Uncertainty estimation method and system for day-ahead scheduling of integrated energy bases

Abstract

This application proposes a method and system for estimating the uncertainty of the day-ahead scheduling plan of a comprehensive energy base. The method includes: obtaining the first prediction data for each moment in each historical similar day corresponding to the day to be scheduled and the uncertainty of each historical similar day to form a training set; using the training set to train a deep neural network model to obtain a trained uncertainty estimation model; obtaining the first prediction data for each moment in each moment of the day to be scheduled, and inputting the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled; wherein, the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value. The technical solution proposed in this application evaluates the uncertainty of the day-ahead scheduling plan in advance to assist scheduling decisions.

Description

Technical Field

[0001] The present application relates to the field of scheduling technology, and in particular to a method and system for estimating uncertainty of a day-ahead scheduling plan for an integrated energy base. Background Art

[0002] To achieve the "dual carbon" goals and build a new power system, large-scale integrated energy bases are a key focus of new energy development. These bases can integrate wind, solar, and thermal power and transmit it via direct current (DC). The operation of these bases requires pre-planned dispatching of wind, solar, and thermal power to avoid insufficient reserve capacity or large-scale curtailment of wind and solar power due to improper configuration. Pre-planned dispatching plans, such as day-ahead dispatching plans, can differ from actual dispatch results on the day of dispatch. Currently, there is no effective method to estimate the potential magnitude of these discrepancies during planning to aid decision-making. Summary of the Invention

[0003] This application provides a method and system for estimating the uncertainty of the day-ahead scheduling plan of an integrated energy base, in order to at least solve the technical problem that there is no effective method to estimate how large the difference may be when making a plan.

[0004] A first embodiment of the present application proposes a method for estimating uncertainty of a day-ahead scheduling plan of an integrated energy base, the method comprising:

[0005] Obtain the first prediction data at each time point on each historical similar day corresponding to the scheduled day and the uncertainty of each historical similar day to form a training set;

[0006] Using the training set to train a deep neural network model to obtain a trained uncertainty estimation model;

[0007] Obtaining first prediction data for each time of the day to be scheduled, and inputting the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled;

[0008] Among them, the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value.

[0009] Preferably, the calculation formula for the uncertainty of each historical similar day is as follows:

[0010]

[0011] Where S i is the uncertainty of the i-th historical similar day, R Pij is the actual photovoltaic output ratio at time j on the i-th historical similar day, is the planned photovoltaic output ratio at time j on the i-th historical similar day, R Wij is the actual wind power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, R Tij is the actual thermal power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, and N is the total number of times on a similar day.

[0012] Preferably, the deep neural network model includes: an input layer, a convolutional neural network (CNN) layer, a long short-term memory (LSTM) artificial neural network (ANN) layer, an attention layer, and an output layer.

[0013] Furthermore, the deep neural network model is trained using the training set to obtain a trained uncertainty estimation model, including:

[0014] The first prediction data at each moment on each historically similar day in the training set is used as the input of the deep neural network model, and the uncertainty of each historically similar day in the training set is used as the output of the deep neural network model to train the model to obtain the trained uncertainty estimation model.

[0015] Furthermore, the training process of the uncertainty estimation model includes:

[0016] Inputting the first prediction data of each time on each historical similar day in the training set into the input layer, and obtaining the input vector corresponding to each of the prediction data through the input layer;

[0017] Inputting the input vector into the CNN layer, and performing feature extraction on the input vector to screen out a target feature vector;

[0018] Obtain the target feature vectors, and input the target feature vectors into the LSTM layer to obtain first output vectors corresponding to the target feature vectors;

[0019] Inputting the first output vectors corresponding to the target feature vectors into the attention layer, and filtering the first output vectors according to the attention weight parameter values ​​of the first output vectors in the attention layer to obtain second output vectors;

[0020] Inputting the second output vector into the output layer to determine the uncertainty prediction value of each historical similar day, and outputting the uncertainty prediction value of each historical similar day;

[0021] According to the difference between the uncertainty of similar historical days in the training set and the predicted value, the model parameters of the deep neural network model are adjusted to obtain the uncertainty estimation model.

[0022] Furthermore, the CNN layer includes a convolution layer and a dropout layer, inputs the input vector into the CNN layer, and performs feature extraction on the input vector to filter out a target feature vector, including:

[0023] Inputting the input vector into the convolution layer to obtain multiple feature vectors of the input vector extracted by the convolution layer;

[0024] The plurality of feature vectors of the input vector are input to the dropout layer to filter out a target feature vector from the plurality of features.

[0025] Furthermore, after inputting the second output vector into the output layer to determine the uncertainty prediction value of each historically similar day and outputting the uncertainty prediction value of each historically similar day, the method further includes:

[0026] The output uncertainty prediction values ​​of each of the historical similar days are subjected to denormalization processing.

[0027] A second embodiment of the present application provides an uncertainty estimation system for a day-ahead scheduling plan of an integrated energy base, the system comprising:

[0028] An acquisition module is used to obtain the first prediction data at each time point on each historical similar day corresponding to the scheduled day and the uncertainty of each historical similar day to form a training set;

[0029] A training module, configured to train a deep neural network model using the training set to obtain a trained uncertainty estimation model;

[0030] an estimation module, configured to obtain first prediction data for each moment of the day to be scheduled, and input the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled;

[0031] Among them, the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value.

[0032] The third aspect embodiment of the present application proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, it implements the uncertainty estimation method of the day-ahead scheduling plan of the integrated energy base as described in the first aspect embodiment.

[0033] The fourth embodiment of the present application proposes a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the uncertainty estimation method of the day-ahead scheduling plan of the integrated energy base as described in the first embodiment is implemented.

[0034] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0035] The present application proposes a method and system for estimating the uncertainty of the day-ahead scheduling plan of a comprehensive energy base, wherein the method comprises: obtaining the first prediction data for each moment in each historical similar day corresponding to the day to be scheduled and the uncertainty of each historical similar day to form a training set; using the training set to train a deep neural network model to obtain a trained uncertainty estimation model; obtaining the first prediction data for each moment in each moment of the day to be scheduled, and inputting the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled; wherein the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value. The technical solution proposed in the present application estimates the uncertainty of the day-ahead scheduling plan in advance to assist scheduling decisions.

[0036] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0038] Figure 1 This is a flow chart of a method for estimating uncertainty of a day-ahead scheduling plan for an integrated energy base provided according to one embodiment of the present application;

[0039] Figure 2 A structural diagram of a schematic diagram of a deep neural network model provided according to one embodiment of the present application;

[0040] Figure 3 A schematic diagram of a dropout layer structure provided according to an embodiment of the present application;

[0041] Figure 4 This is a structural diagram of an uncertainty estimation system for a comprehensive energy base day-ahead scheduling plan provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0043] The present application proposes a method and system for estimating the uncertainty of the day-ahead scheduling plan of a comprehensive energy base, wherein the method includes: obtaining the first prediction data for each moment in each historical similar day corresponding to the day to be scheduled and the uncertainty of each historical similar day to form a training set; using the training set to train a deep neural network model to obtain a trained uncertainty estimation model; obtaining the first prediction data for each moment in each moment of the day to be scheduled, and inputting the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled; wherein the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value. The technical solution proposed in the present application estimates the uncertainty of the day-ahead scheduling plan in advance to assist scheduling decisions.

[0044] The following describes the uncertainty estimation method and system for the integrated energy base day-ahead scheduling plan according to an embodiment of the present application with reference to the accompanying drawings.

[0045] Example 1

[0046] Figure 1 This is a flow chart of a method for estimating uncertainty of a day-ahead scheduling plan for an integrated energy base provided according to one embodiment of the present application. Figure 1 As shown, the method includes:

[0047] Step 1: Obtain the first prediction data at each time point on each historical similar day corresponding to the scheduled day and the uncertainty of each historical similar day to form a training set;

[0048] Among them, the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value.

[0049] It should be noted that the calculation formula for the uncertainty of each historical similar day is as follows:

[0050]

[0051] Where S i is the uncertainty of the i-th historical similar day, R Pij is the actual photovoltaic output ratio at time j on the i-th historical similar day, is the planned photovoltaic output ratio at time j on the i-th historical similar day, RWij is the actual wind power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, R Tij is the actual thermal power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, and N is the total number of times on a similar day.

[0052] For example, S1: sampling the day-ahead power forecast data and historical day-ahead dispatch plan data for each similar day corresponding to the day to be dispatched, i.e., the photovoltaic output ratio forecast value, the wind power output ratio forecast value, and the thermal power output ratio forecast value and the actual output ratio, to create a training set, wherein the training set is composed of the following data:

[0053] 1) Historical day-ahead photovoltaic power forecast data:

[0054]

[0055] Where i represents the sampling point of the day-ahead power forecast data, N is the number of samples of the day-ahead power forecast data. If the 24-hour forecast data is sampled every 15 minutes, N is 96, and M is the total number of samples. If there are 1000 days of operating data, M is 1000.

[0056] 2) Historical day-ahead wind power forecast data:

[0057]

[0058] 3) The day-ahead dispatch output ratio corresponding to the PV power forecast data and wind power forecast data:

[0059]

[0060] In the formula is the planned photovoltaic output ratio, is the planned proportion of wind power output, The proportion of planned thermal power output

[0061]

[0062] 4) The actual wind, solar and thermal output ratios for the next day corresponding to the previous day's dispatch plan:

[0063] R Pij |i=1,2...M;j=1,2...N

[0064] R Wij |i=1,2...M;j=1,2...N

[0065] RTij |i=1,2...M;j=1,2...N

[0066] Where R Pij is the actual photovoltaic output ratio, R Wij is the actual wind power output ratio, R Tij is the actual thermal power output ratio, R Pij +R Wij +R Tij =1;

[0067] 5) Quantification of uncertainty between day-ahead scheduling plan and actual scheduling results:

[0068]

[0069] A sample consists of a data pair formed by the day-ahead photovoltaic and wind power forecast data, the day-ahead dispatch plan data and S. That is:

[0070]

[0071] Among them, the left side is the model input and the right side is the model output target.

[0072] S2. Before training the model, determine the model's loss function. The loss function represents the difference between the model's actual output value and the expected output value:

[0073] X=[x1...x l *** x n ] T

[0074] Where, is the output of the model, M is the number of days in the training set, when When L is less than , it means that there is no error between the uncertainty estimate and the actual uncertainty. The smaller L is, the more accurate the uncertainty estimate is.

[0075] Step 2: Using the training set to train the deep neural network model to obtain a trained uncertainty estimation model;

[0076] The deep neural network model, i.e., the hybrid (LSTM+CNN+attention, LCA) model, includes: an input layer, a convolutional neural network (CNN) layer, a long short-term memory artificial neural network (LSTM) layer, an attention layer, and an output layer. One implementation method for estimating the uncertainty of the day to be scheduled by using the LCA model is to input the first prediction data at each moment of the day to be scheduled into the input layer, convert it into an input vector, input it into the CNN layer for feature extraction, generate a target feature vector, and input the target feature vector into the LSTM layer. The LSTM layer and the attention layer, i.e., the attention layer, predict the predicted value of the uncertainty of the day to be scheduled by learning the rules in the target feature vector extracted by the CNN layer, and output the predicted value through the output layer, such as Figure 2 shown.

[0077] Furthermore, the step 2 specifically includes:

[0078] The first prediction data at each moment on each historically similar day in the training set is used as the input of the deep neural network model, and the uncertainty of each historically similar day in the training set is used as the output of the deep neural network model to train the model to obtain the trained uncertainty estimation model.

[0079] Furthermore, the training process of the uncertainty estimation model includes:

[0080] Step 201: inputting the first prediction data of each time on each historical similar day in the training set into the input layer, and obtaining the input vector corresponding to each of the prediction data through the input layer;

[0081] For example, if the length of the predicted data is m, we can use X=[x1...x l *** x n ] T Represents the input vector.

[0082] Step 202: Input the input vector to the CNN layer, and perform feature extraction on the input vector to filter out a target feature vector;

[0083] It should be noted that the CNN layer includes a convolutional layer and a dropout layer, and the step 202 specifically includes:

[0084] Inputting the input vector into the convolution layer to obtain multiple feature vectors of the input vector extracted by the convolution layer;

[0085] The plurality of feature vectors of the input vector are input to the dropout layer to filter out a target feature vector from the plurality of features.

[0086] In some embodiments, in order to accurately filter out the target feature vector, the CNN layer includes a convolution layer and a dropout layer. The input vector is input into the CNN layer, and features are extracted from the input vector to filter out the target feature vector. One implementation method is: input the input vector into the convolution layer, obtain multiple feature vectors of the input vector extracted by the convolution layer, and input the multiple feature vectors of the input vector into the dropout layer to filter out the target feature vector from multiple features.

[0087] In some exemplary embodiments, when the data dimension of solar power generation is 1-dimensional, the convolution layer is selected as a one-dimensional convolution, then the size of the convolution kernel is 3, and the RELU activation function is used in combination to obtain multiple feature vectors of the input vector.

[0088] In some exemplary embodiments, when the dropout layer is set to 0.2, half of the hidden neurons in the dropout layer network can be temporarily and randomly deleted, and the input and output neurons remain unchanged. Figure 3 The middle circle represents the neurons that have not been deleted, and the circle with a cross represents the neurons that have been deleted, such as Figure 3 The input neurons are then forward propagated through the modified network, and the loss result is backpropagated through the modified network. After this process is completed for a small batch of training samples, the parameters (w, b) corresponding to the neurons that have not been deleted are updated according to the stochastic gradient descent method.

[0089] It is understandable that after obtaining the updated corresponding parameters (w, b), in order to avoid the problem of overfitting of the trained prediction model, in some embodiments, this process can be repeated continuously: the deleted neurons are restored, wherein the deleted neurons remain the same, and the neurons that have not been deleted have been updated, and a subset of half the size is randomly selected from the hidden neurons and temporarily deleted, and the parameters of the deleted neurons are backed up. For a small batch of training samples, the loss is first forward propagated and then back-propagated, and the parameters (w, b) are updated according to the stochastic gradient descent method to solve the problem of overfitting of different networks.

[0090] Among them, w is the parameter weight in the neural network, and b is the bias in the neural network.

[0091] In other embodiments, when the dropout layer is set to 0.2, if the number of neurons is n, 0.2n neurons may be deleted. One implementation of deleting neurons is to change the activation function value of the neuron in the network to 0 with probability p. If the length of the output vector is i, then the target feature vector is H c =[h c1 …h c1 …] T , where the dropout neurons calculate the activation function value in the network One way to calculate is:

[0092] r j (l) ~Benoulli(p)

[0093]

[0094] Among them, the Bernoulli function generates a probability vector r, that is, a vector of 0 and 1 is randomly generated.

[0095] Step 203: Obtain the target feature vectors, and input the target feature vectors into the LSTM layer to obtain first output vectors corresponding to the target feature vectors;

[0096] In some embodiments, the obtained target feature vector is input into the LSTM layer, and the behavioral characteristics of uncertainty are learned through the LSTM layer and the bidirectional long short-term memory artificial neural network (biLSTM) layer structure. If the length of the first output vector is j, the first output vector of the LSTM layer is H L =[h L1 …h Lε …h L ] J , calculate H L One way to calculate is:

[0097]

[0098] H l =max(dropout(L))+b r

[0099] Among them, the LSTM layer needs to be connected to the dropout layer and the maximum pooling layer, Max is the maximum value function in the maximum pooling layer, br is the bias of the pooling layer, L is the output of the LSTM layer, W C, b c are the weight and bias of the LSTM layer respectively.

[0100] Step 204: Inputting the first output vectors corresponding to the target feature vectors into the attention layer, and filtering the first output vectors according to the attention weight parameter values ​​of the first output vectors in the attention layer to obtain second output vectors;

[0101] In some embodiments, after the first output vectors corresponding to the target feature vectors are input into the attention layer, the attention weight parameter values ​​of the first output vectors are distributed according to the weight distribution principle in the attention layer to obtain the attention weight parameter values ​​of the first output vectors, and the first output vectors are screened according to the attention weight parameter values ​​of the first output vectors to obtain the second output vector.

[0102] In some exemplary embodiments, if the length of the second output vector is k, the second output vector S′ is S′=[s1…* ** g i …s k ] T .

[0103] Step 205: Input the second output vector to the output layer to determine the uncertainty prediction value of each historical similar day, and output the uncertainty prediction value of each historical similar day;

[0104] Furthermore, after step 205, the following steps are further included:

[0105] The output uncertainty prediction values ​​of each of the historical similar days are subjected to denormalization processing.

[0106] An exemplary method of denormalization is as follows:

[0107]

[0108] in, is the uncertainty prediction data before denormalization obtained by LCA network prediction, y* is the uncertainty prediction value after denormalization, and y min 、y max They are the minimum and maximum values ​​in the historical output data before normalization.

[0109] Step 206: Adjust the model parameters of the deep neural network model according to the difference between the uncertainty of the historical similar days in the training set and the predicted value to obtain the uncertainty estimation model.

[0110] In some embodiments, after the second output vector is input to the output layer, the output layer obtains the predicted value of uncertainty through the fully connected layer. Assuming that the compensation predicted by the output layer is n, the predicted value of uncertainty Y is An exemplary method for calculating Y is:

[0111] Y=f(W r ·s+b r )

[0112] Among them, W r is the output layer weight, b r is the output layer bias, and f is the activation function of the fully connected layer.

[0113] Step 3: Obtain first prediction data for each moment of the day to be scheduled, and input the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled.

[0114] In summary, the uncertainty estimation method for the day-ahead scheduling plan of an integrated energy base proposed in this embodiment fully utilizes information such as the power forecast result error contained in historical data and the error between the day-ahead scheduling plan and the actual scheduling result, and evaluates its uncertainty when the day-ahead scheduling plan is generated to assist in subsequent scheduling decisions.

[0115] Example 2

[0116] Figure 4 This is a structural diagram of an uncertainty estimation system for a comprehensive energy base day-ahead scheduling plan provided according to one embodiment of the present application, such as Figure 4 As shown, the system includes:

[0117] An acquisition module 100 is used to obtain the first prediction data at each time point on each historical similar day corresponding to the scheduled day and the uncertainty of each historical similar day to form a training set;

[0118] A training module 200 is used to train the deep neural network model using the training set to obtain a trained uncertainty estimation model;

[0119] An estimation module 300 is configured to obtain first prediction data for each moment of the day to be scheduled, and input the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled;

[0120] Among them, the first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value.

[0121] In the embodiment of the present disclosure, the calculation formula for the uncertainty of each historical similar day is as follows:

[0122]

[0123] Where S i is the uncertainty of the i-th historical similar day, RPij is the actual photovoltaic output ratio at time j on the i-th historical similar day, is the planned photovoltaic output ratio at time j on the i-th historical similar day, R Wij is the actual wind power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, R Tij is the actual thermal power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, and N is the total number of times on a similar day.

[0124] In an embodiment of the present disclosure, the deep neural network model includes: an input layer, a convolutional neural network (CNN) layer, a long short-term memory (LSTM) artificial neural network (ANN) layer, an attention layer, and an output layer.

[0125] In the embodiment of the present disclosure, the training module 200 is specifically configured to:

[0126] The first prediction data at each moment on each historically similar day in the training set is used as the input of the deep neural network model, and the uncertainty of each historically similar day in the training set is used as the output of the deep neural network model to train the model to obtain the trained uncertainty estimation model.

[0127] Furthermore, the training process of the uncertainty estimation model includes:

[0128] Inputting the first prediction data of each time on each historical similar day in the training set into the input layer, and obtaining the input vector corresponding to each of the prediction data through the input layer;

[0129] Inputting the input vector into the CNN layer, and performing feature extraction on the input vector to screen out a target feature vector;

[0130] Obtain the target feature vectors, and input the target feature vectors into the LSTM layer to obtain first output vectors corresponding to the target feature vectors;

[0131] Inputting the first output vectors corresponding to the target feature vectors into the attention layer, and filtering the first output vectors according to the attention weight parameter values ​​of the first output vectors in the attention layer to obtain second output vectors;

[0132] Inputting the second output vector into the output layer to determine the uncertainty prediction value of each historical similar day, and outputting the uncertainty prediction value of each historical similar day;

[0133] According to the difference between the uncertainty of similar historical days in the training set and the predicted value, the model parameters of the deep neural network model are adjusted to obtain the uncertainty estimation model.

[0134] It should be noted that the CNN layer includes a convolution layer and a dropout layer. The input vector is input to the CNN layer, and feature extraction is performed on the input vector to screen out the target feature vector, including:

[0135] Inputting the input vector into the convolution layer to obtain multiple feature vectors of the input vector extracted by the convolution layer;

[0136] The plurality of feature vectors of the input vector are input to the dropout layer to filter out a target feature vector from the plurality of features.

[0137] Furthermore, after inputting the second output vector into the output layer to determine the uncertainty prediction value of each historically similar day and outputting the uncertainty prediction value of each historically similar day, the method further includes:

[0138] The output uncertainty prediction values ​​of each of the historical similar days are subjected to denormalization processing.

[0139] In summary, this embodiment proposes an uncertainty estimation system for the day-ahead scheduling plan of an integrated energy base. This method can fully utilize information such as the power forecast result error contained in historical data, the error between the day-ahead scheduling plan and the actual scheduling result, and evaluate its uncertainty when the day-ahead scheduling plan is generated to assist in subsequent scheduling decisions.

[0140] Example 3

[0141] To implement the above embodiments, the present disclosure further proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in the first embodiment is implemented.

[0142] Example 4

[0143] In order to implement the above embodiments, the present disclosure further proposes a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method described in the first embodiment is implemented.

[0144] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0145] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0146] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for estimating uncertainty of a day-ahead scheduling plan for an integrated energy base, characterized in that: The method comprises: Obtain the first prediction data at each time point on each historical similar day corresponding to the scheduled day and the uncertainty of each historical similar day to form a training set; Using the training set to train a deep neural network model to obtain a trained uncertainty estimation model; Obtaining first prediction data for each time of the day to be scheduled, and inputting the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled; The first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value; The calculation formula for the uncertainty of each historical similar day is as follows: Where S i is the uncertainty of the i-th historical similar day, is the actual photovoltaic output ratio at time j on the i-th historical similar day, is the planned photovoltaic output ratio at time j on the i-th historical similar day, is the actual wind power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, R Tij is the actual thermal power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, and N is the total number of times on a similar day.

2. The method according to claim 1, wherein The deep neural network model includes: an input layer, a convolutional neural network (CNN) layer, a long short-term memory (LSTM) artificial neural network (ANN) layer, an attention layer, and an output layer.

3. The method according to claim 2, wherein The deep neural network model is trained using the training set to obtain a trained uncertainty estimation model, including: The first prediction data at each moment on each historically similar day in the training set is used as the input of the deep neural network model, and the uncertainty of each historically similar day in the training set is used as the output of the deep neural network model to train the model to obtain the trained uncertainty estimation model.

4. The method according to claim 3, wherein The training process of the uncertainty estimation model includes: Inputting the first prediction data of each time on each historical similar day in the training set into the input layer, and obtaining the input vector corresponding to each of the prediction data through the input layer; Inputting the input vector into the CNN layer, and performing feature extraction on the input vector to filter out a target feature vector; Obtain the target feature vectors, and input the target feature vectors into the LSTM layer to obtain first output vectors corresponding to the target feature vectors; Inputting the first output vectors corresponding to the target feature vectors into the attention layer, and filtering the first output vectors according to the attention weight parameter values ​​of the first output vectors in the attention layer to obtain second output vectors; Inputting the second output vector into the output layer to determine the uncertainty prediction value of each historical similar day, and outputting the uncertainty prediction value of each historical similar day; According to the difference between the uncertainty of similar historical days in the training set and the predicted value, the model parameters of the deep neural network model are adjusted to obtain the uncertainty estimation model.

5. The method according to claim 4, wherein The CNN layer includes a convolution layer and a dropout layer, inputs the input vector into the CNN layer, and performs feature extraction on the input vector to filter out a target feature vector, including: Inputting the input vector into the convolution layer to obtain multiple feature vectors of the input vector extracted by the convolution layer; The plurality of feature vectors of the input vector are input to the dropout layer to filter out a target feature vector from the plurality of features.

6. The method according to claim 4, wherein After inputting the second output vector into the output layer to determine the uncertainty prediction value of each historically similar day and outputting the uncertainty prediction value of each historically similar day, the method further includes: The output uncertainty prediction values ​​of each of the historical similar days are subjected to denormalization processing.

7. An uncertainty estimation system for day-ahead scheduling of an integrated energy base, characterized by: The system comprises: An acquisition module is used to obtain the first prediction data at each time point on each historical similar day corresponding to the scheduled day and the uncertainty of each historical similar day to form a training set; A training module, configured to train a deep neural network model using the training set to obtain a trained uncertainty estimation model; an estimation module, configured to obtain first prediction data for each moment of the day to be scheduled, and input the first prediction value into the uncertainty estimation model to obtain the uncertainty of the day to be scheduled; The first prediction data includes: wind power prediction value, photovoltaic power prediction value, photovoltaic output ratio prediction value, wind power output ratio prediction value and thermal power output ratio prediction value; The calculation formula for the uncertainty of each historical similar day is as follows: Where S i is the uncertainty of the i-th historical similar day, is the actual photovoltaic output ratio at time j on the i-th historical similar day, is the planned photovoltaic output ratio at time j on the i-th historical similar day, is the actual wind power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, R Tij is the actual thermal power output ratio at time j on the i-th historical similar day, is the planned thermal power output ratio at time j on the i-th historical similar day, and N is the total number of times on a similar day.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the uncertainty estimation method for the day-ahead scheduling plan of the integrated energy base as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the uncertainty estimation method of the day-ahead scheduling plan of the integrated energy base as described in any one of claims 1 to 6 is implemented.

Images