A wind-solar-storage prediction and comparison analysis method and system
By constructing a wind-solar-storage prediction model and conducting real-time comparison and trend analysis, the problems of single prediction models and insufficient dynamic updates in the wind-solar-storage system have been solved, achieving high-precision prediction and forward-looking early warning, and improving the system's operational stability and control efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA ENERGY CONSTR GUANGXI DEV INVESTMENT CO LTD
- Filing Date
- 2026-02-10
- Publication Date
- 2026-05-29
AI Technical Summary
Existing wind, solar and energy storage forecasting methods suffer from problems such as a single forecasting model structure, inability to adapt to wind and solar fluctuations, lack of dynamic judgment criteria for model deployment and update mechanisms, inability to structurally compare forecast results with actual operational data, and difficulty in identifying trend anomalies.
By collecting operating parameters of wind power, photovoltaic and energy storage equipment, a wind-solar-storage prediction model is constructed, generating the current expected value and comparing it with the actual real-time measurement value. A deviation alarm mechanism with grade classification is introduced, and trend warnings are judged by trend slope and over-limit duration. An edge computing node deployment method is used for online prediction and monthly retraining.
It improves prediction accuracy and robustness, realizes the linkage between trend anomaly identification and model performance evaluation of wind, solar and energy storage systems, and enhances the system's forward-looking early warning capability and regulation efficiency.
Smart Images
Figure CN122118675A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of new energy prediction and analysis technology, specifically to a method and system for wind, solar and energy storage prediction and comparative analysis. Background Technology
[0002] To improve the absorption capacity and dispatch efficiency of wind and solar power systems, forecasting technologies have continuously evolved, encompassing various model architectures such as short-term forecasting, ultra-short-term forecasting, and probabilistic forecasting. Meanwhile, to address the intermittency and volatility of renewable energy output, coordinated operation strategies for wind, solar, and energy storage (wind power, photovoltaics, and energy storage) have been extensively studied, integrating next-generation information technologies such as artificial intelligence, big data, and edge computing to enhance the safety and flexibility of grid operation. Currently, various wind, solar, and energy storage dispatching platforms possess basic functions such as multi-source information acquisition, data visualization, and early warning triggering; however, they still face engineering bottlenecks such as poor model generalization ability, large response latency, and inaccurate anomaly identification.
[0003] Current technologies for wind, solar, and energy storage power output prediction and model comparison still have many shortcomings. Traditional prediction methods are mostly based on a single model structure and lack dynamic response mechanisms for time-varying characteristics and nonlinear coupling relationships, leading to a significant increase in prediction errors under strong disturbance scenarios. Existing comparison methods only stay at the level of statistical bias or error indicators, failing to build a unified model performance evaluation system and making it difficult to support the structured management of multiple models. Some technologies ignore the need for continuous evolution of prediction models during deployment, lacking standardized criteria for model switching and dynamic updates, which can easily lead to misdeployment or response lag. Existing methods generally do not consider linking prediction results with real-time monitoring data, making it difficult to identify model failures or operational anomalies in a timely manner. Therefore, there is an urgent need to propose an intelligent method that can integrate trend judgment, hierarchical early warning, and prediction result comparison analysis to improve the prediction accuracy and operational robustness of wind, solar, and energy storage systems. Summary of the Invention
[0004] In view of the above-mentioned problems, the present invention is proposed.
[0005] Therefore, the technical problem solved by this invention is that existing wind, solar and energy storage prediction methods have problems such as: the prediction model structure is simple, it cannot adapt to the fluctuation of wind and solar energy, the model deployment and update mechanism lacks dynamic judgment criteria, the prediction results cannot be compared with the actual operation data in a structured way, and how to achieve the linkage between trend anomaly identification and model performance evaluation.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a wind-solar-storage prediction and comparative analysis method, which includes collecting the operating parameters of wind power, photovoltaic and energy storage equipment, and constructing a wind-solar-storage prediction model.
[0007] The current real-time operating parameters of wind power, photovoltaic and energy storage equipment are input into the wind-solar-storage prediction model to generate the current expected values, which are then compared with the actual real-time measured values to obtain the corresponding differences.
[0008] The difference is used to determine whether the current operating status has deviated. When the difference exceeds the static deviation limit, a deviation alarm mechanism based on the level classification is triggered, and alarm level information is output.
[0009] Analyze the future trend sequence output by the wind, solar and energy storage prediction model to identify whether there is a risk of continuous exceeding the limit, and determine whether a trend warning is triggered based on the trend slope and the duration of exceeding the limit.
[0010] As a preferred embodiment of the wind-solar-storage prediction and comparative analysis method described in this invention, the acquisition of operating parameters of wind power, photovoltaic, and energy storage equipment includes obtaining historical operating parameter data, including wind speed, solar irradiance, bus voltage, IGBT temperature, energy storage SOC, battery temperature, and bearing temperature, from the wind turbine, photovoltaic inverter, and energy storage battery management system, respectively. The historical operating parameters are used to construct a time series input structure at fixed time intervals, and then undergo normalization preprocessing, missing value imputation, and outlier removal.
[0011] As a preferred embodiment of the wind-solar-storage prediction and comparative analysis method described in this invention, the construction of the wind-solar-storage prediction model includes: constructing the wind-solar-storage prediction model using a time-series-based neural network structure; and training the model by minimizing the mean square error between the predicted output and the actual value. The wind-solar-storage prediction model is deployed online using edge computing nodes, and the training process involves retraining on a monthly basis.
[0012] As a preferred embodiment of the wind-solar-storage prediction and comparative analysis method described in this invention, the input of the current real-time operating parameters to the wind-solar-storage prediction model includes: the wind-solar-storage prediction model receiving the current operating parameters of wind power, photovoltaic, and energy storage devices in real time, calculating the expected values at the current moment, subtracting the expected values from the current measured values to obtain the instantaneous deviation value, and comparing the deviation value with the static deviation limit set for the target parameters to obtain the deviation determination result.
[0013] As a preferred embodiment of the wind-solar-storage prediction and comparative analysis method described in this invention, the triggering of the level-based deviation alarm mechanism includes: triggering a level-one alarm and recording log information when the deviation value exceeds a first static deviation limit; triggering a level-two alarm and sending an alarm signal to the monitoring center when the deviation value exceeds a second static deviation limit; and triggering a level-three alarm and adjusting the energy storage power or issuing a control command when the deviation value exceeds a third static deviation limit. The static deviation limits are set based on the historical mean and standard deviation of the target parameters.
[0014] As a preferred embodiment of the wind, solar, and energy storage prediction and comparative analysis method described in this invention, the identification of whether there is a continuous risk of exceeding limits includes: based on the future trend sequence output by the wind, solar, and energy storage prediction model, determining the predicted value at each time point; if three or more consecutive time points exceed the upper and lower limits of the operating parameters, it is determined that there is a trend of exceeding limits. The upper and lower limits are determined based on the historical statistical fluctuation range and safe operation procedures.
[0015] As a preferred embodiment of the wind, solar, and energy storage prediction and comparative analysis method of the present invention, the step of determining whether to trigger a trend warning based on the trend slope and the duration of exceeding the limit includes calculating the ratio of the increase or decrease in the value of consecutive points in the future trend sequence to the time interval, as an estimated value of the trend slope. When the trend slope continuously exceeds a set trend slope threshold, and the duration of the exceeding state is not less than a preset threshold time, it is determined that a trend warning event needs to be triggered, and the trend direction, warning level, and associated measurement point identifier are recorded.
[0016] As a preferred embodiment of the wind, solar and energy storage prediction and comparative analysis method of the present invention, the trend warning event includes generating a structured warning information data packet, the data packet including the warning level, trend direction, expected deviation peak, predicted over-limit start time, duration and target equipment identifier, and uploading the warning data packet to the dispatching platform.
[0017] As a preferred embodiment of the wind, solar, and energy storage prediction and comparative analysis method described in this invention, the structured early warning information data includes a structured early warning information data packet generated by an edge computing device, encapsulated in JSON format, and containing fields including early warning level, trend direction, duration, target device identifier, maximum prediction deviation, and start time. The edge computing device encrypts the data packet and transmits it to the scheduling platform via a network interface, supporting local log recording and remote interface call for coordinated execution of secure scheduling commands.
[0018] Another objective of this invention is to provide a wind-solar-storage prediction and comparison analysis system. This system can collect operating parameters of wind power, photovoltaic, and energy storage devices through a parameter acquisition and model building module, and construct wind-solar-storage prediction models. This solves the problems of existing wind-solar-storage prediction methods, such as the single prediction model structure, inability to adapt to wind and solar fluctuations, lack of dynamic judgment criteria for model deployment and update mechanisms, inability to perform structured comparison of prediction results with actual operating data, and how to link trend anomaly identification with model performance evaluation.
[0019] As a preferred embodiment of the wind, solar and energy storage prediction and comparison analysis system of the present invention, it includes a parameter acquisition and model building module, a real-time comparison and deviation acquisition module, a difference judgment and alarm triggering module, and a trend analysis and early warning generation module.
[0020] The parameter acquisition and model building module is used to collect the operating parameters of wind power, photovoltaic and energy storage equipment, and build wind, solar and energy storage prediction models.
[0021] The real-time comparison and deviation acquisition module is used to input the current real-time operating parameters of wind power, photovoltaic and energy storage equipment into the wind-solar-storage prediction model, generate the current expected value, compare it with the actual real-time measurement value, and obtain the corresponding difference value.
[0022] The difference judgment trigger alarm module is used to determine whether the current operating state has deviated based on the difference. When the difference exceeds the static deviation limit, it triggers a deviation alarm mechanism based on the level classification and outputs alarm level information.
[0023] The trend analysis and early warning generation module is used to analyze the future trend sequence output by the wind, solar and energy storage prediction model, identify whether there is a risk of continuous exceeding the limit, and determine whether to trigger a trend warning based on the trend slope and the duration of exceeding the limit.
[0024] The beneficial effects of this invention are as follows: The wind, solar, and energy storage prediction and comparative analysis method provided by this invention collects and normalizes multi-type operating parameter data to construct a historical feature sequence under a unified dimension, achieving comparability between different physical quantities and effectively eliminating dimensional interference and the influence of numerical span, providing high-quality basic data for subsequent model training and anomaly detection. This data standardization process improves the model's sensitivity to changes in data trends, ultimately achieving the beneficial effects of enhancing modeling accuracy and generalization ability.
[0025] By integrating a short-term forecasting model with a Z-score outlier removal mechanism, a multi-point trend prediction capability for the future state of wind, solar, and energy storage is formed, ensuring that the forecast output is not affected by extreme outliers. This strategy not only improves the smoothness of the forecast curve but also provides a continuous and stable sequence input for subsequent limit-breaking judgments and deviation warnings, ultimately enhancing the accuracy and reliability of future operational trend analysis and prediction.
[0026] By performing Z-score normalization on the difference between the current real-time value and the predicted reference value, and combining it with three levels of static deviation limits, a deviation level determination mechanism is constructed, enhancing the system's step-by-step response capability to minor deviations, moderate anomalies, and severe deviations. Furthermore, through log recording, information reporting, and scheduling linkage measures, the system achieves classified management of deviation anomalies, thereby improving fault perception accuracy and safety control efficiency.
[0027] By analyzing and predicting continuous exceedances in the trend sequence and introducing dual constraints of trend slope and duration, the system can accurately identify slowly evolving potential risks that may lead to accidents. This mechanism not only fills the gap in the coverage of trend-related hidden dangers by static deviation detection in S3, but also provides forward-looking early warning support for the scheduling platform, ultimately realizing the evolution of the system's operational status from responsive to predictive, and effectively enhancing the proactive defense capabilities of the control system. Attached Figure Description
[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0029] Figure 1 This is an overall flowchart of a wind, solar and energy storage prediction and comparative analysis method provided in Embodiment 1 of the present invention. Detailed Implementation
[0030] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0031] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for wind, solar and energy storage prediction and comparative analysis is provided, comprising: S1: Collect operating parameters of wind power, photovoltaic and energy storage equipment, and construct wind, solar and energy storage prediction models.
[0032] Historical operating parameter data, including wind speed, solar irradiance, bus voltage, IGBT temperature, energy storage SOC, battery temperature, and bearing temperature, were obtained from the wind turbine, photovoltaic inverter, and energy storage battery management system, respectively. The historical operating parameters were used to construct a time-series input structure at fixed time intervals, and then underwent normalization preprocessing, missing value imputation, and outlier removal.
[0033] Furthermore, this invention sets a fixed time interval of five minutes, using 12 consecutive sets of samples to form a one-hour data window. Normalization is performed using the Z-score method for each data point. Missing value imputation is performed using the sliding window mean method; for a missing value at a given time point, the average of the two points before and after it is used. Outlier removal uses the three-standard-deviation principle: if the normalized value of a given point exceeds ±3, it is considered an outlier and removed.
[0034] A wind, solar and energy storage prediction model is constructed using a time series-based neural network structure. The model is trained by minimizing the mean square error between the predicted output and the actual value.
[0035] Furthermore, the model uses a multivariate time series Long Short-Term Memory (LSTM) network structure, with the prediction target being the value that should exist at the current moment and the trend prediction for the next 6 time points (i.e., the next 30 minutes).
[0036] Historical data is sampled at fixed intervals of 5 minutes, with a sampling window of 12 steps (i.e., data from the previous hour). Each data sample is input as a two-dimensional matrix, and the model input tensor structure is represented as follows: , in, Indicates the current prediction time point The input tensor. Indicates the first Parameter data at each time point. It represents all parameter data, including multiple measurement points such as wind speed, light intensity, voltage, current, temperature and humidity, and SOC. The sampling window is 12 in this invention. This indicates that each time step contains Data.
[0037] The model structure is set as an input layer, LSTM layer 1, LSTM layer 2, and a fully connected layer. LSTM layer 1 has 128 hidden units, and LSTM layer 2 has 64 hidden units.
[0038] No. The computation process of a layer LSTM is represented as follows: , in, This represents the forget gate vector, which determines the degree to which the state from the previous time step is retained. This represents the input gate vector, which controls the current level of information writing. This represents the output gate vector, which determines the extent to which the current state vector is output to the next layer. This indicates the current state of the cell. Indicates the LSTM layer number. Indicates a hidden state. This represents the input vector of the LSTM at the current time step, which comes from the output of the previous layer. This represents the hidden state of the LSTM at the previous time step. This indicates the cell state at the previous moment. , , and This represents the weight matrix from the previous state to the current gate. , , and This represents the weight matrix of the input to the current gate. , , and This indicates the offset of each door. This represents the Sigmoid activation function. This represents the hyperbolic tangent activation function. It is represented as the Hadamard product (element-wise multiplication).
[0039] The model output consists of two parts: the current expected value prediction (used for comparison) and the future trend sequence output structure. The current expected value prediction is expressed as: , in, This indicates the model's prediction of the current expected value. This indicates that the second layer of the LSTM is at time point The hidden state output. This represents the weight matrix of the current value prediction output layer. This represents the bias term for predicting the output layer based on the current value.
[0040] The output structure of the future trend sequence is represented as follows: , in, Indicates the first Predicted values at each time point. This represents the number of steps to predict the future, which is set to 6 in this invention (i.e., predicting the next 30 minutes, with 1 step every 5 minutes). This represents the decoder function, which receives the hidden state and outputs the predicted sequence.
[0041] Furthermore, the loss function for model training is the mean squared error of the combination of the current expected value and the predicted future trend value: , in, This represents the total loss function. This represents the weighting factor between the current forecast and the trend forecast (0.4 in this invention). This represents the current actual sampled value. Indicates the future number Predicted values for each time point. Indicates the future number Actual values at each point in time.
[0042] The wind, solar and energy storage prediction model uses edge computing nodes for online prediction, and the training process is retrained on a monthly cycle.
[0043] Furthermore, the wind-solar-storage prediction model is deployed as a Docker container in the edge gateway device of the wind farm or photovoltaic power station. The model parameters are stored locally, and real-time operating parameters are received as input. The model calculation is completed through a Python inference engine. The output predicted values are used for subsequent deviation comparison and trend identification. After each inference is completed, the prediction results are uploaded to the main station system cache pool via the MQTT protocol or API interface.
[0044] Furthermore, the monthly training process includes a fixed monthly schedule for extracting historical data from the local database for the past month to construct a training set. The data is divided into a training set (80%), a validation set (10%), and a test set (10%). The Adam optimizer is used for iterative updates, with a learning rate set to 0.001. An early stopping mechanism is implemented on the validation set to prevent overfitting. After model training is complete, its mean squared error (MSE) is evaluated on the test set. If it meets a set threshold, the new model replaces the online deployment model.
[0045] The optimal model update criterion for verifying loss is expressed as follows: , in, Indicates the first The validation set loss for each epoch. This indicates the set acceptable error threshold, which is set to 0.02 in this invention. This represents the set of model parameters corresponding to the minimum loss. This indicates the epoch number during the training process. This represents the optimal epoch number that minimizes the validation loss. After model deployment, the version number is recorded and an inference log is generated for subsequent model iterations.
[0046] It should be noted that S1 aims to construct a joint parameter modeling system for wind, solar, and energy storage with high-precision time series forecasting capabilities. It employs multi-source historical operational data and introduces data processing mechanisms such as standardization, interpolation, and anomaly removal. Based on ensuring data quality, it constructs a long short-term memory network prediction model to achieve a joint output of the current expected values and future trends of key parameters. This design overcomes the limitations of traditional single-point threshold-based judgments by introducing a dual-output structure of expected values and trend sequences. This enables the system to possess time-foresight, adaptive modeling, and continuous multi-point forecasting capabilities, effectively improving the accuracy and reliability of operational status assessment and providing core support for subsequent comparative analysis and trend early warning.
[0047] S2: Input the current real-time operating parameters of wind power, photovoltaic and energy storage equipment into the wind-solar-storage prediction model to generate the current expected values, compare them with the actual real-time measurement values, and obtain the corresponding difference.
[0048] After completing the wind, solar and energy storage prediction model, the model is used to predict real-time operating parameters, and then deviation detection and processing are carried out.
[0049] The wind-solar-storage prediction model receives the current operating parameters of wind power, photovoltaic, and energy storage equipment in real time and calculates the expected values at the current moment. The expected values are subtracted from the current measured values to obtain the instantaneous deviation values, which are then compared with the static deviation limits set for the target parameters to obtain the deviation judgment results.
[0050] Furthermore, after calculating the predicted current value from the model, the system automatically performs a difference calculation between the actual measured value and the predicted value from the model, i.e., subtracting the predicted value from the actual measured value to obtain the instantaneous deviation value. This deviation value can be positive (indicating that the current state of the equipment is higher than the predicted level) or negative (indicating that the current state of the equipment is lower than the predicted level), and it has both directionality and magnitude.
[0051] The calculated instantaneous deviation value is compared point-by-point with the static deviation limit set for the target parameter to determine whether the current operating state is abnormal. The static deviation limit is a threshold range set for each type of target operating parameter. In this invention, it is uniformly set to ±2.0 for the target parameters after Z-score normalization. That is, if the absolute value of the normalized deviation of a parameter exceeds 2.0, it is considered to have a deviation. Otherwise, it is considered to be within the acceptable deviation range.
[0052] S3: Determine whether the current operating status has deviated based on the difference. When the difference exceeds the static deviation limit, trigger the deviation alarm mechanism based on the level classification and output the alarm level information.
[0053] Based on the current instantaneous deviation value, the deviation level is further determined to implement graded early warning management.
[0054] When the deviation exceeds the first static deviation limit, a level 1 alarm is triggered, and log information is recorded. When the deviation exceeds the second static deviation limit, a level 2 alarm is triggered, and an alarm signal is sent to the monitoring center. When the deviation exceeds the third static deviation limit, a level 3 alarm is triggered, adjusting the energy storage power or issuing a control command. The static deviation limits are set based on the historical mean and standard deviation of the target parameters.
[0055] To ensure consistency and portability of the judgment, the deviation value is determined based on Z-score normalized data, and all parameters have been uniformly normalized during the model input stage. The deviation level is set based on the historical mean and standard deviation of the target parameters, using the following static deviation limit setting scheme:
[0056] First static deviation limit (level 1 alarm threshold): The absolute value of the deviation is greater than 1.5.
[0057] Second static deviation limit (secondary alarm threshold): the absolute value of the deviation is greater than 2.0.
[0058] The third static deviation limit (level 3 alarm threshold): the absolute value of the deviation is greater than 3.0.
[0059] The above deviation limits are a unified judgment standard under normalized data, and the numerical values are sourced from the following sources: A deviation of 1.5 standard deviations represents a slight deviation that is still within the observable range. Deviations of 2.0 standard deviations or higher are commonly used as the boundary for judging engineering alarms. A deviation of 3.0 standard deviations is the confidence limit in anomaly identification, and combined with the 99.7% coverage principle in statistics, it is marked as a serious deviation.
[0060] If the absolute value of the instantaneous deviation of a target parameter is greater than 1.5 but not more than 2.0, the system determines it as a slight deviation, triggers a level one alarm, records the abnormal information to the equipment log system, and marks it in the daily inspection report.
[0061] If the absolute value of the deviation is greater than 2.0 but not more than 3.0, the system determines it as a moderate deviation, triggers a level 2 alarm, uploads the alarm data to the monitoring platform in a structured information format in real time, and activates the manual confirmation process.
[0062] If the absolute value of the deviation exceeds 3.0, the system determines it as a severe deviation and triggers a level 3 alarm. The system immediately links with the energy storage system to perform power adjustment operations, or sends a scheduling command suggestion to the control master station to assist in implementing control measures such as load limiting and switching operating modes.
[0063] Alarm level information is generated in real time after each round of judgment and output in a structured data format. Fields include target parameter name, current deviation value, alarm level, trigger time, and response operation identifier. It can be directly transmitted to the control system API or log auditing system.
[0064] S3 introduces a tiered static deviation limit mechanism to achieve refined identification and response control of deviations in key wind, solar, and energy storage parameters. Compared to traditional single-threshold judgment methods, this scheme supports three-level alarm classification (mild, moderate, and severe) and links different processing strategies, exhibiting higher sensitivity and adaptability. Its design enables the system not only to identify sudden anomalies but also to intervene in the early stages of deviation, improving fault prediction and local control capabilities, effectively enhancing system stability and the level of intelligent operation and maintenance.
[0065] S4: Analyze the future trend sequence output by the wind, solar and energy storage prediction model to identify whether there is a risk of continuous exceeding the limit, and determine whether to trigger a trend warning based on the trend slope and the duration of exceeding the limit.
[0066] In addition to instantaneous deviations, trend changes in the prediction results are also analyzed to identify potential risks of continuous exceeding limits.
[0067] Based on the future trend sequence output by the wind, solar, and energy storage prediction model, the predicted value at each time point is judged. If the predicted value exceeds the upper and lower limits of the operating parameters for three or more consecutive time points, it is judged that there is an over-limit trend. The upper and lower limits are determined based on the historical statistical fluctuation range and safe operation procedures.
[0068] The upper and lower limits of the operating parameters are set in this invention as follows: The upper and lower limits for the target parameters are set as follows (based on Z-score normalization): Wind speed and light intensity: upper and lower limits set at ±2.5. Bus voltage, IGBT temperature, battery temperature, and bearing temperature: upper and lower limits set at ±2.0. Energy storage SOC, current, and voltage-related sensitive parameters: upper and lower limits set at ±1.8.
[0069] The ratio of the increase or decrease in value of consecutive points in the future trend sequence to the time interval is calculated as an estimate of the trend slope. When the trend slope continuously exceeds the set trend slope threshold, and the duration of the over-limit state is not less than the preset threshold time, it is determined that a trend warning event needs to be triggered, and the trend direction, warning level, and associated measurement point identifiers are recorded.
[0070] Furthermore, to more accurately identify the strength of trend anomalies, this invention estimates the trend slope of the predicted values at consecutive time points in the future trend sequence. The trend slope is defined as the magnitude of the increase or decrease in value between two consecutive predicted time points divided by the time interval: , in, Indicates the first Segment trend slope estimate. Indicates the first Predicted values at each time point. Indicates the first Predicted values at each time point. The time interval between adjacent prediction time points is set to 5 minutes in this invention.
[0071] Based on the combined results of trend exceeding limits and trend slope judgment, a dual trigger condition for trend early warning is set. If the estimated slope value exceeds the limit for three or more consecutive time periods... If the absolute value is greater than the set threshold of 0.08, it is considered an abnormal trend strength. If the total duration of consecutive predicted values in an out-of-limit state exceeds 15 minutes, the trend persistence condition is met.
[0072] The system only triggers a trend warning event when both the trend slope strength judgment and the continuous prediction limit exceedance judgment are met simultaneously, thus avoiding false alarms or misjudgments of single-point fluctuations.
[0073] Generate a structured early warning information data packet, which includes the early warning level, trend direction, expected peak deviation, predicted start time of exceeding the limit, duration, and target equipment identifier, and upload the early warning data packet to the dispatch platform.
[0074] The structured early warning information data packets are generated by edge computing devices and encapsulated in JSON format. They include fields such as early warning level, trend direction, duration, target device identifier, maximum prediction deviation, and start time. The edge computing devices encrypt the data packets and transmit them to the scheduling platform via a network interface, supporting both local log recording and remote API calls to execute secure scheduling commands.
[0075] It should be noted that S4 identifies continuous limit exceedance risks and abnormal trend slopes by analyzing the future trend sequence output by the wind, solar, and energy storage prediction models, thus achieving forward-looking trend warnings. Its design combines continuous prediction limit exceedances with the intensity of trend changes, avoiding false alarms caused by short-term fluctuations. Simultaneously, it can identify operational anomalies that have not yet occurred but have the potential to evolve. This mechanism breaks through the traditional alarm mode based on current values, possessing stronger predictability and system control preparedness, providing intervention time for the scheduling system, and improving the operational safety and intelligence level of the wind, solar, and energy storage system.
[0076] Furthermore, S3 is a static deviation early warning system, which aims to detect whether the equipment's operating status is abnormal at the current moment. S4 is a trend early warning system, which aims to detect whether there is a continuous abnormal trend in the future. S3 ensures that the system responds promptly when anomalies occur, while S4 ensures that the system can identify potential risks in advance and optimize the control rhythm. Together, they form a dual-track risk perception framework for the stable operation of wind, solar, and energy storage systems, realizing the transformation from emergency response to proactive avoidance.
[0077] Example 2, an embodiment of the present invention, provides a wind, solar and energy storage prediction and comparison analysis system, including a parameter acquisition and model building module, a real-time comparison and deviation acquisition module, a difference judgment and alarm triggering module, and a trend analysis and early warning generation module.
[0078] The parameter acquisition and model building module is used to collect operating parameters of wind power, photovoltaic and energy storage equipment, and build wind, solar and energy storage prediction models.
[0079] The real-time comparison and deviation acquisition module is used to input the current real-time operating parameters of wind power, photovoltaic and energy storage equipment into the wind-solar-storage prediction model, generate the current expected value, compare it with the actual real-time measurement value, and obtain the corresponding difference value.
[0080] The difference judgment and alarm triggering module is used to determine whether the current operating status has deviated based on the difference. When the difference exceeds the static deviation limit, it triggers a deviation alarm mechanism based on the level classification and outputs alarm level information.
[0081] The trend analysis and early warning generation module is used to analyze the future trend sequence output by the wind, solar and energy storage prediction model, identify whether there is a risk of continuous exceeding the limit, and determine whether to trigger a trend warning based on the trend slope and the duration of exceeding the limit.
[0082] This embodiment also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the wind, solar and energy storage prediction and comparison analysis method proposed in the above embodiment.
[0083] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the wind, solar and energy storage prediction and comparison analysis method proposed in the above embodiment.
[0084] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0085] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0086] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0087] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0088] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for predicting and comparing wind, solar, and energy storage, characterized in that, include: Collect operating parameters of wind power, photovoltaic and energy storage equipment, and construct a wind-solar-storage prediction model; The current real-time operating parameters of wind power, photovoltaic and energy storage equipment are input into the wind-solar-storage prediction model to generate the current expected values, which are then compared with the actual real-time measured values to obtain the corresponding differences. The difference is used to determine whether the current operating status has deviated. When the difference exceeds the static deviation limit, a deviation alarm mechanism based on the level classification is triggered, and alarm level information is output. Analyze the future trend sequence output by the wind, solar and energy storage prediction model to identify whether there is a risk of continuous exceeding the limit, and determine whether a trend warning is triggered based on the trend slope and the duration of exceeding the limit.
2. The wind, solar, and energy storage prediction and comparative analysis method as described in claim 1, characterized in that: The collected operating parameters of wind power, photovoltaic, and energy storage equipment include, Historical operating parameter data, including wind speed, irradiance, bus voltage, IGBT temperature, energy storage SOC, battery temperature, and bearing temperature, were obtained from the wind turbine, photovoltaic inverter, and energy storage battery management system, respectively. Historical operating parameters are used to construct a time series input structure at fixed time intervals, and then normalization preprocessing, missing value imputation, and outlier removal are performed.
3. The wind-solar-storage prediction and comparative analysis method as described in claim 2, characterized in that: The construction of the wind, solar and energy storage prediction model includes, A wind-solar-storage prediction model is constructed using a time-series-based neural network structure, and the model is trained by minimizing the mean square error between the predicted output and the actual value. The wind, solar and energy storage prediction model uses edge computing nodes for online prediction, and the training process is retrained on a monthly cycle.
4. The wind-solar-storage prediction and comparative analysis method as described in claim 3, characterized in that: The current real-time operating parameters input into the wind-solar-storage prediction model include: The wind-solar-storage prediction model receives the current operating parameters of wind power, photovoltaic and energy storage equipment in real time and calculates the expected values at the current moment; Subtract the expected value from the current measured value to obtain the instantaneous deviation value, and compare the deviation value with the static deviation limit set for the target parameter to obtain the deviation judgment result.
5. The wind-solar-storage prediction and comparative analysis method as described in claim 4, characterized in that: The triggering of the deviation alarm mechanism based on the classification of levels includes, When the deviation value exceeds the first static deviation limit, a level one alarm is triggered, and log information is recorded. When the deviation value exceeds the second static deviation limit, a level two alarm is triggered, and an alarm signal is sent to the monitoring center. When the deviation exceeds the third static deviation limit, a level three alarm is triggered, and the energy storage power is adjusted or a control command is issued. The static deviation limit is set based on the historical mean and standard deviation of the target parameter.
6. The wind-solar-storage prediction and comparative analysis method as described in claim 5, characterized in that: The identification of whether there is a risk of continuous exceeding limits includes... Based on the future trend sequence output by the wind, solar and energy storage prediction model, the predicted value at each time point is judged. If the predicted value exceeds the upper and lower limits of the operating parameters for three or more consecutive time points, it is judged that there is an over-limit trend. The upper and lower limits are determined based on historical statistical fluctuation ranges and safe operation procedures.
7. The wind-solar-storage prediction and comparative analysis method as described in claim 6, characterized in that: The method of determining whether to trigger a trend warning based on the trend slope and the duration of exceeding the limit includes, Calculate the ratio of the increase or decrease in the value of consecutive points in the future trend sequence to the time interval, and use it as an estimate of the trend slope. When the trend slope continuously exceeds the set trend slope threshold, and the duration of the over-limit state is not less than the preset threshold time, it is determined that a trend warning event needs to be triggered, and the trend direction, warning level and associated measurement point identifier are recorded.
8. The wind, solar, and energy storage prediction and comparative analysis method as described in claim 7, characterized in that: The trend warning events include, Generate a structured early warning information data packet, which includes the early warning level, trend direction, expected peak deviation, predicted start time of exceeding the limit, duration, and target equipment identifier, and upload the early warning data packet to the dispatch platform.
9. The wind-solar-storage prediction and comparative analysis method as described in claim 8, characterized in that: The structured early warning information data includes, The structured early warning information data packet is generated by edge computing devices and encapsulated in JSON format. It includes fields such as early warning level, trend direction, duration, target device identifier, maximum prediction deviation, and start time. Edge computing devices encrypt data packets and transmit them to the scheduling platform via a network interface, supporting the coordinated execution of secure scheduling commands through local log recording and remote call interfaces.
10. A wind-solar-storage prediction and comparative analysis system, employing the wind-solar-storage prediction and comparative analysis method as described in any one of claims 1 to 9, characterized in that: It includes a parameter acquisition and model building module, a real-time comparison and deviation acquisition module, a difference judgment and alarm triggering module, and a trend analysis and early warning generation module; The parameter acquisition and model building module is used to collect the operating parameters of wind power, photovoltaic and energy storage equipment, and build a wind, solar and energy storage prediction model. The real-time comparison and deviation acquisition module is used to input the current real-time operating parameters of wind power, photovoltaic and energy storage equipment into the wind-solar-storage prediction model, generate the current expected value, compare it with the actual real-time measurement value, and obtain the corresponding difference value. The difference judgment trigger alarm module is used to determine whether the current operating state has deviated based on the difference. When the difference exceeds the static deviation limit, it triggers a deviation alarm mechanism based on the level classification and outputs alarm level information. The trend analysis and early warning generation module is used to analyze the future trend sequence output by the wind, solar and energy storage prediction model, identify whether there is a risk of continuous exceeding the limit, and determine whether to trigger a trend warning based on the trend slope and the duration of exceeding the limit.