Microgrid load prediction method and system based on big data

Abstract

The invention belongs to the technical field of power grid load prediction, and particularly provides a micro-grid load prediction method and system based on big data, and the method comprises the steps: collecting micro-grid multi-source data of a preset service area through employing an Internet of Things sensor, carrying out the denoising cleaning of the micro-grid multi-source data, and obtaining the processed multi-source data; performing data-level fusion on the processed multi-source data to obtain fused data; performing multi-scale decomposition on the fused data based on an empirical mode decomposition method combined with multi-scale permutation entropy to obtain a multi-scale sub-data sequence; based on the multi-scale sub-data sequence, constructing a micro-grid load prediction model by adopting a time convolution and time sequence mode attention mechanism combined algorithm; and on the basis of the micro-grid load prediction model, predicting a micro-grid load to be predicted to obtain a load prediction result. According to the technical scheme, the accuracy, robustness and calculation efficiency of load prediction can be remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid load prediction, and in particular relates to a microgrid load prediction method and system based on big data. Background Art

[0002] With the rapid development of smart grid and Internet of Things technologies, microgrids, as an important part of distributed energy systems, have become crucial to the stable operation and optimal dispatch of power systems. Microgrid load forecasting involves a variety of data sources, including power load data, meteorological data, user behavior data, etc. These data are usually multi-dimensional, high-noise, and nonlinear. Therefore, how to effectively integrate these multi-source data and extract useful information from them for load forecasting has become an urgent problem to be solved.

[0003] Traditional load forecasting methods are often based on a single data source or simple data fusion methods, which makes it difficult to accurately capture the complex changes in microgrid loads. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a microgrid load forecasting method and system based on big data, which significantly improves the accuracy, robustness and computational efficiency of load forecasting, and provides strong support for the intelligent scheduling and optimized operation of microgrids.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A microgrid load prediction method based on big data, the method comprising:

[0007] Using IoT sensors to collect multi-source data of microgrids in a preset service area, and denoising and cleaning the multi-source data of the microgrid to obtain processed multi-source data;

[0008] Perform data-level fusion on the processed multi-source data to obtain fused data;

[0009] Performing multi-scale decomposition on the fused data based on an empirical mode decomposition method combined with multi-scale permutation entropy to obtain a multi-scale sub-data sequence;

[0010] Based on the multi-scale sub-data sequence, a microgrid load prediction model is constructed by combining a temporal convolution with a temporal pattern attention mechanism;

[0011] Based on the microgrid load prediction model, the microgrid load to be predicted is predicted to obtain a load prediction result.

[0012] Preferably, the microgrid multi-source data includes historical power load data, meteorological data of a preset service area, social data, and new energy power generation data.

[0013] Preferably, the method for obtaining fused data includes:

[0014] Based on the sensor data, a metadata set of microgrid multi-source data is established, and unbiased estimates of multi-source data are calculated;

[0015] Based on the metadata set, establishing a corresponding data weight set;

[0016] Based on the metadata set, the data weight set and the unbiased estimate of the multi-source data, a total mean square error of the multi-source data is obtained;

[0017] Based on the metadata set, the data weight set, the unbiased estimate and the total mean square error, the conjugate gradient method is used to obtain the optimal weighting factor of each sensor;

[0018] The fused data is obtained based on the historical mean value and the minimum mean square error of each sensor.

[0019] Preferably, the method for obtaining the multi-scale sub-data sequence includes:

[0020] Using independent component analysis method, decorrelation processing is performed on the fused data;

[0021] Add random noise to the fused data after decorrelation processing to obtain noisy data;

[0022] Performing empirical mode decomposition on the noisy data to obtain intrinsic mode function components, and calculating the multi-scale permutation entropy of each intrinsic mode function component;

[0023] The multi-scale sub-data sequence is obtained based on the multi-scale permutation entropy value and frequency of each eigenmode function component.

[0024] Preferably, the method for constructing a microgrid load prediction model includes:

[0025] Extracting deep features of the multi-scale sub-data sequence based on a convolutional layer of a temporal convolutional network;

[0026] Mapping the deep features to prediction values ​​based on a fully connected layer to obtain an initial load prediction result;

[0027] Based on cosine similarity, the attention weight between the current time step of the temporal convolutional network and the past time steps within a preset range is calculated;

[0028] Based on the attention weights, weighted summation is performed on the initial load forecast results outputted in the past time steps to obtain a feature vector of the fused attention mechanism;

[0029] The feature vector of the fusion attention mechanism is combined with the initial load forecast result to obtain the final load forecast result, thus completing the construction of the microgrid load forecasting model.

[0030] The present invention also provides a microgrid load forecasting system based on big data, which is used to implement the method, comprising:

[0031] A multi-source data acquisition module is used to collect multi-source data of microgrids in a preset service area using IoT sensors, and to perform denoising and cleaning on the multi-source data of microgrids to obtain processed multi-source data;

[0032] A data fusion module is used to perform data-level fusion on the processed multi-source data to obtain fused data;

[0033] A multi-scale decomposition module, used for performing multi-scale decomposition on the fused data based on an empirical mode decomposition method combined with multi-scale permutation entropy to obtain a multi-scale sub-data sequence;

[0034] A prediction model building module, used to build a microgrid load prediction model based on the multi-scale sub-data sequence by combining time convolution with a temporal pattern attention mechanism;

[0035] The prediction module is used to predict the microgrid load to be predicted based on the microgrid load prediction model to obtain a load prediction result.

[0036] Preferably, the data fusion module includes:

[0037] A metadata collection unit, used to establish a metadata collection of multi-source data of the microgrid based on sensor data, and calculate an unbiased estimate of the multi-source data;

[0038] A weight set unit, used to establish a corresponding data weight set based on the metadata set;

[0039] A mean square error calculation unit, used for obtaining a total mean square error of the multi-source data based on the metadata set, the data weight set and the unbiased estimate of the multi-source data;

[0040] A weighting factor acquisition unit, configured to obtain an optimal weighting factor of each sensor by using a conjugate gradient method based on the metadata set, the data weight set, the unbiased estimate and the total mean square error;

[0041] The data fusion unit is used to obtain the fused data based on the historical mean value and the minimum mean square error of each sensor.

[0042] Preferably, the multi-scale decomposition module includes:

[0043] A decorrelation processing unit, used for performing decorrelation processing on the fused data by using an independent component analysis method;

[0044] A noise adding unit, used for adding random noise to the fused data after decorrelation processing to obtain noisy data;

[0045] A permutation entropy calculation unit, used for performing empirical mode decomposition on the noisy data to obtain intrinsic mode function components, and calculating the multi-scale permutation entropy of each intrinsic mode function component;

[0046] The multi-scale sequence acquisition unit is used to obtain the multi-scale sub-data sequence based on the multi-scale permutation entropy value and frequency of each intrinsic mode function component.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows: by fusing multi-source data and using multi-scale permutation entropy and empirical mode decomposition methods, the changing patterns of microgrid loads can be captured more comprehensively, thereby improving the accuracy of predictions. De-noising, cleaning and fusion processing of the collected multi-source data can reduce noise interference, improve the reliability and stability of the data, and thus enhance the robustness of the prediction model. The prediction model is constructed by combining the time convolution and time series pattern attention mechanism with an algorithm, which can make full use of the characteristics of time series data, improve computing efficiency and reduce computing costs. Accurate load forecasting results can provide strong support for the intelligent scheduling of microgrids and help achieve the rational allocation and optimal utilization of electricity. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0049] Figure 1 This is a flow chart of a microgrid load prediction method based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] 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.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Embodiment 1

[0053] like Figure 1 As shown, a microgrid load prediction method based on big data includes:

[0054] S1: Use IoT sensors to collect multi-source data of microgrids in a preset service area, and denoise and clean the multi-source data of the microgrid to obtain processed multi-source data; a further implementation method is that the multi-source data of the microgrid includes historical power load data, meteorological data of the preset service area, social data and new energy power generation data.

[0055] S2: Perform data-level fusion on the processed multi-source data to obtain fused data; a further implementation method is that the method for obtaining the fused data includes:

[0056] Based on the data obtained by n sensors, a metadata set of multi-source data of the microgrid is established, and an unbiased estimate of the multi-source data is calculated;

[0057] Based on the metadata set, a corresponding data weight set is established;

[0058] Based on the metadata set, the data weight set and the unbiased estimate of the multi-source data, the total mean square error of the multi-source data is obtained; in this embodiment, the total mean square error The calculation formula is as follows:

[0059] ,

[0060] In the formula, X represents the unbiased estimate, x i 、x j Represents the data collected by sensor i and sensor j in the metadata set, Corresponding to the weights of xi and xj respectively.

[0061] Based on the metadata set, the data weight set, the unbiased estimate and the total mean square error, the conjugate gradient method is used to obtain the optimal weighting factor of each sensor; in this embodiment, the specific process of obtaining the optimal weighting factor is as follows:

[0062] The solution of the optimal weighting factor is expressed as the problem of minimizing the total mean square error, and the total mean square error is used as the objective function;

[0063] Based on preset weight constraints , and obtain the objective function about The gradient is:

[0064] ,in, represents the weighted unbiased estimate;

[0065] Initialize weights and search direction , for k=0,1,2.... until convergence, calculate the gradient , until the gradient reaches 0 and the iteration stops, the optimal weighting factor is obtained. Specifically, the line search method is used to calculate the step size , update the weights , calculate the new gradient , update the search direction ,in, is the conjugate gradient parameter.

[0066] Based on the historical mean and minimum mean square error of each sensor, the fused data is obtained.

[0067] S3: Based on the empirical mode decomposition method combined with multi-scale permutation entropy, the fused data is multi-scaled decomposed to obtain a multi-scale sub-data sequence; a further implementation method is that the method for obtaining the multi-scale sub-data sequence includes:

[0068] The purpose of decorrelation processing on the fused data by using independent component analysis (ICA) is to separate the independent components in the fused data through ICA, thereby removing or reducing the correlation between the components. Specifically, the fused data is centered and whitened to meet the input requirements of ICA. ICA is used to extract independent components from the processed fused data, and the extracted independent components are reconstructed to obtain the fused data after decorrelation processing.

[0069] Random noise is added to the fused data after decorrelation processing to obtain noisy data; in this embodiment, the added random noise is Gaussian white noise.

[0070] Perform empirical mode decomposition on the noisy data to obtain the intrinsic mode function (IMF) components, and calculate the multi-scale permutation entropy of each intrinsic mode function component; specifically, the relationship between the number of extreme points and the number of zero crossings is used as a screening criterion to screen out valid IMF components from the decomposition results: for each IMF component, first calculate the number of its extreme points (including local maxima and minima) and the number of zero crossings. According to the definition of IMF components, a valid IMF component should meet the condition that the number of extreme points and the number of zero crossings are equal or differ by at most one. If the number of extreme points and the number of zero crossings of an IMF component do not meet this condition, it may indicate that the component is not a true IMF, or the decomposition process is not complete, and it is necessary to continue to decompose to obtain a component that better meets the IMF definition. In the EMD process, each time an IMF component is decomposed, it is subtracted from the original signal, and then the remaining signal is decomposed. After each iteration, the above screening criteria are applied to check whether the newly decomposed IMF components are valid.

[0071] In this embodiment, the multi-scale permutation entropy is calculated to evaluate the complexity and randomness of each IMF component. Specifically, a time series is reconstructed for each IMF component to obtain multiple coarse-grained time series. The permutation entropy is calculated for each coarse-grained time series to obtain a multi-scale sub-data sequence.

[0072] Specifically, based on the preset embedding dimension and delay time, the phase space of the coarse-grained time series is reconstructed to obtain a reconstruction matrix. For each row in the reconstruction matrix (representing a phase point), an arrangement pattern is determined according to the size relationship of the elements. The frequency of occurrence of each arrangement pattern is counted, and the arrangement entropy is calculated based on these frequencies.

[0073] Based on the multi-scale permutation entropy and frequency of each intrinsic mode function component, a multi-scale sub-data sequence is obtained. According to the multi-scale permutation entropy and frequency characteristics of the IMF components, the IMF components are divided into different scale categories. The IMF components in the same category are merged to generate a multi-scale sub-data sequence.

[0074] S4: Based on multi-scale sub-data sequences, a microgrid load forecasting model is constructed by combining time convolution with the time series pattern attention mechanism. The basic idea of ​​the time series pattern attention mechanism is that when processing time series data, the model can dynamically pay attention to the key time steps or patterns in the sequence, rather than treating all time steps equally. This is achieved by calculating the attention weight of each time step. The time step with a larger weight has a greater impact on the final prediction or decision.

[0075] A further implementation method is that the method for constructing a microgrid load prediction model includes:

[0076] The convolutional layer based on the temporal convolutional network extracts the deep features of the multi-scale sub-data sequence;

[0077] Based on the fully connected layer, the deep features are mapped to the predicted values ​​to obtain the initial load forecasting results. In this embodiment, the deep residual shrinkage network algorithm is used to improve the residual module of the temporal convolutional network to enhance the robustness and generalization ability of the model. The deep residual shrinkage network algorithm reduces noise interference and highlights important features by introducing soft thresholding and attention mechanisms. Specifically, for each subvector of the deep feature, if its modulus is less than a preset threshold, it is set to zero; otherwise, it is subtracted from the threshold (if positive) or added to the threshold (if negative). This can effectively reduce the impact of noise while retaining important feature information.

[0078] Based on cosine similarity, the attention weight between the current time step of the temporal convolutional network and the past time steps within a preset range is calculated to help the model better capture long-term dependencies in the time series.

[0079] Based on the attention weights, the initial load forecast results output in the past time steps are weighted summed to obtain the feature vector of the fusion attention mechanism;

[0080] The feature vector of the fusion attention mechanism is combined with the initial load forecast result to obtain the final load forecast result and complete the construction of the microgrid load forecasting model.

[0081] S5: Based on the microgrid load prediction model, the microgrid load to be predicted is predicted to obtain a load prediction result.

[0082] Embodiment 2

[0083] The present invention also provides a microgrid load forecasting system based on big data, which is used to implement a method, including:

[0084] A multi-source data acquisition module is used to collect multi-source data of microgrids in a preset service area using IoT sensors, and to denoise and clean the multi-source data of microgrids to obtain processed multi-source data;

[0085] A data fusion module is used to perform data-level fusion on the processed multi-source data to obtain fused data;

[0086] A multi-scale decomposition module is used to perform multi-scale decomposition on the fused data based on an empirical mode decomposition method combined with multi-scale permutation entropy to obtain a multi-scale sub-data sequence;

[0087] The prediction model building module is used to build a microgrid load prediction model based on multi-scale sub-data sequences by combining time convolution with the time series pattern attention mechanism;

[0088] The prediction module is used to predict the microgrid load to be predicted based on the microgrid load prediction model to obtain the load prediction result.

[0089] A further implementation is that the data fusion module includes:

[0090] A metadata collection unit, used to establish a metadata collection of multi-source data of the microgrid based on sensor data, and calculate an unbiased estimate of the multi-source data;

[0091] A weight set unit, used to establish a corresponding data weight set based on the metadata set;

[0092] A mean square error calculation unit, used for obtaining a total mean square error of the multi-source data based on the metadata set, the data weight set and the unbiased estimate of the multi-source data;

[0093] A weighting factor acquisition unit, used for obtaining an optimal weighting factor of each sensor by using a conjugate gradient method based on a metadata set, a data weight set, an unbiased estimate, and a total mean square error;

[0094] The data fusion unit is used to obtain fused data based on the historical mean value and the minimum mean square error of each sensor.

[0095] A further implementation is that the multi-scale decomposition module comprises:

[0096] A decorrelation processing unit, used for decorrelation processing the fused data by using an independent component analysis method;

[0097] A noise adding unit, used for adding random noise to the fused data after decorrelation processing to obtain noisy data;

[0098] A permutation entropy calculation unit is used to perform empirical mode decomposition on the noisy data to obtain intrinsic mode function components and calculate the multi-scale permutation entropy of each intrinsic mode function component;

[0099] The multi-scale sequence acquisition unit is used to obtain a multi-scale sub-data sequence based on the multi-scale permutation entropy value and frequency of each intrinsic mode function component.

[0100] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A microgrid load forecasting method based on big data, characterized in that: The method comprises: Using IoT sensors to collect multi-source data of microgrids in a preset service area, and denoising and cleaning the multi-source data of the microgrid to obtain processed multi-source data; Perform data-level fusion on the processed multi-source data to obtain fused data; Performing multi-scale decomposition on the fused data based on an empirical mode decomposition method combined with multi-scale permutation entropy to obtain a multi-scale sub-data sequence; Based on the multi-scale sub-data sequence, a microgrid load prediction model is constructed by combining a temporal convolution with a temporal pattern attention mechanism; Based on the microgrid load prediction model, the microgrid load to be predicted is predicted to obtain a load prediction result.

2. The microgrid load forecasting method based on big data according to claim 1 is characterized in that: The microgrid multi-source data includes historical power load data, meteorological data of a preset service area, social data, and new energy power generation data.

3. The microgrid load forecasting method based on big data according to claim 1 is characterized in that: Methods for obtaining fused data include: Based on the sensor data, a metadata set of microgrid multi-source data is established, and unbiased estimates of multi-source data are calculated; Based on the metadata set, establishing a corresponding data weight set; Based on the metadata set, the data weight set and the unbiased estimate of the multi-source data, a total mean square error of the multi-source data is obtained; Based on the metadata set, the data weight set, the unbiased estimate and the total mean square error, the conjugate gradient method is used to obtain the optimal weighting factor of each sensor; The fused data is obtained based on the historical mean value and the minimum mean square error of each sensor.

4. The microgrid load forecasting method based on big data according to claim 1 is characterized in that: The method of obtaining the multi-scale sub-data sequence includes: Using independent component analysis method, decorrelation processing is performed on the fused data; Add random noise to the fused data after decorrelation processing to obtain noisy data; Performing empirical mode decomposition on the noisy data to obtain intrinsic mode function components, and calculating the multi-scale permutation entropy of each intrinsic mode function component; The multi-scale sub-data sequence is obtained based on the multi-scale permutation entropy value and frequency of each eigenmode function component.

5. The microgrid load forecasting method based on big data according to claim 1 is characterized in that: The methods for constructing a microgrid load forecasting model include: Extracting deep features of the multi-scale sub-data sequence based on a convolutional layer of a temporal convolutional network; Mapping the deep features to prediction values ​​based on a fully connected layer to obtain an initial load prediction result; Based on cosine similarity, the attention weight between the current time step of the temporal convolutional network and the past time steps within a preset range is calculated; Based on the attention weights, weighted summation is performed on the initial load forecast results outputted in the past time steps to obtain a feature vector of the fused attention mechanism; The feature vector of the fusion attention mechanism is combined with the initial load forecast result to obtain the final load forecast result, thus completing the construction of the microgrid load forecasting model.

6. A microgrid load forecasting system based on big data, used to implement the method described in any one of claims 1 to 5, characterized in that: include: A multi-source data acquisition module is used to collect multi-source data of microgrids in a preset service area using IoT sensors, and to perform denoising and cleaning on the multi-source data of microgrids to obtain processed multi-source data; A data fusion module is used to perform data-level fusion on the processed multi-source data to obtain fused data; A multi-scale decomposition module, used for performing multi-scale decomposition on the fused data based on an empirical mode decomposition method combined with multi-scale permutation entropy to obtain a multi-scale sub-data sequence; A prediction model building module, used to build a microgrid load prediction model based on the multi-scale sub-data sequence by combining time convolution with a temporal pattern attention mechanism; The prediction module is used to predict the microgrid load to be predicted based on the microgrid load prediction model to obtain a load prediction result.

7. The microgrid load forecasting system based on big data according to claim 6 is characterized in that: The data fusion module comprises: A metadata collection unit, used to establish a metadata collection of multi-source data of the microgrid based on sensor data, and calculate an unbiased estimate of the multi-source data; A weight set unit, used to establish a corresponding data weight set based on the metadata set; A mean square error calculation unit, used for obtaining a total mean square error of the multi-source data based on the metadata set, the data weight set and the unbiased estimate of the multi-source data; A weighting factor acquisition unit, configured to obtain an optimal weighting factor of each sensor by using a conjugate gradient method based on the metadata set, the data weight set, the unbiased estimate and the total mean square error; The data fusion unit is used to obtain the fused data based on the historical mean value and the minimum mean square error of each sensor.

8. The microgrid load forecasting system based on big data according to claim 6 is characterized in that: The multi-scale decomposition module comprises: A decorrelation processing unit, used for performing decorrelation processing on the fused data by using an independent component analysis method; A noise adding unit, used for adding random noise to the fused data after decorrelation processing to obtain noisy data; A permutation entropy calculation unit, used for performing empirical mode decomposition on the noisy data to obtain intrinsic mode function components, and calculating the multi-scale permutation entropy of each intrinsic mode function component; The multi-scale sequence acquisition unit is used to obtain the multi-scale sub-data sequence based on the multi-scale permutation entropy value and frequency of each intrinsic mode function component.

Images