ,

[0058]

[0059] P i,max =[p i,max (t)],P i,min =[p i,min (t)],

[0060] R i,max =[r i,max (t)],R i,min =[r i,min (t)],

[0061] S i,max =[s i,max (t)],S i,min =[s i,min (t)],

[0062] E i,max =[e i,max (t)],E i,min =[e i,min (t)],

[0063] By scaling and translating a basic isomorphic polyhedron, the embodiment of the present invention can approximate the actual feasible range of each flexible resource:

[0064]

[0065]

[0066]

[0067] in, is the approximate feasible region of flexible resource i, is the scaling parameter of flexible resource i, R 2T is the translation parameter of flexible resource i, is the isomorphic polyhedron composed of the parameters of the k-th aggregator, are the J of all flexible resources in aggregator k i , K i The average value of the matrix, therefore, they have a similar structure, is the translation parameter and The inner product of Have a similar structure.

[0068] When the feasible regions of flexible resources are similar, the Minkowski sum can be simplified as follows:

[0069]

[0070] Taking into account the influence of uncertainty factors, the calculation model of the maximum inner approximation feasible domain of flexible resources with similar shapes is established as follows:

[0071] In ensuring φ i , Take the maximum condition, so that

[0072]

[0073]

[0074] Among them, ε∈(0,1) represents the acceptable probability of crossing the limit.

[0075] By scaling and translating the basic isomorphic polyhedron The feasible domain of flexible resources can be maximized. Formula (11) consists of multiple constraints, which take into account the uncertainty of parameters and ensure the feasibility of instructions on this basis.

[0076] The above problem is difficult to solve using conventional methods. In order to make this problem solvable, an auxiliary variable is introduced. The maximum inner approximation feasible domain calculation model in formulas (11)-(12) can be transformed into a typical linear programming problem:

[0077] In ensuring α i , Take the minimum condition, so that

[0078] α i ,U i ≥0 (13)

[0079]

[0080]

[0081] Among them, L dim =10T is the dimension of matrix K, U i (m,:) and J i (m,:) is the matrix U i and J i The mth row of . is the matrix K i The mth uncertainty element in is the predicted value of the parameter, is the variance caused by parameter prediction deviation. For energy storage devices with charge and discharge losses, β i (s)| s∈{1,K,2T} Set to 0 to ensure mutual exclusion of charging and discharging behaviors.

[0082] By applying formulas (13)-(15), the scaling parameters and translation parameters of a single flexible resource can be derived. For an aggregator that includes multiple flexible resources, the scaling parameters and translation parameters of its feasible domain can be calculated by formula (10) as follows:

[0083]

[0084] Aggregator’s Aggregation Feasible Domain as follows:

[0085]

[0086]

[0087] in, is the inner proximity aggregation feasible domain of the aggregator k.

[0088] Through the above flexible resource operation model and flexible resource aggregation model, the precise control domain of the flexible resources in the offline training set is calculated, which mainly includes the scaling parameter φ and translation parameter of each flexible resource Provide data support for subsequent neural network prediction of new flexible resource regulation domains.

[0089] In step S103, the feedforward neural network is trained using each aggregate feasible domain and the feasible domain of each basic flexible resource to obtain a credible evaluation model.

[0090] In the actual implementation process, the scaling parameter and the translation parameter have a simple linear relationship with the control domain of the flexible resource. In order to simplify the evaluation model, the result of the neural network evaluation output is set as the scaling parameter φ and the translation parameter of the flexible resource. Specifically, the basic parameters p of the basic flexible resource parameters in the training set and the test set are i,min , p i,max , r i,max The scaling parameter φ and translation parameter of the flexible resource As input, with the minimum loss function as the goal, the parameters v, w, b and s in the neural network are calculated iteratively. When these parameters are iterated, the neural network is trained and the neural network is output. When a new flexible resource is connected, the basic parameter p of the new flexible resource is input. i,min , p i,max , r i,max , the fitting results φ and

[0091] After the data in the offline training set is divided into a training set and a test set in a certain ratio, the data in the training set is input into Figure 4 In the feedforward neural network FNN. FNN is mainly composed of three parts: input layer, hidden layer and output layer. Each layer is fully connected. The input-output relationship is as follows:

[0092] h[1]=f(w[1,1]x[1]+w[2,1]x[2]+w[3,1]x[3]+b[1])

[0093] h[2]=f(w[1,2]x[1]+w[2,2]x[2]+w[3,2]x[3]+b[2])

[0094] Among them, w[1,1], w[1,2], ..., w[3,1], w[3,2] are the weights between layers. The size of the weight indicates the size of the possibility. b[1] and b[2] are the biases between layers. The bias is an important parameter in the model. Introducing the bias after the weight settlement can make more precise adjustments to the input data. f is the activation function of each layer. The setting of the activation function is to free the neural network from linear constraints and limit the output of neurons to a certain range. Generally, relu function, tanh function and sigmoid function are used. In the evaluation of flexible resources, relu function is used between hidden layers, and sigmoid function is used between hidden layers and output layers. The expressions are as follows:

[0095] relu:y=max(0,x)

[0096]

[0097] The propagation process of FNN mainly includes forward propagation and back propagation. The forward propagation process calculates the final result through weights, biases and activation functions; the input part p in the training set i,min , p i,max , r i,max After forward propagation layer by layer calculation, the output is finally obtained and according to and and the true value φ and The difference is used to calculate the loss function.

[0098] In order to minimize the evaluation error, the error can be defined as:

[0099]

[0100] according to Figure 4 It can be further written as:

[0101]

[0102] Then sum the errors of each sample to minimize them:

[0103]

[0104] After the forward propagation is completed, the chain rule is used to reversely calculate the derivative of the loss function Loss for each variable, thereby updating the weight and bias parameters to complete the training process of the neural network.

[0105] The above neural network ensures the accuracy of the evaluation results, but in actual applications, since the power system dispatch instructions will be randomly issued at any point in the feasible domain of each flexible resource, when the bias of the prediction result makes the reported dispatch domain too large, if the dispatch instruction is within the real feasible domain, then the flexible resources can be dispatched normally; once the dispatch instruction exceeds the real feasible domain, the flexible resource dispatch is unreachable, which will cause great harm to the power grid. Therefore, the goal of this problem is not conventional precision fitting, but to constrain the part that exceeds the real value on this basis. In order to solve the problem of unreachable dispatch instructions, the loss function of deep learning is corrected, and a penalty term is introduced in the loss function of the neural network to regulate each round of loop iteration.

[0106] The initial loss function uses the L2 norm loss function:

[0107]

[0108] The L2 norm loss function minimizes the sum of squares of the differences between the estimated value and the target value. It is more sensitive to samples. Due to the use of square terms, outliers have a greater impact on its loss.

[0109] In order to ensure the normal operation of the constraints and make the predicted values ​​more conservative, consider adding a regular penalty term to the initial loss function to constrain the predicted values ​​that exceed the limit. The loss function after adding the regular penalty term is expressed as follows:

[0110]

[0111] Where m is the number of flexible resources in the dataset, Predict the output value for the neural network, is the true value of the predicted parameter of the flexible resource, s is a control switch, and its value is 0 or 1, depending on whether the prediction result needs to be punished. When the predicted value is greater than the actual value, s = 1, when the predicted value is less than or equal to the actual value, s = 0, r is an adjustable hyperparameter that determines the magnitude of the penalty, φ is the penalty function, that is, the regularization term, and its expression is as follows:

[0112]

[0113] Regularization is usually used to prevent model overfitting. Taking the gradient backpropagation into consideration, the L2 regularization selected for φ here only has a multiple impact on the gradient calculation result when the predicted value is greater than the actual value, without other interference. It has no effect on the gradient calculation result when the predicted value is less than the actual value. At the same time, it also makes the predicted result lower, resulting in a conservative predicted value.

[0114] During the neural network training process, the parameters that can be optimized are v, w, b and s. The steepest descent method is used to calculate the gradient of the parameters. Taking w as an example, its expression is as follows:

[0115]

[0116] In each cycle, the optimizable parameters v, w, b and s are calculated in turn, and the parameters are iterated according to the step size (batch) and the learning rate (learning rate). Through the training of the training set, after a certain number of iterations, the parameters v, w, b and s in the neural network are finally determined, and the offline training process of the neural network is completed. The trained neural network is tested with the test set to verify the effectiveness of the neural network.

[0117] In step S104, new flexible resource parameters are dynamically accessed in the online stage, and the new flexible resources are dynamically classified to obtain their corresponding optimal aggregation domains.

[0118] In some embodiments, dynamically classifying new flexible resources to obtain their corresponding optimal aggregation domains includes:

[0119] Calculate proximity based on each aggregate feasible domain and the feasible domain of each flexible resource;

[0120] The K-nearest neighbor algorithm and proximity are used to dynamically classify the new flexible resource parameters and obtain the optimal aggregation domain.

[0121] In the actual implementation process, for newly connected flexible resources, before the scheduling domain evaluation, it is necessary to first divide them into the optimal aggregation domain. Since this classification problem is a simple nonlinear problem, the traditional shallow machine learning method can achieve more accurate classification and avoid the complex model design of deep learning. Specifically, the basic parameters p of the basic flexible resource parameters in the training set and the test set are i,min , p i,max , r i,max The labels of the aggregation domains to which they belong are taken as input, the proximity is calculated according to the basic flexible resource parameters, and the proximity is used as the evaluation criterion to dynamically classify the flexible resources and obtain the aggregation domain with the maximum scaling factor, which is identified as the optimal aggregation domain of the flexible resource.

[0122] Since the k-nearest neighbor algorithm needs to store all training data, it is necessary to randomly select a certain number of flexible resource parameters from the offline training set as the training set, which is generally three to four times the number of prediction sets, and then use the new flexible resource parameters as the test set for label prediction. Since different parameters have different meanings and numerical scales, it is necessary to use the following formula to standardize the data:

[0123]

[0124] The proximity degree is calculated for the standardized new flexible resource. The acquisition of proximity degree depends on the training set data stored in the memory, the set k value and the distance metric. The formula is as follows:

[0125]

[0126] Among them, n is the dimension of the data, x i ,y i are the n-dimensional coordinates of the test point and training sample respectively, p is a variable parameter, and the distance measurement method changes according to its value.

[0127] Preferably, the distance metric usually uses the Euclidean distance (p=2), and the k value can be selected by cross-validation. The appropriate value is determined through multiple experiments. Usually, the k value is less than the square root of the data volume. According to the given distance metric, find the k points closest to x in the training set, and determine the classification result of x according to the categories of these k points. When k=3, the following is obtained: Figure 3 classification effect.

[0128] In step S105, the new flexible resource parameters and the optimal aggregation domain are input into the trust evaluation model for trust evaluation to obtain the trust control domain of the flexible resource.

[0129] like Figure 5 As shown in the figure, during the online evaluation process, after the new flexible resource is classified to obtain its optimal aggregation domain, its parameters are input into the trained neural network (i.e., the trusted evaluation model) corresponding to the optimal aggregation domain, and the trained model is used to quickly calculate the trusted control domain of the new flexible resource.

[0130] In summary, according to the data-driven method for supporting aggregation evaluation of adjustable resource scheduling domains proposed in an embodiment of the present invention, various parameters of flexible resources are obtained, and a rapid evaluation framework combining offline and online is constructed based on the two-stage idea; based on the framework, the newly connected flexible resources are first classified by KNN, and then the trusted regulation domain of the flexible resources is obtained by the neural network. In the offline stage, a multi-aggregator operation model is constructed based on a large number of flexible resources, and the machine learning algorithm is trained to complete the time-consuming work; in the online stage, the dynamically connected flexible resources are quickly classified into the aggregation domain and their feasible domain is evaluated, forming the optimal strategy for aggregation and dimensionality reduction of the adjustable space of a large number of flexible resources, so that the flexible resources can be classified and the feasible domain can be quickly evaluated, which can provide important support for the real-time scheduling and rapid access of flexible resources to the power grid.

[0131] Next, the data-driven device for supporting adjustable resource scheduling domain aggregation evaluation proposed in accordance with an embodiment of the present invention will be described with reference to the accompanying drawings.

[0132] Figure 6 It is a block diagram of a data-driven device supporting adjustable resource scheduling domain aggregation evaluation according to an embodiment of the present invention.

[0133] like Figure 6 As shown, the data-driven device 60 supporting adjustable resource scheduling domain aggregation evaluation includes: an offline generation module 601, an offline calculation module 602, an online classification module 603 and an online trusted evaluation module 604.

[0134] Among them, the offline generation module 601 is used to generate the basic flexible resource parameters and the number of aggregators in the target power grid in the offline stage. The offline calculation module 602 is used to calculate each aggregation feasible domain and the feasible domain of each basic flexible resource according to the basic flexible resource parameters and the number of aggregators. The offline training module 603 is used to train the feedforward neural network using each aggregation feasible domain and the feasible domain of each basic flexible resource to obtain a trusted evaluation model. The online classification module 604 is used to dynamically access new flexible resource parameters in the online stage, and dynamically classify the new flexible resources to obtain their corresponding optimal aggregation domain. The online trusted evaluation module 605 is used to input the new flexible resource parameters and the optimal aggregation domain into the trusted evaluation model for trusted evaluation to obtain the trusted regulation domain of the flexible resource.

[0135] Furthermore, in one embodiment of the present invention, the offline calculation module 602 includes:

[0136] An offline construction unit is used to establish models of the aggregation feasible domain and the flexible resource feasible domain, wherein the models of the aggregation feasible domain and the flexible resource feasible domain include an operation model of the flexible resources and an aggregation model of the flexible resources;

[0137] The offline calculation unit is used for inputting the basic flexible resource parameters and the number of aggregators into the models of the aggregation feasible domain and the flexible resource feasible domain to calculate the feasible domain of each aggregation feasible domain and each basic flexible resource.

[0138] Furthermore, in one embodiment of the present invention, the loss function of the feedforward neural network is expressed as:

[0139]

[0140]

[0141] Where m is the number of new flexible resource parameters, Predict the output value for the neural network, is the true value of the predicted parameter of the flexible resource, s is the control switch, r is the adjustable hyperparameter, and φ is the penalty function.

[0142] Further, in one embodiment of the present invention, the online classification module 603 includes:

[0143] An online calculation unit, used for calculating proximity according to each aggregate feasible domain and the feasible domain of each flexible resource;

[0144] The online dynamic classification unit is used to dynamically classify the new flexible resource parameters using the K nearest neighbor algorithm and proximity to obtain the optimal aggregation domain.

[0145] It should be noted that the aforementioned explanation of the data-driven method embodiment for supporting adjustable resource scheduling domain aggregation evaluation is also applicable to the data-driven device for supporting adjustable resource scheduling domain aggregation evaluation of this embodiment, and will not be repeated here.

[0146] According to the data-driven device supporting aggregation evaluation of adjustable resource scheduling domains proposed in an embodiment of the present invention, various parameters of flexible resources are obtained, and a rapid evaluation framework combining offline and online is constructed according to a two-stage concept; based on the framework, a multi-aggregator operation model is constructed according to a large number of flexible resources in the offline stage, and the machine learning algorithm is trained to complete time-consuming work; in the online stage, the dynamically accessed flexible resources are quickly classified into aggregation domains, and their feasible domains are evaluated, forming an optimal strategy for aggregation and dimensionality reduction of a large number of flexible resource adjustable spaces, so that the flexible resources can be classified and the feasible domain can be quickly evaluated, which can provide important support for the real-time scheduling and rapid access of flexible resources to the power grid.

[0147] Figure 7 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device may include:

[0148] A memory 701 , a processor 702 , and a computer program stored in the memory 701 and executable on the processor 702 .

[0149] When the processor 702 executes the program, the data-driven method for supporting adjustable resource scheduling domain aggregation evaluation provided in the above embodiments is implemented.

[0150] Furthermore, the electronic device further comprises:

[0151] The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0152] The memory 701 is used to store computer programs that can be executed on the processor 702 .

[0153] The memory 701 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0154] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0155] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0156] The processor 702 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0157] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned data-driven method for supporting adjustable resource scheduling domain aggregation evaluation.

[0158] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0159] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0160] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0161] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways as necessary and then storing it in a computer memory.

[0162] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0163] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0164] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0165] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A data-driven method for supporting domain aggregation evaluation of adjustable resource scheduling, characterized in that: The following steps are involved: Generate basic flexible resource parameters and the number of aggregators in the target power grid during the offline phase; Calculating each aggregation feasible domain and each basic flexible resource feasible domain according to the basic flexible resource parameters and the number of aggregators; The feedforward neural network is trained using each of the aggregate feasible domains and the feasible domain of each of the basic flexible resources to obtain a credible evaluation model, wherein the feedforward neural network is trained using each of the aggregate feasible domains and the feasible domain of each of the basic flexible resources to obtain a credible evaluation model, including: Input each of the aggregate feasible domains and the feasible domain of each of the basic flexible resources into the feedforward neural network for forward propagation layer-by-layer calculation to obtain an output value; Reversely calculating the loss function of the feedforward neural network according to the output value to update the weight and bias parameters of the feedforward neural network to obtain the credible assessment model; Dynamically accessing new flexible resource parameters in the online stage, and dynamically classifying the new flexible resources to obtain the optimal aggregation domain corresponding to them, wherein the dynamically classifying the new flexible resources to obtain the optimal aggregation domain corresponding to them includes: Calculating proximity according to each of the aggregated feasible domains and the feasible domain of each of the flexible resources; Dynamically classifying the new flexible resource parameters using a K-nearest neighbor algorithm and the proximity to obtain the optimal aggregation domain; The new flexible resource parameters and the optimal aggregation domain are input into the trusted evaluation model for trusted evaluation to obtain a trusted control domain of the flexible resource.

2. The data-driven method for supporting adjustable resource scheduling domain aggregation evaluation according to claim 1 is characterized in that: The calculating each aggregation feasible domain and the feasible domain of each basic flexible resource according to the basic flexible resource parameters and the number of aggregators includes: Establishing models of an aggregate feasible domain and a flexible resource feasible domain, wherein the models of the aggregate feasible domain and the flexible resource feasible domain include an operation model of flexible resources and an aggregate model of flexible resources; The basic flexible resource parameters and the number of aggregators are input into the models of the aggregation feasible domain and the flexible resource feasible domain to calculate the feasible domain of each aggregation feasible domain and each basic flexible resource.

3. The data-driven method for supporting adjustable resource scheduling domain aggregation evaluation according to claim 1, characterized in that: The loss function expression of the feedforward neural network is: Where m is the number of new flexible resource parameters, Predict output values ​​for the neural network, is the true value of the predicted parameter of the flexible resource, s is the control switch, r is the adjustable hyperparameter, and φ is the penalty function.

4. A data-driven device supporting adjustable resource scheduling domain aggregation evaluation, characterized in that: include: An offline generation module, used to generate basic flexible resource parameters and the number of aggregators in the target power grid during the offline phase; An offline calculation module, used for calculating each aggregation feasible domain and the feasible domain of each basic flexible resource according to the basic flexible resource parameters and the number of aggregators; An offline training module is used to train a feedforward neural network using each of the aggregate feasible domains and the feasible domain of each of the basic flexible resources to obtain a credible evaluation model, wherein the offline training module includes: A forward propagation unit, used for inputting each of the aggregate feasible domains and the feasible domain of each of the basic flexible resources into the feedforward neural network for forward propagation layer-by-layer calculation to obtain an output value; A back propagation unit, used for reversely calculating the loss function of the feedforward neural network according to the output value, so as to update the weight and bias parameters of the feedforward neural network and obtain the credible assessment model; The online classification module is used to dynamically access new flexible resource parameters in the online stage and dynamically classify the new flexible resources to obtain the corresponding optimal aggregation domain, wherein the online classification module includes: An online calculation unit, used for calculating proximity according to each aggregate feasible domain and the feasible domain of each flexible resource; An online dynamic classification unit, used for dynamically classifying the new flexible resource parameters by using a K-nearest neighbor algorithm and the proximity to obtain the optimal aggregation domain; The online trusted evaluation module is used to input the new flexible resource parameters and the optimal aggregation domain into the trusted evaluation model to perform trusted evaluation and obtain the trusted regulation domain of the flexible resource.

5. The data-driven device for supporting adjustable resource scheduling domain aggregation evaluation according to claim 4, characterized in that: The offline computing module comprises: An offline construction unit, used to establish models of the aggregation feasible domain and the flexible resource feasible domain, wherein the models of the aggregation feasible domain and the flexible resource feasible domain include an operation model of the flexible resources and an aggregation model of the flexible resources; The offline calculation unit is used to input the basic flexible resource parameters and the number of aggregators into the models of the aggregation feasible domain and the flexible resource feasible domain to calculate the feasible domain of each aggregation feasible domain and each basic flexible resource.

6. The data-driven device for supporting adjustable resource scheduling domain aggregation evaluation according to claim 4, characterized in that: The loss function expression of the feedforward neural network is: Where m is the number of new flexible resource parameters, Predict output values ​​for the neural network, is the true value of the predicted parameter of the flexible resource, s is the control switch, r is the adjustable hyperparameter, and φ is the penalty function.

7. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a data-driven method for supporting adjustable resource scheduling domain aggregation evaluation as described in any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement a data-driven method for supporting adjustable resource scheduling domain aggregation evaluation as described in any one of claims 1 to 3.

Images