Power grid dispatching methods, equipment, and media based on carbon emission prediction mechanisms

Abstract

This invention discloses a power grid dispatching method, equipment, and medium based on a carbon emission prediction mechanism. The method involves acquiring carbon emission data related to power plants; preprocessing the carbon emission data and using the preprocessed data as sample data; dividing the sample data into training and testing sets; training an improved BP neural network using the training and testing sets; predicting future carbon emissions from power plants using the improved BP neural network; and inputting the future carbon emissions from power plants into an objective function that minimizes carbon emissions, outputting power plant start-up, shutdown, and output plans. The power grid dispatching method, equipment, and medium based on a carbon emission prediction mechanism provided by this invention accurately grasps the future carbon emission capacity per unit of electricity from thermal power plant units through an improved BP neural network, achieving low-carbon dispatching of large power grids under a dual-carbon background, and providing data support and auxiliary decision-making for low-carbon operation.

Description

Technical Field

[0001] This invention relates to a power grid dispatching method, equipment, and medium based on a carbon emission prediction mechanism, belonging to the field of power grid dispatching automation technology. Background Technology

[0002] Carbon emission reduction has become a global hot topic. This has impacted the trading rules of power grid dispatching entities. To ensure the safe and stable operation of power grid dispatching, it is necessary to accurately predict the carbon emissions of each power grid dispatching entity, enabling them to implement accurate dispatching plans based on different carbon emission levels.

[0003] Currently, most carbon emission prediction technologies employ linear or nonlinear theories, taking into account regional carbon emission factors released by relevant departments. These technologies have a wide scope and are applicable to a broad range of carbon emission data. However, they can only serve as trend predictions, and the prediction results have relatively large errors.

[0004] Furthermore, the lack of historical data from individual samples means that the uniqueness of the sample data is not taken into account. Therefore, it is necessary to study methods for accurate carbon emission prediction and to provide effective and reliable data support for low-carbon dispatching, thereby achieving more accurate power grid dispatching plans. Summary of the Invention

[0005] Objective: In order to overcome the shortcomings of the existing technology, this invention provides a power grid dispatching method, equipment and medium based on carbon emission prediction mechanism, which provides data support and auxiliary decision-making for low-carbon dispatching.

[0006] Technical solution: To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] Firstly, a power grid dispatching method based on a carbon emission prediction mechanism includes the following steps:

[0008] Step 1: Obtain carbon emission data related to the power plant.

[0009] Step 2: Preprocess the carbon emission data related to the power plant and use the preprocessed data as sample data.

[0010] Step 3: Divide the sample data into training set and test set.

[0011] Step 4: Train the improved BP neural network using the training set and the test set to obtain the trained improved BP neural network.

[0012] Step 5: Use the trained, improved BP neural network to predict the future carbon emissions of the power plant.

[0013] Step 6: Input the future carbon emissions of the power plant into the objective function with the goal of minimizing carbon emissions, and output the power plant start-up, shutdown and output plans.

[0014] Furthermore, the carbon emission data related to the power plant includes, but is not limited to: power plant units, raw coal conversion factor, data month, fossil combustion carbon emissions, coal consumption, desulfurizer carbonate content, limestone consumption, desulfurizer emission factor, power consumption rate, carbon emissions during the desulfurization process, elemental carbon, unit maintenance status, volatile matter, temperature, lower heating value, purchased electricity, carbon content per unit calorific value, carbon emission factor of purchased electricity, carbon oxidation rate, carbon emissions from purchased electricity, overall carbon emissions, and carbon emission capacity per kilowatt-hour.

[0015] Furthermore, the preprocessing method includes: verifying and correcting erroneous data, and replacing empty data with historical averages.

[0016] Furthermore, the improved BP neural network includes: MobileNetV2, a BP network. A first convolutional layer, MobileNetV2, a second convolutional layer, an average pooling layer, and the BP neural network are sequentially connected.

[0017] Furthermore, the MobileNetV2 network structure is mainly composed of 17 inverse residual units.

[0018] Furthermore, the BP neural network structure is a hierarchical neural network consisting of an input layer, hidden layers, and an output layer. The hidden layers can be expanded into multiple adjacent layers. Neurons in adjacent layers are fully connected, while neurons within each layer are not connected to each other.

[0019] Furthermore, the BP neural network learns in a teacher-guided manner. When a pair of learning patterns is provided to the BP neural network, each neuron receives the input response and generates connection weights. Then, in the direction of reducing the error between the desired output and the actual output, the connection weights are adjusted layer by layer from the output layer through the hidden layers, returning to the input layer. This process is repeated alternately until the global error of the network tends to a given minimum value, thus completing the learning process and obtaining the connection weight parameters.

[0020] Furthermore, the step of training the improved BP neural network using the training set and the test set to obtain the trained improved BP neural network includes:

[0021] The training set is input into the improved BP neural network for training until the improved BP neural network converges, thus obtaining the initial improved BP neural network model.

[0022] The test set is input into the initial improved BP neural network model for testing. When the error function E corresponding to the test set reaches the threshold, the trained improved BP neural network model is obtained.

[0023] Furthermore, the process of inputting the future carbon emissions of the power plant into an objective function with the goal of minimizing carbon emissions, and outputting power plant start-up, shutdown, and output plans, includes:

[0024] Obtain the objective function, and the formula for calculating the objective function is as follows:

[0025]

[0026] Where, δ co2 Let be the objective function. This represents the carbon emissions corresponding to normal operation of the unit. This represents the carbon emissions corresponding to the unit's startup time. M represents the carbon emissions corresponding to the unit's shutdown, where M is the number of power plant units and T is the number of dispatch periods.

[0027] Obtain the predicted value λ of future carbon emissions from power plants co2 The constraint condition st. is constructed, and the formula for calculating st. is as follows:

[0028]

[0029] Under the premise of satisfying the constraint st., the objective function is solved with the goal of minimizing carbon emissions, and the power plant start-up, shutdown and output plans are obtained.

[0030] In a second aspect, a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a power grid dispatching method based on a carbon emission prediction mechanism as described in any of the first aspects.

[0031] Thirdly, a computer device comprising:

[0032] Memory is used to store instructions.

[0033] A processor is configured to execute the instructions, causing the computer device to perform operations of a power grid dispatching method based on a carbon emission prediction mechanism as described in any of the first aspects.

[0034] Beneficial Effects: The grid dispatching method, equipment, and medium based on carbon emission prediction mechanisms provided in this invention address the current mainstream calculation method which uses historical statistical data to obtain the historical average value of the unit's carbon emission capacity per unit of electricity. However, this average-based quantification method is more suitable for large-granularity assessments and cannot provide effective and reliable data support for low-carbon dispatching. This invention, by improving the BP neural network, accurately grasps the future carbon emission capacity per unit of electricity of thermal power plant units, providing data support and auxiliary decision-making for low-carbon operation. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the method of the present invention.

[0036] Figure 2 A schematic diagram of the improved BP neural network of the present invention.

[0037] Figure 3 This is a schematic diagram of the BP neural network model of the present invention.

[0038] Figure 4 The flowchart for the improved BP neural network training of this invention is shown below.

[0039] Figure 5 The diagram shows the interface of the improved BP neural network training software of this invention. Detailed Implementation

[0040] The present invention will be further described below with reference to specific embodiments.

[0041] like Figure 1 As shown, the first embodiment of a power grid dispatching method based on a carbon emission prediction mechanism includes the following steps:

[0042] Step 1: Obtain carbon emission data related to the power plant.

[0043] Step 2: Preprocess the carbon emission data related to the power plant and use the preprocessed data as sample data.

[0044] Step 3: Divide the sample data into training set and test set.

[0045] Step 4: Train the improved BP neural network using the training set and the test set to obtain the trained improved BP neural network.

[0046] Step 5: Use the trained, improved BP neural network to predict the future carbon emissions of the power plant.

[0047] Step 6: Input the future carbon emissions of the power plant into the objective function with the goal of minimizing carbon emissions, and output the power plant start-up, shutdown and output plans.

[0048] Furthermore, in one embodiment, the carbon emission data related to the power plant includes, but is not limited to: power plant units, raw coal conversion factor, data month, fossil combustion carbon emissions, coal consumption, desulfurizer carbonate content, limestone consumption, desulfurizer emission factor, power consumption rate, carbon emissions during the desulfurization process, elemental carbon, unit maintenance status, volatile matter, temperature, lower heating value, purchased electricity, carbon content per unit calorific value, carbon emission factor of purchased electricity, carbon oxidation rate, carbon emissions from purchased electricity, overall carbon emissions, and carbon emission capacity per kilowatt-hour.

[0049] In another embodiment, the preprocessing method includes: verifying and correcting erroneous data, and replacing empty data with historical averages.

[0050] Furthermore, in one embodiment, such as Figure 2 As shown, the improved BP neural network includes: MobileNetV2, a BP network. A first convolutional layer, a MobileNetV2 layer, a second convolutional layer, an average pooling layer, and the BP neural network are sequentially connected.

[0051] Furthermore, in one embodiment, the MobileNetV2 network structure is mainly composed of 17 inverse residual units.

[0052] Furthermore, in one embodiment, such as Figure 3 As shown, the BP neural network structure is a hierarchical neural network consisting of an input layer, hidden layers, and an output layer. The hidden layers can be expanded into multiple adjacent layers. Neurons in adjacent layers are fully connected, while neurons within each layer are not connected to each other.

[0053] Furthermore, in one embodiment, the BP neural network learns in a teacher-instructed manner. When a pair of learning patterns is provided to the BP neural network, each neuron receives the input response of the BP neural network and generates connection weights. Then, in the direction of reducing the error between the desired output and the actual output, each connection weight is modified layer by layer from the output layer through the hidden layers, returning to the input layer. This process is repeated alternately until the global error of the network tends to a given minimum value, thus completing the learning process and obtaining the connection weight parameters.

[0054] Furthermore, in one embodiment, training the improved BP neural network using a training set and a test set to obtain a trained improved BP neural network includes:

[0055] The training set is input into the improved BP neural network for training until the improved BP neural network converges, thus obtaining the initial improved BP neural network model.

[0056] The test set is input into the initial improved BP neural network model for testing. When the error function E corresponding to the test set reaches the threshold, the trained improved BP neural network model is obtained.

[0057] Furthermore, in one embodiment, the step of inputting the future carbon emissions of the power plant into an objective function with the goal of minimizing carbon emissions, and outputting a power plant start-up, shutdown, and output plan, includes:

[0058] Obtain the objective function, and the formula for calculating the objective function is as follows:

[0059]

[0060] Where, δ co2 Let be the objective function. This represents the carbon emissions corresponding to normal operation of the unit. This represents the carbon emissions corresponding to the unit's startup time. M represents the carbon emissions corresponding to the unit's shutdown, where M is the number of power plant units and T is the number of dispatch periods.

[0061] Obtain the predicted value λ of future carbon emissions from power plants co2 The constraint condition st. is constructed, and the formula for calculating st. is as follows:

[0062]

[0063] Under the premise of satisfying the constraint st., the objective function is solved with the goal of minimizing carbon emissions, and the power plant start-up, shutdown and output plans are obtained.

[0064] The second embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a power grid dispatching method based on a carbon emission prediction mechanism as described in any of the first embodiments.

[0065] A third embodiment of a computer device includes:

[0066] Memory is used to store instructions.

[0067] A processor is configured to execute the instructions, causing the computer device to perform operations of a power grid dispatching method based on a carbon emission prediction mechanism as described in any of the first embodiments.

[0068] Example 1:

[0069] like Figure 4 As shown, this invention uses monthly carbon emissions, unit output, and maintenance status data from Fujian thermal power plants over the past six years as the experimental dataset, totaling 12,960 data points. An example of the data is shown in Table 1.

[0070] Example of table data

[0071]

[0072] The collected data undergoes preprocessing. During preprocessing, the validity of the data is checked, and erroneous and empty data are handled. Erroneous data is verified and corrected, and empty data is replaced using the historical mean to correct invalid data.

[0073] The data is divided appropriately. To ensure sufficient training and effective testing of the deep learning model, this invention uses 80% of the dataset as the training set and 20% as the test set.

[0074] Training begins by dividing the data into multiple training samples, and the model network uses a sigmoid transfer function. Through backpropagation error function (t i For the desired output, O i (For the computational output of the network) Continuously adjust the network weights and thresholds to minimize the error function E.

[0075] Using a backpropagation (BP) neural network as the main prediction network has limitations. Due to the inherent sensitivity of BP networks to time series data, the input dimensionality is limited, failing to fully consider other important factors influencing the data results. To address this, the feature extraction layer of the BP network needs to be modified into a pre-sequence feature extraction network primarily based on convolutional operations. Currently, mainstream feature extraction networks include VGGNet, ResNet, and MobileNetV2. However, with the rapid development of neural network technology, VGGNet suffers from vanishing gradients as the network deepens, and its accuracy no longer meets ideal requirements. While ResNet effectively solves the vanishing and exploding gradient problems caused by network depth and extracts more representative features with significantly improved accuracy, it requires a large number of parameters and has a long training time. MobileNetV2, proposed by the Google team in 2018, offers higher accuracy than ResNet, a smaller model size, and solves the problem of depthwise separable convolutional kernels failing to learn features.

[0076] Therefore, this invention ultimately adopts a method of using MobileNetV2 as the preceding feature extraction network and BP network as the following prediction network for carbon emission prediction, which is called M2BPNet network.

[0077] Figure 2In the diagram, (a) represents input data with multiple dimensions such as time dimension, unit output capacity, unit maintenance level, coal consumption, and elemental carbon; (b) represents the MobileNetV2 network structure, which is mainly composed of 17 inverse residual units; and (c) represents the BP network structure, which is a hierarchical neural network composed of an input layer, a hidden layer, and an output layer.

[0078] The BackPropagation (BP) neural network is a hierarchical neural network consisting of an input layer, hidden layers, and an output layer, with the hidden layers potentially expanding to multiple layers. Neurons in adjacent layers are fully connected, while neurons within a single layer are not connected. The network learns by following a teacher-led instruction process. When a learning pattern is provided to the network, each neuron receives the network's input response and generates connection weights. These weights are then adjusted layer by layer from the output layer through each hidden layer, in the direction of reducing the error between the desired and actual output, before returning to the input layer. This process is repeated until the network's global error approaches a given minimum, thus completing the learning process.

[0079] Figure 3 In the middle, O i V represents the output unit. j Indicates hidden unit, δ k This represents the input unit. The connection weight from input unit k to hidden unit j is ω. jk The connection weight from hidden unit j to output unit i is ω. ij ω = (W, w) represents all the connection weights.

[0080] Training begins by dividing the data into multiple training samples, and the model network uses a sigmoid transfer function. Through backpropagation error function (t i For the desired output, O i (For the network's computational output) The network weights and thresholds are continuously adjusted to minimize the error function E. The training process is as follows: Figure 5 The software interface is shown.

[0081] If the training process yields convergence, the training ends; otherwise, the parameters of the BP neural network are optimized. The training parameters are shown in Table 2.

[0082] Table 2 Network Parameters

[0083]

[0084]

[0085] In this invention, 80% of the data is used as training data and 20% as test data; the maximum number of iterations is set to 300 training cycles; the initial learning rate is set to 0.05; the minimum performance gradient is 1e-6; and the maximum number of failed confirmations is 5. The model file is saved once after each training cycle. The input independent variables are the data shown in Table 1, and the output dependent variable is the unit's carbon emission capacity per kilowatt-hour. The experimental prediction results are shown in Table 3.

[0086] Table 3 Experimental Prediction Results

[0087]

[0088] Experimental results show that the highest prediction accuracy was 99.989%, the lowest was 86.71%, and the average prediction accuracy reached 95.53%, which meets the application requirements. The model has been successfully built and is ready for use.

[0089] By inputting the future carbon emissions of power plants into an objective function that optimizes for minimizing carbon emissions, a more accurate power plant start-up, shutdown, and output plan that meets carbon emission requirements is generated, further improving the accuracy of grid dispatch under the carbon emission mechanism.

[0090] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A power grid dispatching method based on a carbon emission prediction mechanism, characterized in that: Includes the following steps: Step 1: Obtain carbon emission data related to the power plant; Step 2: Preprocess the carbon emission data related to the power plant and use the preprocessed data as sample data; Step 3: Divide the sample data into training and test sets; Step 4: Train the improved BP neural network using the training set and the test set to obtain the trained improved BP neural network; Step 5: Use the trained, improved BP neural network to predict the power plant's future carbon emissions; Step 6: Input the future carbon emissions of the power plant into the objective function with the goal of minimizing carbon emissions, and output the power plant start-up and shutdown and output plan; The carbon emission data related to the power plant includes, but is not limited to: power plant units, raw coal conversion factor, data month, fossil combustion carbon emissions, coal consumption, desulfurizer carbonate content, limestone consumption, desulfurizer emission factor, power consumption rate, carbon emissions during the desulfurization process, elemental carbon, unit maintenance status, volatile matter, temperature, lower heating value, purchased electricity, carbon content per unit calorific value, carbon emission factor of purchased electricity, carbon oxidation rate, carbon emissions from purchased electricity, overall carbon emissions, and carbon emission capacity per kilowatt-hour. The preprocessing method includes: verifying and correcting erroneous data, and replacing empty data with historical averages; The improved BP neural network includes: MobileNetV2, a BP network; a first convolutional layer, MobileNetV2, the second convolutional layer, the average pooling layer, and the BP neural network are connected in sequence; The MobileNetV2 network structure consists of 17 inverse residual units. The BP neural network structure is a hierarchical neural network composed of an input layer, hidden layers, and an output layer. The hidden layers can be expanded into multiple adjacent layers. Neurons in adjacent layers are fully connected, while neurons in each layer are not connected. The BP neural network learns by being taught by a teacher. When a pair of learning patterns is provided to the BP neural network, each neuron receives the input response of the BP neural network and generates connection weights. Then, in the direction of reducing the error between the desired output and the actual output, the connection weights are modified layer by layer from the output layer through the hidden layers and back to the input layer. This process is repeated until the global error of the network tends to a given minimum value, thus completing the learning process and obtaining the connection weight parameters. The process of training the improved BP neural network using a training set and a test set to obtain a trained improved BP neural network includes: The training set is input into the improved BP neural network for training until the improved BP neural network converges, thus obtaining the initial improved BP neural network model. The test set is input into the initial improved BP neural network model for testing. When the error function E corresponding to the test set reaches the threshold, the trained improved BP neural network model is obtained. The process of inputting the future carbon emissions of the power plant into an objective function with the goal of minimizing carbon emissions, and outputting power plant start-up, shutdown, and output plans, includes: Obtain the objective function, and the formula for calculating the objective function is as follows: Where, δ co2 Let be the objective function. This represents the carbon emissions corresponding to normal operation of the unit. This represents the carbon emissions corresponding to the unit's startup time. M represents the carbon emissions corresponding to the unit's shutdown, where M is the number of power plant units and T is the number of dispatch periods. Obtain the predicted value λ of future carbon emissions from power plants co2 The constraint condition st. is constructed, and the formula for calculating st. is as follows: Under the premise of satisfying the constraint st., the objective function is solved with the goal of minimizing carbon emissions, and the power plant start-up, shutdown and output plans are obtained.

2. A computer-readable storage medium, characterized in that: It stores a computer program, which, when executed by a processor, implements a power grid dispatching method based on a carbon emission prediction mechanism as described in claim 1.

3. A computer device, characterized in that: include: Memory, used to store instructions; A processor is configured to execute the instructions, causing the computer device to perform the operation of a power grid dispatching method based on a carbon emission prediction mechanism as described in claim 1.

Images