New energy power prediction method, system and device based on power transaction and medium

Abstract

A new energy power prediction method, system, device and medium based on power transaction relates to the technical field of power prediction, and obtains power grid operation data and environmental data around a new energy power station; performing data cleaning on the acquired data to generate original data; preprocessing the original data to optimize the structure of the original data and generate a power index data set; constructing different types of power prediction models by adopting different machine learning methods; on the basis of the power index data set, predicting the generated power of the new energy power station by adopting the power prediction models in sequence; analyzing each type of output prediction results; according to the method, the generated power of the new energy power station can be accurately predicted according to the power utilization conditions of multiple users.

Description

Technical Field

[0001] The present invention relates to the field of power prediction technology, and in particular to a new energy power prediction method, system, device and medium based on power trading. Background Art

[0002] With the continuous increase in the country's new energy installed capacity and the breakthrough progress in large-scale grid-connected technology of new energy, the proportion of wind power and photovoltaic power generation in my country has gradually surpassed coal-fired power to become the main power supply force, and the peak-shaving pressure of the power system has increased sharply.

[0003] However, due to uncertainties such as weather, a large amount of wind or solar energy cannot be fully utilized. Therefore, it is necessary to adjust the load during peak hours to alleviate the peak load pressure. Currently, the grid dispatch center has a relatively simple method for adjusting power demand at the user end, limited to issuing notices to residents to conserve electricity or temporarily closing some industrial enterprises in individual cases. Therefore, for large industrial and commercial customers, the amount of electricity they can participate in market transactions is limited. To obtain more profits, they need to reasonably arrange the timing of their own production activities and maximize their output levels. At the same time, the power department also needs to understand the actual electricity usage of these large industrial and commercial customers to achieve precise power restrictions and thus avoid risks.

[0004] The traditional method currently used is to predict the power and power generation of different types of renewable energy power plants separately. The prediction system cannot achieve integrated prediction and its applicable scenarios are limited.

[0005] A Chinese patent (CN202111411194.0) discloses a short-term wind power prediction method based on SSA optimized BP neural network. The SSA adopted has the characteristics of good stability, strong global search capability and fast convergence speed. It overcomes the shortcomings of BP neural network in the training process, such as long training time and easy to fall into local minimum, improves the global optimization ability of BP neural network algorithm, and improves the accuracy of model prediction by optimizing the weights and thresholds of BP neural network algorithm; however, although SSA helps to escape the local optimal solution, it will also slow down the convergence speed of the entire model training process, thereby affecting the speed of power prediction. Summary of the Invention

[0006] In order to overcome the deficiencies in the background technology, the present invention discloses a method for predicting the power generation of renewable energy based on the electricity consumption of multiple users.

[0007] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0008] New energy power forecasting methods based on power trading include:

[0009] Obtain power grid operation data and environmental data around new energy power stations;

[0010] Perform data cleaning on the acquired data to generate original data;

[0011] Preprocess the raw data to optimize its structure and generate a power index data set;

[0012] Use different machine learning methods to build different types of power prediction models;

[0013] Based on the power index data set, each power prediction model is used in turn to predict the power generation of the new energy power station;

[0014] Analyze the output of various types of prediction results.

[0015] Preferably, the specific steps of data cleaning include:

[0016] Preliminary checks, including checking the integrity of the data and detecting outliers in the data;

[0017] Data synchronization: time series alignment of data after preliminary inspection;

[0018] Data conversion, normalization or standardization of synchronized data;

[0019] Data verification: Verify whether the converted raw data meets the requirements of the power prediction model.

[0020] Preferably, the specific steps of the pretreatment include:

[0021] Feature processing: feature selection, generation, and conversion based on the cleaned raw data;

[0022] Integrate market information, collect real-time market prices, and obtain historical transaction information in the market and form a historical market price sequence by analyzing market mechanisms, quotation rules, and transaction conditions;

[0023] Construct a comprehensive data set, combine the processed feature set with market information, and form a comprehensive power indicator data set containing multi-dimensional information.

[0024] Preferably, the dimension information of the power indicator data set includes a time dimension, a space dimension, a business dimension and an electric quantity indicator dimension;

[0025] Each dimension of information is composed of multi-level business themes, where the first-level business themes include economic factors, weather factors, and meteorological factors; the second-level business themes include wind resources and solar resources; and the third-level business themes include exogenous and endogenous factors.

[0026] The economic factors are exogenous factors, weather factors correspond to the conditions for the use of wind resources, meteorological factors correspond to the conditions for the use of solar resources, and wind resources and solar resources correspond to endogenous factors.

[0027] The exogenous influencing factors and the endogenous influencing factors determine the power supply level of the corresponding dimensional information.

[0028] Preferably, the power prediction model includes a wind-solar ultra-short-term power prediction model, a model based on deep learning, a cluster-based short-term power prediction model, a short-term power prediction model, and a medium- and long-term power prediction model; wherein the wind-solar ultra-short-term power prediction model is suitable for quickly responding to weather changes and is used for real-time power prediction of wind farms and photovoltaic power stations;

[0029] The deep learning-based model is suitable for wind and solar farms with large amounts of data and flat terrain conditions;

[0030] The cluster short-term power forecasting model is applicable to wind and solar farm clusters that are geographically distributed but have similar environmental conditions.

[0031] Short-term power forecasting models are suitable for comprehensive management of wind farms;

[0032] The medium- and long-term power forecast model is suitable for wind farms that need to plan and dispatch power generation capacity in the next few days to months, providing medium- and long-term power forecasts.

[0033] Preferably, the short-term power prediction model includes a regional wind farm cluster centralized power prediction model and a wind-solar hybrid prediction model based on a BP neural network, which are respectively suitable for the integrated management of multiple adjacent wind farms and the simultaneous management of wind power and photovoltaic power stations.

[0034] On the other hand, an embodiment of the present application provides a new energy power prediction system based on power trading, including an acquisition unit, a data processing unit, a model building unit, and a calculation and analysis unit;

[0035] The acquisition unit is composed of various sensors and is used to acquire power grid operation data and environmental data around the new energy power station;

[0036] The data processing unit includes a cleaning module and a preprocessing module, the two modules are used to perform data cleaning and data preprocessing steps respectively;

[0037] The model building unit adopts different machine learning methods to build different types of power prediction models;

[0038] The calculation and analysis unit is used to receive the data output by the data processing unit and to call the power prediction models of the model building unit in sequence.

[0039] Preferably, the model building unit is electrically connected to the acquisition unit and is used to update each power prediction model according to the acquired data.

[0040] On the other hand, an embodiment of the present application provides a computer device, the computer device including a processor and a memory:

[0041] The memory is used to store a computer program and transmit the computer program to the processor;

[0042] The processor is configured to execute the method described above according to instructions in the computer program.

[0043] On the other hand, an embodiment of the present application provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed by a processor, it is used to implement the method described in the above aspects.

[0044] Due to the adoption of the above-mentioned technical solution, the present invention has the following beneficial effects:

[0045] The disclosed method, system, device, and medium for predicting renewable energy power based on electricity trading can optimize the structure of raw data through a preprocessing step, thereby improving data quality and making the data more adaptable, thereby significantly enhancing model performance. Furthermore, through reasonable feature processing, noise and redundancy can be effectively reduced, allowing the model to focus more on learning key features, thereby achieving more accurate prediction results.

[0046] Different types of power prediction models cover a range of algorithmic models, from traditional machine learning to deep learning, designed to meet the power prediction needs of renewable energy power plants at different time scales. Flexible configuration options are provided so that users can create customized prediction solutions suitable for specific scenarios.

[0047] In addition, by clearly dividing different dimensions, we can have a clearer understanding of how each factor affects the operation of the power system, and the interactions between different factors can also be better captured. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic flow chart of a method of the present invention is shown;

[0049] Figure 2 A structural diagram of a multi-level business topic. DETAILED DESCRIPTION

[0050] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0051] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0052] The traditional method currently used is to predict the power and power generation of different types of renewable energy power plants separately. However, the prediction system cannot achieve integrated prediction and its applicable scenarios are limited.

[0053] In view of this, an embodiment of the present invention provides a new energy power prediction method based on electricity trading. The embodiment of the present invention can predict the new energy power generation power according to the electricity consumption of multiple users. The method can be executed by a multi-physics field computing device. The multi-physics field computing device can be implemented in the form of hardware and / or software. The multi-physics field computing device can be configured in a computer device.

[0054] See Figure 1 , Figure 1 The flowchart of the new energy power prediction method based on power trading provided by an embodiment of the present invention is illustrated.

[0055] The embodiment of the present invention provides a new energy power prediction method based on power trading, including:

[0056] S1. Obtaining grid operation data and environmental data around new energy power stations;

[0057] Grid operation data includes grid-side power supply data. New energy power stations generally include wind farms, cluster photovoltaic power stations, regional photovoltaic power stations, and wind-solar power stations. Environmental data surrounding new energy power stations refers to environmental data such as wind speed, sunshine duration, and sunshine intensity in the areas where these power stations are located.

[0058] S2. Clean the acquired data to generate raw data. Data cleaning is important to maintain the authenticity and representativeness of the data and avoid data distortion caused by excessive cleaning. In addition, data cleaning is an iterative process, and the cleaning strategy can be adjusted based on model feedback.

[0059] Specifically, step S2 includes:

[0060] S201. Preliminary inspection, including checking the integrity of the data and detecting outliers in the data;

[0061] Integrity check: Check whether there are missing values ​​in the dataset composed of the original data of power grid data, wind speed, and sunlight. Fill or delete the missing data.

[0062] Outlier detection: using statistical methods (such as boxplots) to identify outliers in a dataset and decide whether to correct, delete, or retain them;

[0063] S202. Data synchronization: Perform time series alignment on the data after preliminary inspection, that is, ensure that data from different sources such as power grid data, wind speed, and sunlight are aligned in time, and perform time series alignment on the data;

[0064] S203. Data conversion, normalization or standardization of the synchronized data;

[0065] Normalization / standardization: convert data into a unified dimension and scale to facilitate model processing;

[0066] Specifically, in a specific scenario, for wind speed and sunlight data, it is necessary to remove erroneous readings due to sensor failure or environmental factors; for power grid data, it is necessary to correct data deviations caused by power system operation or measurement errors;

[0067] S204. Data verification: Verify whether the converted raw data meets the requirements of the power prediction model. As the final stage of data cleaning, it is necessary to verify the effectiveness of data cleaning to ensure that the data quality meets the requirements of the prediction model. According to the needs of the prediction model, new features need to be extracted from the raw data, such as the rate of change of wind speed and the diurnal variation of light intensity.

[0068] S3. Preprocess the raw data to optimize its structure and generate a power indicator dataset;

[0069] In addition, preprocessing can also perform other processing such as filling missing values, removing outliers or normalizing the cleaned raw data to further ensure that the data fed into the power prediction model has high quality.

[0070] Specifically, step S3 includes:

[0071] S301. Feature processing: feature selection, generation and conversion based on the cleaned raw data;

[0072] S302. Integrate market information, collect real-time market prices, and obtain historical transaction information in the market and form a historical market price sequence by analyzing market mechanisms, quotation rules, and transaction conditions;

[0073] S303. Construct a comprehensive data set, combine the processed feature set with market information, and form a comprehensive power index data set containing multi-dimensional information.

[0074] In addition, combined with the Figure 2 , the dimension information of the power indicator data set includes time dimension, space dimension, business dimension and power indicator dimension;

[0075] Each dimension of information is composed of multi-level business themes, where the first-level business themes include economic factors, weather factors, and meteorological factors; the second-level business themes include wind resources and solar resources; and the third-level business themes include exogenous and endogenous factors.

[0076] The economic factors are exogenous factors, weather factors correspond to the conditions for the use of wind resources, meteorological factors correspond to the conditions for the use of solar resources, and wind resources and solar resources correspond to endogenous factors.

[0077] The exogenous and endogenous influencing factors determine the power supply level of the corresponding dimension information;

[0078] It is important to note that the time dimension considers the time span of the data, such as hourly, daily, weekly, etc. The data in the business theme will also change over time, so it needs to be organized according to time series. For example, economic factors (such as electricity price policy) and weather factors (such as wind speed and light intensity) will change over time;

[0079] The spatial dimension refers to data at the geographic level, including provincial, municipal, and county levels. Weather conditions (such as wind speed and sunlight) vary across geographic locations, making the spatial dimension particularly important for wind and solar power generation forecasts. Weather and meteorological factors within business themes exhibit significant spatial variation.

[0080] The business dimension covers various business activities and influencing factors related to the power market, such as economic factors, weather factors, and meteorological factors. These factors together determine the supply level of the power market.

[0081] Among them, economic factors, such as electricity price policies, market demand, and macroeconomic conditions, can be regarded as exogenous influencing factors of electricity quantity indicator variables.

[0082] Weather factors: such as wind speed and light intensity, directly affect the utilization conditions of wind and solar energy resources.

[0083] Meteorological factors: More detailed meteorological information, such as temperature and humidity, will also affect power generation efficiency.

[0084] The electricity indicator dimension directly reflects the operating status of the power system, including key indicators such as actual power generation, power consumption, and grid load.

[0085] Furthermore, all factors within the business theme ultimately impact electricity metrics. For example, weather factors determine the actual generation capacity of renewable energy, while economic factors influence electricity demand.

[0086] S4. Use different machine learning methods to build different types of power prediction models;

[0087] After building and training the power prediction model, a model engineering interface is provided, which lists available power prediction algorithm models in a table format, including ultra-short-term, short-term, medium- to long-term, and probabilistic prediction algorithms. Users can then drag and drop algorithm models from the model library to create a custom algorithm combination, and adjust the order and dependencies between algorithms by dragging and dropping. Finally, click on an algorithm model to view more detailed information, including the model's operation history and parameter configuration. Users can also save and load their custom algorithm combination configurations.

[0088] Specifically, the power prediction model includes a wind and solar ultra-short-term power prediction model, a deep learning-based model, a cluster short-term power prediction model, a short-term power prediction model, and a medium- and long-term power prediction model;

[0089] Among them, the wind and solar ultra-short-term power prediction model is suitable for rapid response to weather changes and is used for real-time power prediction of wind farms and photovoltaic power stations;

[0090] The power forecasting model is based on the FCM and SSA-ELM models. This model uses fuzzy C-means clustering (FCM) to select similar days and combines it with the swarm algorithm to optimize the extreme learning machine (SSA-ELM) to perform ultra-short-term power forecasting of wind farms, targeting the large fluctuations and strong randomness of wind power output.

[0091] Among them, Fuzzy C-means Clustering (FCM) is a data clustering algorithm commonly used to identify natural groupings in data. In wind and solar power forecasting, FCM is used to identify similar weather patterns in historical data, thereby helping the model better understand power output under different weather conditions. Sparrow Search Algorithm Optimized Extreme Learning Machine (SSA-ELM) is an optimization algorithm used to improve the performance of the Extreme Learning Machine (ELM). The Extreme Learning Machine is a fast single-hidden layer feedforward neural network, and the Sparrow Search Algorithm is an efficient optimization tool that can be used to optimize the weights and biases of the ELM, thereby improving the model's prediction accuracy.

[0092] The construction of the wind and solar ultra-short-term power prediction model includes the following steps:

[0093] ① Use FCM to cluster historical wind and solar power data and corresponding weather data to identify different weather patterns. In this process, the input of Fuzzy C-means Clustering (FCM) is: historical wind and solar power data X and corresponding weather data W.

[0094] FCM clustering objective function:

[0095]

[0096] Among them, u ij is the membership of the i-th data point to the j-th class, c j is the cluster center of the jth class, m is the fuzzy factor; the output is the cluster center C and the membership matrix U, and all the original data (X, W) are clustered according to the cluster center and the membership matrix to obtain the processed data set (X', W').

[0097] ②For each cluster, use SSA to optimize the weights and biases of the ELM model, and use the optimized ELM model to predict the power output of each cluster; in this process, the input is the dataset (X', W') processed by FCM;

[0098] The ELM model is:

[0099]

[0100] Among them, β i is the output layer weight, ω i and b i are the weights and biases of the i-th hidden layer node, and g(·) is the activation function;

[0101] The optimization objective of SSA is to minimize Loss(Y′,Y), where Y′ is the predicted output of the ELM model and Y is the actual power output.

[0102] ③ Combine the prediction results of all clusters to obtain the final wind and solar power forecast. In this process, for each FCM clustering result, use SSA-ELM to predict power. The prediction expression is: P′(t) = f′(X′(t), W′(t)), where P′(t) is the predicted power output and f′(·) is the ELM model optimized by SSA.

[0103] When making the final prediction, the prediction results of all clusters are combined and weighted averaged:

[0104]

[0105] where ω j is the weight of the j-th cluster.

[0106] The deep learning-based model is suitable for wind and solar farms with large amounts of data and flat terrain conditions;

[0107] Specifically, the model uses stacked denoising autoencoders to predict wind and solar farm power. Compared with traditional prediction models, this method performs well in improving the accuracy of wind and solar farm power prediction, especially in flat terrain areas.

[0108] The expression for wind and solar farm power prediction using stacked denoising autoencoders (SDAE) based on deep learning model can be described as follows:

[0109] Stacked Denoising Autoencoders (SDAE):

[0110] Input: historical wind farm power data X and corresponding weather data W;

[0111] Denoising Autoencoder (DAE):

[0112] in, is the reconstructed data, f(·) is the encoding-decoding process of DAE, and θ is the parameter of DAE;

[0113] Stacking: Stack multiple DAE layers to form a deep network structure;

[0114] SDAE training objective: minimize the reconstruction error Loss(X,X^).

[0115] Power prediction model:

[0116] Input: dataset (X′, W′) after SDAE pre-training;

[0117] Deep Neural Network (DNN) Model:

[0118] P′(t)=g(X′(t),W′(t),φ)

[0119] Where P′(t) is the predicted power output, g(·) is the DNN model, and φ is the parameter of the DNN;

[0120] Training objective: Minimize the prediction error Loss(P,P′), where P is the actual power output.

[0121] Wind and solar farm power prediction model combining SDAE and DNN:

[0122] Use SDAE to pre-train and extract features from raw data;

[0123] Use DNN to predict the power of data processed by SDAE;

[0124] Final prediction expression: P′(t)=g(f(X′(t),W′(t),θ),φ).

[0125] The cluster short-term power forecasting model is applicable to wind and solar farm clusters that are geographically distributed but have similar environmental conditions.

[0126] Specifically, it uses the binary K-means cluster analysis method to perform sub-region division, cluster feature point optimization, and regional power prediction. This method is suitable for clustered wind and solar power plants.

[0127] The cluster short-term power forecast specifically includes the following steps:

[0128] ①Bipartite K-means clustering

[0129] Input the historical power data X and corresponding feature data F of the wind and solar power plant, and divide the power plant into k sub-regions through binary K-means clustering. The data points in each sub-region have high similarity.

[0130] Minimize the squared error within a subregion:

[0131]

[0132] Among them, S i is the i-th subregion, μ i is the center of the sub-region, and finally outputs the sub-region division results S1, S2, ..., S k .

[0133] ② Cluster feature point optimization

[0134] For each sub-region, select the most representative feature point as the cluster feature point; select the point closest to the sub-region center, or select according to other specific criteria (such as power output stability, weather conditions, etc.); output the cluster feature point set P1, P2, ..., P k .

[0135] ③ Regional power forecast

[0136] For each sub-region, power prediction is performed using its cluster feature points. The prediction expression is P i (t) = h(P i1 ,P i2 ,......P ik ,t), where P i (t) is the predicted power of the i-th sub-region at time t, h is the prediction model, and the prediction model can be a regression model based on historical data, a machine learning model, or a deep learning model.

[0137] ④ Cluster wind and solar power plant power prediction

[0138] Combining the prediction results of all sub-areas, we can get the power prediction of the entire power plant. The final prediction expression is:

[0139]

[0140] Short-term power forecasting models are suitable for comprehensive management of wind farms;

[0141] In addition, the short-term power forecasting model includes a regional wind farm cluster centralized power forecasting model and a wind-solar hybrid forecasting model based on BP neural network, which are respectively suitable for the integrated management of multiple adjacent wind farms and the simultaneous management of wind power and photovoltaic power stations;

[0142] The centralized power prediction model for regional wind farm clusters takes into account the aggregation of data from adjacent wind farms within the region, enabling the prediction model to fully consider the temporal and spatial correlations between sites. Specifically, it includes the following steps:

[0143] ①Data aggregation:

[0144] Input the historical power data P of each wind farm in the area i,t and the corresponding weather data W i,t , where i represents the wind farm number and t represents time. Based on the spatial correlation between wind farms, the data of adjacent wind farms in the region are aggregated. The aggregation method can be a simple arithmetic average, weighted average, or a more complex spatiotemporal model, such as Copula function, spatiotemporal interpolation, etc.

[0145] ② The data set after data aggregation processing (P′ i，t , W′i，t ) Input a pre-built machine learning model, deep learning model, or statistical model to obtain a prediction expression:

[0146]

[0147] Among them, P i,t is the predicted power output, g is the pre-selected prediction model, and θ is the model parameter.

[0148] ③ For each wind farm, use the aggregated data to perform power forecasting. Combine the forecast results of all wind farms to obtain the power forecast for the entire regional wind farm group. The final forecast expression is:

[0149]

[0150] Where N is the number of wind farms in the region.

[0151] The wind-solar hybrid prediction model based on BP neural network combines the prediction of wind power and photovoltaic power into one process, which simplifies the prediction process and realizes the output power prediction of wind farms and photovoltaic power stations in the entire region at the same time.

[0152] The wind-solar hybrid prediction model based on BP (back propagation) neural network is used to simultaneously predict wind power and photovoltaic power, which can be expressed as:

[0153] Input data: historical wind power data P wind,t and photovoltaic power history data P solar,t , the corresponding weather data W t , including wind speed, light intensity, temperature, etc.;

[0154] BP neural network model:

[0155] Input layer: takes historical data of wind power and photovoltaic power as well as weather data as input;

[0156] Hidden layer: multiple neurons, each of which applies an activation function σ to process the input signal;

[0157] Output layer: predicted wind power and photovoltaic power

[0158] Model expression:

[0159]

[0160] Among them, f is the BP neural network model and θ is the model parameter.

[0161] Prediction process:

[0162] Training phase: Use historical data to train the BP neural network and adjust the network weights through the back propagation algorithm to minimize the prediction error;

[0163] Prediction phase: Use the trained model to predict the power of new input data;

[0164] Output power forecast of regional wind farms and photovoltaic power stations:

[0165] For wind farms and photovoltaic power plants in the entire region, the same BP neural network model is used for power prediction;

[0166] The final prediction expression:

[0167] The medium- and long-term power forecast model is suitable for wind farms that need to plan and dispatch power generation capacity in the next few days to months, providing medium- and long-term power forecasts.

[0168] The model uses graph convolutional networks (GCN), wind speed differential fitting (DF) and particle swarm optimization (PSO) to improve the accuracy of medium- and long-term wind farm power generation forecasting.

[0169] The medium- and long-term wind power prediction model based on graph convolutional network (GCN), wind speed differential fitting (DF) and particle swarm optimization algorithm (PSO) can be expressed as:

[0170] Wind speed differential fitting (DF):

[0171] Input: Historical wind speed data V t ;

[0172] Difference fitting: perform differential processing on wind speed data to capture the changing trend of wind speed;

[0173] Differential expression: ΔV t =V t -V t-1 .

[0174] Graph Convolutional Network (GCN):

[0175] Input: Wind speed data ΔV after differential processing t and other related features X t ;

[0176] GCN model: extracts spatial features from wind speed data through graph convolutional layers;

[0177] GCN expression: H t =GCN(ΔV t ,X t ,A), where H t are the extracted features and A is the adjacency matrix of the graph.

[0178] Particle Swarm Optimization (PSO):

[0179] Optimization goal: minimize power prediction error;

[0180] Optimization parameters: weights and biases of the GCN model;

[0181] PSO expression: Find the optimal solution by iteratively updating the particle positions.

[0182] Medium- and long-term wind power forecast model:

[0183] Combine the features Ht extracted by GCN and other related features X t , power prediction is performed using the GCN model optimized by PSO;

[0184] Prediction expression:

[0185] in, is the predicted wind power, and θ is the model parameter optimized by PSO.

[0186] In addition, the power prediction models mentioned above can be various types of supervised recursive neural network models, such as support vector machines, random forests, generalized linear regression models, and gradient boosting decision trees. After the modeling is completed, automatic iterative learning of the model can be achieved by dynamically updating data samples during subsequent task scheduling and operation.

[0187] In addition, during the training process of each model, the system can implement a training task and model online interface that can display the progress and status of the current model training and calculation. Users can manually start or pause model training and calculation tasks, or set up automated training and calculation plans to ensure that tasks are automatically executed when needed. Users can view historical training and calculation task records, including completion time and performance indicators. After the task is completed, users can view detailed training and calculation reports, including training time, model performance, and corresponding model indicator data.

[0188] S5. Based on the power index data set, each power prediction model is used in turn to predict the power generation of the new energy power station;

[0189] S6. Analyze the output prediction results of various types;

[0190] The result post-processing module is mainly used to analyze the output of various types of prediction results and output the results with the best accuracy for users to choose;

[0191] In addition, because new energy power stations are different from traditional thermal power generation and are restricted by many meteorological or regional factors, there are often many types of new energy power stations. For different types of new energy power stations, under the same conditions, the factors affecting their power generation are also different; therefore, by constructing different types of power prediction models, after sequential calculations, the most accurate calculation result is selected as the power generation power of the new energy power station.

[0192] The above is a detailed description of an embodiment of a new energy power prediction method based on electricity trading provided by the present invention. The following is a detailed description of an embodiment of a new energy power prediction system based on electricity trading provided by the present invention.

[0193] The embodiment of the present invention provides a structure of a new energy power prediction system based on power trading, which includes an acquisition unit, a data processing unit, a model building unit and a calculation and analysis unit;

[0194] The acquisition unit is composed of various sensors and is used to acquire power grid operation data and environmental data around the new energy power station;

[0195] The data processing unit includes a cleaning module and a preprocessing module, the two modules are used to perform data cleaning and data preprocessing steps respectively;

[0196] The model building unit adopts different machine learning methods to build different types of power prediction models; in addition, the model building unit is electrically connected to the acquisition unit and is used to update each power prediction model according to the acquired data;

[0197] The calculation and analysis unit is used to receive the data output by the data processing unit and to call the power prediction models of the model building unit in sequence.

[0198] Some embodiments of the present invention further provide an electronic device, the electronic device including a memory and a processor;

[0199] The memory is used to store programs;

[0200] The processor executes the program to implement the steps of the new energy power prediction method based on power trading in any of the above embodiments.

[0201] Some embodiments of the present invention further provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the new energy power prediction method based on electricity trading in any of the above embodiments.

[0202] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, electronic devices, and computer storage media can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0203] In several embodiments provided by the present invention, it is understood that each box in the flow chart or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved.

[0204] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, electronic devices, computer storage media and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0205] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0206] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0207] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the method described in each embodiment of the present invention through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name: Read-Only Memory, English abbreviation: ROM), random access memory (full name: Random Access Memory, English abbreviation: RAM), magnetic disk or optical disk, etc. Various media that can store program codes.

[0208] The parts of the present invention that are not described in detail are prior art. It is obvious to those skilled in the art that the present invention is not limited to the details of the above-mentioned exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is limited by the appended claims rather than the above description. Therefore, it is intended that all changes that fall within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any figure marks in the claims should not be regarded as limiting the claims involved.

Claims

1. A new energy power prediction method based on power trading, characterized by: include: Obtain power grid operation data and environmental data around new energy power stations; Clean the acquired data to generate original data; Preprocess the raw data to optimize the structure of the raw data and generate a power index data set; Use different machine learning methods to build different types of power prediction models; Based on the power index data set, each power prediction model is used in turn to predict the power generation of the new energy power station; Analyze the output of each type of prediction results.

2. The new energy power prediction method based on power trading as claimed in claim 1 is characterized by: The specific steps of data cleaning include: Preliminary checks, including checking the integrity of the data and detecting outliers in the data; Data synchronization, time series alignment of data after preliminary inspection; Data conversion, normalization or standardization of synchronized data; Data verification: Verify whether the converted raw data meets the requirements of the power prediction model.

3. The new energy power prediction method based on power trading as claimed in claim 1 is characterized by: The specific steps of the pretreatment include: Feature processing: feature selection, generation and conversion based on the cleaned raw data; Integrate market information, collect real-time market prices, and obtain historical transaction information in the market and form a historical market price sequence by analyzing market mechanisms, quotation rules, and transaction conditions; Construct a comprehensive data set, combine the processed feature set with market information, and form a comprehensive power indicator data set containing information in multiple dimensions.

4. The new energy power prediction method based on power trading as claimed in claim 3 is characterized by: The dimension information of the power index data set includes time dimension, space dimension, business dimension and power index dimension; Each dimension of information is composed of multi-level business themes, among which the first-level business themes include economic factors, weather factors and meteorological factors, the second-level business themes include wind resources and solar resources, and the third-level business themes include exogenous factors and endogenous factors. The economic factors are exogenous influencing factors, the weather factors correspond to the conditions for the utilization of wind resources, the meteorological factors correspond to the conditions for the utilization of solar energy resources, and the wind resources and solar energy resources correspond to the endogenous influencing factors. The exogenous influencing factors and endogenous influencing factors determine the power supply level of the corresponding dimensional information.

5. The new energy power prediction method based on power trading as claimed in claim 1 is characterized by: The power prediction model includes a wind-solar ultra-short-term power prediction model, a model based on deep learning, a cluster-based short-term power prediction model, a short-term power prediction model, and a medium- and long-term power prediction model; wherein the wind-solar ultra-short-term power prediction model is suitable for rapid response to weather changes and is used for instant power prediction of wind farms and photovoltaic power stations; The deep learning-based model is suitable for wind and solar farms with large amounts of data and flat terrain conditions; The cluster-based short-term power forecasting model is applicable to wind and solar farm clusters that are geographically widely distributed but have similar environmental conditions; The short-term power forecast model is suitable for comprehensive management of wind farms; The medium- and long-term power forecast model is suitable for wind farms that need to plan and dispatch power generation capacity in the next few days to months, providing medium- and long-term power forecasts.

6. The new energy power prediction method based on power trading as claimed in claim 5 is characterized by: The short-term power prediction model includes a regional wind farm cluster centralized power prediction model and a wind-solar hybrid prediction model based on a BP neural network, which are respectively suitable for the comprehensive management of multiple adjacent wind farms and the simultaneous management of wind power and photovoltaic power stations.

7. A new energy power prediction system based on power trading, characterized by: It includes an acquisition unit, a data processing unit, a model building unit and a calculation and analysis unit; The acquisition unit is composed of various sensors and is used to acquire power grid operation data and environmental data around the new energy power station; The data processing unit includes a cleaning module and a preprocessing module, and the two modules are used to perform data cleaning and data preprocessing steps respectively; The model building unit adopts different machine learning methods to build different types of power prediction models; The calculation and analysis unit is used to receive the data output by the data processing unit, and to call the power prediction models of the model building unit in sequence.

8. The new energy power prediction system based on power trading as claimed in claim 7, characterized in that: The model building unit is electrically connected to the acquisition unit and is used to update each power prediction model according to the acquired data.

9. A computer device, characterized in that: The computer device comprises a processor and a memory: The memory is used to store a computer program and transmit the computer program to the processor; The processor is configured to execute the method according to any one of claims 1 to 6 according to the instructions in the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and the computer program is used to implement the method according to any one of claims 1 to 6 when executed by a processor.

Images