Distributed energy resource mining method and device and storage medium

Abstract

The invention discloses a distributed energy resource mining method and device and a storage medium. The method comprises the steps that multiple distributed energy resource data are collected from distributed energy resource equipment; performing data cleaning on the distributed energy resource data; if the data cleaning is completed, mining interactive operation features among the distributed energy resource data in multiple dimensions to construct a multi-dimensional feature matrix; and inputting the multi-dimensional feature matrix of the plurality of historical moments into a preset time sequence model to generate the power of the next moment in the future. According to the embodiment of the invention, the coupling of multi-dimensional interactive operation characteristics is realized, and the diversity and dynamic change characteristics of the distributed energy resources are used, so that the data processing efficiency is effectively improved in the face of massive distributed energy resources, the accuracy of predicting the future power of the distributed energy resource equipment is improved, and the prediction efficiency of the future power of the distributed energy resource equipment is improved. And a detailed and accurate basis is provided for management personnel to formulate reasonable operation strategies and decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power, and in particular to a method, device and storage medium for mining distributed energy resources. Background Art

[0002] With the widespread integration of distributed energy resources (DERs) such as distributed power sources (photovoltaic, wind power, etc.), distributed energy storage, and controllable loads into the power system, the operating characteristics of the power system have changed significantly.

[0003] These distributed energy resources are characterized by dispersion, intermittency, and randomness. They have complex interactive relationships with each other and with traditional power grids, which are of vital importance to the optimal scheduling, stability analysis, and reliability assessment of power systems.

[0004] At present, research on distributed energy resources mostly focuses on the characteristic analysis of a single type of energy resource or simple integrated operation. Traditional modeling methods are often based on physical models and empirical formulas, which are not suitable for the diversity and dynamic change characteristics of distributed energy resources. As a result, when faced with massive distributed energy resources, the data processing efficiency of traditional modeling methods is low. Summary of the Invention

[0005] In view of this, the present invention provides a method, device and storage medium for mining distributed energy resources, so as to improve the efficiency of processing data of distributed energy resources.

[0006] A first aspect of the present invention provides a method for mining distributed energy resources, comprising:

[0007] Collect various distributed energy resource data from distributed energy resource devices;

[0008] performing data cleaning on the distributed energy resource data;

[0009] If the data cleaning is completed, then the interactive operation characteristics between the distributed energy resource data are mined in multiple dimensions to construct a multi-dimensional feature matrix;

[0010] The multidimensional feature matrix of multiple historical moments is input into a preset time series model to generate the power of the next moment in the future.

[0011] A second aspect of the present invention provides a distributed energy resource mining device, comprising:

[0012] A data acquisition module is used to collect various distributed energy resource data from distributed energy resource devices;

[0013] A data cleaning module, configured to clean the distributed energy resource data;

[0014] A multi-dimensional feature matrix construction module is used to mine the interactive operation characteristics between the distributed energy resource data in multiple dimensions to construct a multi-dimensional feature matrix after the data cleaning is completed;

[0015] The power prediction module is used to input the multidimensional feature matrix of multiple historical moments into a preset time series model to generate the power at the next moment in the future.

[0016] A third aspect of the present invention provides an electronic device, comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the distributed energy resource mining method as described in the first aspect above.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the distributed energy resource mining method as described in the first aspect above.

[0021] A fifth aspect of the present invention provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the distributed energy resource mining method as described in the first aspect above.

[0022] In this embodiment, various distributed energy resource data are collected from distributed energy resource devices; the distributed energy resource data is cleaned; and once the data cleaning is complete, the interactive operating characteristics between the distributed energy resource data are mined in multiple dimensions to construct a multidimensional feature matrix; the multidimensional feature matrix at multiple historical moments is input into a preset time series model to generate the power at the next moment in the future. This embodiment achieves the coupling of multidimensional interactive operating characteristics and utilizes the diversity and dynamic change characteristics of distributed energy resources to effectively improve data processing efficiency when faced with massive distributed energy resources, thereby improving the accuracy of predicting the future power of distributed energy resource devices and providing managers with a detailed and accurate basis for formulating reasonable operating strategies and decisions.

[0023] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 This is a flow chart of a distributed energy resource mining method provided in Example 1 of the present invention.

[0026] Figure 2 This is a structural diagram of a distributed energy resource mining device provided in Example 2 of the present invention.

[0027] Figure 3 This is a structural diagram of an electronic device provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can cover sequential implementations other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Example 1

[0031] See also Figure 1, shows a flow chart of a distributed energy resource mining method provided by the first embodiment of the present invention. The method can be executed by a distributed energy resource mining device. The distributed energy resource mining device can be implemented in the form of hardware and / or software. The distributed energy resource mining device can be configured in an electronic device. Figure 1 As shown, the method includes:

[0032] Step 101: Collect a variety of distributed energy resource data from distributed energy resource devices.

[0033] In this embodiment, sensors, smart meters, distributed energy management systems and other equipment can be used to collect various distributed energy resource data from distributed energy resource equipment such as distributed photovoltaic power stations, distributed wind farms, distributed energy storage systems, controllable loads, microgrids, integrated energy, power grids, etc., for example, power data (such as power, voltage, current, charge and discharge status, state of charge, etc.), electricity price data (such as time-of-use electricity prices, etc.) and environmental data (such as temperature, wind speed, solar irradiance, etc.), etc.

[0034] Assume that distributed energy resource data of n time steps are collected, which is recorded as D = {d1, d2, ..., d n}, where d i Refers to the distributed energy resource data at time i.

[0035] Step 102: Clean the distributed energy resource data.

[0036] In this embodiment, the distributed energy resource data may be pre-processed to achieve data cleaning of the distributed energy resource data and improve the quality of the distributed energy resource data.

[0037] In the specific implementation, data cleaning includes outlier processing and missing value processing.

[0038] Based on the statistical method, assuming that the distributed energy resource data obeys a certain probability distribution, the mean and standard deviation of the parameter values in the distributed energy resource data can be calculated.

[0039] The average values of the parameters in the distributed energy resource data are:

[0040]

[0041] The standard deviation of the parameter values in the distributed energy resource data is:

[0042]

[0043] Where μ is the mean value of the parameter in the distributed energy resource data, σ is the standard deviation of the parameter in the distributed energy resource data, and x iis the i-th parameter value in the distributed energy resource data, and n is the number of parameter values in the distributed energy resource data.

[0044] The standard deviation of the parameter value in the distributed energy resource data is taken as a multiple value (such as 3 times the value) as the abnormal threshold to determine the normal range of the parameter value in the distributed energy resource data.

[0045] If the absolute value of the difference between a parameter value in the distributed energy resource data and the average value of the parameter value in the distributed energy resource data is greater than the abnormal threshold, the parameter value in the distributed energy resource data is determined to be an abnormal value and the abnormal value is filtered out.

[0046] For example, in the case where |x j When -μ|>3σ, it can be considered that x j is an outlier, where x j The jth parameter value in the distributed energy resource data, μ is the mean value of the parameter value in the distributed energy resource data, and σ is the standard deviation of the parameter value in the distributed energy resource data.

[0047] Distributed energy resource data has time series characteristics and correlation, so the Lagrange interpolation method can be used to supplement the missing parameter values in distributed energy resource data.

[0048] It is known that there are n points (the horizontal axis is time, the vertical axis is the parameter value of distributed energy resource data) (x0, y0), (x1, y1), ..., (x n-1 ,y n-1 ), let the interpolation polynomial L(x) be:

[0049] L(x)=y0l0(x)+y1l1(x)+…+y n-1 l n-1 (x)

[0050] Among them, y i is the ordinate of the known point; l i (x) is the Lagrange basis function, for n+1 points x0,x1,…,x n , the calculation formula of the Lagrange basis function is:

[0051]

[0052] By using the interpolation polynomial L(x), we can find the value of n-1 ], and the y value corresponding to any x in , realizes the interpolation calculation between known data points.

[0053] Step 103: If data cleaning is completed, interactive operation characteristics between distributed energy resource data are mined in multiple dimensions to construct a multi-dimensional feature matrix.

[0054] In this embodiment, multiple dimensions in which distributed energy resource data are associated in terms of interactive relationships and operating characteristics can be screened out, the interactive operating characteristics between distributed energy resource data can be mined from multiple dimensions, and a structured multidimensional feature matrix can be constructed using the interactive operating characteristics.

[0055] In one embodiment of the present invention, the parameter values of the distributed energy resource data include power values and communication data, and step 103 may include the following steps:

[0056] Step 1031: Mining interactive operation features between power values in the dimension of time series to obtain multiple first indicator values.

[0057] In this embodiment, the interactive operation characteristics between power values can be mined in the dimension of time series to obtain multiple first indicator values. At this time, the multiple first indicator values are used to quantify the dynamic fluctuation characteristics of the power of distributed energy resource equipment.

[0058] In practical applications, the plurality of first indicator values include a fluctuation rate of the power value, an autocorrelation coefficient of the power value, an average value of the power value, and a standard deviation of the power value.

[0059] The fluctuation rate of power value reflects the intensity of the change of power value at adjacent time points. The fluctuation rate of power value is:

[0060]

[0061] Among them, V is the fluctuation rate of power value, P t is the power value at time t, P t+1 is the power value at time t+1, and N is the number of power sampling points.

[0062] The autocorrelation coefficient of the power value represents the autocorrelation coefficient of the k-step lag, which is used to measure the periodicity of the sequence composed of power values. The autocorrelation coefficient of the power value is:

[0063]

[0064] Among them, ρ k is the autocorrelation coefficient of the power value, P t is the power value at time t, P t+1 is the power value at time t+1, is the average power value, and N is the number of power sampling points.

[0065] The average power value is used to reflect the average processing level. The average power value is:

[0066]

[0067] Where μ is the average power value, P t is the power value at time t, and N is the number of power sampling points.

[0068] The standard deviation of the power value is used to quantify the degree of dispersion of the power value. The standard deviation of the power value is:

[0069]

[0070] Among them, σ is the standard deviation of the power value, μ is the average value of the power value, P t is the power value at time t, and N is the number of power sampling points.

[0071] Step 1032: Use the communication data to mine interactive operation characteristics between distributed energy resource devices in the dimension of spatial association to obtain multiple second indicator values.

[0072] In this embodiment, communication data can be used to mine the interactive operation characteristics between distributed energy resource devices in the dimension of spatial association to obtain multiple second indicator values. At this time, the multiple second indicator values represent the topological relationship and communication quality between distributed energy resource devices, which can support large-scale device access and has high scalability.

[0073] In a specific implementation, the multiple second indicator values include the communication weights between the distributed energy resource devices and the average topological distances between the distributed energy resource devices.

[0074] The communication weight between distributed energy resource devices is used to represent the association weight between two distributed energy resource devices. The larger the communication weight, the closer the association. The communication weight between distributed energy resource devices is:

[0075]

[0076] Among them, W ij is the communication weight between the i-th distributed energy resource device and the j-th distributed energy resource device, d ij is the topological distance between the i-th distributed energy resource device and the j-th distributed energy resource device, is the communication delay from the i-th distributed energy resource device to the j-th distributed energy resource device (in ms), is the communication delay threshold (i.e., the maximum communication delay allowed by the power system), α and β are weight coefficients, and α+β=1.

[0077] The average topological distance of distributed energy resource devices is:

[0078]

[0079] in, is the average topological distance of the i-th distributed energy resource device, D i Set for device The i-th distributed energy resource device in n is the number of distributed energy resource devices, is the adjacent device of the i-th distributed energy resource device, and the adjacent device is other distributed energy resource devices whose weight with the i-th distributed energy resource device is greater than the weight threshold, that is, θ is the weight threshold, such as θ = 0.3, is the number of adjacent devices, D j is the jth adjacent device, d ij is the topological distance between the i-th distributed energy resource device and the j-th distributed energy resource device.

[0080] Step 1033: Mining the interactive operation characteristics between power values in the dimension of energy flow to obtain multiple third indicator values.

[0081] In this embodiment, the interactive operation characteristics between power values are mined in the dimension of energy flow to obtain multiple third indicator values. At this time, the multiple third indicator values are used to describe the matching relationship of energy interaction between distributed energy resource devices.

[0082] In practical applications, the multiple third index values include similarity of power curves, phase difference, and energy complementarity index.

[0083] The similarity of the power curves is:

[0084]

[0085] Among them, S ij is the similarity of the power curves, and its value range is (0,1]. The greater the similarity, the more similar the power curves are. i (t) is the power value of the i-th distributed energy resource device at time t, P j (t') is the power value of the jth distributed energy resource device at time t', π is the dynamic time warping alignment path, which allows nonlinear time axis matching (such as aligning sequences of different lengths), and min is the function that takes the minimum value.

[0086] The phase difference is:

[0087] ΔT ij =argmax Δt Corr(P i (t),P j (t+Δt))

[0088] Where, ΔT ijis the phase difference (i.e., time offset) between the i-th distributed energy resource device and the j-th distributed energy resource device, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t+Δt) is the power value of the jth distributed energy resource device at time t+Δt (sequence after time offset); argmax Δt To find the cross-correlation function Corr(P i (t),P j (t+Δt)) The time offset Δt of the maximum output value;

[0089] The cross-correlation function is:

[0090]

[0091] Among them, μ i is the average power value of the i-th distributed energy resource device, μ j is the average power value of the jth distributed energy resource device, σ i is the standard deviation of the power value of the i-th distributed energy resource device, σ j is the standard deviation of the power value of the jth distributed energy resource device, and T is the number of time sampling points.

[0092] The energy complementarity index is:

[0093]

[0094] Among them, C ij is the energy complementarity index between the i-th distributed energy resource device and the j-th distributed energy resource device, and its value range is [0,1]. The larger the energy complementarity index is, the stronger the output complementarity of the two distributed energy resource devices is. i (t) is the power value of the i-th distributed energy resource device at time t, P j (t) is the power value of the jth distributed energy resource device at time t, min(P i (t),P j (t)) is the i (t) and P j The smallest one among (t) is used to quantify the overlapping output, where T is the number of time sampling points.

[0095] Step 1034: Mining interactive operation characteristics between distributed energy resource data in the dimension of the external environment to obtain multiple fourth indicator values.

[0096] In this embodiment, the interactive operation characteristics between distributed energy resource data are mined in the dimension of the external environment to obtain a plurality of fourth indicator values.

[0097] In a specific implementation, the fourth indicator value includes temperature, wind speed, light amplitude, real-time electricity price and predicted electricity price.

[0098] At this time, the fourth index value can be expressed as:

[0099] E=[T,W,G,λ,λ']

[0100] Among them, E is the set of fourth indicator values, T is temperature (unit: °C), which affects the efficiency and load demand of distributed energy resource equipment; W is wind speed (unit: m / s), which is used for wind farm output prediction; G is sunlight amplitude (unit: W / ㎡), which is used for photovoltaic output calculation; λ is the real-time electricity price (unit: RMB / kWh), which affects the economic dispatch strategy; λ' is the predicted electricity price (unit: RMB / kWh), which is used to optimize the operation plan for future time periods.

[0101] Of course, the above-mentioned first, second, third, and fourth index values are merely examples. When implementing this embodiment, other first, second, third, and fourth index values may be set based on actual circumstances, and this embodiment does not limit this. Furthermore, in addition to the above-mentioned first, second, third, and fourth index values, those skilled in the art may also adopt other first, second, third, and fourth index values based on actual needs, and this embodiment does not limit this either.

[0102] Step 1035: Construct a multi-dimensional feature matrix using the multiple first index values, the multiple second index values, the multiple third index values, and the multiple fourth index values.

[0103] In this embodiment, a plurality of first indicator values, a plurality of second indicator values, a plurality of third indicator values, and a plurality of fourth indicator values may be integrated to construct a structured multi-dimensional feature matrix.

[0104] For designated distributed energy resource equipment D i , the power fluctuation rate V i , the autocorrelation coefficient of power value The average power value μ i The standard deviation of the power value σ i Constructed with a length of L t The first eigenvector of Right now, in, is the set of real numbers.

[0105] For designated distributed energy resource equipment D i , for distributed energy resource equipment D i With adjacent device D j (Adjacent device Dj A collection of neighboring devices ) communication weight W between ij Find the average value and get the average communication weight

[0106] The average communication weight and the average topological distance Construct the second eigenvector Right now, in, is the number of neighboring devices.

[0107] For designated distributed energy resource equipment D i Filter interactive devices Among them, the interactive device is the distributed energy resource device D i Other distributed energy resource devices D whose power curves have similarities greater than the similarity threshold j ,in, Among them, S ij D is a distributed energy resource device i and distributed energy resource equipment D j The similarity of the power curves between them is expressed as γ, where γ is the similarity threshold, such as γ = 0.5.

[0108] Distributed energy resource equipment D i Between interactive devices j The average value S of the similarity of the power curve ij , and get the average similarity in, A collection of interactive devices. is the number of interactive devices.

[0109] Distributed energy resource equipment D i Interactive device D j The phase difference ΔT ij Find the average value and get the average phase difference

[0110] Distributed energy resource equipment D i Interactive device D j Energy complementarity index C ij Calculate the average value and get the average energy complementarity index

[0111] The third eigenvector is constructed by the average similarity, average phase difference and average energy complementarity index Right now,

[0112] For designated distributed energy resource equipment D i, the temperature T, wind speed W, light amplitude G, real-time electricity price λ and predicted electricity price λ' are used to construct the fourth eigenvector Right now,

[0113] The first eigenvector The second eigenvector The third eigenvector With the fourth eigenvector Construct device feature vector F i .

[0114]

[0115] Multiple device feature vectors F i Stacking forms a multi-dimensional feature matrix F,

[0116] Furthermore, the multidimensional feature matrix can be updated according to the time window. In this process, the environmental variable E at the next time t can be calculated. (t+1) and the environment variable E at the previous time t+1 (t) The change between |E (t+1) -E (t) ||2.

[0117] If the environmental change is greater than the change threshold θ, that is, ||E (t+1) -E (t) ||2>θ, then calculate the increment ΔF of the interactive operation feature at the previous moment (t) .

[0118] Calculate the multidimensional feature matrix F at the previous moment (t) Increment ΔF of interactive operation characteristics (t) The sum between them is used as the power F at the next moment (t+1) , that is, F (t+1) =F (t) +ΔF (t) .

[0119] If the environmental change is less than or equal to the change threshold θ, that is, ||E (t+1) -E (t) ||2≤θ, then the multidimensional feature matrix F at the previous moment (t) Set to the power F at the next moment (t+1) , that is, F (t+1) =F (t) .

[0120] In this embodiment, the multi-dimensional feature matrix can be updated based on the real-time features of environmental mutations, and has strong dynamic adaptability.

[0121] In addition, the multidimensional feature matrix can be normalized using methods such as Min-Max Normalization to convert data of different dimensions into the same numerical range [0,1] to eliminate the impact of the dimension on subsequent analysis.

[0122] Exemplarily, the min-max normalization method is:

[0123]

[0124] Among them, x is the original data, x min is the minimum value of the data, x max is the maximum value of the data, x norm The data are normalized.

[0125] Step 104: Input the multidimensional feature matrix of multiple historical moments into a preset time series model to generate the power at the next moment in the future.

[0126] In this embodiment, the multi-dimensional feature matrix of multiple historical moments (time steps) can be input into a preset time series model to generate the power of the next moment (time step) in the future.

[0127] In practical applications, a sample set is constructed for a time series model, and the sample set includes multiple multi-character feature matrices.

[0128] Assume that the sample data points of the multidimensional feature matrix p are n (if the time span is one year and the time step is one hour, then the value of n is 8760), recorded as the sample set The normalized multidimensional feature matrix is divided into 80:20 ratios, and the first 0.8n multidimensional feature matrices are used as the training set. The last 0.2n multidimensional feature matrices are used as validation sets

[0129] Assume that each sample contains a multidimensional feature matrix of the past m time steps (if m = 24, it is one day's data) to predict the power of the next time step.

[0130] For the training set, the samples are constructed as follows:

[0131] Let X train is the input feature (i.e., the multidimensional feature matrix of the past m time steps), Y train is the output label (i.e. the power of the next time step), then:

[0132]

[0133] Where j represents the j-th multidimensional feature matrix, j = 1, 2, ..., 0.8nm; Xtrain [j] is an m×14 matrix, each row represents a time step, and each column represents a feature.

[0134] Use the same method to construct the sample X of the validation set val (input features), Y val (label), where j = 0.8n+1, 0.8n+2, ..., 0.2nm.

[0135] At this time, X train The shape is (0.8nm,m), Y train The shape is (0.8nm,1), X val The shape is (0.2nm,m), Y val The shape is (0.2nm,1).

[0136] Considering that time series models require input data to have a specific shape, usually [number of samples, time steps, number of features], we can transform X train The shape of is adjusted to (0.8nm,m,14), and X val The shape of the image is adjusted to (0.2nm,m,14).

[0137] Exemplarily, the time series model includes an input layer, a hidden layer, and an output layer; wherein the hidden layer is a two-layer long short-term memory network (LSTM), and the output layer is a fully connected layer.

[0138] The input layer receives data of shape (m, 14), that is, each sample contains m time steps and each time step has 14 features.

[0139] The first hidden layer is an LSTM, assuming it contains h1=64 neurons.

[0140] The processing of LSTM can be modeled as follows:

[0141]

[0142] Among them, i t is the input gate; f t For the forget gate; t is the output gate; is a candidate memory unit; C t is the memory unit; h t is the hidden state; x t is the input of the current time step, h t-1 is the hidden state of the previous time step, C t-1is the memory cell for the previous time step, σ is the sigmoid function, tanh is the hyperbolic tangent function, ⊙ represents element-wise multiplication, and W and b are the weight matrix and bias vector. The first LSTM hidden layer returns a sequence, meaning that each time step has an output, allowing subsequent hidden layers to process the complete time series information.

[0143] The second hidden layer is also an LSTM layer, assuming it contains 32 neurons. This layer does not return a sequence, but outputs the hidden state of the last time step to further extract high-level features of the time series.

[0144] The output layer is a fully connected layer containing 1 neuron, which maps the output of the previous layer to the final predicted value through linear transformation. Let the output of the previous layer be h last , the predicted value of the output layer Where W o is the weight matrix, b o is the bias vector.

[0145] During the training process, you can adjust the configuration parameters of the time series model; these configuration parameters include the number of hidden layers (such as 2-layer LSTM, 3-layer LSTM, etc.) and the number of nodes in each hidden layer (such as 32 nodes in the first LSTM layer, 16 nodes in the first LSTM layer, 64 nodes in the first LSTM layer, 32 nodes in the first LSTM layer, etc.).

[0146] Adaptive Moment Estimation (Adam) is used to train the time series model under various configuration parameters based on the training set.

[0147] Adam is a comprehensive use of the first-order moment estimation and second-order moment estimation of the gradient. The steps of using Adam to update the parameters θ of the time series model (including the weights and biases of the LSTM layer and the output layer) are as follows:

[0148] 1) Calculate the gradient g t

[0149]

[0150] Among them, θ is the parameter of the new time series model, is the gradient at time t-1 in the time series model, and J is the loss function, such as the mean square error (MSE):

[0151]

[0152] Where N is the number of samples, y i is the true value, is the predicted value.

[0153] 2) Calculate the first-order moment estimate m t

[0154] m t =β1m t-1 +(1-β1)g t

[0155] Where β1 is the hyperparameter of the first-order momentum, such as β1 = 0.9.

[0156] 3) Calculate the second-order moment estimate v t

[0157] v t =β2v t-1 +(1-β2)g t 2

[0158] Where β2 is the hyperparameter of the second-order momentum, such as β2 = 0.999.

[0159] 4) Modified first-order moment estimation

[0160]

[0161] 5) Modified second-order moment estimation

[0162]

[0163] 6) Update parameter θ t

[0164]

[0165] Among them, the learning rate α=0.001, ∈ is a small constant, such as ∈=10 -8 , to prevent the denominator from being zero.

[0166] During the training process, the input feature X in the training set train and label Y train Input the time series model and train it for T=1000 iterations. In each iteration, the model calculates the predicted value based on the current parameters. in accordance with With Y train The MSE is calculated and the Adam optimization algorithm calculates the gradient based on the loss and updates the parameters of the time series model.

[0167] If the training is completed, the mean square error (MSE) of the time series model under each configuration parameter is verified based on the validation set.

[0168] Compare the mean square error (MSE) under various configuration parameters and determine the configuration parameters with the smallest mean square error for the time series model.

[0169] Then, when applying the time series model, the multidimensional feature matrix of multiple historical moments is received in the input layer; in the first layer of the long short-term memory network, features are extracted from the multidimensional feature matrix of multiple moments to obtain the first hidden features of multiple time steps; in the second layer of the long short-term memory network, features are extracted from the first hidden features of multiple time steps to obtain the second hidden features of the last time step; in the fully connected layer, the second hidden features of the last time step are mapped to the power of the next future moment.

[0170] In this embodiment, LSTM is applied to the mining of distributed energy resource data, which has the following advantages:

[0171] (1) Mining hidden features: It can automatically mine hidden features and relationships from a large amount of multi-source heterogeneous distributed energy resource data;

[0172] (2) Adapt to different data patterns: It can automatically extract patterns of different distributed energy resource data and perform adaptive learning;

[0173] (3) Inter-resource interaction analysis: quantified the mutual influence between distributed energy resource data and improved the accuracy of the interactive operation characteristic model of distributed energy resource equipment;

[0174] (4) Forecasting and decision support: It can accurately predict the future operating status of distributed energy resources and provide detailed and accurate basis for managers to formulate reasonable operating strategies and make decisions.

[0175] In this embodiment, various distributed energy resource data are collected from distributed energy resource devices; the distributed energy resource data is cleaned; and once the data cleaning is complete, the interactive operating characteristics between the distributed energy resource data are mined in multiple dimensions to construct a multidimensional feature matrix; the multidimensional feature matrix at multiple historical moments is input into a preset time series model to generate the power at the next moment in the future. This embodiment achieves the coupling of multidimensional interactive operating characteristics and utilizes the diversity and dynamic change characteristics of distributed energy resources to effectively improve data processing efficiency when faced with massive distributed energy resources, thereby improving the accuracy of predicting the future power of distributed energy resource devices and providing managers with a detailed and accurate basis for formulating reasonable operating strategies and decisions.

[0176] Example 2

[0177] See also Figure 2 , shows a schematic structural diagram of a distributed energy resource mining device provided by the second embodiment of the present invention. Figure 2 As shown, the device includes:

[0178] The data collection module 201 is used to collect various distributed energy resource data from distributed energy resource devices;

[0179] A data cleaning module 202 is used to clean the distributed energy resource data;

[0180] A multi-dimensional feature matrix construction module 203 is configured to mine the interactive operation characteristics between the distributed energy resource data in multiple dimensions to construct a multi-dimensional feature matrix after the data cleaning is completed;

[0181] The power prediction module 204 is configured to input the multi-dimensional feature matrix of multiple historical moments into a preset time series model to generate the power at the next moment in the future.

[0182] In one embodiment of the present invention, the data cleaning module 202 includes:

[0183] an outlier processing module, configured to calculate an average value and a standard deviation of the parameter values in the distributed energy resource data; take a multiple of the standard deviation of the parameter values in the distributed energy resource data as an outlier threshold; if the absolute value of the difference between a certain parameter value in the distributed energy resource data and the average value of the parameter values in the distributed energy resource data is greater than the outlier threshold, determine that the parameter value is an outlier, and filter out the outlier;

[0184] The missing value processing module is used to supplement the missing parameter values in the distributed energy resource data using the Lagrange interpolation method.

[0185] In one embodiment of the present invention, the parameter values of the distributed energy resource data include power values and communication data;

[0186] The multi-dimensional feature matrix construction module 203 includes:

[0187] A first indicator value mining module is used to mine the interactive operation characteristics between the power values in the dimension of time series to obtain multiple first indicator values;

[0188] a second indicator value mining module, configured to use the communication data to mine interactive operation characteristics between the distributed energy resource devices in a spatial correlation dimension to obtain a plurality of second indicator values;

[0189] a third indicator value mining module, configured to mine interactive operation characteristics between the power values in the dimension of energy flow to obtain a plurality of third indicator values;

[0190] a fourth indicator value mining module, configured to mine interactive operation characteristics between the distributed energy resource data in the dimension of the external environment to obtain a plurality of fourth indicator values;

[0191] The indicator value fusion module is used to construct a multi-dimensional feature matrix by combining the plurality of the first indicator values, the plurality of the second indicator values, the plurality of the third indicator values and the plurality of the fourth indicator values.

[0192] In one embodiment of the present invention, the plurality of first indicator values include a volatility of the power value, an autocorrelation coefficient of the power value, an average value of the power value, and a standard deviation of the power value;

[0193] The fluctuation rate of the power value is:

[0194]

[0195] Wherein, V is the fluctuation rate of the power value, P t is the power value at time t, P t+1 is the power value at time t+1, and N is the number of power sampling points;

[0196] The autocorrelation coefficient of the power value is:

[0197]

[0198] Among them, ρ k is the autocorrelation coefficient of the power value, P t is the power value at time t, P t+1 is the power value at time t+1, is the average value of the power value, and N is the number of power sampling points;

[0199] The plurality of second indicator values include communication weights between the distributed energy resource devices and average topological distances between the distributed energy resource devices;

[0200] The communication weights between the distributed energy resource devices are:

[0201]

[0202] Among them, W ij is the communication weight between the i-th distributed energy resource device and the j-th distributed energy resource device, d ij is the topological distance between the i-th distributed energy resource device and the j-th distributed energy resource device, is the communication delay from the i-th distributed energy resource device to the j-th distributed energy resource device, is the communication delay threshold, α and β are weight coefficients;

[0203] The average topological distance of the distributed energy resource devices is:

[0204]

[0205] in, is the average topological distance of the i-th distributed energy resource device, is an adjacent device of the i-th distributed energy resource device, wherein the adjacent device is another distributed energy resource device whose weight with the i-th distributed energy resource device is greater than a weight threshold, is the number of adjacent devices, D j is the jth adjacent device, d ij is the topological distance between the i-th distributed energy resource device and the j-th distributed energy resource device;

[0206] The plurality of third index values include similarity of power curves, phase difference, and energy complementarity index;

[0207] The similarity of the power curves is:

[0208]

[0209] Among them, S ij is the similarity of the power curve, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t') is the power value of the jth distributed energy resource device at time t', π is the dynamic time warping alignment path, and min is the function of taking the minimum value;

[0210] The phase difference is

[0211] ΔT ij =argmax Δt Corr(P i (t),P j (t+Δt))

[0212] Where, ΔT ij is the phase difference between the i-th distributed energy resource device and the j-th distributed energy resource device, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t+Δt) is the power value of the jth distributed energy resource device at time t+Δt; argmax Δt To find the cross-correlation function Corr(P i (t),Pj (t+Δt)) The time offset Δt of the maximum output value;

[0213] The cross-correlation function is:

[0214]

[0215] Among them, μ i is the average value of the power value of the i-th distributed energy resource device, μ j is the average power value of the jth distributed energy resource device, σ i is the standard deviation of the power value of the i-th distributed energy resource device, σ j is the standard deviation of the power value of the j-th distributed energy resource device, and T is the number of time sampling points;

[0216] The energy complementation index is:

[0217]

[0218] Among them, C ij is the energy complementarity index between the i-th distributed energy resource device and the j-th distributed energy resource device, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t) is the power value of the jth distributed energy resource device at time t, min(P i (t),P j (t)) is the i (t) and P j (t) takes the smallest one, where T is the number of time sampling points;

[0219] The fourth indicator value includes temperature, wind speed, light amplitude, real-time electricity price and predicted electricity price.

[0220] In one embodiment of the present invention, the indicator value fusion module includes:

[0221] A first eigenvector construction module is configured to construct, for a specified distributed energy resource device, a fluctuation rate of the power value, an autocorrelation coefficient of the power value, an average value of the power value, and a standard deviation of the power value into a first eigenvector;

[0222] A second eigenvector construction module is configured to average the communication weights between the designated distributed energy resource device and the adjacent device to obtain an average communication weight; and construct a second eigenvector using the average communication weight and the average topological distance;

[0223] A third feature vector construction module is used to screen interactive devices for the specified distributed energy resource devices; the interactive devices are other distributed energy resource devices whose similarity to the power curve of the distributed energy resource device is greater than a similarity threshold; the similarities of the power curves between the distributed energy resource devices and the interactive devices are averaged to obtain an average similarity; the phase differences between the distributed energy resource devices and the interactive devices are averaged to obtain an average phase difference; the energy complementarity indexes between the distributed energy resource devices and the interactive devices are averaged to obtain an average energy complementarity index; the average similarity, the average phase difference and the average energy complementarity index are used to construct a third feature vector;

[0224] a fourth eigenvector construction module, configured to construct a fourth eigenvector for the specified distributed energy resource device by combining the temperature, the wind speed, the light amplitude, the real-time electricity price, and the predicted electricity price;

[0225] a device feature vector construction module, configured to construct a device feature vector by combining the first feature vector, the second feature vector, the third feature vector, and the fourth feature vector;

[0226] The device feature vector stacking module is used to stack multiple device feature vectors to form a multi-dimensional feature matrix.

[0227] In one embodiment of the present invention, the multi-dimensional feature matrix construction module 203 further includes:

[0228] A change calculation module is used to calculate the change between the environmental variables at the next moment and the environmental variables at the previous moment;

[0229] an increment calculation module, configured to calculate the increment of the interactive operation feature at the previous moment if the change amount is greater than a change threshold;

[0230] A first feature matrix updating module is configured to calculate the sum of the multidimensional feature matrix at the previous moment and the increment of the interactive operation feature as the multidimensional feature matrix at the next moment;

[0231] The second feature matrix updating module is configured to set the multidimensional feature matrix at the previous moment as the multidimensional feature matrix at the next moment if the change amount is less than or equal to the change threshold.

[0232] In one embodiment of the present invention, the time series model includes an input layer, a hidden layer, and an output layer; the hidden layer is a two-layer long short-term memory network, and the output layer is a fully connected layer;

[0233] The power prediction module 204 includes:

[0234] A multi-dimensional feature matrix receiving module, configured to receive the multi-dimensional feature matrix at multiple historical moments in the input layer;

[0235] A first hidden feature extraction module is configured to extract features from the multidimensional feature matrix at multiple moments in the first layer of the long short-term memory network to obtain first hidden features at multiple time steps;

[0236] A second hidden feature extraction module is used to extract features from the first hidden features of multiple time steps in the second layer of the long short-term memory network to obtain a second hidden feature of the last time step;

[0237] A power mapping module is used to map the second hidden feature of the last time step into the power of the next future moment in the fully connected layer.

[0238] In one embodiment of the present invention, it further comprises:

[0239] A set construction module, used to construct a training set and a validation set for the time series model;

[0240] A configuration parameter adjustment module, configured to adjust configuration parameters of the time series model; the configuration parameters include the number of hidden layers and the number of nodes in each hidden layer;

[0241] A time series model training module, configured to train the time series model under each of the configuration parameters using adaptive moment estimation based on the training set;

[0242] A time series model verification module is used to verify the mean square error of the time series model under each of the configuration parameters based on the verification set if the training is completed;

[0243] The configuration parameter application module is used to determine the configuration parameters that minimize the mean square error when applying the time series model.

[0244] The distributed energy resource mining device provided in the embodiment of the present invention can execute the distributed energy resource mining method provided in any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the distributed energy resource mining method.

[0245] Example 3

[0246] See also Figure 3, which shows a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, blade servers, mainframe computers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0247] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. Various programs and data required for the operation of the electronic device 10 can also be stored in the RAM 13. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0248] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0249] Processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any other suitable processors, controllers, microcontrollers, etc. Processor 11 executes the various methods and processes described above, such as the distributed energy resource mining method.

[0250] In some embodiments, the distributed energy resource mining method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the distributed energy resource mining method described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to execute the distributed energy resource mining method in any other appropriate manner (e.g., by means of firmware).

[0251] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0252] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0253] In the context of the present invention, computer-readable storage media can be tangible media that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage media can include but are not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage media can be machine-readable signal media. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0254] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0255] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0256] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0257] Example 4

[0258] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the distributed energy resource mining method provided by any embodiment of the present invention.

[0259] The computer program product may be implemented by writing computer program code for performing the operations of the present invention in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0260] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0261] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for mining distributed energy resources, characterized in that: include: Collect various distributed energy resource data from distributed energy resource devices; performing data cleaning on the distributed energy resource data; If the data cleaning is completed, then the interactive operation characteristics between the distributed energy resource data are mined in multiple dimensions to construct a multi-dimensional feature matrix; The multidimensional feature matrix of multiple historical moments is input into a preset time series model to generate the power of the next moment in the future.

2. The method according to claim 1, characterized in that The data cleaning of the distributed energy resource data includes: Calculating the mean and standard deviation of the parameter values in the distributed energy resource data; Taking multiple values of the standard deviation of the parameter value in the distributed energy resource data as an abnormality threshold; If the absolute value of the difference between a certain parameter value in the distributed energy resource data and the average value of the parameter value in the distributed energy resource data is greater than the abnormal threshold, the parameter value is determined to be an abnormal value, and the abnormal value is filtered out; The Lagrange interpolation method is used to supplement the missing parameter values in the distributed energy resource data.

3. The method according to claim 1, characterized in that The parameter values of the distributed energy resource data include power value and communication data; The mining of interactive operation characteristics between the distributed energy resource data in multiple dimensions to construct a multi-dimensional feature matrix includes: Mining interactive operation features between the power values in the dimension of time series to obtain a plurality of first indicator values; Using the communication data to mine interactive operation characteristics between the distributed energy resource devices in a spatial correlation dimension to obtain a plurality of second indicator values; Mining the interactive operation characteristics between the power values in the dimension of energy flow to obtain multiple third indicator values; Mining interactive operation characteristics between the distributed energy resource data in the dimension of the external environment to obtain a plurality of fourth indicator values; A multidimensional feature matrix is constructed by using the plurality of first index values, the plurality of second index values, the plurality of third index values, and the plurality of fourth index values.

4. The method according to claim 3, characterized in that The plurality of first indicator values include a volatility of the power value, an autocorrelation coefficient of the power value, an average value of the power value, and a standard deviation of the power value; The fluctuation rate of the power value is: Wherein, V is the fluctuation rate of the power value, P t is the power value at time t, P t+1 is the power value at time t+1, and N is the number of power sampling points; The autocorrelation coefficient of the power value is: Among them, ρ k is the autocorrelation coefficient of the power value, P t is the power value at time t, P t+1 is the power value at time t+1, is the average value of the power value, and N is the number of power sampling points; The plurality of second indicator values include communication weights between the distributed energy resource devices and average topological distances between the distributed energy resource devices; The communication weights between the distributed energy resource devices are: Among them, W ij is the communication weight between the i-th distributed energy resource device and the j-th distributed energy resource device, d ij is the topological distance between the i-th distributed energy resource device and the j-th distributed energy resource device, is the communication delay from the i-th distributed energy resource device to the j-th distributed energy resource device, is the communication delay threshold, α and β are weight coefficients; The average topological distance of the distributed energy resource devices is: in, is the average topological distance of the i-th distributed energy resource device, is an adjacent device of the i-th distributed energy resource device, wherein the adjacent device is another distributed energy resource device whose weight with the i-th distributed energy resource device is greater than a weight threshold, is the number of adjacent devices, D j is the jth adjacent device, d ij is the topological distance between the i-th distributed energy resource device and the j-th distributed energy resource device; The plurality of third index values include similarity of power curves, phase difference, and energy complementarity index; The similarity of the power curves is: Among them, S ij is the similarity of the power curve, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t') is the power value of the jth distributed energy resource device at time t', π is the dynamic time warping alignment path, and min is the function of taking the minimum value; The phase difference is ΔT ij =argmax Δt Corr(P i (t),P j (t+Δt)) Where, ΔT ij is the phase difference between the i-th distributed energy resource device and the j-th distributed energy resource device, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t+Δt) is the power value of the jth distributed energy resource device at time t+Δt; argmax Δt To find the cross-correlation function Corr(P i (t),P j (t+Δt)) The time offset Δt of the maximum output value; The cross-correlation function is: Among them, μ i is the average value of the power value of the i-th distributed energy resource device, μ j is the average power value of the jth distributed energy resource device, σ i is the standard deviation of the power value of the i-th distributed energy resource device, σ j is the standard deviation of the power value of the j-th distributed energy resource device, and T is the number of time sampling points; The energy complementation index is: Among them, C ij is the energy complementarity index between the i-th distributed energy resource device and the j-th distributed energy resource device, P i (t) is the power value of the i-th distributed energy resource device at time t, P j (t) is the power value of the jth distributed energy resource device at time t, min(P i (t),P j (t)) is the i (t) and P j (t) takes the smallest one, where T is the number of time sampling points; The fourth indicator value includes temperature, wind speed, light amplitude, real-time electricity price and predicted electricity price.

5. The method according to claim 4, characterized in that The step of constructing a multidimensional feature matrix using the plurality of first indicator values, the plurality of second indicator values, the plurality of third indicator values, and the plurality of fourth indicator values includes: For the specified distributed energy resource device, constructing the fluctuation rate of the power value, the autocorrelation coefficient of the power value, the average value of the power value and the standard deviation of the power value into a first eigenvector; For the designated distributed energy resource device, averaging the communication weights between the distributed energy resource device and the adjacent device to obtain an average communication weight; Constructing a second eigenvector by combining the average communication weight and the average topological distance; Screening interactive devices for the designated distributed energy resource device; the interactive devices are other distributed energy resource devices whose similarity to the power curve of the distributed energy resource device is greater than a similarity threshold; averaging the similarities of the power curves between the distributed energy resource device and the interactive device to obtain an average similarity; averaging the phase differences between the distributed energy resource device and the interactive device to obtain an average phase difference; averaging the energy complementarity indexes between the distributed energy resource device and the interactive device to obtain an average energy complementarity index; Constructing a third eigenvector by using the average similarity, the average phase difference and the average energy complementarity index; For the designated distributed energy resource device, construct a fourth eigenvector using the temperature, the wind speed, the light amplitude, the real-time electricity price, and the predicted electricity price; Constructing a device feature vector by combining the first feature vector, the second feature vector, the third feature vector, and the fourth feature vector; A plurality of the device feature vectors are stacked to form a multi-dimensional feature matrix.

6. The method according to claim 3, characterized in that The mining of interactive operation characteristics between the distributed energy resource data in multiple dimensions to construct a multi-dimensional feature matrix further includes: Calculate the change between the environmental variables at the next moment and the environmental variables at the previous moment; If the change amount is greater than the change threshold, then calculating the increment of the interactive operation feature at the previous moment; Calculating the sum of the multidimensional feature matrix at the previous moment and the increment of the interactive operation feature as the multidimensional feature matrix at the next moment; If the change amount is less than or equal to the change threshold, the multidimensional feature matrix at the previous moment is set as the multidimensional feature matrix at the next moment.

7. The method according to any one of claims 1 to 6, characterized in that The time series model includes an input layer, a hidden layer and an output layer; the hidden layer is a two-layer long short-term memory network, and the output layer is a fully connected layer; The step of inputting the multidimensional feature matrix of multiple historical moments into a preset time series model to generate the power at the next future moment includes: Receiving the multidimensional feature matrix at multiple historical moments in the input layer; In the first layer of the long short-term memory network, features are extracted from the multidimensional feature matrix at multiple moments to obtain first hidden features at multiple time steps; In the second layer of the long short-term memory network, features are extracted from the first hidden features of multiple time steps to obtain the second hidden features of the last time step; In the fully connected layer, the second hidden feature of the last time step is mapped to the power of the next future moment.

8. The method according to claim 7, characterized in that Also includes: Constructing a training set and a validation set for the time series model; Adjusting configuration parameters of the time series model; the configuration parameters include the number of hidden layers and the number of nodes in each hidden layer; Training the time series model under each of the configuration parameters using adaptive moment estimation according to the training set; If the training is completed, the mean square error of the time series model under each of the configuration parameters is verified based on the verification set; Determine the configuration parameters that minimize the mean square error when applying the time series model.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor. The computer program is executed by the at least one processor to enable the at least one processor to execute the distributed energy resource mining method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the distributed energy resource mining method according to any one of claims 1 to 8 is implemented.

Images