Multi-node energy consumption prediction method for transformer substation

Abstract

The invention relates to the technical field of transformer substations, in particular to a transformer substation multi-node energy consumption prediction method, which comprises the following steps that: a data acquisition and integration module collects energy consumption data, meteorological data and equipment state data from multiple channels of a transformer substation monitoring system, a meteorological system and historical energy consumption records, and integrates the energy consumption data, the meteorological data and the equipment state data; extracting features helpful for energy consumption prediction from the original data; by accurately predicting the energy consumption of each node of the transformer substation, the energy distribution of the electric power system can be better managed, unnecessary energy waste is reduced, the energy utilization efficiency is improved, the energy consumption prediction is beneficial to optimizing the operation strategy of the electric power system, reducing the use of fossil fuel, and further reducing the emission of greenhouse gases. The method has positive influence on environmental protection and climate change coping, energy conservation and emission reduction are achieved, meanwhile, the operation cost of a power system can be reduced, economic benefits are improved, and power waste and economic losses caused by too high energy consumption can be avoided through prediction and optimization.

Description

Technical Field

[0001] The invention relates to the technical field of transformer substations, in particular to a method for predicting multi-node energy consumption of transformer substations. Background Art

[0002] A substation is a place in the power system that transforms voltage and current, receives electric energy and distributes electric energy. The substation in a power plant is a step-up substation, which boosts the electric energy generated by the generator and feeds it into the high-voltage power grid. With the continuous growth of global energy demand and the improvement of environmental protection awareness, the effective use of energy and energy conservation and emission reduction have become the focus of social attention. As an important part of the power system, the energy consumption forecast of the substation is of great significance to improving energy utilization efficiency and power system stability.

[0003] However, the current substation energy consumption prediction methods have some shortcomings. The prediction accuracy is insufficient. When processing complex time series data, the prediction accuracy of the existing prediction methods needs to be improved. The model generalization ability is poor. The performance of the model on new data is unstable, and the generalization ability needs to be strengthened. The computational efficiency is low. The model training and prediction process takes a long time, and the computational efficiency needs to be improved. It lacks real-time performance. The existing prediction methods are difficult to meet the needs of real-time prediction. Therefore, it is necessary to propose a substation multi-node energy consumption prediction method. Summary of the invention

[0004] In view of the problems in the prior art, the present invention provides a method for predicting multi-node energy consumption of a substation.

[0005] The technical solution adopted by the present invention to solve the technical problem is: a method for predicting energy consumption of multiple nodes in a substation, S1, data collection and integration module collects energy consumption data, meteorological data, and equipment status data from multiple channels such as substation monitoring system, meteorological system, and historical energy consumption records, and integrates them to extract features that are helpful for energy consumption prediction from the original data; S2, the feature engineering module extracts features that are helpful for energy consumption prediction from the original data, such as time features, load features, and meteorological features; S3, model selection and construction module selects a suitable machine learning or deep learning model for energy consumption prediction, selects a suitable prediction model according to the nature of the problem and the characteristics of the data, and constructs the model; S4, the model training and optimization module uses historical data to train the model and optimizes the model through cross-validation and hyperparameter optimization methods; S5, the prediction and evaluation module uses the trained model to predict energy consumption and evaluates the prediction results through actual data; S6, visualization and decision support module visualizes the prediction results and provides intuitive energy consumption information and decision support for decision makers.

[0006] The data acquisition and integration module includes data acquisition and data preprocessing; The data collection is modularized by data entry through monitoring data, meteorological data, and historical data; The data preprocessing performs overall data processing operations through data cleaning, data standardization and data integration.

[0007] The feature engineering module includes feature selection, feature construction and feature conversion; The feature selection includes correlation analysis and principal component analysis; The correlation analysis is to analyze the relationship between each feature and energy consumption, and select features that are highly correlated with energy consumption; The principal component analysis reduces the dimensionality of high-dimensional data and retains the most important features; The feature construction includes time features, derivative features and feature conversion; The time characteristics are compared by multiple data of date, time and season, and these characteristics may have an impact on energy consumption; The derived features are new features calculated based on the original features, including daily average and weekly average of energy consumption; The feature conversion includes encoding conversion and function conversion; The encoding conversion converts the categorical features into numerical features; The functional conversion applies a mathematical function to convert the feature, such as a logarithm or an exponential.

[0008] In the model selection and construction module, statistical models, machine learning models and deep learning models are superimposed on each other, and a data operation library is built through front-end data collection and processed features to perform transportation library operations; Then the grid coordination search searches for the optimal parameters by traversing all parameter combinations. At the same time, the random search randomly searches in the parameter space to find the optimal parameters. In addition, the Bayesian optimization is applied to sample the parameter space based on the probability model to find the optimal parameters. Finally, the genetic algorithm is used to simulate the natural selection process to optimize the parameters through iterative selection and crossover.

[0009] The model training and optimization module operates through data segmentation: the collected data set is divided into a training set, a validation set and a test set; the training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the model performance.

[0010] Model training: Use the training set data to train the model, and be careful to avoid overfitting during the training process; Model validation: Use validation set data to validate the model, monitor the performance of the model through evaluation indicators, and adjust model parameters or select a more appropriate model based on the validation results; Model evaluation: Use the test set data to evaluate the final model to determine the generalization ability of the model; Model adjustment: Adjust the model according to the evaluation results, including parameter adjustment and model structure optimization, until the model meets the expected performance requirements.

[0011] The visualization and decision support module is displayed through forecast data and supports operators to make modifications and adjustments.

[0012] A method for using a multi-node energy consumption prediction method for a substation comprises the following steps: Step 1: Extract energy consumption data and meteorological data of each node of the substation from the database, perform data cleaning, standardization and feature extraction, select features that are helpful for energy consumption prediction based on correlation analysis and feature importance evaluation, and select a suitable prediction model based on the nature of the problem and the characteristics of the data; Step 2: Find the optimal parameters of the model through cross-validation and parameter optimization methods, train the model using the training set data, validate the model using the validation set data, adjust the model to reduce overfitting, and use the test set data to evaluate the generalization ability of the model; Step 3: Analyze the prediction results of the model, determine the accuracy, robustness and efficiency of the model, deploy the trained model to actual applications, and perform real-time energy consumption prediction Beneficial effects of the present invention: 1. The multi-node energy consumption prediction method of a substation described in the present invention can better manage the energy distribution of the power system and reduce unnecessary energy waste by accurately predicting the energy consumption of each node of the substation, thereby improving energy utilization efficiency. Energy consumption prediction helps to optimize the operation strategy of the power system, reduce the use of fossil fuels, and thus reduce greenhouse gas emissions. It has a positive impact on environmental protection and climate change response. While saving energy and reducing emissions, it can also reduce the operating cost of the power system and improve economic benefits. Through prediction and optimization, power waste and economic losses caused by excessive energy consumption can be avoided.

[0013] 2. The present invention discloses a method for predicting multi-node energy consumption in a substation. The method can help achieve balanced distribution of the load in the power system, avoid situations where some nodes are overloaded while other nodes are idle, and improve the stability and reliability of the system. Through energy consumption prediction, energy consumption anomalies can be discovered in time, and preventive maintenance can be performed on equipment to reduce the risk of failures and extend the service life of equipment. Energy consumption prediction provides a basis for scheduling of the power system, and the allocation of power resources can be adjusted according to the prediction results to achieve a more efficient and economical scheduling strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the overall process structure of the present invention; Figure 2 It is a schematic diagram of the data collection and integration module process structure of the present invention; Figure 3 It is a schematic diagram of the process structure of the feature engineering module of the present invention. DETAILED DESCRIPTION

[0015] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0016] See also Figure 1-3 , a multi-node energy consumption prediction method for a substation according to the present invention includes a data acquisition and integration module that collects energy consumption data, meteorological data, and equipment status data from multiple channels including a substation monitoring system, a meteorological system, and historical energy consumption records, and integrates the data to extract features that are helpful for energy consumption prediction from the original data; The feature engineering module extracts features that are helpful for energy consumption prediction from the raw data, such as time features, load features, and meteorological features; The model selection and construction module selects a suitable machine learning or deep learning model for energy consumption prediction, selects a suitable prediction model based on the nature of the problem and the characteristics of the data, and constructs the model; The model training and optimization module uses historical data to train the model and optimizes the model through cross-validation and hyperparameter optimization methods; The prediction and evaluation module uses the trained model to predict energy consumption and evaluates the prediction results through actual data; The visualization and decision support module visualizes the prediction results and provides intuitive energy consumption information and decision support to decision makers.

[0017] The data acquisition and integration module includes data acquisition and data preprocessing; The data collection is modularized by data entry through monitoring data, meteorological data, and historical data; The data preprocessing performs overall data processing operations through data cleaning, data standardization and data integration.

[0018] A method for predicting multi-node energy consumption in a substation according to claim 1, characterized in that: the feature engineering module includes feature selection, feature construction and feature conversion; The feature selection includes correlation analysis and principal component analysis; The correlation analysis is to analyze the relationship between each feature and energy consumption, and select features that are highly correlated with energy consumption; The principal component analysis reduces the dimensionality of high-dimensional data and retains the most important features; The feature construction includes time features, derivative features and feature conversion; The time characteristics are compared by multiple data of date, time and season, and these characteristics may have an impact on energy consumption; The derived features are new features calculated based on the original features, including daily average and weekly average of energy consumption; The feature conversion includes encoding conversion and function conversion; The encoding conversion converts the categorical features into numerical features; The functional conversion applies a mathematical function to convert the feature, such as a logarithm or an exponential.

[0019] In the model selection and construction module, statistical models, machine learning models and deep learning models are superimposed on each other, and a data operation library is built through front-end data collection and processed features to perform transportation library operations; Statistical models: such as linear regression and support vector machines, which are suitable for dealing with linear problems; Machine learning models: such as random forests and gradient boosted trees, which can handle nonlinear problems; Deep learning models: such as recurrent neural networks, long short-term memory networks, and convolutional neural networks, which are suitable for processing time series data; Among them, the long short-term memory network is a special RNN architecture that can learn long-term dependencies. In energy consumption prediction, LSTM can process time series data and capture historical energy consumption patterns. The running code of the example is: model=Sequential() model.add(LSTM(units=50,return_sequences=True,input_shape=(X_train.shape[1],1))) model.add(LSTM(units=50)) model.add(Dense(1)) model.compile(optimizer='adam', loss='mean_squared_error') model.fit(X_train,y_train,epochs=50,batch_size=32,validation_data=(X_val, y_val)) Among them, the gradient boosting tree is an ensemble learning algorithm based on the decision tree. It minimizes the loss function by iteratively training the decision tree. The running code of the example is: model=GradientBoostingRegressor(n_estimators=100,learning_rate=0.1,max_depth=3) model.fit(X_train, y_train) y_pred = model.predict(X_test) The swarm intelligence optimization algorithm with integrated adaptive weight adjustment combines integrated learning and swarm intelligence optimization methods to improve prediction performance by adaptively adjusting weights. The running code of the example is: while not_converged: adjust_weights() predict_with_intelligent_agents() evaluate_performance() Based on the above algorithm: Prediction accuracy: The prediction accuracy of the model is evaluated by calculating the mean square error and mean absolute error indicators; Generalization ability: Use the test set data to evaluate the generalization ability of the model to ensure that the model can make accurate predictions on unseen data; Computational efficiency: evaluate the computational complexity and running time of the model to ensure the efficiency of the model in practical applications; Robustness: Analyze the performance of the model under different conditions, such as different time periods and different weather conditions At the same time, the selected model should be able to explain the intrinsic reasons for changes in energy consumption to a certain extent. The model should have high prediction accuracy and reduce prediction errors. The model should have a certain generalization ability and be able to make accurate predictions for unseen data. The model construction and prediction process should have high computational efficiency.

[0020] Then the grid coordination search searches for the optimal parameters by traversing all parameter combinations. At the same time, the random search randomly searches in the parameter space to find the optimal parameters. In addition, the Bayesian optimization is applied to sample the parameter space based on the probability model to find the optimal parameters. Finally, the genetic algorithm is used to simulate the natural selection process to optimize the parameters through iterative selection and crossover.

[0021] The model training and optimization module operates through data segmentation: the collected data set is divided into a training set, a validation set and a test set; the training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the model performance.

[0022] Model training: Use the training set data to train the model, and be careful to avoid overfitting during the training process; Model validation: Use validation set data to validate the model, monitor the performance of the model through evaluation indicators, and adjust model parameters or select a more appropriate model based on the validation results; Model evaluation: Use the test set data to evaluate the final model to determine the generalization ability of the model; Model adjustment: Adjust the model according to the evaluation results, including parameter adjustment and model structure optimization, until the model meets the expected performance requirements.

[0023] The visualization and decision support module is displayed through forecast data and supports operators to make modifications and adjustments.

[0024] Experimental data Dataset Introduction The data set used in the experiment contains detailed energy consumption records from a regional substation, spanning a year. The data set specifically includes the following: Energy consumption data: hourly energy consumption data of each node in the substation in the past year; Meteorological data: temperature, humidity, wind speed weather information within the corresponding time; Equipment status data: operating status information of main equipment in the substation, such as load rate and switch status; The data set was preprocessed, including data cleaning, missing value filling, and outlier processing, to ensure the quality and integrity of the data.

[0025] Experimental environment and tools The experiment was conducted in the following environment: Operating system: Ubuntu 18.04 LTS Programming language: Python 3.7 Machine learning libraries: TensorFlow 2.3, scikit-learn 0.24.2 Data processing libraries: pandas 1.1.5, NumPy 1.19.2 Visualization tools: matplotlib 3.3.4, seaborn 0.11.1 The following main tools were used in the experiment: Jupyter Notebook: used for writing and executing experimental codes; TensorBoard: used to visualize the model training process and results; scikit-learn: used for model training, validation, and testing; Experimental results analysis The analysis of experimental results mainly includes the following aspects: Model prediction performance: Use the test set data to evaluate the prediction performance of the model and calculate the MSE and MAE indicators.

[0026] Draw a comparison chart between actual energy consumption and predicted energy consumption to intuitively display the prediction effect of the model.

[0027] Model generalization ability: Analyze the prediction performance of the model in different time periods and evaluate its generalization ability; compare the prediction results of different models and select the model with the best performance.

[0028] Model robustness: Test model performance under different meteorological conditions and equipment states to analyze the robustness of the model; perform sensitivity analysis on the model to determine the sensitivity of the model to the input data.

[0029] Model efficiency: Record and analyze the model’s training time and prediction time to evaluate the model’s computational efficiency.

[0030] Here is an example of the experimental results: Model performance indicators: MSE: 0.12 MAE: 0.08 Model generalization ability: The prediction accuracy on the test set is comparable to that on the training set, indicating that the model has good generalization ability.

[0031] The prediction error of the model under different meteorological conditions is within an acceptable range.

[0032] Model robustness: The model is insensitive to slight changes in the input data and shows good robustness.

[0033] Model efficiency: Model training time: about 2 hours; Model prediction time: 0.5 seconds per piece of data on average; Through these experiments and data analysis, the performance of the proposed multi-node energy consumption prediction algorithm can be evaluated, providing a basis for the optimization and application of the algorithm.

[0034] Optimization strategy Feature engineering optimization: Use more advanced feature selection techniques, such as model-based feature selection, to extract the most influential features and introduce new features, such as trend characteristics of historical energy consumption, the impact of holidays and special events.

[0035] Model structure optimization: For deep learning models, you can try different network architectures, such as adding or reducing hidden layers and adjusting the number of neurons. For machine learning models, you can try different algorithm combinations, such as random forests and gradient boosting trees in ensemble learning methods.

[0036] Hyperparameter tuning: Use more efficient hyperparameter optimization methods, such as Bayesian optimization and genetic algorithms, to find the best hyperparameter combination.

[0037] Data augmentation: Improve the generalization ability of the model by synthesizing new data samples or using data augmentation techniques to expand the dataset.

[0038] Model simplification: Reduce the complexity of the model, such as reducing the number of layers and neurons in the neural network to reduce the computational burden, and use model pruning techniques to remove unimportant weights and neurons.

[0039] Computing resource optimization: Use parallel computing and distributed computing to accelerate the model training and prediction process, and use specialized hardware accelerators such as GPUs or TPUs; Algorithm optimization: Use fast algorithms, such as fast Fourier transform (FFT), to speed up the data processing process, implement vectorized operations of the algorithm, and reduce the use of loops and iterations.

[0040] Optimize the model's generalization ability. The generalization ability of the model is the key to ensuring that the model performs well on unknown data. The following are some optimization strategies: Regularization: Apply regularization techniques such as L1 regularization, L2 regularization, or dropout to reduce the risk of overfitting.

[0041] Data augmentation: Data augmentation is used to simulate different data distributions and improve the performance of the model on unseen data.

[0042] Ensemble learning: Use ensemble learning methods, such as Bagging and Boosting, to combine the prediction results of multiple models to improve the overall generalization ability.

[0043] Transfer learning: Using transfer learning technology, the knowledge learned from one task is applied to another related task to improve the generalization ability of the model on the new task.

[0044] Through the above optimization strategy, the performance of the substation multi-node energy consumption prediction method can be effectively improved. In practical applications, the appropriate optimization method should be selected according to the specific situation to achieve the best balance between prediction accuracy, prediction speed and model generalization ability.

[0045] Specific use: Install the Python programming environment, ensure that the Python version meets the model operation requirements, install necessary machine learning and data processing libraries, such as TensorFlow, scikit-learn, and pandas, import the pre-trained energy consumption prediction model from the provided model file, extract the energy consumption data and meteorological data of each node of the substation from the database, perform data cleaning, standardization, and feature extraction, select features that are helpful for energy consumption prediction based on correlation analysis and feature importance evaluation, and select a suitable prediction model based on the nature of the problem and the characteristics of the data; find the optimal parameters of the model through cross-validation and parameter optimization methods, train the model with training set data, verify the model with validation set data, adjust the model to reduce overfitting, and use the test set data to evaluate the generalization ability of the model; analyze the prediction results of the model to determine the accuracy, robustness, and efficiency of the model, deploy the trained model to actual applications, perform real-time energy consumption prediction, output the prediction results to the specified file or database for further analysis and use, and use visualization tools to visualize the prediction results, such as drawing a comparison chart between actual energy consumption and predicted energy consumption.

[0046] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for predicting multi-node energy consumption in a substation, characterized in that: include: S1, data collection and integration module collects energy consumption data, meteorological data, and equipment status data from multiple channels such as substation monitoring system, meteorological system, and historical energy consumption records, and integrates them to extract features that are helpful for energy consumption prediction from the original data; S2, the feature engineering module extracts features that are helpful for energy consumption prediction from the original data, such as time features, load features, and meteorological features; S3, model selection and construction module selects a suitable machine learning or deep learning model for energy consumption prediction, selects a suitable prediction model according to the nature of the problem and the characteristics of the data, and constructs the model; S4, the model training and optimization module uses historical data to train the model and optimizes the model through cross-validation and hyperparameter optimization methods; S5, the prediction and evaluation module uses the trained model to predict energy consumption and evaluates the prediction results through actual data; S6, visualization and decision support module visualizes the prediction results and provides intuitive energy consumption information and decision support for decision makers.

2. A method for predicting multi-node energy consumption of a substation according to claim 1, characterized in that: The data acquisition and integration module includes data acquisition and data preprocessing; The data collection is modularized by data entry through monitoring data, meteorological data, and historical data; The data preprocessing performs overall data processing operations through data cleaning, data standardization and data integration.

3. A method for predicting multi-node energy consumption of a substation according to claim 1, characterized in that: The feature engineering module includes feature selection, feature construction and feature conversion; The feature selection includes correlation analysis and principal component analysis; The correlation analysis is to analyze the relationship between each feature and energy consumption, and select features that are highly correlated with energy consumption; The principal component analysis reduces the dimensionality of high-dimensional data and retains the most important features; The feature construction includes time features, derivative features and feature conversion; The time characteristics are compared by multiple data of date, time and season, and these characteristics may have an impact on energy consumption; The derived features are new features calculated based on the original features, including daily average and weekly average of energy consumption; The feature conversion includes encoding conversion and function conversion; The encoding conversion converts the categorical features into numerical features; The functional conversion applies a mathematical function to convert the feature, such as a logarithm or an exponential.

4. A method for predicting multi-node energy consumption of a substation according to claim 1, characterized in that: In the model selection and construction module, statistical models, machine learning models and deep learning models are superimposed on each other, and a data operation library is built through front-end data collection and processed features to perform transportation library operations; Then the grid coordination search searches for the optimal parameters by traversing all parameter combinations. At the same time, the random search randomly searches in the parameter space to find the optimal parameters. In addition, the Bayesian optimization is applied to sample the parameter space based on the probability model to find the optimal parameters. Finally, the genetic algorithm is used to simulate the natural selection process to optimize the parameters through iterative selection and crossover.

5. A method for predicting multi-node energy consumption of a substation according to claim 1, characterized in that: The model training and optimization module operates by data segmentation: the collected data set is divided into a training set, a validation set and a test set; The training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the model performance; Model training: Use the training set data to train the model, and be careful to avoid overfitting during the training process; Model validation: Use validation set data to validate the model, monitor the performance of the model through evaluation indicators, and adjust model parameters or select a more appropriate model based on the validation results; Model evaluation: Use the test set data to evaluate the final model to determine the generalization ability of the model; Model adjustment: Adjust the model according to the evaluation results, including parameter adjustment and model structure optimization, until the model meets the expected performance requirements.

6. A method for predicting multi-node energy consumption of a substation according to claim 1, characterized in that: The visualization and decision support module is displayed through forecast data and supports operators to make modifications and adjustments.

7. A method for using the method for predicting multi-node energy consumption in a substation according to any one of claims 1 to 6, characterized in that: The following steps are involved: Step 1: Extract energy consumption data and meteorological data of each node of the substation from the database, perform data cleaning, standardization and feature extraction, select features that are helpful for energy consumption prediction based on correlation analysis and feature importance evaluation, and select a suitable prediction model based on the nature of the problem and the characteristics of the data; Step 2: Find the optimal parameters of the model through cross-validation and parameter optimization methods, train the model using the training set data, validate the model using the validation set data, adjust the model to reduce overfitting, and use the test set data to evaluate the generalization ability of the model; Step 3: Analyze the prediction results of the model, determine the accuracy, robustness and efficiency of the model, and deploy the trained model to practical applications to perform real-time energy consumption prediction.

Images