Multi-data prediction method and device for micro-grid, and medium

Abstract

The invention discloses a multi-data prediction method and device for a micro-grid and a medium, and relates to the technical field of micro-grids. The method comprises the steps that original data are collected and preprocessed, and photovoltaic power generation data, building power utilization data and electric vehicle charging data are obtained; introducing meteorological data, and calculating the grey correlation degree of the meteorological data and the photovoltaic power generation data to obtain strongly correlated meteorological variables; performing empirical mode decomposition on the building power consumption data and the electric vehicle charging data, and reconstructing the data into stationary sub-signals; and inputting the meteorological variables and the stationary sub-signals into an Informer model, and performing prediction through a sparse self-attention mechanism and a self-attention distillation mechanism. According to the invention, EMD is utilized to decompose a building electrical load sequence and an electric vehicle charging load sequence, so that the nonlinear influence is weakened; and in combination with a strong self-attention mechanism of an Informer algorithm, a dependency relationship in a long-time sequence is captured, so that the robustness and adaptability of the model are enhanced.

Description

Technical Field

[0001] The present application relates to the field of microgrid technology, and in particular to a multi-data prediction method, device and medium for a microgrid. Background Art

[0002] Concepts such as DC buildings, flexible buildings, and zero-energy buildings have become the focus of research in recent years. Considering the high proportion of new energy structures in buildings with a variety of distributed development, the application of building electrification has become an inevitable trend. Photovoltaic, Energy Storage, DC, Flexible, PEDF (Photovoltaic, Energy Storage, DC, Flexible, PEDF) is a key strategy that uses solar photovoltaic power generation and energy storage technology to achieve seamless connection between power generation, storage and consumption. Through flexible scheduling, PEDF improves energy utilization efficiency, reduces energy waste, and enhances the reliability and stability of the power system.

[0003] The prediction models of traditional photovoltaic storage direct-flexible power distribution systems, such as LSTM and Transformer, are difficult to effectively capture the long-range spatiotemporal correlation between input and output, resulting in the accumulation of prediction errors; time series data such as building power load and electric vehicle charging load are highly nonlinear and non-stationary, and direct prediction is prone to deviations; there are complex correlations between multiple variables such as photovoltaic meteorological factors and load characteristics, and redundant information increases the complexity of the model and affects the prediction efficiency; the computational complexity of the traditional self-attention mechanism in long-sequence prediction grows quadratically, which limits the expansion of actual application scenarios.

[0004] Through the above analysis, the problems and defects of the prior art are as follows: The prediction model of the photovoltaic storage direct-current flexible distribution system in the existing technology has poor correlation capture and large nonlinear influence, and cannot capture the dependency in long time series. The prediction error accumulates greatly and the efficiency is poor. Summary of the invention

[0005] The embodiments of the present application provide a multi-data prediction method, device and medium for a microgrid, which can solve the problems in the prior art of the prediction model of the photovoltaic storage direct flexible distribution system, such as poor correlation capture, large nonlinear influence, inability to capture dependencies in long time series, large prediction error accumulation and poor efficiency.

[0006] In the first aspect, an embodiment of the present application provides a multi-data prediction method for a microgrid, the method comprising: collecting and preprocessing raw data to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data; introducing meteorological data, calculating the grey correlation between meteorological data and photovoltaic power generation data, and obtaining strongly correlated meteorological variables; performing empirical mode decomposition on building electricity consumption data and electric vehicle charging data, and reconstructing them into stationary sub-signals; inputting meteorological variables and stationary sub-signals into an Informer model, and performing predictions through a sparse self-attention mechanism and a self-attention distillation mechanism.

[0007] In one implementation of the present application, raw data is collected and preprocessed to obtain photovoltaic power generation data, building electricity consumption data and electric vehicle charging data, specifically including: collecting raw data according to a preset time, and calculating the mathematical expectation and standard deviation of the raw data at each time point according to the time point of the raw data; setting confidence probability and limit values, and using a dynamic statistical threshold method to identify outliers in the raw data; calculating the mean of adjacent time points of the outlier, and replacing the outlier with the mean to obtain first data; using standard deviation normalization to perform a linear transformation on the first data to obtain photovoltaic power generation data, building electricity consumption data and electric vehicle charging data.

[0008] In one implementation of the present application, meteorological data is introduced, and the grey correlation between meteorological data and photovoltaic power generation data is calculated to obtain strongly correlated meteorological variables, specifically including: photovoltaic power generation data includes photovoltaic power generation power, and the grey correlation between meteorological data and photovoltaic power generation power is calculated; key meteorological variables are screened according to the correlation threshold, and the correlation threshold is dynamically adjusted according to the season; key meteorological variables and corresponding photovoltaic power generation power are combined into a first feature set.

[0009] In one implementation of the present application, empirical mode decomposition is performed on building electricity consumption and electric vehicle charging sequences and reconstructed into steady sub-signals, specifically including: performing empirical mode decomposition on building electricity consumption data and electric vehicle charging data to generate multiple inherent modal components; dividing the inherent modal components to obtain high-frequency modes and low-frequency modes; performing modal summation on high-frequency modes and low-frequency modes to obtain a steady time series signal, and the steady time series signal is the second feature set.

[0010] In one implementation of the present application, empirical mode decomposition is performed on the building power load and electric vehicle charging load sequences to generate multiple intrinsic mode components, specifically including: dividing the local area according to the time point, identifying the local maximum and minimum points in the building power load data and the electric vehicle charging data; using the cubic spline interpolation function to connect the maximum and minimum points respectively to form an upper envelope and a lower envelope respectively; taking the average of the upper envelope and the lower envelope, iteratively separating the intrinsic mode components until the residual term is a monotonic function, and obtaining multiple intrinsic mode components.

[0011] In one implementation of the present application, meteorological variables and stationary sub-signals are input into the Informer model, and prediction is performed through a sparse self-attention mechanism and a self-attention distillation mechanism, specifically including: compressing redundant calculations and extracting long sequence key features of the first feature set and the second feature set through a sparse self-attention mechanism and a self-distillation mechanism in the encoder; using a one-step generative reasoning method to infer long sequence key features in the decoder, and outputting prediction results for a preset period, the prediction results including the peak power of photovoltaic power generation, the predicted peak of building load, and the peak of electric vehicle charging load.

[0012] In one implementation of the present application, after the decoder uses one-step generative reasoning to output the prediction results for a preset period, the method also includes: dynamically adjusting the charging threshold of the energy storage system according to the peak power of photovoltaic power generation; and presetting the energy storage capacity according to the peak charging load of the electric vehicle.

[0013] In one implementation of the present application, after the meteorological variables and the stationary sub-signals are input into the Informer model and predicted through the sparse self-attention mechanism and the self-attention distillation mechanism, the method also includes: performing incremental learning updates on the Informer model; collecting real-time data, verifying the deviation between the prediction result and the real-time data through a sliding window mechanism, and performing feedback calibration based on the deviation.

[0014] In the second aspect, an embodiment of the present application also provides a multi-data prediction device for a microgrid, the device comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: collect raw data and pre-process it to obtain photovoltaic power generation data, building electricity consumption data and electric vehicle charging data; introduce meteorological data, calculate the grey correlation between meteorological data and photovoltaic power generation data, and obtain strongly correlated meteorological variables; perform empirical mode decomposition on building electricity consumption data and electric vehicle charging data, and reconstruct them into stationary sub-signals; input the meteorological variables and stationary sub-signals into the Informer model, and perform predictions through sparse self-attention mechanism and self-attention distillation mechanism.

[0015] On the third aspect, an embodiment of the present application also provides a multi-data prediction non-volatile computer storage medium for a microgrid, storing computer executable instructions, and the computer executable instructions are set to: collect and pre-process raw data to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data; introduce meteorological data, calculate the grey correlation between meteorological data and photovoltaic power generation data, and obtain strongly correlated meteorological variables; perform empirical mode decomposition on building electricity consumption data and electric vehicle charging data, and reconstruct them into stationary sub-signals; input meteorological variables and stationary sub-signals into the Informer model, and perform predictions through sparse self-attention mechanism and self-attention distillation mechanism.

[0016] The embodiment of the present application provides a multi-data prediction method, device and medium for a microgrid. The model analyzes the correlation between photovoltaic meteorological factor variables through GRA, reduces the spatial complexity of input data and eliminates redundant information; uses EMD to decompose the building power load and electric vehicle charging load series, weakens their nonlinear effects; combines the powerful self-attention mechanism of the Informer algorithm to capture the dependencies in long time series, greatly enhancing the robustness and adaptability of the model. Effectively improve the accuracy and stability of multi-data prediction. Based on the actual data of photovoltaic output power, building power load and electric vehicle charging load, the model performance is verified using indicators such as mean square error (MSE), mean absolute error (MAE), and root mean square error (RMSE), which proves the significant effect of the proposed method in improving prediction ability and practicality, and provides strong data support for the power planning and application of photovoltaic, energy storage, DC, Flexible (PEDF) distribution system, which has important practical significance and application potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic diagram of a flow chart of a multi-data prediction method for a microgrid provided in an embodiment of the present application; Figure 2 A solar-storage direct-current flexible power distribution system architecture diagram of a microgrid multi-data prediction method provided in an embodiment of the present application; Figure 3 A logical schematic diagram of a GRA / EMD-Informer integrated prediction model of a multi-data prediction method for a microgrid provided in an embodiment of the present application; Figure 4 A schematic diagram of the overall framework of an Informer model of a multi-data prediction method for a microgrid provided in an embodiment of the present application; Figure 5 A schematic diagram of the internal structure of a multi-data prediction device for a microgrid provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0019] The embodiments of the present application provide a multi-data prediction method, device and medium for a microgrid, which solve the problems in the prior art of the prediction model of the photovoltaic storage direct-current flexible distribution system, such as poor correlation capture, large nonlinear influence, inability to capture dependencies in long time series, large prediction error accumulation and poor efficiency.

[0020] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0021] Figure 1 A flowchart of a multi-data prediction method for a microgrid provided in an embodiment of the present application. Figure 1 As shown, a multi-data prediction method for a microgrid provided in an embodiment of the present application specifically includes the following steps: Step 10: Collect raw data and preprocess them to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data.

[0022] In this step, in order to optimize the energy storage capacity configuration and scheduling strategy of the “PV-storage-direct-flexible” distribution system, Figure 2 As shown in the figure, photovoltaic power generation data: radiation, component temperature, ambient temperature and relative humidity are used as input variables, actual power generation power is used as the target value, typical office building electricity consumption data: active power, reactive power and temperature are used as load characteristics, and the power average of each stage of electric vehicle load data is used as data feature.

[0023] As an optional embodiment, raw data is collected and preprocessed to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data, which may specifically include: Step 101: collecting raw data according to a preset time, and calculating the mathematical expectation and standard deviation of each time point of the raw data according to the time point of the raw data; Understandably, reference Figure 3First of all, the original data may be disturbed by external factors. At the same time, the photovoltaic power output power and load itself also have certain volatility. Despite these short-term instabilities, the overall data change trend basically follows the law of time series changes. Through appropriate data processing and technical means, these fluctuations can be effectively dealt with to ensure the accuracy and reliability of the prediction model.

[0024] In this step, for example, the sampling interval is 15 minutes, and the 3σ statistical discrimination method is used to set the confidence probability and limit value to identify whether the difference between the data at a certain time point in the collected original data and the previous and next time points exceeds the limit. All the sequence data in one day are { }, is the measured value at the i-th moment of a certain day, the mathematical expectation E(I) and the standard deviation Calculation formula:

[0025]

[0026] Step 102: Set the confidence probability and limit value, and use the dynamic statistical threshold method to identify outliers in the original data.

[0027] In this step, referring to the mathematical expectation and standard deviation, the power value at a time point in a day is usually (E(i)-3 (i), E(i)+3 ) range, if it is not within this range, it is judged as an abnormal point.

[0028] Step 103: Calculate the mean of the time points adjacent to the outlier, and replace the outlier with the mean to obtain the first data.

[0029] In this step, when an abnormal data point is identified, the average value algorithm is used to replace the abnormality x. The specific calculation is shown in the formula:

[0030] Step 104: linearly transform the first data using standard deviation standardization to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data.

[0031] In this step, in order to optimize the energy storage capacity configuration and scheduling strategy of PEDF and solve the problems of slow convergence, gradient disappearance, gradient explosion and overfitting in model training, the standard deviation standardization (StandardScaler) method is used to perform linear transformation on the original data; in view of the fact that the original data of the PEDF distribution system is noisy and unstable, accompanied by extreme maximum and minimum values, by calculating the mean and standard deviation of the training set, these statistics are used as the mean and variance of the population to ensure that the test data set also adopts the same transformation rules to maintain data consistency. The conversion function calculation formula is:

[0032] in is the mean of all sample data, is the standard deviation of all sample data, calculated as follows:

[0033]

[0034] This conversion method not only maps the original data to a smaller specific interval, but also retains the relative order and distribution characteristics of the data without changing its arrangement order; it effectively alleviates the impact of extreme values ​​in the data on model training, improves the generalization ability and prediction accuracy of the model, and thus enhances the power planning and application effect of the PEDF distribution system, demonstrating its wide applicability and important value in the field of smart grids. After the original data is replaced with outliers for detection and standardization, the new data set can be divided into training set, test set and validation set.

[0035] Step 20: Introduce meteorological data, calculate the grey correlation between meteorological data and photovoltaic power generation data, and obtain strongly correlated meteorological variables.

[0036] In this step, Grey Relation Analysis (GRA) is used to obtain the independent variable information with the best impact on the dependent variable in the complex independent variable relationship, which plays a decision-making effect in screening the optimal independent variable.

[0037] As an optional embodiment, meteorological data is introduced, and the grey correlation between meteorological data and photovoltaic power generation data is calculated to obtain strongly correlated meteorological variables, which may specifically include: Step 201: Photovoltaic power generation data includes photovoltaic power generation power, and the grey correlation between meteorological data and photovoltaic power generation power is calculated.

[0038] In this step, we first determine the reference series that reflects the system behavior characteristics and the comparison series that affects the system behavior. The meteorological data is subtracted from the real-time photovoltaic power value and the absolute value is taken. i(k)=|x0(t k )-x i (t k )|, calculate the correlation coefficient between the reference series and the comparison series. Each difference has its corresponding correlation coefficient. The size of the difference can be used as a measure of the degree of correlation. Calculate the correlation coefficient of each meteorological factor and photovoltaic power separately. :

[0039] Among them, MIN2 and MAX2 are the minimum difference between the two levels and the maximum difference between the two levels respectively. , is an adjustable coefficient, and its value is generally (0, 1); because the correlation coefficient is the correlation degree between the comparison series and the reference series at each moment, and the information is too scattered to be convenient for overall comparison, the correlation coefficients at each moment are concentrated into one value, that is, the average value is calculated, and the correlation degree is calculated as follows:

[0040] Step 202: Filter key meteorological variables according to the correlation threshold, and the correlation threshold is dynamically adjusted according to the season.

[0041] Step 203: combining the key meteorological variables and the corresponding photovoltaic power generation power into a first feature set.

[0042] Step 30: Perform empirical mode decomposition on the building electricity consumption data and electric vehicle charging data and reconstruct them into stable sub-signals.

[0043] As an optional embodiment, empirical mode decomposition is performed on the building electricity consumption and electric vehicle charging sequences to reconstruct them into stable sub-signals, which may specifically include: Step 301: empirical mode decomposition is performed on the building electricity consumption data and the electric vehicle charging data to generate multiple intrinsic mode components.

[0044] In this step, in order to reduce the error caused by directly predicting non-stationary time series data, EMD is used to decompose and reconstruct the original signal. The signal is decomposed according to the time scale characteristics of the data itself, without pre-setting the basis function. It can meet the needs of converting nonlinear and non-stationary signal sequences such as office building power load and electric vehicle charging station load into stationary linear data, and can provide useful information of the original signal sequence, which promotes the improvement of model prediction accuracy.

[0045] As an optional embodiment, empirical mode decomposition is performed on the building power load and electric vehicle charging load sequences to generate multiple intrinsic mode components, which may specifically include: Step 3011: Divide the local area according to the time point, and identify the local maximum and minimum points in the building power load data and electric vehicle charging data.

[0046] In this step, all the maximum and minimum points are found according to the original data sequence x(t).

[0047] Step 3012: Use a cubic spline interpolation function to connect the maximum value point and the minimum value point respectively to form an upper envelope and a lower envelope respectively.

[0048] In this step, the cubic spline interpolation function is used to fit the upper envelope Us(t) and lower envelope Vs(t) of the original data.

[0049] Step 3013: average the upper envelope and the lower envelope, iteratively separate the eigenmode components until the residual term is a monotonic function, and obtain multiple intrinsic mode components.

[0050] In this step, the mean envelope m(t) is obtained according to Us(t) and Vs(t):

[0051] The intermediate signal c(t) is obtained by subtracting the mean envelope from the original data sequence. c(t) is the intrinsic modal function (IMF):

[0052] To determine whether c(t) satisfies IMF, the absolute difference between the number of extreme points and the number of zero-crossing points in the entire data segment is less than 1; the average value of the upper envelope formed by the maximum points and the lower envelope formed by the minimum points at any time is zero; that is, if m(t) still has negative local maxima and positive local minima, it means that this is not an intrinsic mode function yet, and the above steps need to be repeated for screening; when c(t) satisfies the above two conditions, it is an IMF component, let c(t)=IMF1, and then subtract the original signal from IMF1 as the new original signal. Repeat the above steps until the nth order IMF component or residual is found. When is a monotonic function or a constant, the decomposition is completed, and finally n IMF components that meet the requirements and a residual r(t) can be obtained:

[0053]

[0054] Step 302: Divide the natural modal components to obtain high-frequency modes and low-frequency modes.

[0055] Step 303: performing modal summation on the high-frequency mode and the low-frequency mode to obtain a stationary time series signal, where the stationary time series signal is the second feature set.

[0056] Step 40: Input meteorological variables and stationary sub-signals into the Informer model and make predictions through the sparse self-attention mechanism and self-attention distillation mechanism.

[0057] As an optional embodiment, meteorological variables and stationary sub-signals are input into the Informer model, and prediction is performed through a sparse self-attention mechanism and a self-attention distillation mechanism, which may specifically include: Step 401: In the encoder, redundant calculations are compressed through a sparse self-attention mechanism and a self-distillation mechanism, and long sequence key features of the first feature set and the second feature set are extracted.

[0058] The calculation formula of the self-attention mechanism is as follows:

[0059] Writing this in probability form:

[0060] In the formula: , , are the i-th row of Q, K, and V respectively, , Then, we use KL divergence to establish a measurement standard, measure and evaluate the query sparsity, select the elite items with the highest scores, calculate the dot product of their query and key, and fill the remaining lazy queries with the mean of the input of the Self-Attention layer to ensure that the input and output sequence lengths of each ProSparseSelf-attention layer are consistent:

[0061]

[0062]

[0063]

[0064] Where: and Q are sparse matrices of the same size.

[0065] like Figure 4 As shown in the figure, the encoder processes a large amount of long sequence inputs, extracts the long-term dependencies of robust long sequence inputs through Self-attention Distilling, privileges the dominant features with dominant features by using distilling, adds convolutional pooling operations between adjacent attention blocks, and downsamples the features, as shown in the green part. The distillation process from layer j to layer (j+1) is shown in the formula:

[0066] Where: Represents the multi-head self-attention mechanism; convld represents the ELU activation function of the convolution filter in the time dimension; the decoder consists of two identical multi-attention layers. The gray part in the input is the timestamp information, the dark yellow is a start sequence, and the 0 part is the placeholder of the target sequence. After the input is sent to the decoder, the target sequence is obtained in the placeholder. Among them, the vector input to the decoder is:

[0067] The hidden attention mechanism of the decoder does not select a specific token as the starting sequence. It uses generative reasoning to generate the target dimension sequence in one step, and a fully connected layer obtains the final output.

[0068] Step 402: Using one-step generative reasoning long sequence key features in the decoder, the prediction results for a preset period are output, and the prediction results include the photovoltaic power generation peak, the building load prediction peak, and the electric vehicle charging load peak.

[0069] The above content can be used to obtain forecast data on building power load, photovoltaic output power, and electric vehicle charging load, which can be used as a reference for formulating microgrid system charging and discharging plans and capacity configuration strategies for energy storage systems.

[0070] As an optional embodiment, after the decoder uses one-step generative reasoning to output the prediction results for a preset period, the method may also include: Step 50: dynamically adjusting the charging threshold of the energy storage system according to the peak power of photovoltaic power generation; presetting the energy storage capacity according to the peak charging load of the electric vehicle.

[0071] As an optional embodiment, after the meteorological variables and the stationary sub-signal are input into the Informer model and prediction is performed through the sparse self-attention mechanism and the self-attention distillation mechanism, the method may further include: Step 60: performing incremental learning updates on the Informer model; collecting real-time data, verifying the deviation between the prediction result and the real-time data through the sliding window mechanism, and performing feedback calibration based on the deviation.

[0072] In summary, the embodiment of the present application integrates a multi-data prediction model based on grey relational analysis (GRA) and integrated empirical mode decomposition (EEMD), combined with the Informer algorithm, for the optimization of energy storage capacity configuration and scheduling strategy in the photovoltaic storage direct flexible power distribution system. By calculating the load hour by hour and improving the prediction accuracy, using GRA to evaluate the correlation between variables, using EEMD to separate the different frequency components of the original data, and using the Informer algorithm with self-attention distillation mechanism to capture the precise long-range correlation between input and output, the spatiotemporal complexity is effectively reduced.

[0073] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, the embodiment of this application also provides a multi-data prediction device for a microgrid, whose structure is as follows: Figure 5 shown.

[0074] Figure 5 This is a schematic diagram of the internal structure of a multi-data prediction device for a microgrid provided in an embodiment of the present application. Figure 5 As shown, the device includes: at least one processor 501; and, a memory 502 communicatively connected to the at least one processor; Among them, the memory 502 stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor 501 so that at least one processor 501 can: collect raw data and preprocess it to obtain photovoltaic power generation data, building electricity consumption data and electric vehicle charging data; introduce meteorological data, calculate the grey correlation between meteorological data and photovoltaic power generation data, and obtain strongly correlated meteorological variables; perform empirical mode decomposition on building electricity consumption data and electric vehicle charging data, and reconstruct them into stationary sub-signals; input meteorological variables and stationary sub-signals into the Informer model, and predict them through sparse self-attention mechanism and self-attention distillation mechanism.

[0075] Some embodiments of the present application provide corresponding Figure 1 A non-volatile computer storage medium for multi-data prediction of a microgrid stores computer executable instructions, and the computer executable instructions are set to: collect original data and preprocess them to obtain photovoltaic power generation data, building power consumption data and electric vehicle charging data; introduce meteorological data, calculate the grey correlation between meteorological data and photovoltaic power generation data, and obtain strongly correlated meteorological variables; perform empirical mode decomposition on building power consumption data and electric vehicle charging data, and reconstruct them into stationary sub-signals; input meteorological variables and stationary sub-signals into an Informer model, and predict them through a sparse self-attention mechanism and a self-attention distillation mechanism.

[0076] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the IoT device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0077] The system and medium provided in the embodiments of the present application correspond one-to-one to the method. Therefore, the system and medium also have similar beneficial technical effects to the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the system and medium will not be repeated here.

[0078] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0079] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0080] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0082] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0083] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0084] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0085] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0086] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A multi-data prediction method for a microgrid, characterized in that: The method comprises: Collect raw data and pre-process them to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data; Introducing meteorological data, calculating the grey correlation between the meteorological data and photovoltaic power generation data, and obtaining strongly correlated meteorological variables; Performing empirical mode decomposition on the building electricity consumption data and the electric vehicle charging data to reconstruct them into stable sub-signals; The meteorological variables and stationary sub-signals are input into the Informer model and predicted through the sparse self-attention mechanism and the self-attention distillation mechanism.

2. A multi-data prediction method for a microgrid according to claim 1, characterized in that: Collect and pre-process the raw data to obtain photovoltaic power generation data, building electricity consumption data and electric vehicle charging data, including: Collecting raw data according to a preset time, and calculating the mathematical expectation and standard deviation of each time point of the raw data according to the time point of the raw data; Setting confidence probabilities and limits, and using a dynamic statistical threshold method to identify outliers in the raw data; Calculate the mean of the time points adjacent to the abnormal value, and replace the abnormal value with the mean to obtain the first data; The first data are linearly transformed by standard deviation standardization to obtain photovoltaic power generation data, building electricity consumption data and electric vehicle charging data.

3. A multi-data prediction method for a microgrid according to claim 2, characterized in that: Meteorological data are introduced, and the grey correlation between the meteorological data and photovoltaic power generation data is calculated to obtain strongly correlated meteorological variables, including: The photovoltaic power generation data includes photovoltaic power generation, and the grey correlation between the meteorological data and the photovoltaic power generation is calculated; Screening key meteorological variables according to a correlation threshold, wherein the correlation threshold is dynamically adjusted according to the season; The key meteorological variables and the corresponding photovoltaic power generation power are combined into a first feature set.

4. A multi-data prediction method for a microgrid according to claim 3, characterized in that: The building electricity consumption and electric vehicle charging sequence are subjected to empirical mode decomposition and reconstructed into stable sub-signals, specifically including: Performing empirical mode decomposition on the building electricity consumption data and the electric vehicle charging data to generate a plurality of intrinsic mode components; Dividing the natural modal components to obtain high-frequency modes and low-frequency modes; Modal summation is performed on the high-frequency mode and the low-frequency mode to obtain a steady time series signal, and the steady time series signal is the second feature set.

5. A multi-data prediction method for a microgrid according to claim 4, characterized in that: The empirical mode decomposition of the building power load and electric vehicle charging load series is performed to generate multiple intrinsic mode components, including: Divide the local area according to the time point, and identify the local maximum and minimum points in the building power consumption data and the electric vehicle charging data; Using a cubic spline interpolation function to connect the maximum value point and the minimum value point respectively to form an upper envelope and a lower envelope respectively; The upper envelope and the lower envelope are averaged, and the eigenmode components are separated iteratively until the residual term is a monotonic function, thereby obtaining a plurality of eigenmode components.

6. A multi-data prediction method for a microgrid according to claim 4, characterized in that: The meteorological variables and stationary sub-signals are input into the Informer model, and prediction is performed through the sparse self-attention mechanism and the self-attention distillation mechanism, specifically including: In the encoder, redundant calculations are compressed and key features of long sequences of the first feature set and the second feature set are extracted through a sparse self-attention mechanism and a self-distillation mechanism; The decoder uses one-step generative reasoning to infer the long sequence key features and outputs prediction results for a preset period, wherein the prediction results include a photovoltaic power generation peak, a building load prediction peak, and an electric vehicle charging load peak.

7. A multi-data prediction method for a microgrid according to claim 6, characterized in that: After the decoder uses one-step generative reasoning to output the prediction result of the preset period, the method further includes: Dynamically adjust the charging threshold of the energy storage system according to the peak power of photovoltaic power generation; The energy storage capacity is preset according to the peak charging load of the electric vehicle.

8. A multi-data prediction method for a microgrid according to claim 6, characterized in that: After the meteorological variables and stationary sub-signals are input into the Informer model and predicted by the sparse self-attention mechanism and the self-attention distillation mechanism, the method further includes: Performing incremental learning updates on the Informer model; Collect real-time data, verify the deviation between the prediction result and the real-time data through a sliding window mechanism, and perform feedback calibration based on the deviation.

9. A multi-data prediction device for a microgrid, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: Collect raw data and pre-process them to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data; Introducing meteorological data, calculating the grey correlation between the meteorological data and photovoltaic power generation data, and obtaining strongly correlated meteorological variables; Performing empirical mode decomposition on the building electricity consumption data and the electric vehicle charging data to reconstruct them into stable sub-signals; The meteorological variables and stationary sub-signals are input into the Informer model and predicted through the sparse self-attention mechanism and the self-attention distillation mechanism.

10. A non-volatile computer storage medium for multi-data prediction of a microgrid, storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Collect raw data and pre-process them to obtain photovoltaic power generation data, building electricity consumption data, and electric vehicle charging data; Introducing meteorological data, calculating the grey correlation between the meteorological data and photovoltaic power generation data, and obtaining strongly correlated meteorological variables; Performing empirical mode decomposition on the building electricity consumption data and the electric vehicle charging data to reconstruct them into stable sub-signals; The meteorological variables and stationary sub-signals are input into the Informer model and predicted through the sparse self-attention mechanism and the self-attention distillation mechanism.

Images