Distributed photovoltaic and load operation scene extraction method and device

Abstract

The invention relates to the technical field of power system stability analysis, and particularly provides a distributed photovoltaic and load operation scene extraction method and device, and the method comprises the steps: enabling an independent variable set of a to-be-extracted scene region to serve as the input of a pre-trained SOM neural network model, and obtaining a scene type of the to-be-extracted scene region; taking the independent variable set of the to-be-extracted scene area as a pre-trained BLS model corresponding to the scene type, and obtaining a distributed photovoltaic and load prediction result of the to-be-extracted scene area; integrating the extended set, and taking the extended set as the input of a pre-trained MBLS probability prediction model to obtain a distributed photovoltaic and load prediction result corresponding to each preset probability quantile; and selecting a distributed photovoltaic and load operation scene of the scene area to be extracted. The extraction result provided by the invention better reflects the characteristics in reality, so that the uncertainty information of the prediction object is effectively described, and reliable support is provided for operation and scheduling of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power system stability analysis, and in particular to a method and device for extracting operating scenarios of distributed photovoltaics and loads. Background Art

[0002] To improve the accuracy and computational efficiency of deterministic forecasts of photovoltaic and load power generation, extensive research has been conducted on these forecasting methods. Some researchers have proposed point forecasting models based on width-based learning systems, achieving relatively satisfactory forecasting results. However, point forecasts only provide a fixed value for photovoltaic output at the time of prediction and fail to capture the randomness and volatility of photovoltaic power generation. Other research has applied quantile regression to the probabilistic forecasting of renewable energy output and power load. The general idea is to calculate a series of quantiles of wind power or photovoltaic output at future points in time to fully describe the fluctuation range and probability distribution of renewable energy power generation. Therefore, width-based learning methods can extend photovoltaic and load forecasting models to probabilistic forecasting.

[0003] However, existing research on probabilistic prediction has not considered the cross-problem of prediction confidence, resulting in an inability to effectively describe the uncertainty information of the predicted object in reality. Summary of the Invention

[0004] In order to overcome the above-mentioned defects, the present invention proposes a method and device for extracting operating scenarios of distributed photovoltaics and loads.

[0005] In a first aspect, a method for extracting operating scenarios of distributed photovoltaics and loads is provided, the method comprising:

[0006] Using the independent variable set of the scene area to be extracted as the input of the pre-trained SOM neural network model, and obtaining the scene type of the scene area to be extracted output by the pre-trained SOM neural network model;

[0007] Using the independent variable set of the scene area to be extracted as a pre-trained BLS model corresponding to the scene type, obtaining a distributed photovoltaic and load forecast result of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type;

[0008] Integrate the preset probability quantiles, the independent variable set of the scene area to be extracted, and the distributed photovoltaic and load forecast results to obtain an extended set, and use the extended set as the input of the pre-trained MBLS probability forecast model to obtain the distributed photovoltaic and load forecast results corresponding to each preset probability quantile output by the pre-trained MBLS probability forecast model;

[0009] The distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles.

[0010] Preferably, the set of independent variables includes: timestamp, meteorological data and historical distributed photovoltaic and load power data.

[0011] Preferably, the training process of the pre-trained SOM neural network model includes:

[0012] Construct training data using a pre-acquired set of historical independent variables;

[0013] The initial SOM neural network model is trained using the training data to obtain the pre-trained SOM neural network model.

[0014] Preferably, the training process of the pre-trained BLS model includes:

[0015] The pre-acquired historical independent variable set is used as the input of the pre-trained SOM neural network model to obtain the scenario type output by the pre-trained SOM neural network model;

[0016] The independent variable set corresponding to each scenario category and its corresponding distributed photovoltaic and load actual results are used as the training data of the initial BLS model corresponding to each scenario category to train, and the pre-trained BLS model corresponding to each scenario category is obtained.

[0017] Preferably, the preset probability quantile is τ k ,k=1,2,…,K, K is the number of preset probability quantiles.

[0018] Furthermore, the extended set is as follows:

[0019]

[0020] In the above formula, τ K is the K-th preset probability quantile, x(T) is the set of independent variables corresponding to time T, The distributed photovoltaic and load forecast results for the scene area to be extracted are output by the pre-trained BLS model corresponding to the scene type at time T.

[0021] Furthermore, the training process of the pre-trained MBLS probability prediction model includes:

[0022] Utilize pre-acquired extended set historical data and its corresponding distributed photovoltaic and load actual results to construct training data;

[0023] The initial MBLS probability prediction model is trained using the training data to obtain the pre-trained MBLS probability prediction model.

[0024] Furthermore, the characteristic node model of the MBLS probability prediction model is as follows:

[0025] F i =φ(X N W Fi +β Fi ),i=1,2,...,N g

[0026] In the above formula, F i is the i-th feature node model of the MBLS probability prediction model, φ is the mapping function, X N is an extended set that does not contain the preset probability quantiles, W Fi is the weight matrix from the input layer to the i-th feature node layer, β Fi is the bias term corresponding to the i-th feature node model, N g is the number of feature map groups.

[0027] Furthermore, the enhanced node model of the MBLS probability prediction model is as follows:

[0028]

[0029] In the above formula, E is the enhancement node of the MBLS probability prediction model, ζ is the activation function, F is the model vector of each feature node of the MBLS probability prediction model, and W E is the weight matrix of feature mapping to the enhanced node layer, X M is an extended set containing only the preset probability quantiles, V M For X M To the weight matrix of the enhanced node layer, β E To enhance the node bias term, is the Nth order of the MBLS probability prediction model g Feature node model.

[0030] Furthermore, the output of the MBLS probability prediction model is as follows:

[0031]

[0032] In the above formula, is the output of the MBLS probability prediction model, r is the Huber norm form of the ramp function, W N is the weight matrix from each feature node model vector F to the output layer of the MBLS probability prediction model, W Mis the weight matrix from the enhanced node model E to the output layer of the MBLS probability prediction model, and β is the output layer bias term.

[0033] Furthermore, the Huber norm of the ramp function is as follows:

[0034]

[0035] In the above formula, π is the function variable, is the Huber function of the combination of l1 and l2 norms, P cap is the rated capacity of the PV system or the upper limit of the capacity of the load line, and a is a hyperparameter.

[0036] Preferably, the distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles, including:

[0037] Clustering the distributed photovoltaic and load forecast results corresponding to each preset probability quantile to obtain clustering results;

[0038] Using the cluster center in the clustering result as a reduced scene;

[0039] The quality of the scenario set is evaluated for each reduction scenario, and the reduction scenario with the highest score is used as the distributed photovoltaic and load operation scenario in the scenario area to be extracted.

[0040] Furthermore, the evaluation indicators for the scene set quality evaluation of each reduced scene include: average number of fluctuations, average fluctuation amplitude, accumulated power and maximum power.

[0041] In a second aspect, a distributed photovoltaic and load operation scene extraction device is provided, the distributed photovoltaic and load operation scene extraction device comprising:

[0042] A first analysis module is configured to use the independent variable set of the scene area to be extracted as an input of a pre-trained SOM neural network model to obtain a scene type of the scene area to be extracted output by the pre-trained SOM neural network model;

[0043] A second analysis module is configured to use the independent variable set of the scene area to be extracted as a pre-trained BLS model corresponding to the scene type, and obtain a distributed photovoltaic and load forecast result of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type;

[0044] The third analysis module is used to integrate the preset probability quantiles, the independent variable set of the scene area to be extracted, and the distributed photovoltaic and load forecast results to obtain an extended set, and use the extended set as the input of the pre-trained MBLS probability prediction model to obtain the distributed photovoltaic and load forecast results corresponding to each preset probability quantile output by the pre-trained MBLS probability prediction model;

[0045] The fourth analysis module is configured to select a distributed photovoltaic and load operation scenario in a scene area to be extracted based on the distributed photovoltaic and load forecast results corresponding to each preset probability quantile.

[0046] Preferably, the set of independent variables includes: timestamp, meteorological data and historical distributed photovoltaic and load power data.

[0047] Preferably, the training process of the pre-trained SOM neural network model includes:

[0048] Construct training data using a pre-acquired set of historical independent variables;

[0049] The initial SOM neural network model is trained using the training data to obtain the pre-trained SOM neural network model.

[0050] Preferably, the training process of the pre-trained BLS model includes:

[0051] The pre-acquired historical independent variable set is used as the input of the pre-trained SOM neural network model to obtain the scenario type output by the pre-trained SOM neural network model;

[0052] The independent variable set corresponding to each scenario category and its corresponding distributed photovoltaic and load actual results are used as the training data of the initial BLS model corresponding to each scenario category to train, and the pre-trained BLS model corresponding to each scenario category is obtained.

[0053] Preferably, the preset probability quantile is τ k ,k=1,2,…,K, K is the number of preset probability quantiles.

[0054] Furthermore, the extended set is as follows:

[0055]

[0056] In the above formula, τ K is the K-th preset probability quantile, x(T) is the set of independent variables corresponding to time T, The distributed photovoltaic and load forecast results for the scene area to be extracted are output by the pre-trained BLS model corresponding to the scene type at time T.

[0057] Furthermore, the training process of the pre-trained MBLS probability prediction model includes:

[0058] Utilize pre-acquired extended set historical data and its corresponding distributed photovoltaic and load actual results to construct training data;

[0059] The initial MBLS probability prediction model is trained using the training data to obtain the pre-trained MBLS probability prediction model.

[0060] Furthermore, the characteristic node model of the MBLS probability prediction model is as follows:

[0061] F i =φ(X N W Fi +β Fi ),i=1,2,...,N g

[0062] In the above formula, F i is the i-th feature node model of the MBLS probability prediction model, φ is the mapping function, X N is an extended set that does not contain the preset probability quantiles, W Fi is the weight matrix from the input layer to the i-th feature node layer, β Fi is the bias term corresponding to the i-th feature node model, N g is the number of feature map groups.

[0063] Furthermore, the enhanced node model of the MBLS probability prediction model is as follows:

[0064]

[0065] In the above formula, E is the enhancement node of the MBLS probability prediction model, ζ is the activation function, F is the model vector of each feature node of the MBLS probability prediction model, and W E is the weight matrix of feature mapping to the enhanced node layer, X M is an extended set containing only the preset probability quantiles, V M For X M To the weight matrix of the enhanced node layer, β E To enhance the node bias term, is the Nth order of the MBLS probability prediction model g Feature node model.

[0066] Furthermore, the output of the MBLS probability prediction model is as follows:

[0067]

[0068] In the above formula, is the output of the MBLS probability prediction model, r is the Huber norm form of the ramp function, W N is the weight matrix from each feature node model vector F to the output layer of the MBLS probability prediction model, W M is the weight matrix from the enhanced node model E to the output layer of the MBLS probability prediction model, and β is the output layer bias term.

[0069] Furthermore, the Huber norm of the ramp function is as follows:

[0070]

[0071] In the above formula, π is the function variable, is the Huber function of the combination of l1 and l2 norms, P cap is the rated capacity of the PV system or the upper limit of the capacity of the load line, and a is a hyperparameter.

[0072] Preferably, the distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles, including:

[0073] Clustering the distributed photovoltaic and load forecast results corresponding to each preset probability quantile to obtain clustering results;

[0074] Using the cluster center in the clustering result as a reduced scene;

[0075] The quality of the scenario set is evaluated for each reduction scenario, and the reduction scenario with the highest score is used as the distributed photovoltaic and load operation scenario in the scenario area to be extracted.

[0076] Furthermore, the evaluation indicators for the scene set quality evaluation of each reduced scene include: average number of fluctuations, average fluctuation amplitude, accumulated power and maximum power.

[0077] In a third aspect, a computer device is provided, comprising: one or more processors;

[0078] The processor is configured to execute one or more programs;

[0079] When the one or more programs are executed by the one or more processors, the method for extracting operating scenarios of distributed photovoltaics and loads is implemented.

[0080] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed, the method for extracting operating scenarios of distributed photovoltaics and loads is implemented.

[0081] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0082] The present invention provides a method and device for extracting operating scenarios of distributed photovoltaics and loads, comprising: using the independent variable set of the scene area to be extracted as the input of a pre-trained SOM neural network model to obtain the scene type of the scene area to be extracted output by the pre-trained SOM neural network model; using the independent variable set of the scene area to be extracted as the pre-trained BLS model corresponding to the scene type to obtain the distributed photovoltaic and load prediction results of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type; integrating preset probability quantiles, the independent variable set of the scene area to be extracted and the distributed photovoltaic and load prediction results to obtain an extended set, and using the extended set as the input of a pre-trained MBLS probability prediction model to obtain the distributed photovoltaic and load prediction results corresponding to each preset probability quantile output by the pre-trained MBLS probability prediction model; and selecting the distributed photovoltaic and load operating scenarios of the scene area to be extracted from the distributed photovoltaic and load prediction results corresponding to each preset probability quantile. The technical solution provided by the present invention solves the quantile crossing problem by training the MBLS probability prediction model through quantile regression, so that the operation scenario extraction results better reflect the characteristics in reality, thereby effectively characterizing the uncertainty information of the prediction object, and providing more reliable information support for power system planning, operation, stability analysis and control, and trading, which is of great significance to improving the overall energy efficiency of the system and promoting the consumption of new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 This is a flow chart of the main steps of the method for extracting operating scenarios of distributed photovoltaics and loads according to an embodiment of the present invention. DETAILED DESCRIPTION

[0084] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0085] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments 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 shall fall within the scope of protection of the present invention.

[0086] As disclosed in the background technology, in order to improve the accuracy and computational efficiency of deterministic predictions of photovoltaic and load power generation, existing research has conducted extensive research on photovoltaic and load prediction methods. Some scholars have proposed a point prediction model based on a width learning system and achieved relatively ideal prediction results. However, point prediction only gives a fixed value of photovoltaic output at the time to be predicted, and cannot reflect the randomness and volatility of photovoltaic power generation. Other research work applies quantile regression to the probabilistic prediction of renewable energy output and power load. The overall idea is to calculate a series of quantiles of wind power or photovoltaic output at future moments to fully describe the fluctuation range and probability distribution of renewable energy power generation. Therefore, the method based on width learning can expand the photovoltaic and load prediction model to probabilistic prediction.

[0087] However, existing research on probabilistic prediction has not considered the cross-problem of prediction confidence, resulting in an inability to effectively describe the uncertainty information of the predicted object in reality.

[0088] In order to improve the above problems, the present invention provides a method and device for extracting operating scenarios of distributed photovoltaics and loads, including: using the independent variable set of the scene area to be extracted as the input of a pre-trained SOM neural network model to obtain the scene type of the scene area to be extracted output by the pre-trained SOM neural network model; using the independent variable set of the scene area to be extracted as the pre-trained BLS model corresponding to the scene type to obtain the distributed photovoltaic and load prediction results of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type; integrating the preset probability quantiles, the independent variable set of the scene area to be extracted and the distributed photovoltaic and load prediction results to obtain an extended set, and using the extended set as the input of a pre-trained MBLS probability prediction model to obtain the distributed photovoltaic and load prediction results corresponding to each preset probability quantile output by the pre-trained MBLS probability prediction model; selecting the distributed photovoltaic and load operating scenarios of the scene area to be extracted from the distributed photovoltaic and load prediction results corresponding to each preset probability quantile. The technical solution provided by the present invention solves the quantile crossing problem by training the MBLS probability prediction model through quantile regression, so that the operation scenario extraction results better reflect the characteristics in reality, thereby effectively characterizing the uncertainty information of the prediction object, and providing more reliable information support for power system planning, operation, stability analysis and control, and trading, which is of great significance to improving the overall energy efficiency of the system and promoting the consumption of new energy.

[0089] The above scheme is described in detail below.

[0090] Example 1

[0091] See attached Figure 1 , Figure 1FIG. 1 is a flow chart showing the main steps of a method for extracting operating scenarios of distributed photovoltaic and loads according to an embodiment of the present invention. Figure 1 As shown, the method for extracting the operating scenarios of distributed photovoltaic and load in the embodiment of the present invention mainly includes the following steps:

[0092] Step S101: using the independent variable set of the scene area to be extracted as the input of the pre-trained SOM neural network model, and obtaining the scene type of the scene area to be extracted output by the pre-trained SOM neural network model;

[0093] Step S102: using the independent variable set of the scene area to be extracted as a pre-trained BLS model corresponding to the scene type, and obtaining a distributed photovoltaic and load forecast result of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type;

[0094] Step S103: Integrate the preset probability quantiles, the independent variable set of the scene area to be extracted, and the distributed photovoltaic and load forecast results to obtain an extended set, and use the extended set as the input of the pre-trained MBLS probability prediction model to obtain the distributed photovoltaic and load forecast results corresponding to each preset probability quantile output by the pre-trained MBLS probability prediction model;

[0095] Step S104: selecting a distributed photovoltaic and load operation scenario in a scene area to be extracted from the distributed photovoltaic and load forecast results corresponding to each preset probability quantile.

[0096] In this embodiment, the independent variable set includes: timestamp, meteorological data, and historical distributed photovoltaic and load power data. The meteorological data can be meteorological data such as solar irradiance and temperature.

[0097] In this embodiment, the training process of the pre-trained SOM neural network model includes:

[0098] Construct training data using a pre-acquired set of historical independent variables;

[0099] The initial SOM neural network model is trained using the training data to obtain the pre-trained SOM neural network model.

[0100] During the training process of the pre-trained SOM neural network model, relative quantization error (Qe) and topological error (Te) are introduced to represent the quality of SOM clustering. The final classification is automatically selected according to the maximum CHI (Calinski-Harabasz Index), and the category to which each time t belongs is output and recorded.

[0101] In this embodiment, the training process of the pre-trained BLS model includes:

[0102] The pre-acquired historical independent variable set is used as the input of the pre-trained SOM neural network model to obtain the scenario type output by the pre-trained SOM neural network model;

[0103] The independent variable set corresponding to each scenario category and its corresponding distributed photovoltaic and load actual results are used as the training data of the initial BLS model corresponding to each scenario category to train, and the pre-trained BLS model corresponding to each scenario category is obtained.

[0104] In this embodiment, the preset probability quantile is τ k ,k=1,2,…,K, K is the number of preset probability quantiles.

[0105] In one embodiment, the extended set is as follows:

[0106]

[0107] In the above formula, τ K is the K-th preset probability quantile, x(T) is the set of independent variables corresponding to time T, The distributed photovoltaic and load forecast results for the scene area to be extracted are output by the pre-trained BLS model corresponding to the scene type at time T.

[0108] In one embodiment, the training process of the pre-trained MBLS probability prediction model includes:

[0109] Utilize pre-acquired extended set historical data and its corresponding distributed photovoltaic and load actual results to construct training data;

[0110] The initial MBLS probability prediction model is trained using the training data to obtain the pre-trained MBLS probability prediction model.

[0111] In one embodiment, the characteristic node model of the MBLS probability prediction model is as follows:

[0112] F i =φ(X N W Fi +β Fi ),i=1,2,...,N g

[0113] In the above formula, F i is the i-th feature node model of the MBLS probability prediction model, φ is the mapping function, X N is an extended set that does not contain the preset probability quantiles, W Fiis the weight matrix from the input layer to the i-th feature node layer, β Fi is the bias term corresponding to the i-th feature node model, N g is the number of feature map groups.

[0114] In one embodiment, the enhanced node model of the MBLS probability prediction model is as follows:

[0115]

[0116] In the above formula, E is the enhancement node of the MBLS probability prediction model, ζ is the activation function, F is the model vector of each feature node of the MBLS probability prediction model, and W E is the weight matrix of feature mapping to the enhanced node layer, X M is an extended set containing only the preset probability quantiles, V M For X M To the weight matrix of the enhanced node layer, β E To enhance the node bias term, is the Nth order of the MBLS probability prediction model g Feature node model.

[0117] In one embodiment, the output of the MBLS probability prediction model is as follows:

[0118]

[0119] In the above formula, is the output of the MBLS probability prediction model, r is the Huber norm form of the ramp function, W N is the weight matrix from each feature node model vector F to the output layer of the MBLS probability prediction model, W M is the weight matrix from the enhanced node model E to the output layer of the MBLS probability prediction model, and β is the output layer bias term.

[0120] In one embodiment, the Huber norm of the ramp function is as follows:

[0121]

[0122] In the above formula, π is the function variable, is the Huber function of the combination of l1 and l2 norms, P cap is the rated capacity of the PV system or the upper limit of the capacity of the load line, and a is a hyperparameter.

[0123] The Adam optimization algorithm is used to solve the training MBLS parameters θ={V M ,W M ,WN ,β E ,β}. Set the learning rate α of the Adam optimization algorithm to 0.001, the exponential decay rate β1 of the first-order moment estimate to 0.9, the exponential decay rate β2 of the second-order moment estimate to 0.999, and the constant ε to 10 -8 The training set is used to iteratively train the MBLS parameters θ, and the test set is used to verify the effectiveness of the obtained MBLS model parameters.

[0124] In this embodiment, the distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles, including:

[0125] Clustering the distributed photovoltaic and load forecast results corresponding to each preset probability quantile to obtain clustering results;

[0126] Using the cluster center in the clustering result as a reduced scene;

[0127] The quality of the scenario set is evaluated for each reduction scenario, and the reduction scenario with the highest score is used as the distributed photovoltaic and load operation scenario in the scenario area to be extracted.

[0128] In one embodiment, the evaluation indicators for the scene set quality evaluation of each reduced scene include: average number of fluctuations, average fluctuation amplitude, accumulated power and maximum power.

[0129] Example 2

[0130] Based on the same inventive concept, the present invention also provides a distributed photovoltaic and load operation scene extraction device, the distributed photovoltaic and load operation scene extraction device comprising:

[0131] A first analysis module is configured to use the independent variable set of the scene area to be extracted as an input of a pre-trained SOM neural network model to obtain a scene type of the scene area to be extracted output by the pre-trained SOM neural network model;

[0132] A second analysis module is configured to use the independent variable set of the scene area to be extracted as a pre-trained BLS model corresponding to the scene type, and obtain a distributed photovoltaic and load forecast result of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type;

[0133] The third analysis module is used to integrate the preset probability quantiles, the independent variable set of the scene area to be extracted, and the distributed photovoltaic and load forecast results to obtain an extended set, and use the extended set as the input of the pre-trained MBLS probability prediction model to obtain the distributed photovoltaic and load forecast results corresponding to each preset probability quantile output by the pre-trained MBLS probability prediction model;

[0134] The fourth analysis module is configured to select a distributed photovoltaic and load operation scenario in a scene area to be extracted based on the distributed photovoltaic and load forecast results corresponding to each preset probability quantile.

[0135] Preferably, the set of independent variables includes: timestamp, meteorological data and historical distributed photovoltaic and load power data.

[0136] Preferably, the training process of the pre-trained SOM neural network model includes:

[0137] Construct training data using a pre-acquired set of historical independent variables;

[0138] The initial SOM neural network model is trained using the training data to obtain the pre-trained SOM neural network model.

[0139] Preferably, the training process of the pre-trained BLS model includes:

[0140] The pre-acquired historical independent variable set is used as the input of the pre-trained SOM neural network model to obtain the scenario type output by the pre-trained SOM neural network model;

[0141] The independent variable set corresponding to each scenario category and its corresponding distributed photovoltaic and load actual results are used as the training data of the initial BLS model corresponding to each scenario category to train, and the pre-trained BLS model corresponding to each scenario category is obtained.

[0142] Preferably, the preset probability quantile is τ k , k=1,2,…,K, K is the number of preset probability quantiles.

[0143] Furthermore, the extended set is as follows:

[0144]

[0145] In the above formula, τ K is the K-th preset probability quantile, x(T) is the set of independent variables corresponding to time T, The distributed photovoltaic and load forecast results for the scene area to be extracted are output by the pre-trained BLS model corresponding to the scene type at time T.

[0146] Furthermore, the training process of the pre-trained MBLS probability prediction model includes:

[0147] Utilize pre-acquired extended set historical data and its corresponding distributed photovoltaic and load actual results to construct training data;

[0148] The initial MBLS probability prediction model is trained using the training data to obtain the pre-trained MBLS probability prediction model.

[0149] Furthermore, the characteristic node model of the MBLS probability prediction model is as follows:

[0150] F i =φ(X N W Fi +β Fi ), i=1,2,...,N g

[0151] In the above formula, F i is the i-th feature node model of the MBLS probability prediction model, φ is the mapping function, X N is an extended set that does not contain the preset probability quantiles, W Fi is the weight matrix from the input layer to the i-th feature node layer, β Fi is the bias term corresponding to the i-th feature node model, N g is the number of feature map groups.

[0152] Furthermore, the enhanced node model of the MBLS probability prediction model is as follows:

[0153]

[0154] In the above formula, E is the enhancement node of the MBLS probability prediction model, ζ is the activation function, F is the model vector of each feature node of the MBLS probability prediction model, and W E is the weight matrix of feature mapping to the enhanced node layer, X M is an extended set containing only the preset probability quantiles, V M For X M To the weight matrix of the enhanced node layer, β E To enhance the node bias term, is the Nth order of the MBLS probability prediction model g Feature node model.

[0155] Furthermore, the output of the MBLS probability prediction model is as follows:

[0156]

[0157] In the above formula, is the output of the MBLS probability prediction model, r is the Huber norm form of the ramp function, W N is the weight matrix from each feature node model vector F to the output layer of the MBLS probability prediction model, W Mis the weight matrix from the enhanced node model E to the output layer of the MBLS probability prediction model, and β is the output layer bias term.

[0158] Furthermore, the Huber norm of the ramp function is as follows:

[0159]

[0160] In the above formula, π is the function variable, is the Huber function of the combination of l1 and l2 norms, P cap is the rated capacity of the PV system or the upper limit of the capacity of the load line, and a is a hyperparameter.

[0161] Preferably, the distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles, including:

[0162] Clustering the distributed photovoltaic and load forecast results corresponding to each preset probability quantile to obtain clustering results;

[0163] Using the cluster center in the clustering result as a reduced scene;

[0164] The quality of the scenario set is evaluated for each reduction scenario, and the reduction scenario with the highest score is used as the distributed photovoltaic and load operation scenario in the scenario area to be extracted.

[0165] Furthermore, the evaluation indicators for the scene set quality evaluation of each reduced scene include: average number of fluctuations, average fluctuation amplitude, accumulated power and maximum power.

[0166] Example 3

[0167] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a distributed photovoltaic and load operation scenario extraction method in the above embodiment.

[0168] Example 4

[0169] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a method for extracting operating scenarios of distributed photovoltaics and loads in the above embodiment.

[0170] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0171] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0172] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0173] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for extracting operating scenarios of distributed photovoltaics and loads, characterized in that: The method comprises: Using the independent variable set of the scene area to be extracted as the input of the pre-trained SOM neural network model, and obtaining the scene type of the scene area to be extracted output by the pre-trained SOM neural network model; Using the independent variable set of the scene area to be extracted as a pre-trained BLS model corresponding to the scene type, obtaining a distributed photovoltaic and load forecast result of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type; Integrate the preset probability quantiles, the independent variable set of the scene area to be extracted, and the distributed photovoltaic and load forecast results to obtain an extended set, and use the extended set as the input of the pre-trained MBLS probability forecast model to obtain the distributed photovoltaic and load forecast results corresponding to each preset probability quantile output by the pre-trained MBLS probability forecast model; The distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles.

2. The method according to claim 1, wherein The independent variable set includes: timestamp, meteorological data and historical distributed photovoltaic and load power data.

3. The method according to claim 1, wherein The training process of the pre-trained SOM neural network model includes: Construct training data using a pre-acquired set of historical independent variables; The initial SOM neural network model is trained using the training data to obtain the pre-trained SOM neural network model.

4. The method according to claim 1, wherein The training process of the pre-trained BLS model includes: The pre-acquired historical independent variable set is used as the input of the pre-trained SOM neural network model to obtain the scenario type output by the pre-trained SOM neural network model; The independent variable set corresponding to each scenario category and its corresponding distributed photovoltaic and load actual results are used as the training data of the initial BLS model corresponding to each scenario category to train, and the pre-trained BLS model corresponding to each scenario category is obtained.

5. The method according to claim 1, wherein The preset probability quantile is τ k ,k=1,2,…,K, K is the number of preset probability quantiles.

6. The method according to claim 5, wherein The extension set is as follows: In the above formula, τ K is the K-th preset probability quantile, x(T) is the set of independent variables corresponding to time T, The distributed photovoltaic and load forecast results for the scene area to be extracted are output by the pre-trained BLS model corresponding to the scene type at time T.

7. The method according to claim 6, wherein The training process of the pre-trained MBLS probability prediction model includes: Utilize pre-acquired extended set historical data and its corresponding distributed photovoltaic and load actual results to construct training data; The initial MBLS probability prediction model is trained using the training data to obtain the pre-trained MBLS probability prediction model.

8. The method according to claim 7, wherein The characteristic node model of the MBLS probability prediction model is as follows: F i =φ(X N W Fi +b Fi ),i=1,2,...,N g In the above formula, F i is the i-th feature node model of the MBLS probability prediction model, φ is the mapping function, X N is an extended set that does not contain the preset probability quantiles, W Fi is the weight matrix from the input layer to the i-th feature node layer, β Fi is the bias term corresponding to the i-th feature node model, N g is the number of feature map groups.

9. The method according to claim 8, wherein The enhanced node model of the MBLS probability prediction model is as follows: In the above formula, E is the enhancement node of the MBLS probability prediction model, ζ is the activation function, F is the model vector of each feature node of the MBLS probability prediction model, and W E is the weight matrix of feature mapping to the enhanced node layer, X M is an extended set containing only the preset probability quantiles, V M For X M To the weight matrix of the enhanced node layer, β E To enhance the node bias term, is the Nth order of the MBLS probability prediction model g Feature node model.

10. The method according to claim 9, wherein The output of the MBLS probability prediction model is as follows: In the above formula, is the output of the MBLS probability prediction model, r is the Huber norm form of the ramp function, W N is the weight matrix from each feature node model vector F to the output layer of the MBLS probability prediction model, W M is the weight matrix from the enhanced node model E to the output layer of the MBLS probability prediction model, and β is the output layer bias term.

11. The method according to claim 10, wherein The Huber norm form of the ramp function is as follows: In the above formula, π is the function variable, is the Huber function of the combination of l1 and l2 norms, P cap is the rated capacity of the PV system or the upper limit of the capacity of the load line, and a is a hyperparameter.

12. The method according to claim 1, wherein The distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles, including: Clustering the distributed photovoltaic and load forecast results corresponding to each preset probability quantile to obtain clustering results; Using the cluster center in the clustering result as a reduced scene; The quality of the scenario set is evaluated for each reduction scenario, and the reduction scenario with the highest score is used as the distributed photovoltaic and load operation scenario in the scenario area to be extracted.

13. The method according to claim 12, wherein: The evaluation indicators for the scene set quality evaluation of each reduced scene include: average number of fluctuations, average fluctuation amplitude, accumulated power and maximum power.

14. A distributed photovoltaic and load operation scene extraction device, characterized in that: The device comprises: A first analysis module is configured to use the independent variable set of the scene area to be extracted as an input of a pre-trained SOM neural network model to obtain a scene type of the scene area to be extracted output by the pre-trained SOM neural network model; A second analysis module is configured to use the independent variable set of the scene area to be extracted as a pre-trained BLS model corresponding to the scene type, and obtain a distributed photovoltaic and load forecast result of the scene area to be extracted output by the pre-trained BLS model corresponding to the scene type; The third analysis module is used to integrate the preset probability quantiles, the independent variable set of the scene area to be extracted, and the distributed photovoltaic and load forecast results to obtain an extended set, and use the extended set as the input of the pre-trained MBLS probability prediction model to obtain the distributed photovoltaic and load forecast results corresponding to each preset probability quantile output by the pre-trained MBLS probability prediction model; The fourth analysis module is configured to select a distributed photovoltaic and load operation scenario in a scene area to be extracted based on the distributed photovoltaic and load forecast results corresponding to each preset probability quantile.

15. The device according to claim 14, wherein The independent variable set includes: timestamp, meteorological data and historical distributed photovoltaic and load power data.

16. The device according to claim 14, wherein The training process of the pre-trained SOM neural network model includes: Construct training data using a pre-acquired set of historical independent variables; The initial SOM neural network model is trained using the training data to obtain the pre-trained SOM neural network model.

17. The device according to claim 14, wherein The training process of the pre-trained BLS model includes: The pre-acquired historical independent variable set is used as the input of the pre-trained SOM neural network model to obtain the scenario type output by the pre-trained SOM neural network model; The independent variable set corresponding to each scenario category and its corresponding distributed photovoltaic and load actual results are used as the training data of the initial BLS model corresponding to each scenario category to train, and the pre-trained BLS model corresponding to each scenario category is obtained.

18. The device according to claim 14, wherein The preset probability quantile is τ k ,k=1,2,…,K, K is the number of preset probability quantiles.

19. The device according to claim 18, wherein The extension set is as follows: In the above formula, τ K is the K-th preset probability quantile, x(T) is the set of independent variables corresponding to time T, The distributed photovoltaic and load forecast results for the scene area to be extracted are output by the pre-trained BLS model corresponding to the scene type at time T.

20. The device according to claim 19, wherein The training process of the pre-trained MBLS probability prediction model includes: Utilize pre-acquired extended set historical data and its corresponding distributed photovoltaic and load actual results to construct training data; The initial MBLS probability prediction model is trained using the training data to obtain the pre-trained MBLS probability prediction model.

21. The device according to claim 20, characterized in that The characteristic node model of the MBLS probability prediction model is as follows: F i =φ(X N W Fi +b Fi ),i=1,2,...,N g In the above formula, F i is the i-th feature node model of the MBLS probability prediction model, φ is the mapping function, X N is an extended set that does not contain the preset probability quantiles, W Fi is the weight matrix from the input layer to the i-th feature node layer, β Fi is the bias term corresponding to the i-th feature node model, N g is the number of feature map groups.

22. The device according to claim 21, wherein The enhanced node model of the MBLS probability prediction model is as follows: In the above formula, E is the enhancement node of the MBLS probability prediction model, ζ is the activation function, F is the model vector of each feature node of the MBLS probability prediction model, and W E is the weight matrix of feature mapping to the enhanced node layer, X M is an extended set containing only the preset probability quantiles, V M For X M To the weight matrix of the enhanced node layer, β E To enhance the node bias term, is the Nth order of the MBLS probability prediction model g Feature node model.

23. The device according to claim 22, wherein The output of the MBLS probability prediction model is as follows: In the above formula, is the output of the MBLS probability prediction model, r is the Huber norm form of the ramp function, W N is the weight matrix from each feature node model vector F to the output layer of the MBLS probability prediction model, W M is the weight matrix from the enhanced node model E to the output layer of the MBLS probability prediction model, and β is the output layer bias term.

24. The device according to claim 23, wherein The Huber norm form of the ramp function is as follows: In the above formula, π is the function variable, is the Huber function of the combination of l1 and l2 norms, P cap is the rated capacity of the PV system or the upper limit of the capacity of the load line, and a is a hyperparameter.

25. The device according to claim 14, wherein The distributed photovoltaic and load operation scenarios of the scene area to be extracted are selected from the distributed photovoltaic and load forecast results corresponding to the preset probability quantiles, including: Clustering the distributed photovoltaic and load forecast results corresponding to each preset probability quantile to obtain clustering results; Using the cluster center in the clustering result as a reduced scene; The quality of the scenario set is evaluated for each reduction scenario, and the reduction scenario with the highest score is used as the distributed photovoltaic and load operation scenario in the scenario area to be extracted.

26. The device according to claim 25, characterized in that The evaluation indicators for the scene set quality evaluation of each reduced scene include: average number of fluctuations, average fluctuation amplitude, accumulated power and maximum power.

27. A computer device, characterized in that: include: one or more processors; The processor is configured to execute one or more programs; When the one or more programs are executed by the one or more processors, the method for extracting operating scenarios of distributed photovoltaics and loads as described in any one of claims 1 to 13 is implemented.

28. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed, the method for extracting operating scenarios of distributed photovoltaics and loads as described in any one of claims 1 to 13 is implemented.

Images