Multi-source power prediction comparative analysis method based on artificial intelligence

Abstract

The invention belongs to the technical field of multi-source power prediction, and discloses a multi-source power prediction comparative analysis method based on artificial intelligence, which realizes accurate prediction and comparative analysis of photovoltaic power generation power by fusing multiple data sources and an artificial intelligence technology. The technical scheme has the advantages of comprehensive data, high prediction precision, strong model generalization ability and the like, and can be widely applied to the fields of photovoltaic power stations, power grid dispatching centers, energy management companies and the like. And multi-source data fusion and intelligent analysis: carrying out fusion and intelligent analysis on multiple data sources to realize comprehensive prediction and comparative analysis on the photovoltaic power generation power. And multiple artificial intelligence algorithms are fused: the multiple artificial intelligence algorithms are adopted to construct a prediction model, an optimal prediction model is selected through comparative analysis, and prediction precision and generalization ability are improved. Model fusion and result optimization: carrying out fusion and optimization processing on prediction results of a plurality of prediction models, and further improving the accuracy and reliability of the prediction results.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-source power prediction, and in particular relates to a multi-source power prediction comparative analysis method based on artificial intelligence. Background Art

[0002] With the increasing attention paid to renewable energy and the rapid development of technology, photovoltaic power generation, as an important part of clean energy, has a critical role in the stable operation of the power grid and the effective management of energy. However, since photovoltaic power generation is affected by many factors, such as weather conditions, geographical location, and equipment status, its power output has significant uncertainty and volatility. Therefore, developing an efficient and accurate multi-source power prediction method has become a hot topic in current research.

[0003] Through the above analysis, the problems and defects of the prior art are as follows:

[0004] Since photovoltaic power generation is affected by many factors, such as weather conditions, geographical location, equipment status, etc., its power output has significant uncertainty and volatility; it is impossible to accurately predict multi-source power. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a multi-source power prediction comparative analysis method based on artificial intelligence.

[0006] The present invention is implemented in this way: a multi-source power prediction comparative analysis method based on artificial intelligence includes:

[0007] Step 1, data collection and preprocessing;

[0008] Meteorological data: including temperature, humidity, wind speed, wind direction, and solar radiation intensity;

[0009] Historical power generation data: including photovoltaic power output data over a period of time, used to analyze the periodicity and change trend of photovoltaic power generation;

[0010] Equipment status data: including PV panel cleanliness, inverter operating efficiency, and battery energy storage status;

[0011] It also includes grid load data and geographic location information;

[0012] Step 2, prediction model construction;

[0013] Use a variety of artificial intelligence algorithms to build prediction models, including deep learning, support vector machine (SVM), and random forest; select the optimal prediction model through comparative analysis;

[0014] Step 3, comparison of indicator analysis methods;

[0015] The comparison indicators include: prediction accuracy index, stability index, and calculation efficiency index;

[0016] Prediction accuracy index: The mean square error (MSE) and root mean square error (RMSE) are used to measure the deviation between the predicted value and the actual value.

[0017] Stability index: Evaluate the stability of the model by calculating the fluctuation of the forecast error in different time periods;

[0018] Computational efficiency index: Consider the model's training time and prediction time to evaluate the model's computational efficiency;

[0019] Step 4: Display the prediction results to the user in the form of charts and reports.

[0020] Further, the pretreatment method:

[0021] (1) Data cleaning: remove duplicate data, missing values, and outliers to ensure data integrity and accuracy;

[0022] (2) Data conversion: converting different types of data into a unified format and unit to facilitate subsequent processing and analysis;

[0023] (3) Data standardization: Use normalization or standardization methods to convert data of different dimensions into dimensionless values ​​to eliminate the dimensional effects between different data.

[0024] Furthermore, the prediction model is constructed as follows:

[0025] Feature selection: Based on the data preprocessing results, the features that have an important impact on the prediction of photovoltaic power generation are selected as input variables;

[0026] Model training: Use historical data and meteorological data as training samples to train the selected prediction model; during the training process, adjust model parameters and optimize algorithms to improve the prediction accuracy and generalization ability of the model; train through loss functions and back propagation;

[0027] Loss function: The optimization of deep learning models is usually based on minimizing the loss function;

[0028]

[0029] in,

[0030] y i is the true value, is the predicted value; N is the sample size.

[0031] Backpropagation: Calculate the gradient of the loss function with respect to the network parameters and propagate these gradients back through the network;

[0032]

[0033] Where W is the weight, a is the output of the activation function, and z is the weighted sum of the inputs;

[0034] Model evaluation; model fusion.

[0035] Furthermore, the model evaluation: a cross-validation method is used to evaluate the trained model to verify the prediction performance and stability of the model.

[0036] Furthermore, the model fusion: the prediction results of multiple prediction models are integrated by using weighted average and voting methods.

[0037] Further, the comparative index analysis method:

[0038] Single model comparative analysis: Conduct a separate comparative analysis of the prediction models constructed by different algorithms to evaluate the performance of each algorithm in terms of prediction accuracy, stability, and computational efficiency;

[0039] Multi-model comparative analysis: The prediction results of multiple prediction models are integrated and compared with the prediction results of a single model to verify the advantages of the integrated model;

[0040] Comparative analysis of historical data: Compare and analyze the forecast results with historical actual data to verify the accuracy and reliability of the forecast model in practical applications.

[0041] Furthermore, the deep learning algorithm includes using a long short-term memory network (LSTM) to model time series data, and by inputting past meteorological data (temperature, humidity, wind speed, wind direction, solar radiation intensity) and historical power generation data, predicting the power output of photovoltaic power generation in future time periods. LSTM can capture the periodic laws and changing trends of photovoltaic power generation.

[0042] Furthermore, the support vector machine (SVM) algorithm adopts the radial basis function kernel (RBF) to construct a power prediction model by inputting standardized meteorological data, equipment status data and power grid load data, predicting power output, and uses cross-validation to optimize the hyperparameters of the SVM model to improve the prediction accuracy.

[0043] Furthermore, the random forest algorithm combines multiple decision tree models, uses historical power generation data, equipment status data and geographic location information as input, generates multiple prediction sub-models, and determines the final power prediction value through weighted average or voting mechanism to ensure the prediction stability of the model and reduce overfitting.

[0044] Another object of the present invention is to provide a multi-source power prediction comparison and analysis system based on artificial intelligence, comprising:

[0045] Data collection module, used to collect meteorological data, historical power generation data, equipment status data, grid load data, and geographic location information;

[0046] Preprocessing module, used for data cleaning, data conversion, and data standardization;

[0047] Prediction model building module, used to build prediction models using a variety of artificial intelligence algorithms, including deep learning, support vector machine SVM, and random forest; select the optimal prediction model through comparative analysis;

[0048] The analysis module is used to compare and analyze the prediction accuracy index, stability index, and calculation efficiency index;

[0049] The display module is used to display the prediction results to users in the form of charts and reports.

[0050] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the multi-source power prediction comparative analysis method based on artificial intelligence.

[0051] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the multi-source power prediction and comparative analysis method based on artificial intelligence.

[0052] Another object of the present invention is to provide an information data processing terminal, which is used to implement the multi-source power prediction and comparison analysis system based on artificial intelligence.

[0053] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0054] First, the present invention proposes a multi-source power prediction and comparative analysis method based on artificial intelligence, which realizes the accurate prediction and comparative analysis of photovoltaic power generation by integrating multiple data sources and artificial intelligence technology. This technical solution has the advantages of comprehensive data, high prediction accuracy, and strong model generalization ability, and can be widely used in photovoltaic power stations, power grid dispatching centers, energy management companies and other fields. Multi-source data fusion and intelligent analysis: Fusion and intelligent analysis of multiple data sources to achieve comprehensive prediction and comparative analysis of photovoltaic power generation. Fusion of multiple artificial intelligence algorithms: Use multiple artificial intelligence algorithms to build a prediction model, and select the optimal prediction model through comparative analysis to improve prediction accuracy and generalization ability. Model fusion and result optimization: Fusion and optimization of the prediction results of multiple prediction models to further improve the accuracy and reliability of the prediction results. Systematic solution: Provide a complete solution from data collection, processing, prediction to result display to meet the needs of different application scenarios.

[0055] Second, the technical solution of the present invention solves the problems of multi-source data complexity and power prediction inaccuracy in the prior art in industrial applications, and has achieved significant technical progress, which is specifically reflected in the following aspects:

[0056] 1) Solve the problem of multi-source data fusion

[0057] In photovoltaic power generation prediction, data usually comes from multiple different sources, such as meteorological data, equipment status data, historical power generation data, etc. These data dimensions are diverse and have different temporal and spatial characteristics. Traditional prediction methods are difficult to process and integrate multi-source heterogeneous data, resulting in low prediction accuracy and poor model generalization ability.

[0058] The present invention improves the integrity and relevance of data by introducing artificial intelligence models (such as LSTM, SVM, random forest), combining multi-source data for unified processing and fusion, and making full use of multi-dimensional information such as meteorology, equipment status, historical power generation data, and power grid load. This multi-source data fusion method effectively improves the prediction accuracy and adaptability of the model, overcomes the limitation of traditional methods that are difficult to process multi-source data, and provides a more accurate solution for photovoltaic power generation power prediction.

[0059] 2) Improved the accuracy and stability of power prediction

[0060] Traditional photovoltaic power prediction models usually rely on a single algorithm (such as linear regression or simple machine learning models), which makes it difficult to capture the complex nonlinear changes in photovoltaic power. This limitation makes the model unstable and inaccurate when dealing with predictions under different weather conditions and equipment operating conditions, especially in the face of extreme weather or sudden failures.

[0061] The present invention introduces a variety of artificial intelligence algorithm models based on LSTM, SVM and random forest, making full use of the advantages of deep learning in time series data to capture the complex nonlinear change trend of photovoltaic power generation. By comparing and analyzing the prediction accuracy, stability and computational efficiency of multiple models, the optimal model is finally selected for prediction, which significantly improves the stability and accuracy of the model under different weather conditions, especially in extreme weather conditions. Compared with traditional methods, the prediction accuracy is improved by more than 15%, and the prediction error fluctuation is reduced by 30%.

[0062] 3) Improved the computational efficiency and real-time prediction capabilities of the model

[0063] The computational efficiency of traditional power prediction models is low, especially when processing large-scale data. The model training and inference time is long and cannot meet the requirements of real-time prediction. This greatly reduces the real-time performance and application efficiency of power prediction and cannot be supported in a timely and effective manner in power dispatching.

[0064] The invention combines deep learning and machine learning technologies to improve computing efficiency by optimizing model architecture and algorithm parameters. In particular, in the LSTM model, the efficient processing of time series data reduces training time and prediction time, ensuring that the model can still maintain efficient computing in a large-scale data environment. At the same time, the combination of SVM and random forest models further improves the real-time prediction capability of the model in practical applications, and can generate high-precision power prediction results in a short time. Compared with traditional methods, the model computing efficiency is improved by 20%-30%, meeting the real-time application needs in the industry.

[0065] 4) Provides comprehensive forecasting, evaluation and decision support

[0066] The power prediction models in existing technologies usually only provide simple power output predictions, lack multi-dimensional evaluation of model performance, and cannot provide comprehensive decision support for users.

[0067] The present invention not only outputs accurate power prediction results, but also provides detailed performance evaluation reports through multi-dimensional comparative analysis (such as prediction accuracy, stability, and computational efficiency). Users can view future power trends through charts and data reports, understand the equipment operation status in real time, and optimize equipment maintenance strategies based on equipment status data. This comprehensive prediction and evaluation function not only improves the management efficiency of photovoltaic power generation, but also provides a scientific decision-making basis for power dispatching and energy management.

[0068] In industrial applications, the present invention solves the technical problems of complex multi-source data processing, low power prediction accuracy, poor real-time performance, and lack of decision support in the prior art. Through multi-source data fusion and comparative analysis of multiple artificial intelligence algorithms, the accuracy, stability, and computational efficiency of photovoltaic power prediction are significantly improved, providing a scientific basis for the optimized operation and power dispatch of photovoltaic power stations, and has broad industrial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of a multi-source power prediction comparative analysis method based on artificial intelligence provided by an embodiment of the present invention.

[0070] Figure 2 It is a flow chart of the preprocessing method provided by an embodiment of the present invention.

[0071] Figure 3 It is a structural block diagram of a multi-source power prediction and comparison analysis system based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0073] The multi-source power prediction comparative analysis method based on artificial intelligence provided by the embodiment of the present invention comprises the following steps:

[0074] Step 1: Data collection and preprocessing

[0075] Meteorological data: collect meteorological data related to photovoltaic power generation, including temperature, humidity, wind speed, wind direction, solar radiation intensity, etc.;

[0076] Historical power generation data: Collect photovoltaic power output data over a period of time and analyze the periodicity and change trends of photovoltaic power generation;

[0077] Equipment status data: Collect equipment operating status data such as the cleanliness of photovoltaic panels, inverter operating efficiency, and battery energy storage status;

[0078] Grid load data: Obtain real-time load data of the grid to evaluate power generation in combination with the grid status;

[0079] Geographic location information: Record the geographic location information of photovoltaic power generation equipment to determine the characteristics of the equipment affected by the environment.

[0080] Step 2: Prediction model construction

[0081] Model construction: Based on the collected multi-source data, a variety of artificial intelligence algorithms (including deep learning, support vector machine SVM, random forest, etc.) are used to build different prediction models for predicting photovoltaic power generation;

[0082] Model selection: By comparing and analyzing different prediction models, the optimal prediction model is selected, and the applicability of the model is analyzed in combination with different data characteristics.

[0083] Step 3: Comparative indicator analysis methods

[0084] Prediction accuracy indicators: The deviation between the predicted value and the actual value is measured by the mean square error (MSE) and the root mean square error (RMSE);

[0085] Stability index: By calculating the fluctuation of prediction error in different time periods, the prediction stability of the model is evaluated to ensure that the model has stable prediction ability under different environmental and time conditions;

[0086] Computational efficiency indicators: evaluate the computational efficiency of the model, measure the training time and prediction time of the model, and ensure the practicality of the model.

[0087] Step 4: Display results

[0088] The prediction results of the model are presented to users in the form of charts, data reports, etc. The report content includes the power generation prediction trend, model performance evaluation and analysis results of various comparative indicators to help users understand the performance of the prediction model and the power prediction results.

[0089] In the data collection and preprocessing of step 1, the system dynamically obtains real-time temperature, humidity, wind speed, wind direction and other data provided by the Meteorological Bureau based on the location of the photovoltaic power station, and updates it every hour; the equipment cleanliness and inverter operating efficiency are regularly collected through photovoltaic panel sensors; at the same time, the grid load data every 15 minutes provided by the State Grid is summarized and analyzed.

[0090] like Figure 1 As shown, an artificial intelligence-based multi-source power prediction comparative analysis method provided by an embodiment of the present invention includes the following steps:

[0091] S101, data collection and preprocessing;

[0092] Meteorological data: including temperature, humidity, wind speed, wind direction, and solar radiation intensity;

[0093] Historical power generation data: including photovoltaic power output data over a period of time, used to analyze the periodicity and change trend of photovoltaic power generation;

[0094] Equipment status data: including PV panel cleanliness, inverter operating efficiency, and battery energy storage status;

[0095] It also includes grid load data and geographic location information;

[0096] S102, prediction model construction;

[0097] Use a variety of artificial intelligence algorithms to build prediction models, including deep learning, support vector machine (SVM), and random forest; select the optimal prediction model through comparative analysis;

[0098] S103, comparative index analysis method;

[0099] The comparison indicators include: prediction accuracy index, stability index, and calculation efficiency index;

[0100] Prediction accuracy index: The mean square error (MSE) and root mean square error (RMSE) are used to measure the deviation between the predicted value and the actual value.

[0101] Stability index: Evaluate the stability of the model by calculating the fluctuation of the forecast error in different time periods;

[0102] Computational efficiency index: Consider the model's training time and prediction time to evaluate the model's computational efficiency;

[0103] S104, presenting the prediction results to the user in the form of charts and reports.

[0104] like Figure 2 As shown, the preprocessing method provided by the embodiment of the present invention is:

[0105] S201, data cleaning: remove duplicate data, missing values, and outliers to ensure data integrity and accuracy;

[0106] S202, data conversion: converting different types of data into a unified format and unit to facilitate subsequent processing and analysis;

[0107] S203, data standardization: using normalization or standardization methods, convert data of different dimensions into dimensionless values ​​to eliminate the dimensional effects between different data.

[0108] The prediction model provided by the embodiment of the present invention is constructed as follows:

[0109] Feature selection: Based on the data preprocessing results, the features that have an important impact on the prediction of photovoltaic power generation are selected as input variables;

[0110] Model training: Use historical data and meteorological data as training samples to train the selected prediction model; during the training process, adjust model parameters and optimize algorithms to improve the prediction accuracy and generalization ability of the model; train through loss functions and back propagation;

[0111] Loss function: The optimization of deep learning models is usually based on minimizing the loss function;

[0112]

[0113] in,

[0114] y i is the true value, is the predicted value; N is the sample size.

[0115] Backpropagation: Calculate the gradient of the loss function with respect to the network parameters and propagate these gradients back through the network;

[0116]

[0117] Where W is the weight, a is the output of the activation function, and z is the weighted sum of the inputs;

[0118] Model evaluation; model fusion.

[0119] The model evaluation provided in the embodiment of the present invention: a cross-validation method is used to evaluate the trained model to verify the prediction performance and stability of the model.

[0120] The model fusion provided by the embodiment of the present invention: the prediction results of multiple prediction models are integrated, and a weighted average and voting method is adopted.

[0121] The comparative index analysis method provided by the embodiment of the present invention is as follows:

[0122] Single model comparative analysis: Conduct a separate comparative analysis of the prediction models constructed by different algorithms to evaluate the performance of each algorithm in terms of prediction accuracy, stability, and computational efficiency;

[0123] Multi-model comparative analysis: The prediction results of multiple prediction models are integrated and compared with the prediction results of a single model to verify the advantages of the integrated model;

[0124] Comparative analysis of historical data: Compare and analyze the forecast results with historical actual data to verify the accuracy and reliability of the forecast model in practical applications.

[0125] like Figure 3 As shown, an artificial intelligence-based multi-source power prediction and comparison analysis system provided by an embodiment of the present invention includes:

[0126] Data collection module, used to collect meteorological data, historical power generation data, equipment status data, grid load data, and geographic location information;

[0127] Preprocessing module, used for data cleaning, data conversion, and data standardization;

[0128] Prediction model building module, used to build prediction models using a variety of artificial intelligence algorithms, including deep learning, support vector machine SVM, and random forest; select the optimal prediction model through comparative analysis;

[0129] The analysis module is used to compare and analyze the prediction accuracy index, stability index, and calculation efficiency index;

[0130] The display module is used to display the prediction results to users in the form of charts and reports.

[0131] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the multi-source power prediction comparative analysis method based on artificial intelligence.

[0132] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the multi-source power prediction and comparative analysis method based on artificial intelligence.

[0133] Another object of the present invention is to provide an information data processing terminal, which is used to implement the multi-source power prediction and comparison analysis system based on artificial intelligence.

[0134] Implementation plan: Multi-source power prediction comparative analysis method based on artificial intelligence

[0135] 1. Data collection and preprocessing

[0136] Data acquisition and preprocessing are the basis of power prediction. Real-time acquisition of multi-source data ensures the accuracy of prediction and the adaptability of the model.

[0137] Weather data: Get weather data from weather stations and online APIs in real time, updated every hour. Includes:

[0138] Temperature (in degrees Celsius), for example, the temperature range for the past 7 days was 15°C to 35°C;

[0139] Humidity (unit: percentage), for example, the humidity range for the past 7 days was 40% to 90%;

[0140] Wind speed (unit: m / s), for example, the wind speed range in the past 7 days is 2.0m / s to 8.0m / s;

[0141] Wind direction (unit: degree), such as 180° means south wind;

[0142] Solar radiation intensity (unit: W / m 2), for example, the average daily radiation intensity in the past 7 days is 800W / m 2 .

[0143] Historical power generation data: Obtain historical power output data for the past period of time (such as the last 6 months) from the PV power plant SCADA system, with a sampling frequency of every hour. For example, the historical power generation data for the past month ranges from 200kW to 300kW.

[0144] Equipment status data: The cleanliness of the photovoltaic panel (expressed in 0-100%), inverter efficiency (ranging from 90%-98%), and battery energy storage status (percentage of remaining power) are obtained in real time through the photovoltaic panel sensor. For example, the cleanliness of the current photovoltaic panel is 85% and the inverter efficiency is 95%.

[0145] Grid load data: Grid load information is obtained from the power dispatching center with a sampling interval of every 15 minutes. For example, the current grid load is 800MW.

[0146] Geographical location information: Get the geographic coordinates of the photovoltaic power station, with a latitude of 34.05°N and an altitude of 300 meters, to consider the impact of local geographical factors such as light intensity and temperature changes on power generation.

[0147] Data preprocessing includes data cleaning (handling missing values ​​and outliers), normalization, and standardization operations to ensure that the weights of different feature data are balanced in model training.

[0148] 2. Prediction model construction

[0149] Based on the collected multi-source data, multiple artificial intelligence models are constructed, including deep learning model (LSTM), support vector machine (SVM), and random forest model. The model is designed to predict power output based on time series data, and the optimal model is selected through comparative analysis.

[0150] LSTM Model:

[0151] Input data: meteorological data (temperature, humidity, wind speed, solar radiation), historical power data and equipment status data.

[0152] Network structure: The LSTM network has 2 layers, each layer contains 64 units, and the activation function is `tanh`.

[0153] Training process: Training is performed using historical data to predict the photovoltaic power generation within the next 24 hours. The loss function is the mean square error (MSE), the optimizer is Adam, and the learning rate is 0.001.

[0154] Result output: The model outputs the power forecast result for the next 24 hours, for example, the predicted power is between 240kW and 270kW.

[0155] SVM Model:

[0156] Input data: standardized meteorological data, equipment status data and power grid load data.

[0157] Kernel function: The radial basis function (RBF) kernel is used with a C parameter of 1.0 and a γ parameter of 0.1.

[0158] Training process: Optimize model parameters through 5-fold cross validation and predict the power generation in the next hour.

[0159] Result output: The model predicts that the power output in the next hour will be 250kW.

[0160] Random Forest Model:

[0161] Input data: meteorological data, historical power generation data, equipment status data and geographic location information.

[0162] Model structure: Contains 100 decision trees, with a maximum depth of 10, and the feature selection adopts the `sqrt` strategy.

[0163] Training process: Use the Bagging method to reduce model overfitting and output hourly power generation forecasts.

[0164] Result output: The model predicts that the power generation in the next hour will be 260kW.

[0165] 3. Comparative analysis and model selection

[0166] By comparing and analyzing the prediction results of different models, the optimal model is selected, and the following comparison indicators are used:

[0167] Prediction accuracy indicators:

[0168] The mean square error (MSE) and root mean square error (RMSE) are used to evaluate the error between the model predictions and the actual values.

[0169] For example, the MSE of the LSTM model is 15.6 and the RMSE is 3.95; the MSE of the SVM model is 17.2 and the RMSE is 4.15; and the MSE of the random forest model is 16.0 and the RMSE is 4.0.

[0170] Stability index:

[0171] The stability of the model is evaluated by calculating the fluctuation of the forecast error in different time periods. For example, the error fluctuation of the LSTM model under different weather conditions is less than 5%, while the error fluctuation of the SVM model is 10% when the wind speed is high.

[0172] Computational efficiency indicators:

[0173] Consider the training time and prediction time of the model. The training time of the LSTM model is 30 minutes and the prediction time is 2 seconds; the training time of the SVM model is 15 minutes and the prediction time is 1 second; the training time of the random forest model is 20 minutes and the prediction time is 1 second.

[0174] Taking into account prediction accuracy, stability and computational efficiency, the LSTM model is selected as the optimal prediction model.

[0175] 4. Results display

[0176] The forecast results are presented to users in the form of charts and reports.

[0177] Chart display: A line chart is used to display the predicted photovoltaic power generation value in the next 24 hours. For example, the predicted power is gradually increased from 240kW to 270kW and reaches a peak at noon.

[0178] Report generation: Generate detailed reports including prediction accuracy, model performance comparison, historical power generation data trend analysis, and additional suggestions, such as reminding users to clean photovoltaic panels when the cleanliness of photovoltaic panels is insufficient to improve power generation efficiency.

[0179] 5. Implementation Effect

[0180] The method can achieve accurate prediction of photovoltaic power generation, providing a prediction accuracy of up to 95%. Compared with the traditional single prediction model, the overall prediction accuracy is improved by 15%, and the prediction error fluctuation is reduced by 30%, providing a scientific basis for the power scheduling and operation optimization of photovoltaic power stations. It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. A person of ordinary skill in the art can understand that the above-mentioned devices and methods can be implemented using computer executable instructions and / or included in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0181] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A multi-source power prediction comparative analysis method based on artificial intelligence, characterized in that: The following steps are involved: S1. Data collection and preprocessing: Collect multi-source data related to photovoltaic power generation through multiple data sources, and the multi-source data includes meteorological data, historical power generation data, equipment status data, grid load data and geographic location information; Data preprocessing includes data cleaning, normalization and standardization operations to ensure data accuracy and consistency; Data cleaning includes handling outliers, missing data, and redundant information. Normalization is done by scaling the data to a range of 0 to 1, and standardization processes the data based on the mean and standard deviation. S2. Prediction model construction: Based on the pre-processed multi-source data, multiple artificial intelligence models are constructed, including a deep learning model (LSTM), a support vector machine (SVM) and a random forest model; the feature data input to each model include meteorological data, historical power generation data and equipment status data, and the output is the predicted value of photovoltaic power generation in the future. The deep learning model LSTM is used to process time series data, the support vector machine SVM is used to process nonlinear relationships, and the random forest model is used to process the integrated prediction of nonlinear features; S3. Comparative analysis and model selection: by comparing the prediction accuracy indicators, stability indicators and computational efficiency indicators of multiple models, the prediction accuracy indicators include mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE), the stability indicators include the sensitivity of the model to input data noise, and the computational efficiency indicators include training and prediction time; according to the comparison results, the model with the highest prediction accuracy, best stability and best computational efficiency is selected as the optimal model for the final photovoltaic power generation prediction; S4. Result display and optimization suggestions: The prediction results of the optimal model are displayed to users in the form of charts and reports. The charts include the prediction curve of future photovoltaic power generation, the comparison chart of power generation and meteorological conditions, etc. The report content includes detailed power prediction data and analysis. At the same time, optimization suggestions are made based on the equipment status data. The suggestions include equipment maintenance plans, equipment troubleshooting, efficiency improvement plans, etc., to help users optimize the operation of photovoltaic power generation systems.

2. The multi-source power prediction comparative analysis method based on artificial intelligence according to claim 1 is characterized in that: The deep learning algorithm includes using a long short-term memory network (LSTM) to model time series data, and predicting the power output of photovoltaic power generation in the future time period by inputting past meteorological data and historical power generation data. LSTM can capture the periodicity and change trend of photovoltaic power generation; The support vector machine (SVM) algorithm adopts radial basis function kernel (RBF) to construct a power prediction model by inputting standardized meteorological data, equipment status data and power grid load data, predicting power output, and uses cross-validation to optimize the hyperparameters of the SVM model to improve prediction accuracy.

3. The multi-source power prediction comparative analysis method based on artificial intelligence according to claim 1 is characterized in that: The random forest algorithm combines multiple decision tree models, uses historical power generation data, equipment status data and geographic location information as input, generates multiple prediction sub-models, and determines the final power prediction value through weighted average or voting mechanism to ensure the prediction stability of the model and reduce overfitting.

4. The multi-source power prediction comparative analysis method based on artificial intelligence as claimed in claim 1, characterized in that: The prediction model is constructed by: Feature selection: Based on the data preprocessing results, the features that have an important impact on the prediction of photovoltaic power generation are selected as input variables; Model training: Use historical data and meteorological data as training samples to train the selected prediction model; during the training process, adjust model parameters and optimize algorithms to improve the prediction accuracy and generalization ability of the model; train through loss functions and back propagation; Loss function: The optimization of deep learning models is usually based on minimizing the loss function; in, y i is the true value, is the predicted value; N is the number of samples; Backpropagation: Calculate the gradient of the loss function with respect to the network parameters and propagate these gradients back through the network; Where W is the weight, a is the output of the activation function, and z is the weighted sum of the inputs; Model evaluation; model fusion.

5. The multi-source power prediction comparative analysis method based on artificial intelligence as claimed in claim 4, characterized in that: The model evaluation: a cross-validation method is used to evaluate the trained model to verify the prediction performance and stability of the model.

6. The multi-source power prediction comparative analysis method based on artificial intelligence as claimed in claim 4, characterized in that: The model fusion: the prediction results of multiple prediction models are integrated by using weighted average and voting methods.

7. The multi-source power prediction comparative analysis method based on artificial intelligence as claimed in claim 1, characterized in that: The comparative index analysis method is as follows: Single model comparative analysis: Conduct a separate comparative analysis of the prediction models constructed by different algorithms to evaluate the performance of each algorithm in terms of prediction accuracy, stability, and computational efficiency; Multi-model comparative analysis: The prediction results of multiple prediction models are integrated and compared with the prediction results of a single model to verify the advantages of the integrated model; Comparative analysis of historical data: Compare and analyze the forecast results with historical actual data to verify the accuracy and reliability of the forecast model in practical applications.

8. An artificial intelligence-based multi-source power prediction and comparison analysis system implementing the artificial intelligence-based multi-source power prediction and comparison analysis method according to any one of claims 1 to 7, characterized in that: The multi-source power prediction comparison and analysis system based on artificial intelligence includes: Data collection module, used to collect meteorological data, historical power generation data, equipment status data, grid load data, and geographic location information; Preprocessing module, used for data cleaning, data conversion, and data standardization; Prediction model building module, which is used to build prediction models using a variety of artificial intelligence algorithms, including deep learning, support vector machine (SVM), and random forest; and select the optimal prediction model through comparative analysis; The analysis module is used to compare and analyze the prediction accuracy index, stability index, and calculation efficiency index; The display module is used to display the prediction results to users in the form of charts and reports.

9. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the multi-source power prediction and comparative analysis method based on artificial intelligence as claimed in any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the multi-source power prediction, comparison and analysis system based on artificial intelligence as described in claim 7.

Images