Distributed photovoltaic short-term power prediction method and system based on data relevance

Abstract

The invention discloses a distributed photovoltaic short-term power prediction method and system based on data relevance, and belongs to the technical field of power prediction. Comprising the following steps: acquiring historical power data and corresponding operation influence characteristics based on preset time granularity; calculating relevance between the historical power data and the operation influence characteristics and generating confidence, screening reference data according to the relevance and the confidence, and filling missing items in the historical power data according to the reference data so as to update the historical power data; the updated historical power data and the corresponding operation influence characteristics are input into a trained short-term power prediction model for processing, and a short-term power prediction result is obtained; wherein when the short-term power prediction model is trained, model parameters of the short-term power prediction model are determined through a subtraction optimization algorithm. The accuracy and robustness of distributed photovoltaic short-term power prediction can be improved, and the problem that prediction precision is influenced by data quality and needs to be improved in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power prediction, and in particular to a distributed photovoltaic short-term power prediction method and system based on data correlation. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] Distributed photovoltaic short-term power forecast refers to the forecast of the power of distributed photovoltaic power stations in the next few days, which is crucial for ensuring the stable operation of the power grid, optimizing scheduling plans and improving the utilization rate of renewable energy.

[0004] At present, short-term power prediction of distributed photovoltaic power stations is mostly carried out through machine learning algorithms and deep learning algorithms. For example, the weather forecast, historical power data and other information of distributed photovoltaic power stations are processed through neural network models such as Transformer model and LSTM network to obtain short-term power prediction results.

[0005] The prediction accuracy of the neural network model depends on high-quality historical data, but due to reasons such as facility configuration, network failures and human operation, the historical power data of distributed photovoltaic power stations may be missing in large areas. In this case, historical power data is generally processed by using mean filling or direct elimination, ignoring the factors affecting power and the correlation between them, resulting in quality problems in the processed historical power data, affecting the accuracy of short-term power prediction.

[0006] At the same time, since distributed photovoltaic power is affected by many factors, such as light intensity, ambient temperature and season, and these factors have high nonlinearity and uncertainty, the existing neural network model is prone to overfitting when predicting distributed photovoltaic short-term power. Summary of the invention In order to address the deficiencies in the prior art, the present invention provides a distributed photovoltaic short-term power prediction method, system, electronic device, computer-readable storage medium and computer program product based on data correlation, which comprehensively considers the impact of operating influencing characteristics on power and the correlation with power development to fill historical data, ensure the data quality of historical data, and improve the accuracy of distributed photovoltaic short-term power prediction.

[0007] In a first aspect, the present invention provides a distributed photovoltaic short-term power prediction method based on data correlation; A distributed photovoltaic short-term power prediction method based on data correlation, comprising: Based on the preset time granularity, obtain historical power data and corresponding operation impact characteristics; Calculating the correlation between the historical power data and the operation influencing feature and generating a confidence level, screening benchmark data according to the correlation and the confidence level, and filling in missing items in the historical power data according to the benchmark data to update the historical power data; Inputting the updated historical power data and the corresponding operation impact characteristics into the trained short-term power prediction model for processing to obtain a short-term power prediction result; Wherein, when training the short-term power prediction model, the model parameters of the short-term power prediction model are determined by a subtraction optimization algorithm.

[0008] In some implementations, the calculating the correlation between the historical power data and the operation influencing feature and generating a confidence level, screening benchmark data according to the correlation and the confidence level, and filling in missing items in the historical power data according to the benchmark data specifically includes: Calculate the correlation between historical power data and each operation influencing feature, and select multiple operation influencing features with high power correlation according to the correlation; Evaluate the confidence of the operational impact features with high power correlation, and use the operational impact feature with the highest confidence as the benchmark feature; According to the benchmark characteristic value corresponding to the power missing item in the historical power data, the power data with a difference from the benchmark characteristic value within a certain range is screened in the historical power data and weighted averaged to obtain the power value corresponding to the power missing item.

[0009] In some embodiments, the model parameters of the short-term power prediction model are determined by the subtraction optimization algorithm as follows: with the goal of minimizing the mean square error loss function, the position of the individual in the search space is updated by calculating the subtraction mean of the individual to iterate until the termination condition is met, and the model parameters are output.

[0010] In some embodiments, the short-term power prediction model includes multiple residual connected stack prediction units, the stack prediction unit includes multiple residual connected prediction blocks, the prediction block includes multiple fully connected layers, a forward feedback module and a backward feedback module, and the forward feedback module and the backward feedback module are arranged in parallel.

[0011] In some embodiments, the fully connected layer is used to process input data and predict the forward expansion coefficient and the backward expansion coefficient; the forward feedback module is used to generate input feedback according to the backward expansion coefficient, and the backward feedback module is used to generate predicted output according to the forward expansion coefficient. In some implementations, the feedforward module includes a linear layer and a bias layer connected in sequence.

[0012] In a second aspect, the present invention provides a distributed photovoltaic short-term power prediction system based on data correlation; A distributed photovoltaic short-term power prediction system based on data correlation, comprising: The acquisition module is configured to: acquire historical power data and corresponding operation impact characteristics based on a preset time granularity; A data filling module is configured to: calculate the correlation between the historical power data and the operation influencing feature and generate a confidence level, filter benchmark data according to the correlation and the confidence level, and fill in missing items in the historical power data according to the benchmark data to update the historical power data; The short-term power prediction module is configured to: input the updated historical power data and the corresponding operation impact characteristics into the trained short-term power prediction model for processing to obtain a short-term power prediction result; Wherein, when training the short-term power prediction model, the model parameters of the short-term power prediction model are determined by a subtraction optimization algorithm.

[0013] In a third aspect, the present invention provides an electronic device; An electronic device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned distributed photovoltaic short-term power prediction method based on data correlation.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium; A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned distributed photovoltaic short-term power prediction method based on data association.

[0015] In a fifth aspect, the present invention provides a computer program product; A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned distributed photovoltaic short-term power prediction method based on data association.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The technical solution provided by the present invention realizes the dynamic correction of the power time series based on the correlation between the operation impact characteristics and the power, fully exploits the data characteristics, so that the supplemented and improved power time series is close to the actual operation status of the distributed photovoltaic power station, ensures the validity and authenticity of the historical power data, and thus improves the accuracy of the short-term power forecast of the distributed photovoltaic forecast.

[0017] 2. The technical solution provided by the present invention optimizes the model parameters of the short-term power prediction model through a subtraction optimization algorithm, thereby preventing the short-term power prediction model from falling into the local optimum during the training process, preventing overfitting, and improving the prediction accuracy of short-term power prediction.

[0018] 3. The technical solution provided by the present invention adopts a short-term power prediction model formed by a stack of forward feedback modules, backward feedback modules and fully connected layers, which can effectively extract trend and periodic characteristics in time series, adapt to seasonal changes and trend changes in distributed photovoltaic power, and provide high-precision prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0020] Figure 1 A schematic diagram of a process flow of a distributed photovoltaic short-term power prediction method based on data association provided by an embodiment of the present invention; Figure 2 A schematic diagram of a network architecture of a short-term power prediction model provided by an embodiment of the present invention; Figure 3 A schematic diagram of the system architecture of a distributed photovoltaic short-term power prediction system based on data correlation provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0022] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0024] Embodiment 1 The present invention provides a distributed photovoltaic short-term power prediction method based on data association. Next, combined with Figure 1-Figure 2 , a distributed photovoltaic short-term power prediction method based on data correlation disclosed in this embodiment is described in detail. The distributed photovoltaic short-term power prediction method based on data correlation includes the following steps: S1. Based on a preset time granularity, obtain historical power data and corresponding operation impact characteristics.

[0025] In this embodiment, historical power data of the past two months and corresponding operation influencing characteristics are obtained. Here, the operation influencing characteristics include light intensity, ambient temperature, ambient humidity and weather.

[0026] S2. Calculate the correlation between the historical power data and the operation impact characteristics and generate confidence, filter the benchmark data according to the correlation and confidence, and fill in the missing items in the historical power data according to the benchmark data to update the historical power data.

[0027] In order to accurately fill in the missing power items in the historical power data, it is necessary to comprehensively consider the influence of factors such as light intensity, ambient temperature, ambient humidity and power, as well as the data reference value based on which the data quality screening and filling is based. As an implementation method, S2 specifically includes: S201, calculate the correlation between historical power data and each operation influencing feature, and select multiple operation influencing features with high power correlation according to the correlation. The correlation is expressed as follows: ; In the formula, Indicates relevance, Indicates i power observations, represents the mean of the power observations, Indicates i Operational impact characteristic observations, represents the mean of the observed values ​​of the operation influencing feature. In this embodiment, the operation influencing feature may be light intensity, ambient temperature, ambient humidity and / or weather.

[0028] S202. Evaluate the confidence of the operation-influencing features with a high correlation with power, and use the operation-influencing feature with the highest confidence as the benchmark feature; the confidence is expressed as: ; In the formula, represents the weight parameter, represents the mean of the operational impact characteristics, represents the standard deviation of the running impact characteristic, Indicates the number of missing items that affect the running feature, The number of data points representing the impact of the run on the feature.

[0029] S203 . According to the reference characteristic value corresponding to the power missing item in the historical power data, power data having a difference with the reference characteristic value within a certain range is selected from the historical power data and weighted averaged as the power value corresponding to the power missing item.

[0030] In this embodiment, the missing power items in the historical power data are filled by screening the benchmark values ​​based on correlation and confidence, so as to provide the subsequent distributed photovoltaic short-term power forecast as much as possible.

[0031] S3. Input the updated historical power data and the corresponding operation impact characteristics into the trained short-term power prediction model for processing to obtain the short-term power prediction result.

[0032] In order to fully exploit the information contained in the time series data, in this embodiment, the short-term power prediction model is designed as a plurality of residual connected stack prediction units, the stack prediction unit includes a plurality of residual connected prediction blocks, the prediction block includes 4 fully connected layers, a forward feedback module and a backward feedback module, the forward feedback module and the backward feedback module are arranged in parallel, the forward feedback module includes a linear layer and a bias layer connected in sequence, and the backward feedback module includes a linear layer and a bias layer connected in sequence. The updated historical power data and the corresponding operation impact features are input into the first stack prediction unit for step-by-step processing, and the input of each of the remaining stack prediction units is the residual output from the previous stack prediction unit, and the output of each stack prediction unit is aggregated into a final prediction output in a multi-level accumulation manner, which is the short-term power prediction result.

[0033] Next, the input data is represented as ,Here, Indicates the number of operation-affecting feature types, The data processing flow of any prediction block is as follows:

[0034] In the formula, represents the first fully connected layer, Indicates the first output data.

[0035]

[0036] In the formula, represents the second fully connected layer, Indicates the second output data.

[0037]

[0038] In the formula, represents the third fully connected layer, Indicates the third output data.

[0039]

[0040] In the formula, represents the fourth fully connected layer, Indicates the fourth output data.

[0041] ; In the formula, represents the first linear projection layer, Represents the backward expansion coefficient.

[0042] ; In the formula, represents the second linear layer, represents the positive expansion coefficient, Indicates a superscript.

[0043] ; In the formula, Indicates The prediction output of the prediction block, represents the prediction basis vector, express quantity information.

[0044] ; In the formula, Indicates input feedback, express The i elements, Represents the reverse prediction basis vector. In this embodiment, input feedback is used to give constraints on the function space of the prediction block input data. In order to improve the generalization ability of the model and avoid overfitting, as an implementation method, when training the short-term power prediction model, the model parameters of the short-term power prediction model are determined by a subtraction optimization algorithm with the goal of minimizing the mean square error loss function. The specific process is as follows: Step 1: Optimize the search space of the problem, randomly initialize the position of the search particle in the search space, and the initial particle swarm is expressed as: ; ; Where A is the overall matrix, A i It is i individuals, d represents the individual's first d dimension, N represents the number of individuals, m represents the number of decision variables, represents a random number in the interval [0, 1], Indicates The upper bound of the dimensional decision variable, Indicates The lower bound of the dimensional decision variable.

[0045] Each search particle corresponds to a solution to the optimization problem, and the set of their fitness function values ​​is represented by the vector F The formula is as follows: ; In the formula, F represents the fitness function, Indicates The fitness value corresponding to the search particle.

[0046] Step 2: Update the position of the individual in the search space by calculating the subtraction mean of the individual to iterate until the termination condition is met and output the model parameters.

[0047] Specifically, any particle The displacement in the search space is calculated by each particle The position update formula is expressed as follows: ; In the formula, Indicates i Search Agents The new position of , N is the total number of search agents, Represents a vector of dimension m.

[0048] If the updated position is better, replace the original position with the following formula, otherwise keep it as it is; expressed as: ; In the formula, express The corresponding objective function value is express The corresponding objective function value.

[0049] Embodiment 2 This embodiment discloses a distributed photovoltaic short-term power prediction system based on data association, including: The acquisition module is configured to: acquire historical power data and corresponding operation impact characteristics based on a preset time granularity; A data filling module is configured to: calculate the correlation between the historical power data and the operation influencing feature and generate a confidence level, filter benchmark data according to the correlation and the confidence level, and fill in missing items in the historical power data according to the benchmark data to update the historical power data; The short-term power prediction module is configured to: input the updated historical power data and the corresponding operation impact characteristics into the trained short-term power prediction model for processing to obtain a short-term power prediction result; Wherein, when training the short-term power prediction model, the model parameters of the short-term power prediction model are determined by a subtraction optimization algorithm.

[0050] It should be noted that the acquisition module, data filling module and short-term power prediction module correspond to the steps in Embodiment 1, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in Embodiment 1. It should be noted that the modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0051] Embodiment 3 Embodiment 3 of the present invention provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, the steps of the above-mentioned distributed photovoltaic short-term power prediction method based on data correlation are completed.

[0052] Embodiment 4 Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned distributed photovoltaic short-term power prediction method based on data correlation are completed.

[0053] Embodiment 5 Embodiment 5 of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned distributed photovoltaic short-term power prediction method based on data correlation.

[0054] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0055] 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.

[0056] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0057] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A distributed photovoltaic short-term power prediction method based on data correlation, characterized in that: include: Based on the preset time granularity, obtain the historical power data of distributed photovoltaics and the corresponding operation impact characteristics; Calculating the correlation between the historical power data and the operation influencing feature and generating a confidence level, screening benchmark data according to the correlation and the confidence level, and filling in missing items in the historical power data according to the benchmark data to update the historical power data; Inputting the updated historical power data and the corresponding operation impact characteristics into the trained short-term power prediction model for processing to obtain a short-term power prediction result; Wherein, when training the short-term power prediction model, the model parameters of the short-term power prediction model are determined by a subtraction optimization algorithm.

2. The distributed photovoltaic short-term power prediction method based on data association as claimed in claim 1, characterized in that: The calculating the correlation between the historical power data and the operation influencing feature and generating a confidence level, selecting benchmark data according to the correlation and the confidence level, and filling in missing items in the historical power data according to the benchmark data specifically includes: Calculate the correlation between historical power data and each operation influencing feature, and select multiple operation influencing features with high power correlation according to the correlation; Evaluate the confidence of the operational impact features with high power correlation, and use the operational impact feature with the highest confidence as the benchmark feature; According to the benchmark characteristic value corresponding to the power missing item in the historical power data, the power data with a difference from the benchmark characteristic value within a certain range is screened in the historical power data and weighted averaged to obtain the power value corresponding to the power missing item.

3. The distributed photovoltaic short-term power prediction method based on data association as claimed in claim 1, characterized in that: The method of determining the model parameters of the short-term power prediction model by the subtraction optimization algorithm is as follows: with the goal of minimizing the mean square error loss function, the position of the individual in the search space is updated by calculating the subtraction mean of the individual to iterate until the termination condition is met, and the model parameters are output.

4. The distributed photovoltaic short-term power prediction method based on data association as claimed in claim 1, characterized in that: The short-term power prediction model includes a plurality of residually connected stack prediction units, the stack prediction unit includes a plurality of residually connected prediction blocks, the prediction block includes a plurality of fully connected layers, a forward feedback module and a backward feedback module, and the forward feedback module and the backward feedback module are arranged in parallel.

5. The distributed photovoltaic short-term power prediction method based on data association as claimed in claim 4 is characterized in that: The fully connected layer is used to process the input data and predict the forward expansion coefficient and the backward expansion coefficient; the forward feedback module is used to generate input feedback according to the backward expansion coefficient, and the backward feedback module is used to generate a predicted output according to the forward expansion coefficient.

6. The distributed photovoltaic short-term power prediction method based on data association as claimed in claim 4, characterized in that: The forward feedback module includes a linear layer and a bias layer connected in sequence.

7. A distributed photovoltaic short-term power prediction system based on data correlation, characterized in that: include: The acquisition module is configured to: acquire historical power data and corresponding operation impact characteristics based on a preset time granularity; A data filling module is configured to: calculate the correlation between the historical power data and the operation influencing feature and generate a confidence level, filter benchmark data according to the correlation and the confidence level, and fill in missing items in the historical power data according to the benchmark data to update the historical power data; The short-term power prediction module is configured to: input the updated historical power data and the corresponding operation impact characteristics into the trained short-term power prediction model for processing to obtain a short-term power prediction result; Wherein, when training the short-term power prediction model, the model parameters of the short-term power prediction model are determined by a subtraction optimization algorithm.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the distributed photovoltaic short-term power prediction method based on data correlation according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the distributed photovoltaic short-term power prediction method based on data correlation described in any one of claims 1-6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the distributed photovoltaic short-term power prediction method based on data correlation described in any one of claims 1-6 are implemented.

Images