New energy consumption prediction and early warning method, system, device and medium

Abstract

The invention relates to the technical field of new energy power systems, in particular to a new energy consumption prediction and early warning method, system, equipment and medium, and the method comprises the steps: collecting and preprocessing multi-source data in real time, and constructing a feature vector to support hierarchical collaborative prediction: firstly, training a new energy power generation prediction model based on historical weather and power generation data; secondly, combining load characteristics and influence factors to establish a load prediction model, and finally combining real-time parameters of a power grid to construct a consumption capability prediction model; early warning is triggered by dynamically comparing the generated power with the consumption capability predicted value, and a source-grid-load-storage cooperative control strategy is generated to execute regulation and control; the data coupling relation is deeply mined through the hierarchical modeling architecture, accurate quantification of the consumption potential and risk prospective early warning are achieved, the prediction timeliness and the power grid toughness are remarkably improved, power abandoning is effectively restrained, and the operation economical efficiency is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new energy power system, and particularly relates to a new energy consumption prediction and early warning method, system equipment and medium. BACKGROUND

[0002] With the emphasis on environmental protection, new energy power generation such as wind power generation and solar power generation has been widely used. However, new energy power generation has the characteristics of intermittency, volatility and uncertainty, which brings great challenges to the stable operation of the power grid and the effective consumption of new energy. Accurate prediction of new energy consumption capacity and timely warning are of great significance to the planning, scheduling and operation of the power grid.

[0003] At present, the existing new energy consumption prediction and early warning method has the problems of low prediction accuracy, untimely warning and the like, and cannot meet the actual demand. Therefore, a more efficient and accurate new energy consumption prediction and early warning system and method are urgently needed. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a new energy consumption prediction and early warning method and system to solve the problems of insufficient prediction accuracy of power grid consumption capacity, delayed warning response and invalidation of abandoned power risk control caused by the intermittency and volatility of new energy power generation.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a new energy consumption prediction and early warning method, comprising:

[0008] Real-time collection of new energy related data and transmission of data;

[0009] Preprocessing and data conversion of the collected new energy related data;

[0010] Feature extraction of the processed data to generate a feature vector;

[0011] Training of the first prediction model, the second prediction model and the third prediction model using historical data to determine the parameters of the models, and using the trained prediction models to perform real-time prediction of new energy consumption capacity in the next time period;

[0012] Comparison and judgment of the two types of prediction results of real-time prediction, and issuance of a warning signal;

[0013] Generation of a corresponding control strategy according to the warning signal, and execution of a control operation.

[0014] As a preferred scheme of the new energy consumption prediction and early warning method, the real-time collection of new energy related data and data transmission comprises:

[0015] The new energy related data comprises new energy power generation data, power grid data, load data and meteorological data.

[0016] The new energy related data is collected in real time through sensors, smart phones, hardware devices of meteorological monitoring equipment and data interfaces, and data transmission is performed.

[0017] As a preferred scheme of the new energy consumption prediction and early warning method, the preprocessing and data conversion of the collected new energy related data comprise:

[0018] The data cleaning is performed by using a statistical method.

[0019] The missing values are filled by using interpolation.

[0020] The data conversion is performed to convert the data into time series and numerical data.

[0021] As a preferred scheme of the new energy consumption prediction and early warning method, the feature extraction of the processed data to generate a feature vector comprises:

[0022] The feature extraction comprises time features, meteorological features, load features and power grid features.

[0023] High-order features are generated through feature combination and transformation.

[0024] As a preferred scheme of the new energy consumption prediction and early warning method, the training of the first prediction model, the second prediction model and the third prediction model by using historical data to determine the parameters of the models, and the real-time prediction of the new energy consumption capacity in the next time period by using the trained prediction model comprise:

[0025] The new energy power generation prediction model is established by using historical new energy power generation data and meteorological data.

[0026] The load prediction model is established by using historical load data and related influencing factors.

[0027] The new energy consumption capacity prediction model is established according to the new energy power generation prediction result, the load prediction result and the power grid data.

[0028] The beneficial effects of the preferred technical solution are: through hierarchical modeling and collaborative prediction mechanism, the accuracy and timeliness of new energy consumption capacity prediction are significantly improved. Specifically, the new energy generation prediction model is trained using historical new energy generation data and meteorological data, the load prediction model is constructed combining historical load data and multi-dimensional influencing factors, and finally the new energy consumption capacity prediction model is established by combining real-time operation parameters of the power grid. This hierarchical prediction architecture fully explores the coupling relationship between data, guarantees the professionalism of each sub-model, and accurately quantifies the power grid consumption potential through multi-source information fusion; the model training process fully learns the historical operation rules, so that the real-time prediction result can dynamically reflect the real carrying capacity of the power grid, provide high reliability decision basis for early warning and active regulation, effectively reduce the risk of new energy curtailment, and enhance the resilience and economy of power grid operation.

[0029] As a preferred scheme of the new energy consumption prediction and early warning method, the real-time predicted two types of prediction results are compared and judged, and a warning signal is issued, including:

[0030] The warning signal includes sound alarm, light alarm, short message notification and email notification.

[0031] The warning judgment condition includes that when the actual new energy generation power is greater than the new energy consumption capacity prediction value minus the preset safety margin, the warning signal is issued.

[0032] The beneficial effects of the preferred technical solution are: through the dynamic threshold judgment mechanism of the preset safety margin, multi-level warning is triggered before the real-time new energy generation power approaches the power grid consumption limit, effectively overcoming the traditional early warning hysteresis problem; the four-dimensional stereoscopic alarm channel of sound, light, short message and email is fused to ensure that power grid dispatchers can immediately perceive the risk in different scenarios; the design intelligently associates prediction data with actual operation status, avoiding false positives leading to frequent operations, and accurately capturing critical risk points to gain key decision-making time window for dispatchers, thereby significantly reducing the probability of new energy curtailment, improving the power grid consumption flexibility and operation safety margin, and ultimately realizing the dual protection of efficient utilization of new energy and stable operation of the power grid.

[0033] As a preferred scheme of the new energy consumption prediction and early warning method, the real-time predicted two types of prediction results are compared and judged, and a warning signal is issued, including:

[0034] The control strategy includes adjusting the new energy generation power, adjusting the load, adjusting the power grid operation mode, and increasing the charge and discharge power of the energy storage device.

[0035] An optimization algorithm is used to establish an optimization model and solve it.

[0036] The beneficial effects of the preferred technical solutions are: the multi-dimensional coordinated control strategy is intelligently generated, and the dynamic response capability of the power grid to new energy fluctuation is significantly improved; the system automatically triggers the optimization algorithm based on the early warning signal, coordinates and adjusts the four core means of new energy power generation regulation, load demand management, power grid operation mode optimization and energy storage charging and discharging control, and forms a globally optimal decision scheme; under the premise of ensuring the safe and stable operation of the power grid, the phenomenon of new energy curtailment is maximally inhibited, the economic operation efficiency and system regulation potential are taken into account, the new energy consumption capacity and the power grid resilience are simultaneously strengthened, and efficient and reliable decision support is provided for dispatchers.

[0037] In a second aspect, the present application provides a new energy consumption prediction and early warning system, comprising:

[0038] A data acquisition module acquires relevant data of new energy power generation data, power grid data, load data and meteorological data, and transmits the data to a data preprocessing module;

[0039] The data preprocessing module cleans, converts and normalizes the collected raw data, removes noise, outliers and missing values in the data, converts data of different formats and units into a unified format and unit, and normalizes the data to a certain range;

[0040] The feature engineering module extracts useful features from the preprocessed data to improve the accuracy and generalization ability of the prediction model;

[0041] The prediction model module includes a new energy power generation prediction sub-module, a load prediction sub-module and a new energy consumption capacity prediction sub-module;

[0042] The early warning module compares the new energy consumption capacity prediction result with the actual new energy power generation, and issues an early warning signal when the actual new energy power generation exceeds the new energy consumption capacity of the power grid;

[0043] The control strategy generation module generates a corresponding control strategy according to the early warning signal and the actual operation of the power grid, and improves the new energy consumption capacity of the power grid.

[0044] In a third aspect, the present application provides an electronic device, comprising:

[0045] A memory and a processor;

[0046] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions, which realize the steps of a new energy consumption prediction and early warning method when the computer executable instructions are executed by the processor.

[0047] In a fourth aspect, the present application provides a computer readable storage medium storing computer executable instructions, which, when executed by a processor, implement the steps of the new energy consumption prediction and early warning method.

[0048] Compared with the prior art, the present application has the following beneficial effects: the present application collects multi-source heterogeneous data in real time, and solves the problems of noise and missing of field data by using a standardized preprocessing process, thereby providing high reliability input for prediction; a high-order feature vector with strong correlation with new energy output is constructed based on feature engineering, thereby significantly improving the adaptability of the prediction model to complex scenarios; a new energy generation, load and consumption capacity prediction model based on collaborative training is used to accurately predict the power grid consumption bottleneck in the future period, and a multi-level early warning mechanism is triggered through a dynamic safety margin threshold; finally, a source-grid-load-storage collaborative control strategy is generated according to the early warning result, thereby forming a "prediction-early warning-control" closed-loop management at the power grid dispatching end; the problem of power curtailment caused by random fluctuations of new energy is effectively solved, the dispatching response delay is reduced to the minute level in a real power grid scenario, the consumption efficiency and system stability are simultaneously improved, and practical technical support is provided for high-proportion new energy grid connection BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 The whole flowchart of the new energy consumption prediction and early warning method according to one embodiment of the present application.

[0051] Figure 2 The system principle diagram of the prediction model module of the new energy consumption prediction and early warning method according to one embodiment of the present application.

[0052] Figure 3 The system principle diagram of the new energy consumption prediction and early warning system according to one embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings in the specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the protection scope of the present application.

[0054] Embodiment 1, refer toFigure 1 For an embodiment of the present application, a new energy consumption prediction and early warning method is provided, comprising:

[0055] S1: Real-time collection of new energy related data and transmission of data;

[0056] S2: Preprocessing and data conversion of the collected new energy related data;

[0057] S3: Feature extraction of the processed data to generate a feature vector;

[0058] S4: Training of the first, second and third prediction models using historical data to determine the parameters of the models, and real-time prediction of the new energy consumption capacity in the next time period using the trained prediction models;

[0059] S5: Comparison and judgment of the two types of prediction results of the real-time prediction, and issuance of a warning signal;

[0060] S6: Generation of a corresponding control strategy according to the warning signal and execution of a control operation.

[0061] It should be noted that the inherent intermittency and volatility of new energy power generation lead to dynamic changes in grid consumption capacity, and traditional methods are difficult to effectively avoid the phenomenon of curtailment due to single data source, insufficient prediction accuracy and lagging control response. Grid dispatchers lack predictive means when facing sudden changes in new energy output and can only take emergency measures such as cutting machines, which not only affects the utilization rate of clean energy but also threatens the stable operation of the system.

[0062] Therefore, in view of the above problems of new energy random fluctuation being difficult to capture, consumption bottleneck early warning lagging behind, and lack of coordination of control strategies, through steps S1-S6, an engineering-level solution is achieved through a complete technical closed loop: real-time collection and standardized processing of multi-source heterogeneous data to ensure input quality; feature engineering deeply mines the meteorological-grid-load coupling relationship to support high-precision prediction; a multi-model architecture trained in coordination accurately quantifies the consumption capacity limit in the future period; a multi-level early warning mechanism based on dynamic safety thresholds triggers an intervention window in advance; and finally a source-grid-load-storage multi-dimensional coordinated control strategy is generated to dynamically balance the supply and demand of the system in a minute-level dispatching cycle, thereby suppressing the risk of curtailment from the root cause and improving the resilience of the grid.

[0063] Embodiment 2, refer to Figures 1-2 For an embodiment of the present application, based on the above embodiment, a new energy consumption prediction and early warning method is provided.

[0064] In the present application, the real-time collection of new energy related data and transmission of data in step S1 includes: real-time collection of related data, collected once every 15 minutes;

[0065] Real-time power, installed capacity, historical power generation data of each new energy power station (such as wind power plant, solar photovoltaic power station, etc.); grid data including node voltage, line flow, transformer parameter, grid structure of the power grid; load data including real-time load power, historical load data, load characteristics of each region; meteorological data including wind speed, wind direction, solar radiation intensity, temperature, humidity, air pressure; the data acquisition module collects the above data in real time through sensors, smart meters, hardware devices of meteorological monitoring equipment, and data interfaces with other systems.

[0066] In an optional embodiment, the real-time collection of new energy related data in step S1 and the transmission of data can also be through a data pre-screening mechanism of the edge computing layer. By deploying an edge computing unit on the side of the new energy power station, local real-time analysis is performed on the original sensor data (such as wind speed mutation, solar radiation sudden change), and only abnormal data or key trend change data exceeding the threshold range are uploaded to the main system, thereby reducing the communication bandwidth pressure and improving the transmission timeliness of high value data; for example, when the wind speed sensor of the wind power plant detects a step change of 5 m / s or more lasting 10 seconds, the edge device immediately packages the power fluctuation data of this period for preferential transmission, ensuring that the main system quickly captures the output mutation characteristics.

[0067] In another optional embodiment, the real-time collection of new energy related data in step S1 and the transmission of data can also be through multi-source meteorological data fusion verification to enhance data reliability. Spatial correlation verification is performed on meteorological radar satellite data, adjacent power station meteorological station data and local sensor data, and local micro-meteorological environment monitoring deviation is corrected through a weighted average algorithm; for example, for a mountain photovoltaic power station, the solar radiation intensity data of three reference stations in the surrounding area are fused, and the terrain elevation model is combined to dynamically compensate for the radiation jump caused by cloud cover, to generate site-level corrected meteorological parameters, thereby improving the data quality of new energy output prediction from the source.

[0068] In the embodiments of the present application, the pre-processing and data conversion of the collected new energy related data in step S2 include: processing the collected data to obtain pre-processed data;

[0069] Abnormal values are identified and removed using statistical methods (such as mean, median, standard deviation, etc.), and missing values are filled in using interpolation methods (such as linear interpolation, polynomial interpolation, spline interpolation, etc.); data conversion includes converting time data into time stamps or time series, and converting classification data into numerical data (such as using one-hot encoding, label encoding, etc.); data normalization includes normalizing the data to the [0, 1] interval using the minimum-maximum normalization method, and is expressed as:

[0070]

[0071] wherein x is the original data, x min is the minimum value of the data, x max is the maximum value of the data, is the normalized data.

[0072] In an alternative embodiment, the preprocessing and data conversion of the collected new energy related data in step S2 can also enhance data integrity through a field station group collaborative interpolation mechanism: in view of the geographical dispersion characteristics of the new energy power station cluster, when a single station has continuous missing meteorological data due to equipment failure, the real-time wind speed and irradiance data of adjacent stations are automatically selected based on a spatial correlation model (such as a K-nearest neighbor algorithm) to calculate a replacement value, and then a time series interpolation method is used for dynamic filling.

[0073] In another alternative embodiment, the preprocessing and data conversion of the collected new energy related data in step S2 can also adapt to the changes in the power grid operation state through dynamic incremental normalization: in view of the significant difference between load data and new energy output in summer and winter, a sliding time window (such as 72 hours) is used to dynamically update the minimum and maximum values of the normalization parameters, instead of global static normalization; for example, when a cold wave strikes and causes a sudden increase in load, the system automatically uses the load extreme value of the last 3 days as a new benchmark for data scaling, avoiding the distortion of the current data distribution by the historical annual extreme value, and improving the sensitivity of the prediction model in high volatility scenarios.

[0074] In the embodiments of the present application, the feature extraction of the processed data in step S3 to generate a feature vector includes: extracting features to generate a feature vector.

[0075] Time features (such as hours, days, weeks, months, seasons, etc.), meteorological features (such as wind speed, wind direction, solar radiation intensity, etc., which are related to new energy power generation), load features (such as peak value, valley value, change trend, etc. of load), grid features (such as node voltage, line flow, etc. of the grid, which are related to new energy consumption), in addition, new features can also be generated through feature combination and feature transformation methods, such as combining wind speed and wind direction into wind power density features, and combining solar radiation intensity and temperature into photovoltaic output related features, etc.

[0076] In an alternative embodiment, the feature extraction of the processed data in step S3 to generate a feature vector can also construct a regional wind power density matrix by fusing multi-wind farm group meteorological data. Specifically, it includes: obtaining real-time wind speed, wind direction and air density data of each wind farm in the target area, calculating the wind power density value of each grid point through vector synthesis; combining terrain elevation data to perform spatial interpolation correction on the wind power density, and generating a dynamic feature layer reflecting the regional wind resource distribution intensity.

[0077] In another optional implementation, in step S3, feature extraction is performed on the processed data, and the feature vector can be generated by constructing a photovoltaic power output attenuation time series feature. Specifically, this includes collecting power station operating parameters such as component temperature, dust accumulation thickness, and aging years, and coupling them with solar radiation intensity features; calculating the light intensity-temperature attenuation coefficient based on a physical model, superimposing the transmittance attenuation factor caused by dust accumulation, and generating a dynamic attenuation correction feature vector.

[0078] In this embodiment of the application, step S4 uses historical data to train the first prediction model, the second prediction model, and the third prediction model to determine the parameters of the model. Using the trained prediction model, the real-time prediction of the renewable energy absorption capacity in the next time period includes: updating and training the prediction model every morning using data from the previous 7 days; making a real-time prediction every 15 minutes to obtain the predicted values ​​of renewable energy power generation, load power, and renewable energy absorption capacity for the next hour.

[0079] like Figure 1 Steps four and five Figure 2 As shown, it specifically includes a new energy power generation forecasting submodule, a load forecasting submodule, and a new energy absorption capacity forecasting submodule;

[0080] The new energy power generation prediction submodule utilizes historical new energy power generation data and meteorological data to establish a new energy power generation prediction model, predicting the power generation capacity of new energy sources in the future. It employs machine learning algorithms (such as support vector machines, random forests, gradient boosting trees, etc.), time series analysis algorithms (such as ARIMA, SARIMA, etc.), or deep learning algorithms (such as recurrent neural networks, long short-term memory networks, etc.). Taking the long short-term memory network LSTM as an example, its model structure includes an input layer, a hidden layer, and an output layer. The hidden layer is an LSTM unit, which controls the memory and forgetting of information through forget gates, input gates, and output gates, represented as follows:

[0081] Forgotten Gate:

[0082] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0083] Input Gate:

[0084] i t =σ(W i ·{h t-1 ,x t ]+b i )

[0085] Candidate cell status:

[0086]

[0087] Cell state update:

[0088]

[0089] Output gate:

[0090] o t = sigma(W o * [h t-1 , x t ] + b o )

[0091] Hidden layer output:

[0092] h t = o t * tanh(C t )

[0093] where sigma is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, W f , W i , W C , W o are weight matrices, b f , b i , b C , b o are bias vectors, h t-1 is the hidden layer output at the previous time, x t is the input at the current time, f t is the forget gate output, i t is the input gate output, is the candidate cell state, C t is the cell state at the current time, o t is the output gate output, h t is the hidden layer output at the current time.

[0094] Load prediction submodule: using historical load data and related influencing factors (such as time, weather, economic indicators, etc.), a load prediction model is established to predict the load power in the future period of time;

[0095] New energy consumption capacity prediction submodule: according to the new energy generation prediction result, the load prediction result and the grid data, a new energy consumption capacity prediction model is established to predict the new energy power that can be consumed by the grid in the future period of time; this model can consider factors such as the transmission capacity, peak regulation capacity, reserve capacity of the grid, and use optimization algorithms (such as linear programming, nonlinear programming, mixed integer programming, etc.) or machine learning algorithms for modeling.

[0096] It should be noted that the new energy power generation prediction sub-module adopts an adaptive time series modeling technology (such as LSTM), and through the dynamic memory mechanism of the forgetting gate, the input gate and the output gate, the strong nonlinear correlation between the meteorological data (wind speed, irradiance) and the historical power generation is accurately captured. The model automatically updates the parameters by fusing the latest data of nearly 7 days in the early morning every day, ensures the generalization ability to the weather mutation scene, realizes the minute-level rolling prediction of the future 1-hour power generation, and the load prediction sub-module fuses multi-dimensional influencing factors (time period, economic indicators, weather conditions), and mines the internal law of load change through a machine learning algorithm; the prediction result not only reflects the benchmark electricity demand, but also accurately quantifies the load fluctuation caused by extreme temperature or special events, providing key input for the calculation of the accommodation capacity; the new energy accommodation capacity prediction sub-module is a core decision engine, which deeply fuses the power generation prediction, the load prediction and the real-time power grid parameters (power transmission capacity, standby margin); based on the optimization algorithm, the physical constraints of the power grid (such as line thermal stability limit, peak regulation capacity) are modeled, and the safe accommodation threshold of the whole network is dynamically calculated.

[0097] In an optional implementation, the first prediction model, the second prediction model and the third prediction model are trained by using historical data in step S4, and the parameters of the model can also be determined by a meteorological event driven incremental training mechanism: when an extreme meteorological event such as a typhoon or a sandstorm is monitored, the system automatically triggers the model incremental training process; under the premise of retaining the basic model architecture (such as the LSTM gating mechanism), only the high-frequency sampling data (such as 5-minute-level wind speed mutation records, photovoltaic output sudden drop data) of 24 hours before and after the event are used for local parameter fine-tuning, so that the new energy power generation prediction sub-module quickly adapts to the nonlinear characteristics under extreme working conditions, and avoids the prediction deviation lag problem caused by traditional daily batch training.

[0098] In another optional implementation, the first prediction model, the second prediction model and the third prediction model are trained by using historical data in step S4, and the parameters of the model can also be enhanced by a load feature recursive screening mechanism: during the training process of the load prediction sub-module, the recursive feature elimination (RFE) algorithm is embedded. Based on the correlation analysis of historical load data and influencing factors (such as holiday flags, industrial user production plans), redundant features (such as economic indicators with a correlation with the load curve lower than a threshold) are dynamically removed, and key feature combinations (such as temperature-time cross terms) are retained.

[0099] In an optional embodiment, the real-time prediction of the new energy consumption capacity in the next time period in step S4 using the trained prediction model can also improve the accuracy through a multi-time scale rolling correction mechanism: on the basis of performing 1-hour prediction every 15 minutes, a 5-minute super-short-term correction link is superimposed; for example, when the new energy power generation prediction submodule detects that the actual irradiance drops by 20% within 5 minutes, the online recalculation of the consumption capacity prediction submodule is triggered immediately, and the latest power grid topology data (such as line information due to fault tripping) and real-time load fluctuation value are used to generate a corrected consumption capacity threshold within 10 seconds to cope with sudden scenarios such as photovoltaic cloud blocking.

[0100] In another optional embodiment, the real-time prediction of the new energy consumption capacity in the next time period in step S4 using the trained prediction model can also be optimized through a regional collaborative consumption margin allocation mechanism: after the new energy consumption capacity prediction submodule calculates the total consumption threshold, the total consumption capacity is decomposed to each new energy power station cluster based on the grid structure characteristics (such as the inter-regional tie line capacity, node voltage sensitivity) using a distributed optimization algorithm; for example, a wind power station cluster at the end of the power grid is allocated a lower consumption weight, and a photovoltaic cluster close to the load center and equipped with energy storage is allocated a higher margin, so as to balance the consumption pressure and grid safety risk from the spatial dimension.

[0101] In the embodiments of the present application, the comparison and judgment of the two types of prediction results predicted in real time in step S5 and the issuance of the early warning signal include: comparing the real-time predicted new energy power generation with the predicted value of the new energy consumption capacity, and judging whether to issue an early warning signal;

[0102] The early warning signal is issued in various forms such as sound alarm, light alarm, short message notification, and email notification to remind the power grid dispatchers to take corresponding measures; the judgment condition of the early warning is set as: when the actual new energy power generation P new is greater than the predicted value of the new energy consumption capacity P cap minus a certain safety margin ΔP, that is, P new >P cap -ΔP, the early warning signal is issued.

[0103] It should be noted that the traditional new energy consumption early warning mechanism often causes response lag due to rigid judgment conditions. When the actual power generation power breaks through the consumption limit, an alarm is triggered, forcing dispatch personnel to take extreme measures such as cutting off the machine, which not only aggravates the loss of abandoned electricity but also threatens the safety of the power grid. Step S5 innovatively introduces a dynamic threshold mechanism for safety margin. By comparing the difference between the new energy power generation power and the predicted value of the consumption capacity in real time, early warning is given before the power generation power approaches the theoretical consumption limit, effectively seizing the golden window period for risk disposal. Simultaneously integrating sound and light alarms, short messages and emails into four-dimensional alarm channels ensures that dispatch personnel can perceive the risk situation with zero delay regardless of whether they are in the control center, on-site inspection or remote office scenarios. Intelligent coupling of high-precision prediction data and real-time operation status not only avoids false alarms causing invalid dispatch disturbances, but also accurately locates the critical risk point, providing sufficient decision-making time for starting load regulation, energy storage charging and discharging and other collaborative control strategies.

[0104] In an optional embodiment, the comparison and judgment of the two types of prediction results predicted in real time in step S5, and the issuance of the early warning signal can also be achieved through a dynamic partition margin threshold linkage mechanism: according to the grid topology, the new energy consumption sensitive area (such as a wind power cluster area, a photovoltaic concentration area) is divided, and a safety margin threshold is independently set for each area. When the actual power generation power of a certain area exceeds the predicted value of the consumption capacity of the area minus the area-specific margin (such as the wind power area margin = basic value ± transmission channel utilization correction amount), a partition early warning signal is automatically triggered and the risk source is located. For example, when the cross-regional transmission channel is close to full load, the safety margin threshold of the sending end wind farm is automatically reduced, the local power abandonment risk is early warned, and the flow animation warning graph of the associated transmission path is pushed to the dispatch platform.

[0105] In another optional embodiment, the comparison and judgment of the two types of prediction results predicted in real time in step S5, and the issuance of the early warning signal can also be achieved through a load change trend prediction enhancement mechanism: the future 15-minute load change slope output by the load prediction sub-module is included in the early warning condition. When the new energy power generation power has not broken through the current consumption capacity threshold, but the load prediction curve shows a steep downward trend (such as a prediction slope less than -10 MW / min) and the new energy output continues to rise, the early warning level is automatically upgraded and a pre-intervention instruction is generated; for example, if it is predicted that the concentrated shutdown of industrial parks during lunchtime will cause a sharp drop in load, even if the current power has not reached the threshold, the system will start a yellow early warning 15 minutes in advance, and the control strategy module will pre-load the energy storage charging scheme.

[0106] In the embodiments of the present application, the generation of the corresponding control strategy according to the early warning signal in step S6, and the execution of the control operation include: generating a control strategy according to the early warning signal and executing the control operation;

[0107] Adjusting new energy power generation (such as active power control of new energy power station), adjusting load (such as implementing demand response measures, guiding users to adjust power load), adjusting power grid operation mode (such as adjusting transformer tap, switching capacitor bank, adjusting line power flow, etc.), increasing the charging and discharging power of energy storage device; the generation of control strategy adopts optimization algorithm, and the optimization model is established and solved to minimize new energy curtailment rate, maximize power grid operation economy and stability.

[0108] It should be noted that the synchronous driving of new energy power station active power regulation (suppressing output fluctuation), demand response load adjustment (shifting peak pressure), power grid topology optimization (such as capacitor bank switching and power flow redistribution) and energy storage charging and discharging strategy (dynamic compensation of charging and discharging power) form a four-dimensional coordinated intervention of source-grid-load-storage; the Pareto optimal strategy is generated by solving the optimization model with the multi-objective constraints of minimizing curtailment rate and maximizing economy and stability, to ensure that the adjustment process takes into account the power grid safety boundary and resource utilization efficiency; the control strategy is directly connected to the power grid control system to realize second-level execution, and the warning window is converted into effective action time, for example, when wind power suddenly increases, the power station output is quickly reduced and the energy storage charging is started, and the industrial load demand response is activated at the same time, to avoid the waste of clean energy caused by traditional generator tripping operation.

[0109] In an optional embodiment, according to the warning signal, the corresponding control strategy is generated in step S6, and the execution of the control operation can also be realized by dynamically matching the differentiated control strategy according to the risk level of consumption: when the new energy curtailment risk is in the low threshold interval (such as the initial stage of wind power sudden increase), the energy storage charging and demand response load adjustment are preferentially started, and the excess power is stored to realize flexible consumption by shifting peak load; if the risk escalates to the high threshold interval (such as the sudden surge of photovoltaic output superimposed with low load), the new energy power station active power is quickly reduced and the power grid power flow is actively optimized (such as adjusting transformer tap to release power transmission capacity), forming a "soft regulation-rigid control" hierarchical response mechanism to minimize the amount of direct generator tripping under the premise of ensuring power grid safety.

[0110] In an optional embodiment, according to the warning signal, the corresponding control strategy is generated in step S6, and the execution of the control operation can also be realized by a multi-time scale rolling optimization mechanism to enhance dynamic adaptability: based on the real-time updated new energy output fluctuation characteristics (such as wind power continuous climbing), the optimization model parameters are rolled and corrected with a period of 5 minutes, and the control strategy combination is dynamically adjusted. For example, in the scenario of continuous wind power enhancement, if the forecast shows that the consumption gap is still expanding after the first round of control starts the energy storage charging, the power grid operation mode adjustment (such as switching capacitor bank to improve voltage stability) and load side precise peak cutting (such as reducing high energy consumption industrial load) are automatically superimposed, the optimal consumption target is approached through multi-round strategy iteration, and the problem of insufficient or excessive adjustment of single control is avoided.

[0111] In summary, the present application solves the problem of consumption in the high proportion of new energy grid-connected scenario by constructing a "data-aware-intelligent prediction-dynamic early warning-collaborative control" whole-chain technology system. In the data-aware layer, relying on real-time collection of multi-source heterogeneous data and edge computing pre-screening mechanism, high timeliness and high reliability input basis is provided for the prediction model; in the intelligent prediction layer, adaptive time series modeling and physical constraint fusion technology are adopted to accurately quantify the dynamic change trend of new energy output, load demand and power grid consumption capacity; in the dynamic early warning layer, the safety margin threshold and partition linkage mechanism are innovatively introduced to realize the prospective capture and stereoscopic alarm of critical risk; in the collaborative control layer, the four-dimensional strategy optimization of source, network, load and storage and the risk grading response mechanism are adopted to form the minute-level closed-loop regulation and control capability.

[0112] Embodiment 3, the above is a schematic scheme of a new energy consumption prediction and early warning method. It should be noted that the technical scheme of the system for new energy consumption prediction and early warning and the technical scheme of the above new energy consumption prediction and early warning method belong to the same concept. The technical scheme of the system for new energy consumption prediction and early warning in this embodiment is not described in detail. The details can be referred to the description of the technical scheme of the above new energy consumption prediction and early warning method.

[0113] As shown in Figure 3 , the present embodiment also provides a new energy consumption prediction and early warning system, comprising:

[0114] A data acquisition module acquires relevant data of new energy power generation data, power grid data, load data and meteorological data, and transmits the data to a data preprocessing module;

[0115] Specifically, install wind speed sensors, solar radiation sensors, intelligent electric meters and other equipment to collect real-time power generation data of wind power plants and solar photovoltaic power stations in a certain area, node voltage and line flow data of the power grid, load data of each region and meteorological data. The data is transmitted to the data preprocessing module through the wireless communication module;

[0116] The data preprocessing module cleans, converts and normalizes the collected raw data, removes noise, outliers and missing values in the data, converts data of different formats and units to a unified format and unit, and normalizes the data to a certain range;

[0117] Specifically, the collected data is cleaned to remove outliers, and linear interpolation method is used to fill in missing values; time data is converted to time stamp, and classification data is one-hot encoded; the data is normalized by using the minimum-maximum normalization method;

[0118] A feature engineering module extracts useful features from the preprocessed data to improve the accuracy and generalization ability of the prediction model.

[0119] Specifically, time features (such as hours, days, weeks, seasons, etc.), meteorological features (such as wind speed, wind direction, solar radiation intensity, etc.), load features (such as peak load, valley load, change trend, etc.), and grid features (such as node voltage, line flow, etc.) are extracted, and new features such as wind power density and photovoltaic output related features are generated.

[0120] A prediction model module includes a new energy generation prediction sub-module, a load prediction sub-module, and a new energy consumption capacity prediction sub-module.

[0121] Specifically, the new energy generation prediction sub-module uses an LSTM network model, with historical meteorological data and new energy generation data as input, and future 1-hour new energy generation power prediction values as output. The parameters of the LSTM network are optimized through training data to determine the best network structure (such as the number of hidden layer neurons, training batch size, training rounds, etc.).

[0122] The load prediction sub-module uses a random forest algorithm, with historical load data, time features, and meteorological features as input, and future 1-hour load power prediction values as output. The best parameters of the random forest (such as the number of decision trees, maximum depth, etc.) are determined through cross-validation.

[0123] The new energy consumption capacity prediction sub-module uses a linear programming model, with the power transmission capacity, peak regulation capacity, and reserve capacity of the grid as constraint conditions, and maximizes the new energy consumption as the target. The input is the new energy generation prediction value, the load prediction value, and the grid data, and the output is the new energy consumption capacity prediction value.

[0124] The early warning module compares the new energy consumption capacity prediction result with the actual new energy generation power, and issues a warning signal when the actual new energy generation power exceeds the new energy consumption capacity of the grid.

[0125] Specifically, the safety margin ΔP is set to 10MW, and when the actual new energy generation power is greater than the new energy consumption capacity prediction value minus 10MW, a sound and light alarm signal is issued.

[0126] The control strategy generation module generates corresponding control strategies according to the warning signal and the actual operation of the grid to improve the new energy consumption capacity of the grid.

[0127] Specifically, when the warning signal is issued, control strategies such as adjusting the active power of new energy power stations and switching capacitor banks are generated and sent to the grid control system for execution.

[0128] The embodiment also provides an electronic device suitable for new energy consumption prediction and early warning, comprising a memory and a processor; the memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions to realize the method for new energy consumption prediction and early warning proposed in the above embodiment.

[0129] The embodiment also provides a storage medium having a computer program stored thereon, and the computer program is executed by a processor to realize the method for new energy consumption prediction and early warning proposed in the above embodiment.

[0130] The storage medium proposed in the embodiment and the method for new energy consumption prediction and early warning proposed in the above embodiment belong to the same inventive concept, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0131] From the above description about the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, and the computer software product can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH memory, a hard disk or an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method of each embodiment of the present application.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for new energy consumption prediction and early warning, characterized in that, The application relates to a new energy consumption capacity prediction method and device. Real-time new energy related data is collected and transmitted; The collected new energy related data is preprocessed and converted; Features are extracted from the processed data to generate a feature vector; Historical data is used to train a first prediction model, a second prediction model and a third prediction model, determine the parameters of the models, and use the trained prediction models to predict the new energy consumption capacity in the next time period in real time; The two types of real-time prediction results are compared and judged, and a warning signal is issued; According to the warning signal, a corresponding control strategy is generated, and a control operation is performed.

2. The method of claim 1, wherein the method comprises: The real-time collection of new energy related data and the transmission of data includes: New energy related data includes new energy power generation data, power grid data, load data and weather data; Real-time new energy related data is collected through sensors, smart phones, weather monitoring equipment and data interfaces, and data transmission is performed.

3. The method of claim 2, wherein the method further comprises: The collected new energy related data is preprocessed and converted, including: Data cleaning is performed using statistical methods; Missing values are filled in by interpolation; Data conversion is performed to convert the data into time series and numerical data.

4. The method of claim 3, wherein the new energy consumption prediction and early warning method is characterized by, The processed data is subjected to feature extraction to generate a feature vector, including: Feature extraction includes time features, weather features, load features and power grid features; High-order features are generated through feature combination and transformation.

5. The method of claim 4, wherein the new energy consumption prediction and early warning method is characterized by, The historical data is used to train a first prediction model, a second prediction model and a third prediction model, determine the parameters of the models, and use the trained prediction models to predict the new energy consumption capacity in the next time period in real time, including: A new energy power generation prediction model is established using historical new energy power generation data and weather data; A load prediction model is established using historical load data and related influencing factors; A new energy consumption capacity prediction model is established based on the new energy power generation prediction results, the load prediction results and the power grid data.

6. The method of new energy consumption prediction and early warning according to claim 5, characterized in that, The two types of real-time prediction results are compared and judged, and a warning signal is issued, including: The warning signal includes sound alarm, light alarm, SMS notification and email notification; The warning judgment condition includes that when the actual new energy power generation power is greater than the new energy consumption capacity prediction value minus the preset safety margin, a warning signal is issued.

7. The method of new energy consumption prediction and early warning according to claim 6, characterized in that, According to the warning signal, a corresponding control strategy is generated, and a control operation is performed, including: The control strategy includes adjusting the new energy power generation power, adjusting the load, adjusting the power grid operation mode and increasing the charge-discharge power of the energy storage device; An optimization model is established and solved using an optimization algorithm.

8. A new energy consumption prediction and early warning system, applying the method of any one of claims 1-7, characterized in that, The application relates to a new energy consumption capacity prediction method and device. A data collection module collects new energy power generation data, power grid data, load data and weather related data, and transmits the data to a data preprocessing module; The data preprocessing module cleans, converts and normalizes the collected raw data, removes noise, outliers and missing values in the data, converts data of different formats and units into a unified format and unit, and normalizes the data to a certain range; The feature engineering module extracts useful features from the preprocessed data to improve the accuracy and generalization ability of the prediction model; The prediction model module comprises a new energy power generation prediction sub-module, a load prediction sub-module and a new energy consumption capacity prediction sub-module; The early warning module compares the new energy consumption capacity prediction result with the actual new energy power generation, and issues a warning signal when the actual new energy power generation exceeds the new energy consumption capacity of the power grid; The control strategy generation module generates a corresponding control strategy according to the warning signal and the actual operation condition of the power grid, and improves the new energy consumption capacity of the power grid. 9.An electronic device, comprising: a memory and a processor; The memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions, which realize the steps of the new energy consumption prediction and early warning method in any one of claims 1 to 7 when executed by the processor. 10.A computer readable storage medium storing computer executable instructions, which realize the steps of the new energy consumption prediction and early warning method in any one of claims 1 to 7 when executed by a processor.

Images