Multi-element load pixel matrix energy system regulation and control method and device based on artificial intelligence

Abstract

The invention discloses a multi-element load pixel matrix energy system regulation and control method and device based on artificial intelligence. The method comprises the steps of obtaining to-be-processed source load data of an energy system; wherein the to-be-processed source load data comprises historical output sequence data of a source side and historical load sequence data of a load side; inputting the to-be-processed source load data into the source load joint prediction model, so that the source load joint prediction model converts the historical output sequence data into a wind-light output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, extracts to-be-processed features from the wind-light output pixel matrix and the multi-element load pixel matrix, and outputs the to-be-processed features to the wind-light output pixel matrix and the multi-element load pixel matrix; fusing and sharing each feature in the to-be-processed features to obtain source load prediction data; and performing source-load combined regulation and control on the energy system according to the source-load prediction data. By implementing the method, the coupling relationship between the source side and the load side can be analyzed, synchronous prediction of the source side and the load side of the energy system is realized, and the accuracy and efficiency of source side and load side prediction data are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system prediction, and in particular to a method and device for regulating a multi-load pixel matrix energy system based on artificial intelligence. Background Art

[0002] With the increase in the penetration rate of new energy on the energy supply side and the widespread access of flexible loads such as electric vehicles and energy storage batteries on the energy consumption side, the uncertainty of the energy system has increased significantly, posing a challenge to the stable operation of the energy system; therefore, achieving accurate prediction of new energy power generation on the energy side and load demand on the load side in the energy system to improve the power balance of the energy system is the key to ensuring the efficient operation of the power system.

[0003] Traditional prediction methods mainly use time series analysis or machine learning, such as autoregressive model (ARIMA), support vector machine (SVM), long short-term memory neural network (LSTM), etc. to independently predict the output of energy side or the load capacity of load side. However, the traditional method of independently predicting the output of energy side or the load capacity of load side ignores the coupling relationship between the energy side and the load side in the energy system, resulting in low prediction accuracy of the energy side and the load side, which in turn leads to unstable operation of the energy system when the energy system is regulated according to the prediction data. Summary of the invention

[0004] The embodiments of the present invention provide a method and device for controlling a multi-load pixel matrix energy system based on artificial intelligence, which can solve the problem of low accuracy of prediction data caused by unilateral prediction of the source side or the load side in the prior art, improve the accuracy and prediction efficiency of the source-load prediction data, and further improve the power balance and operation stability of the energy system when the energy system is controlled according to the source-load prediction data.

[0005] An embodiment of the present invention provides a multi-load pixel matrix energy system control method based on artificial intelligence, comprising:

[0006] Obtaining source and load data to be processed of the energy system; wherein the source and load data to be processed include: historical output sequence data of the source side and historical load sequence data of the load side; the historical output sequence data include: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data include: historical electric load sequence data, historical heat load sequence data and historical cooling load sequence data;

[0007] Input the source-load data to be processed into the source-load joint prediction model, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-element load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain the source-load prediction data; wherein the features to be processed include: wind power output features, photovoltaic output features, electric load features, thermal load features, and cold load features; the source-load prediction data include: wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value, and cold load prediction value;

[0008] The energy system is jointly regulated with respect to source and load according to the source-load prediction data.

[0009] Furthermore, the source load data to be processed also includes: historical meteorological data;

[0010] Before the source-load data to be processed is input into the source-load joint prediction model, the method further includes:

[0011] Determine the first Pearson correlation coefficient and the first Spearman rank correlation coefficient of the historical meteorological data and the historical wind power output sequence data according to the historical meteorological data, the historical wind power output sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula;

[0012] According to the historical meteorological data, the historical photovoltaic output sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the second Pearson correlation coefficient and the second Spearman rank correlation coefficient of the historical meteorological data and the historical photovoltaic output sequence data are determined;

[0013] Determine the third Pearson correlation coefficient and the third Spearman rank correlation coefficient between the historical meteorological data and the historical electric load sequence data according to the historical meteorological data, the historical electric load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula;

[0014] According to the historical meteorological data, the historical heat load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the fourth Pearson correlation coefficient and the fourth Spearman rank correlation coefficient of the historical meteorological data and the historical heat load sequence data are determined;

[0015] According to the historical meteorological data, the historical cooling load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient of the historical meteorological data and the historical cooling load sequence data are determined;

[0016] The feature combinations whose absolute values ​​are less than the preset correlation threshold value among the first Pearson correlation coefficient, the first Spearman rank correlation coefficient, the second Pearson correlation coefficient, the second Spearman rank correlation coefficient, the third Pearson correlation coefficient, the third Spearman rank correlation coefficient, the fourth Pearson correlation coefficient, the fourth Spearman rank correlation coefficient, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient are eliminated; wherein the feature combination is a combination of historical meteorological data and historical output series data or historical load series data.

[0017] Furthermore, the source-load joint prediction model includes a data conversion sub-model and a prediction sub-model;

[0018] The source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-load pixel matrix, extracts the features to be processed from the wind-solar output pixel matrix and the multi-load pixel matrix respectively, fuses and shares the features in the features to be processed, and obtains the source-load prediction data, including:

[0019] The data conversion sub-model places the historical wind power output sequence data, the historical photovoltaic output sequence data, the historical electric load sequence data, the historical heat load sequence data and the historical cold load sequence data in independent channels according to the distribution pattern of the static image, and obtains the historical wind power output sequence channel, the historical photovoltaic output sequence channel, the historical electric load sequence channel, the historical heat load sequence channel and the historical cold load sequence channel;

[0020] The historical wind power output sequence channel and the historical photovoltaic output sequence channel are stacked vertically to obtain a wind and solar output pixel matrix, and the historical electric load sequence channel, the historical heat load sequence channel and the historical cooling load sequence channel are stacked vertically to obtain a multi-load pixel matrix;

[0021] The prediction sub-model extracts the features to be processed from the wind and solar output pixel matrix and the multi-load pixel matrix, fuses and shares the features to be processed, and obtains the source-load prediction data.

[0022] Furthermore, the prediction sub-model includes: a feature extraction layer, a feature fusion layer, a sharing layer and an output layer; wherein the feature extraction layer includes a first convolutional neural network, a second convolutional neural network and a fully connected network;

[0023] The features to be processed are extracted from the wind and solar power output pixel matrix and the multi-element load pixel matrix, and each feature in the features to be processed is integrated and shared to obtain source-load prediction data, including:

[0024] The first convolutional neural network extracts features from the wind and solar output pixel matrix to obtain the first high-dimensional features;

[0025] The second convolutional neural network extracts features from the multivariate load pixel matrix to obtain the second high-dimensional features;

[0026] The fully connected network extracts features from historical meteorological data to obtain the third highest dimensional features;

[0027] Obtaining a feature to be processed according to the first high-dimensional feature, the second high-dimensional feature, and the third high-dimensional feature, and inputting the feature to be processed into a feature fusion layer, so that the feature fusion layer fuses the first high-dimensional feature, the second high-dimensional feature, and the third high-dimensional feature to obtain a fused feature;

[0028] Inputting the fused features into the shared layer, so that the shared layer performs forward feature extraction and reverse feature extraction on the fused features respectively to obtain forward features and reverse features;

[0029] The forward features, the reverse features and the fusion features are input into the output layer, so that the output layer outputs the source load prediction data according to the forward features, the reverse features and the fusion features.

[0030] Furthermore, the shared layer is composed of a bidirectional long short-term memory network.

[0031] Furthermore, the construction of the source-load joint prediction model includes:

[0032] Acquire a training sample set of the energy system; wherein the training sample set includes: historical wind power output sequence data samples, historical photovoltaic output sequence data samples, historical electric load sequence data samples, historical heat load sequence data samples, historical cooling load sequence data samples and historical meteorological data samples;

[0033] An initial source-load joint prediction model is constructed, and the initial source-load joint prediction model is trained with the training sample set until the initial source-load joint prediction model reaches a preset convergence condition, thereby generating a source-load joint prediction model.

[0034] Furthermore, after obtaining the training sample set of the energy system, it also includes:

[0035] Data cleaning and data normalization are performed on each sample in the training sample set.

[0036] Based on the above method embodiment, the present invention provides a corresponding device embodiment;

[0037] An embodiment of the present invention provides a multi-load pixel matrix energy system control device based on artificial intelligence, including: a data acquisition module, a prediction module and a control module;

[0038] The data acquisition module is used to acquire the source-load data to be processed of the energy system; wherein the source-load data to be processed includes: historical output sequence data of the source side and historical load sequence data of the load side; the historical output sequence data includes: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data includes: historical electric load sequence data, historical heat load sequence data and historical cooling load sequence data;

[0039] The prediction module is used to input the source-load data to be processed into the source-load joint prediction model, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-element load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain the source-load prediction data; wherein the features to be processed include: wind power output features, photovoltaic output features, electric load features, thermal load features and cold load features; the source-load prediction data include: wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value and cold load prediction value;

[0040] The control module is used to perform source-load joint control on the energy system according to the source-load prediction data.

[0041] Another embodiment of the present invention provides a device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the artificial intelligence-based multi-load pixel matrix energy system control method described in the above-mentioned embodiment of the invention.

[0042] Another embodiment of the present invention provides a medium, which includes a stored computer program, wherein when the computer program is running, the device where the medium is located is controlled to execute an artificial intelligence-based multi-load pixel matrix energy system control method described in the above-mentioned invention embodiment.

[0043] The following beneficial effects are achieved by implementing the present invention:

[0044] The present invention provides a method and device for regulating an energy system based on a multi-load pixel matrix of artificial intelligence. The method obtains historical output sequence data of the source side and historical load sequence data of the load side in the energy system as source-load data to be processed, and then inputs the source-load joint prediction model according to the source-load data to be processed, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-load pixel matrix respectively, and fuses and shares the features to be processed to obtain source-load prediction data; by simultaneously obtaining data on both sides of the source and load when obtaining data, the source and load are considered at the same time. The influence of the data on both sides of the load side on the source-load prediction data is analyzed, and the source-side data and the load-side data are processed separately through the source-load joint prediction model, and then feature extraction, fusion and sharing are performed, so that the source-load joint prediction model can fully extract the coupling relationship between the source-side data and the load-side data, and then the wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value and cooling load prediction value are output as a whole after fusion and sharing of the features, so that the prediction of the source-side data is not separated from the load-side data, and the prediction of the load-side data is not separated from the source-side data. The accuracy and prediction efficiency of the source-load prediction data are improved by combining the two, and then the power balance and operation stability of the energy system are improved when the energy system is regulated according to the source-load prediction data. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flow chart of a method for controlling a multi-load pixel matrix energy system based on artificial intelligence provided by one embodiment of the present invention.

[0046] Figure 2 It is a schematic diagram of a source-load joint prediction model provided by an embodiment of the present invention.

[0047] Figure 3 It is a schematic diagram of the process of feature extraction performed by the feature extraction layer provided in one embodiment of the present invention.

[0048] Figure 4 Schematic diagram of the network structure of BiLSTM provided by one embodiment of the present invention.

[0049] Figure 5 It is a schematic diagram of the MHA layer structure provided by an embodiment of the present invention.

[0050] Figure 6 It is a structural schematic diagram of an artificial intelligence-based multi-load pixel matrix energy system control device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in 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 creative work are within the scope of protection of the present invention.

[0052] In order to better understand the technical solution of the present invention, the professional terms involved in the present invention are explained as follows:

[0053] (1) CNN: Convolutional neural network;

[0054] (2) BiLSTM: Bi-directional long short-term memory, bidirectional long short-term memory neural network;

[0055] (3)MHA: Multi-head attention;

[0056] (4)MTL: Multi-task laerning.

[0057] like Figure 1 As shown, a method for controlling a multi-load pixel matrix energy system based on artificial intelligence is provided in one embodiment of the present invention, comprising:

[0058] Step S1: Obtain the source-load data to be processed of the energy system; wherein the source-load data to be processed includes: historical output sequence data of the source side and historical load sequence data of the load side; the historical output sequence data includes: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data includes: historical electric load sequence data, historical heat load sequence data and historical cooling load sequence data;

[0059] Step S2: inputting the source-load data to be processed into the source-load joint prediction model, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-element load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain source-load prediction data; wherein the features to be processed include: wind power output features, photovoltaic output features, electric load features, thermal load features, and cold load features; the source-load prediction data include: wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value, and cold load prediction value;

[0060] Step S3: performing source-load joint regulation on the energy system according to the source-load prediction data.

[0061] For step S1, in a preferred embodiment, the source load data to be processed also includes: historical meteorological data; before the source load data to be processed is input into the source load joint prediction model, it also includes: determining the first Pearson correlation coefficient and the first Spearman rank correlation coefficient of the historical meteorological data and the historical wind power output sequence data according to the historical meteorological data, the historical wind power output sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula; determining the second Pearson correlation coefficient and the second Spearman rank correlation coefficient of the historical meteorological data and the historical photovoltaic output sequence data according to the historical meteorological data, the historical photovoltaic output sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula; determining the third Pearson correlation coefficient and the third Spearman rank correlation coefficient of the historical meteorological data and the historical electric load sequence data according to the historical meteorological data, the historical electric load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula; determining the third Pearson correlation coefficient and the third Spearman rank correlation coefficient of the historical meteorological data and the historical electric load sequence data according to the historical meteorological data , historical heat load sequence data, Pearson correlation coefficient calculation formula and Spearman rank correlation coefficient calculation formula, determine the fourth Pearson correlation coefficient and the fourth Spearman rank correlation coefficient between the historical meteorological data and the historical heat load sequence data; determine the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient between the historical meteorological data and the historical cooling load sequence data according to the historical meteorological data, the historical cooling load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula; among the first Pearson correlation coefficient, the first Spearman rank correlation coefficient, the second Pearson correlation coefficient, the second Spearman rank correlation coefficient, the third Pearson correlation coefficient, the third Spearman rank correlation coefficient, the fourth Pearson correlation coefficient, the fourth Spearman rank correlation coefficient, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient, the feature combination whose absolute value is less than the preset correlation threshold is eliminated; wherein, the feature combination is a combination of historical meteorological data and historical output sequence data or historical load sequence data.

[0062] Specifically, the historical output sequence data of the source side of the energy system, the historical load sequence data of the load side, and the historical meteorological data that have an impact on the source side and load side data are obtained; wherein the historical output sequence data includes: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data includes: historical electric load sequence data, historical thermal load sequence data, and historical cooling load sequence data. Preferably, after obtaining the above data, the obtained data is cleaned and normalized in turn; wherein the method of data cleaning and data normalization is implemented by the following formula:

[0063]

[0064] Among them, x i is the i-th data after filling; x i-2 、x i-1 、x i+1 and x i+2 They are the i-2, i-1, i+1 and i+2 data respectively; is the normalized ith data, x min is the minimum value in the data, x max is the maximum value in the data.

[0065] For the above processed data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula are used to analyze the factors affecting the wind and solar power output and multi-load fluctuations, and screen out the features with strong correlation. Specifically, the Pearson correlation coefficient and the Spearman rank correlation coefficient are calculated respectively according to the historical meteorological data and the historical wind power output sequence data, the historical photovoltaic output sequence data, the historical electric load sequence data, the historical heat load sequence data and the historical cold load sequence data, and the first Pearson correlation coefficient, the first Spearman rank correlation coefficient, the second Pearson correlation coefficient, the second Spearman rank correlation coefficient, the third Pearson correlation coefficient, the third Spearman rank correlation coefficient, the fourth Pearson correlation coefficient, the fourth Spearman rank correlation coefficient, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient are obtained. Exemplarily, based on historical meteorological data, historical wind power output series data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the first Pearson correlation coefficient and the first Spearman rank correlation coefficient of the historical meteorological data and the historical wind power output series data are determined, which can be calculated by the following formula. The remaining calculations can be calculated using the following formulas, which are not listed here one by one.

[0066]

[0067]

[0068] Among them, R represents the wind and solar power output variable or load variable, which in this example represents the historical wind power output sequence data; S is the meteorological factor variable, that is, the historical meteorological data; is the Pearson correlation coefficient of variables R and S. In this example, it is the first Pearson correlation coefficient of historical meteorological data and historical wind power output series data; cov(R,S) is the covariance of variables R and S; σ R is the standard deviation of variable R; σ S is the standard deviation of variable S; r i is the i-th data of variable R; s i is the i-th data of variable S; is the average value of each data of variable R; is the average value of each data of variable S; n is the total number of data; is the Spearman rank correlation coefficient of variables R and S. In this example, it represents the first Spearman rank correlation coefficient of historical meteorological data and historical wind power output series data. Rand(r i ) is r i rank value; Rand(s i ) is s i rank value of .

[0069] After completing the calculation of the first Pearson correlation coefficient, the first Spearman rank correlation coefficient, the second Pearson correlation coefficient, the second Spearman rank correlation coefficient, the third Pearson correlation coefficient, the third Spearman rank correlation coefficient, the fourth Pearson correlation coefficient, the fourth Spearman rank correlation coefficient, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient, the feature combinations whose absolute values ​​in the above calculations are less than the preset correlation threshold are eliminated, thereby achieving the processing of screening out features with low correlation.

[0070] It should be noted that the Pearson correlation coefficient and the Spearman rank correlation coefficient can indicate the direction and degree of the trend of change between two variables, and the value range is [-1,1]. When it is close to 1, it means that there is a strong positive correlation between the two variables; when it is close to -1, it means that there is a strong negative correlation; and when the value is close to 0, it means that the correlation is very low.

[0071] For step S2, the processed source load data to be processed is input into Figure 2 In the source-load joint prediction model shown.

[0072] In a preferred embodiment, the source-load joint prediction model includes a data conversion sub-model and a prediction sub-model; the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain the source-load prediction data, including: the data conversion sub-model converts the historical wind power output sequence data, the historical photovoltaic output sequence data, the historical electric load sequence data, the historical thermal load sequence data and the historical The cold load sequence data are placed in independent channels to obtain historical wind power output sequence channels, historical photovoltaic output sequence channels, historical electric load sequence channels, historical thermal load sequence channels and historical cold load sequence channels; the historical wind power output sequence channels and historical photovoltaic output sequence channels are vertically stacked to obtain a wind-solar output pixel matrix, and the historical electric load sequence channels, historical thermal load sequence channels and historical cold load sequence channels are vertically stacked to obtain a multi-load pixel matrix; the prediction sub-model extracts the features to be processed from the wind-solar output pixel matrix and the multi-load pixel matrix, and fuses and shares the features to be processed to obtain the source-load prediction data.

[0073] Specifically, the processing of the data to be processed by the source-load joint prediction model is mainly divided into two aspects: data processing and feature extraction prediction. Data processing is implemented by the data conversion sub-model in the source-load joint prediction model, which mainly realizes channel stacking through the distribution pattern of the static image to obtain a pixel matrix. First, the data conversion sub-model places the historical wind power output sequence data and the historical photovoltaic output sequence data in independent channels according to the static image distribution pattern. After obtaining the historical wind power output sequence channel and the historical photovoltaic output sequence channel, the two channels are stacked vertically to form a two-channel wind and solar output pixel matrix, as shown below:

[0074]

[0075] Among them, W represents wind power output; V represents photovoltaic power output; M represents different times of the day; and M is the time step.

[0076] Similarly, according to the static graph distribution mode, the historical electric load sequence data, the historical heat load sequence data, and the historical cooling load sequence data are placed in independent channels respectively. After obtaining the historical electric load sequence channel, the historical heat load sequence channel, and the historical cooling load sequence channel, the three channels are stacked vertically to form a three-channel multivariate load pixel matrix, as shown below:

[0077]

[0078] Among them, E is the electrical load; H is the heat load; C is the cooling load.

[0079] In the above pixel matrix, in the same channel, the horizontal distribution of load reflects the temporal correlation between loads in different time periods within the day, while the vertical distribution reflects the periodicity of loads at the same time point for several consecutive days; different channels reflect the cross-correlation between various loads.

[0080] The wind and solar output pixel matrix, multi-element load pixel matrix and historical meteorological data generated above are transmitted to the prediction sub-model.

[0081] In a preferred embodiment, the prediction sub-model includes: a feature extraction layer, a feature fusion layer, a sharing layer and an output layer; wherein the feature extraction layer includes a first convolutional neural network, a second convolutional neural network and a fully connected network; the extracting of the features to be processed from the wind and solar output pixel matrix and the multi-element load pixel matrix, fusing and sharing the features to be processed, and obtaining the source-load prediction data, includes: the first convolutional neural network extracts the features of the wind and solar output pixel matrix to obtain the first high-dimensional features; the second convolutional neural network extracts the features of the multi-element load pixel matrix to obtain the second high-dimensional features; the fully connected network extracts the features of the historical meteorological data Perform feature extraction to obtain a third high-dimensional feature; obtain a feature to be processed according to the first high-dimensional feature, the second high-dimensional feature and the third high-dimensional feature, and input the feature to be processed into a feature fusion layer, so that the feature fusion layer fuses the first high-dimensional feature, the second high-dimensional feature and the third high-dimensional feature to obtain a fused feature; input the fused feature into a shared layer, so that the shared layer performs forward feature extraction and reverse feature extraction on the fused feature respectively to obtain a forward feature and a reverse feature; input the forward feature, the reverse feature and the fused feature into an output layer, so that the output layer outputs source-load prediction data according to the forward feature, the reverse feature and the fused feature.

[0082] In a preferred embodiment, the shared layer is composed of a bidirectional long short-term memory network.

[0083] Specifically, Figure 2 As shown, the prediction sub-model includes a feature extraction layer, a feature fusion layer, a sharing layer and an output layer; wherein the feature extraction layer includes a first convolutional neural network (i.e. Figure 2 The first CNN), the second convolutional neural network (i.e. Figure 2 The second CNN) and a fully connected network; the feature fusion layer is Figure 2 The fusion layer in is , the shared layer is a BiLSTM (bidirectional long short-term memory network) shared layer, and the output layer includes an MHA layer and several fully connected layers.

[0084] like Figure 3As shown in the figure, the CNN network includes a convolution layer, a batch normalization layer, an activation function layer, a pooling layer, and a fully connected layer; the activation function of the CNN network uses the ReLU function, and the pooling layer uses the maximum pooling operation; the fully connected network is composed of multiple fully connected layers. After the first CNN network performs multiple convolution and pooling operations on the wind and solar output pixel matrix, the first high-dimensional feature (i.e. Figure 3 The second CNN network performs multiple convolution and pooling operations on the multi-load pixel matrix to extract the second high-dimensional features of the multi-load pixel matrix (i.e. Figure 3 The fully connected network extracts features from historical meteorological data and obtains the third high-dimensional feature (i.e. Figure 3 The first, second and third high-dimensional features extracted above are input into the fusion layer for feature fusion to obtain fusion features. The fusion features are input into the BiLSTM shared layer.

[0085] like Figure 4 The figure shows a BiLSTM network structure. BiLSTM is a network structure formed by connecting the forward and reverse LSTM networks. This structure can realize the forward feature extraction and reverse feature extraction of the time series at the same time, making the neural network's feature extraction more global and complete. When the fused features are input to this layer, BiLSTM performs forward feature extraction and reverse feature extraction on the fused features respectively to obtain forward features and reverse features. The corresponding extraction formulas are shown below:

[0086]

[0087] Among them, y t is the output result of the BiLSTM network, which is a feature output sequence containing forward features and reverse features; is the output result of the forward LSTM network, i.e., the forward feature; is the output result of the reverse LSTM network, i.e., the reverse feature; W f is the weight coefficient of the forward output result; W b is the weight coefficient of the reverse output result; b is the bias term; σ is the activation function.

[0088] The extracted features and sequence data are input into the MHA layer, which combines multiple fully connected layers to execute multiple attention mechanisms in parallel, thereby realizing multi-task learning to output source load prediction data. Figure 5 The figure shows a schematic diagram of the MHA layer structure provided by the present invention. First, the output of the above BiLSTM network is regarded as the query Q, and the output sequence of the entire BiLSTM network is used as the key K and value V of the MHA. The query Q, key K and value V are mapped to m independent linear spaces respectively, that is:

[0089]

[0090] Then, in each attention head, the scaled dot product attention mechanism is used to evaluate and calculate the respective attention weight scores, namely:

[0091]

[0092] Finally, after concatenating the results of multiple attention heads, the final multi-head attention is obtained through linear transformation, namely:

[0093]

[0094] Among them, R o , and are the parameter matrices that map the query Q, key K, and value V to the linear space corresponding to the i-th attention head; Q i , K i and V i H are the results of mapping the query Q, key K and value V to the linear space corresponding to the i-th attention head; i is the attention score of the ith attention head; softmax(·) is the normalized exponential function; K i is the transpose of ; l is a scaling factor; MHA(Q,K,V) is the output of the final multi-head attention, i.e., the source load prediction data; R o A parameter matrix that maps the multi-head attention representation to the original vector space.

[0095] The corresponding wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value and cooling load prediction value are obtained through the above processing. For the construction of the above-mentioned source-load joint prediction model, in a preferred embodiment, the construction of the source-load joint prediction model includes: obtaining a training sample set of the energy system; wherein the training sample set includes: historical wind power output sequence data samples, historical photovoltaic output sequence data samples, historical electric load sequence data samples, historical thermal load sequence data samples, historical cooling load sequence data samples and historical meteorological data samples; constructing an initial source-load joint prediction model, training the initial source-load joint prediction model with the training sample set until the initial source-load joint prediction model reaches a preset convergence condition, and generating a source-load joint prediction model.

[0096] In a preferred embodiment, after obtaining the training sample set of the energy system, the method further includes: performing data cleaning and data normalization processing on each sample in the training sample set.

[0097] Specifically, a training sample set including historical wind power output sequence data samples, historical photovoltaic output sequence data samples, historical electric load sequence data samples, historical heat load sequence data samples, historical cold load sequence data samples and historical meteorological data samples is obtained. Preferably, the preprocessing steps in the above step S1 are used to perform data cleaning and data normalization on each data sample to eliminate abnormal data and fill in missing data. An initial source-load joint prediction model is constructed, and the model structure and the role of each structure in the training are not repeated here. The initial source-load joint prediction model is trained with the above training sample set. For each iterative training, a data sample segment at time t-1 in the sample can be selected as input, and a data sample segment at time t can be selected as output to train the initial source-load joint prediction model until the initial source-load joint prediction model reaches a preset convergence condition, that is, when the overall error is minimized or the maximum number of iterations is reached, a source-load joint prediction model is generated.

[0098] For step S3, the source side and the load side of the energy system at the corresponding prediction time are jointly regulated according to the wind power output prediction value, the photovoltaic output prediction value, the electric load prediction value, the thermal load prediction value and the cooling load prediction value.

[0099] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0100] like Figure 6 As shown, an embodiment of the present invention provides a multi-load pixel matrix energy system control device based on artificial intelligence, including: a data acquisition module, a prediction module and a control module;

[0101] The data acquisition module is used to acquire the source-load data to be processed of the energy system; wherein the source-load data to be processed includes: historical output sequence data of the source side and historical load sequence data of the load side; the historical output sequence data includes: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data includes: historical electric load sequence data, historical heat load sequence data and historical cooling load sequence data;

[0102] The prediction module is used to input the source-load data to be processed into the source-load joint prediction model, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-element load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain the source-load prediction data; wherein the features to be processed include: wind power output features, photovoltaic output features, electric load features, thermal load features and cold load features; the source-load prediction data include: wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value and cold load prediction value;

[0103] The control module is used to perform source-load joint control on the energy system according to the source-load prediction data.

[0104] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0105] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0106] Based on the above method item embodiments, the present invention provides corresponding device item embodiments.

[0107] An embodiment of the present invention provides a device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements an artificial intelligence-based multi-load pixel matrix energy system control method as described in any one of the present invention.

[0108] The device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The device may include, but is not limited to, a processor and a memory.

[0109] The processor may be a central processing unit (CPU), 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. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the device, and various interfaces and lines are used to connect various parts of the entire device.

[0110] The memory can be used to store the computer program, and the processor realizes various functions of the device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0111] Based on the above method item embodiments, the present invention provides corresponding medium item embodiments.

[0112] An embodiment of the present invention provides a medium, which includes a stored computer program, wherein when the computer program is running, the device where the medium is located is controlled to execute any one of the artificial intelligence-based multi-load pixel matrix energy system control methods described in the present invention.

[0113] The medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0114] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-load pixel matrix energy system control method based on artificial intelligence, characterized in that: include: Obtaining source and load data to be processed of the energy system; wherein the source and load data to be processed include: historical output sequence data of the source side and historical load sequence data of the load side; the historical output sequence data include: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data include: historical electric load sequence data, historical heat load sequence data and historical cooling load sequence data; Input the source-load data to be processed into the source-load joint prediction model, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-element load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain the source-load prediction data; wherein the features to be processed include: wind power output features, photovoltaic output features, electric load features, thermal load features, and cold load features; the source-load prediction data include: wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value, and cold load prediction value; The energy system is jointly regulated with respect to source and load according to the source-load prediction data.

2. The method for controlling a multi-load pixel matrix energy system based on artificial intelligence as claimed in claim 1, characterized in that: The source load data to be processed also includes: historical meteorological data; Before the source-load data to be processed is input into the source-load joint prediction model, the method further includes: Determine the first Pearson correlation coefficient and the first Spearman rank correlation coefficient of the historical meteorological data and the historical wind power output sequence data according to the historical meteorological data, the historical wind power output sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula; According to the historical meteorological data, the historical photovoltaic output sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the second Pearson correlation coefficient and the second Spearman rank correlation coefficient of the historical meteorological data and the historical photovoltaic output sequence data are determined; Determine the third Pearson correlation coefficient and the third Spearman rank correlation coefficient between the historical meteorological data and the historical electric load sequence data according to the historical meteorological data, the historical electric load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula; According to the historical meteorological data, the historical heat load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the fourth Pearson correlation coefficient and the fourth Spearman rank correlation coefficient of the historical meteorological data and the historical heat load sequence data are determined; According to the historical meteorological data, the historical cooling load sequence data, the Pearson correlation coefficient calculation formula and the Spearman rank correlation coefficient calculation formula, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient of the historical meteorological data and the historical cooling load sequence data are determined; The feature combinations whose absolute values ​​are less than the preset correlation threshold value among the first Pearson correlation coefficient, the first Spearman rank correlation coefficient, the second Pearson correlation coefficient, the second Spearman rank correlation coefficient, the third Pearson correlation coefficient, the third Spearman rank correlation coefficient, the fourth Pearson correlation coefficient, the fourth Spearman rank correlation coefficient, the fifth Pearson correlation coefficient and the fifth Spearman rank correlation coefficient are eliminated; wherein the feature combination is a combination of historical meteorological data and historical output series data or historical load series data.

3. The artificial intelligence-based multi-load pixel matrix energy system control method according to claim 2, characterized in that: The source-load joint prediction model includes a data conversion sub-model and a prediction sub-model; The source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-load pixel matrix, extracts the features to be processed from the wind-solar output pixel matrix and the multi-load pixel matrix respectively, fuses and shares the features in the features to be processed, and obtains the source-load prediction data, including: The data conversion sub-model places the historical wind power output sequence data, the historical photovoltaic output sequence data, the historical electric load sequence data, the historical heat load sequence data and the historical cold load sequence data in independent channels according to the distribution pattern of the static image, and obtains the historical wind power output sequence channel, the historical photovoltaic output sequence channel, the historical electric load sequence channel, the historical heat load sequence channel and the historical cold load sequence channel; The historical wind power output sequence channel and the historical photovoltaic output sequence channel are stacked vertically to obtain a wind and solar output pixel matrix, and the historical electric load sequence channel, the historical heat load sequence channel and the historical cooling load sequence channel are stacked vertically to obtain a multi-load pixel matrix; The prediction sub-model extracts the features to be processed from the wind and solar output pixel matrix and the multi-load pixel matrix, fuses and shares the features to be processed, and obtains the source-load prediction data.

4. The artificial intelligence-based multi-load pixel matrix energy system control method according to claim 3, characterized in that: The prediction sub-model includes: a feature extraction layer, a feature fusion layer, a sharing layer and an output layer; wherein the feature extraction layer includes a first convolutional neural network, a second convolutional neural network and a fully connected network; The features to be processed are extracted from the wind and solar power output pixel matrix and the multi-element load pixel matrix, and each feature in the features to be processed is integrated and shared to obtain source-load prediction data, including: The first convolutional neural network extracts features from the wind and solar output pixel matrix to obtain the first high-dimensional features; The second convolutional neural network extracts features from the multivariate load pixel matrix to obtain the second high-dimensional features; The fully connected network extracts features from historical meteorological data to obtain the third highest dimensional features; Obtaining a feature to be processed according to the first high-dimensional feature, the second high-dimensional feature, and the third high-dimensional feature, and inputting the feature to be processed into a feature fusion layer, so that the feature fusion layer fuses the first high-dimensional feature, the second high-dimensional feature, and the third high-dimensional feature to obtain a fused feature; Inputting the fused features into the shared layer, so that the shared layer performs forward feature extraction and reverse feature extraction on the fused features respectively to obtain forward features and reverse features; The forward features, the reverse features and the fusion features are input into the output layer, so that the output layer outputs source load prediction data according to the forward features, the reverse features and the fusion features.

5. The artificial intelligence-based multi-load pixel matrix energy system control method according to claim 4, characterized in that: The shared layer is composed of a bidirectional long short-term memory network.

6. The artificial intelligence-based multi-load pixel matrix energy system control method according to claim 4, characterized in that: The construction of the source-load joint prediction model includes: Acquire a training sample set of the energy system; wherein the training sample set includes: historical wind power output sequence data samples, historical photovoltaic output sequence data samples, historical electric load sequence data samples, historical heat load sequence data samples, historical cooling load sequence data samples and historical meteorological data samples; An initial source-load joint prediction model is constructed, and the initial source-load joint prediction model is trained with the training sample set until the initial source-load joint prediction model reaches a preset convergence condition, thereby generating a source-load joint prediction model.

7. The artificial intelligence-based multi-load pixel matrix energy system control method according to claim 5, characterized in that: After obtaining the training sample set of the energy system, it also includes: Data cleaning and data normalization are performed on each sample in the training sample set.

8. A multi-load pixel matrix energy system control device based on artificial intelligence, characterized in that: include: Data acquisition module, prediction module and control module; The data acquisition module is used to acquire the source-load data to be processed of the energy system; wherein the source-load data to be processed includes: historical output sequence data of the source side and historical load sequence data of the load side; the historical output sequence data includes: historical wind power output sequence data and historical photovoltaic output sequence data; the historical load sequence data includes: historical electric load sequence data, historical heat load sequence data and historical cooling load sequence data; The prediction module is used to input the source-load data to be processed into the source-load joint prediction model, so that the source-load joint prediction model converts the historical output sequence data into a wind-solar output pixel matrix, converts the historical load sequence data into a multi-element load pixel matrix, and extracts the features to be processed from the wind-solar output pixel matrix and the multi-element load pixel matrix respectively, and fuses and shares the features in the features to be processed to obtain the source-load prediction data; wherein the features to be processed include: wind power output features, photovoltaic output features, electric load features, thermal load features and cold load features; the source-load prediction data include: wind power output prediction value, photovoltaic output prediction value, electric load prediction value, thermal load prediction value and cold load prediction value; The control module is used to perform source-load joint control on the energy system according to the source-load prediction data.

9. A device, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements an artificial intelligence-based multi-load pixel matrix energy system control method as described in any one of claims 1 to 7.

10. A medium, characterized in that The medium includes a stored computer program, wherein when the computer program is running, the device where the medium is located is controlled to execute an artificial intelligence-based multi-load pixel matrix energy system control method as described in any one of claims 1 to 7.

Images