Electricity and coal purchase quantity prediction method and system

Abstract

The invention belongs to the technical field of electricity and coal purchase quantity prediction, and discloses an electricity and coal purchase quantity prediction method and system. The method comprises the following steps: constructing an electricity and coal purchase quantity prediction model and a power plant operation optimization model based on a cloud data center; based on the data acquisition device, acquiring real-time electricity-coal purchasing associated data, performing data dimension reduction to obtain dimension-reduced real-time electricity-coal purchasing associated data, and sending the dimension-reduced real-time electricity-coal purchasing associated data to the power plant server; on the basis of the power plant server, according to the dimension-reduced real-time electricity-coal purchasing associated data, an electricity-coal purchasing quantity prediction model is used to predict the electricity-coal purchasing quantity; and based on the power plant server, generating a power plant operation optimization strategy by using the power plant operation optimization model according to the real-time electricity and coal purchase quantity prediction result. According to the method, the problems of insufficient prediction accuracy, poor model generalization ability, limited data processing ability and lack of dynamic adaptability in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power coal purchase quantity prediction, and in particular relates to a power coal purchase quantity prediction method and system. Background Art

[0002] With the rapid development of my country's economy, the demand for electricity is growing, and the demand for thermal coal in power plants is also increasing. Accurately predicting the purchase volume of thermal coal is of great significance for power plants to reasonably arrange production plans and reduce inventory costs.

[0003] The existing coal purchase volume forecasting technology has the following defects:

[0004] 1) Insufficient prediction accuracy: Existing prediction methods for power coal purchase volume mainly rely on historical data and statistical models, such as time series analysis and regression analysis. These methods often fail to fully capture the nonlinear relationship between power coal purchase volume and a variety of complex factors (such as market supply and demand, seasonal changes, power plant operation, etc.), resulting in low prediction accuracy.

[0005] 2) Poor model generalization ability: The traditional coal purchase volume prediction model may perform well on a specific data set, but when faced with new data environments or changes in application scenarios, the model's generalization ability is insufficient and it is difficult to maintain the accuracy of the prediction;

[0006] 3) Limited data processing capabilities: Existing technologies are often inefficient in processing high-dimensional data and large amounts of real-time data; data preprocessing steps may be too simple or insufficient to extract key features, resulting in poor prediction model performance;

[0007] 4) Lack of dynamic adaptability: Existing methods for predicting the purchase volume of power coal often lack the ability to adjust the operation of power plants in real time, and are unable to quickly optimize the operating status of purchasing power plants based on real-time changes in the purchase volume. Summary of the invention

[0008] In order to solve the problems of insufficient prediction accuracy, poor model generalization ability, limited data processing capability and lack of dynamic adaptability in the prior art, the present invention aims to provide a method and system for predicting the purchase volume of electric coal.

[0009] The technical solution adopted by the present invention is:

[0010] A method for predicting the purchase volume of electric coal comprises the following steps:

[0011] Based on the cloud data center, a coal purchase forecasting model and a power plant operation optimization model are built and deployed to all power plant servers connected to the cloud data center;

[0012] Based on the data collection device, the real-time coal purchase related data of the current power plant is collected, and the data dimension is reduced to obtain the real-time coal purchase related data after dimension reduction, and the data is sent to the power plant server;

[0013] Based on the power plant server, according to the real-time coal purchase related data after dimensionality reduction, the coal purchase volume prediction model is used to predict the coal purchase volume, and the real-time coal purchase volume prediction result is obtained;

[0014] Based on the power plant server, according to the real-time coal purchase forecast results, the power plant operation optimization model is used to generate the power plant operation optimization strategy, and the real-time power plant operation optimization strategy is obtained.

[0015] Furthermore, real-time power coal procurement related data include historical power coal procurement volume data, real-time power plant operation data, real-time climate and environmental data, real-time market supply and demand data, real-time alternative energy data, real-time power coal inventory data and real-time power plant transportation data.

[0016] Furthermore, based on the cloud data center, a coal purchase quantity prediction model and a power plant operation optimization model are constructed and deployed to all power plant servers connected to the cloud data center, including the following steps:

[0017] Based on the cloud data center, historical coal purchase related data of several historical power plants are collected and preprocessed to obtain several preprocessed historical coal purchase related data;

[0018] Performing data dimensionality reduction on a number of pre-processed historical coal purchase related data to obtain a principal component set and a number of dimensionality-reduced historical coal purchase related data;

[0019] Based on several historical coal purchase related data after dimensionality reduction, a coal purchase volume prediction model is constructed using a deep learning algorithm, and several historical coal purchase volume prediction results are generated;

[0020] Based on several historical coal purchase forecast results, a reinforcement learning algorithm is used to build a power plant operation optimization model and generate several historical power plant operation optimization experiences;

[0021] Extracting first model metadata of a power coal purchase quantity prediction model and second model metadata of a power plant operation optimization model;

[0022] sending the first model metadata and the second model metadata to all power plant servers connected to the cloud data center;

[0023] Based on the power plant server, the model is reconstructed according to the first model metadata to obtain a reconstructed power coal purchase quantity prediction model;

[0024] Based on the power plant server, the model is reconstructed according to the second model metadata to obtain a reconstructed power plant operation optimization model.

[0025] Furthermore, the PCA method is used to perform data dimensionality reduction on a number of preprocessed historical coal purchase related data to obtain a principal component set and a number of reduced-dimensional historical coal purchase related data.

[0026] Furthermore, the power coal procurement volume prediction model is constructed based on the RF-Attention-BiLSTM algorithm, and the power coal procurement volume prediction model includes a key feature extraction module constructed based on the RF algorithm, an attention weight module constructed based on the Attention mechanism, and a power coal procurement volume prediction module constructed based on the BiLSTM algorithm.

[0027] Furthermore, the power plant operation optimization model is constructed based on the DQN algorithm, and the power plant operation optimization model includes a deep Q network, an intelligent agent, and an experience replay pool.

[0028] Furthermore, based on the data collection device, the real-time coal purchase related data of the current power plant is collected, and the data dimension reduction is performed to obtain the real-time coal purchase related data after dimension reduction, and send it to the power plant server, including the following steps:

[0029] Based on the data collection device, the real-time coal purchase related data of the current power plant is collected and preprocessed to obtain the real-time coal purchase related data after preprocessing;

[0030] According to the principal component set, the pre-processed real-time coal purchase related data is subjected to data dimension reduction to obtain the reduced-dimensional real-time coal purchase related data;

[0031] The real-time coal procurement related data after dimension reduction is sent to the power plant server.

[0032] Further, based on the power plant server, according to the real-time coal purchase related data after dimensionality reduction, using the coal purchase quantity prediction model, the coal purchase quantity prediction is performed to obtain the real-time coal purchase quantity prediction result, including the following steps;

[0033] Based on the power plant server, the real-time coal purchase related data after dimension reduction is input into the coal purchase quantity prediction model;

[0034] According to the real-time coal purchase related data after dimensionality reduction, the key feature extraction module of the coal purchase quantity prediction model is used to extract M real-time key features;

[0035] According to the attention weights preset by the attention weight module of the electric coal purchase volume prediction model, feature splicing is performed on the M real-time key features to obtain real-time splicing features;

[0036] According to the real-time splicing features, the electric coal procurement volume prediction module of the electric coal procurement volume prediction model is used to predict the electric coal procurement volume and obtain the real-time electric coal procurement volume prediction result.

[0037] Furthermore, based on the power plant server, according to the real-time power coal purchase quantity prediction result, the power plant operation optimization model is used to generate the power plant operation optimization strategy, and the real-time power plant operation optimization strategy is obtained, including the following steps;

[0038] Based on the power plant server, the real-time coal purchase forecast results are input into the power plant operation optimization model;

[0039] According to the real-time prediction result of the power coal purchase volume, the state space of the power plant operation optimization model is updated to obtain an updated state space;

[0040] According to the real-time power coal purchase volume forecast result, the corresponding historical power plant operation optimization experience is retrieved from the experience playback pool of the power plant operation optimization model to obtain several matching historical power plant operation optimization experiences;

[0041] According to a number of matching historical power plant operation optimization experiences, the action space of the power plant operation optimization model is updated to obtain an updated action space;

[0042] Use the updated state space as the input of the deep Q network and use the agent to generate the Q value of each possible action in the updated action space;

[0043] Use the reward function to obtain the reward value of each possible action in the updated action space, and update the Q value of the possible action according to the reward value to obtain the updated Q value of each possible action;

[0044] Repeat the above Q-value update steps until the number of iterations reaches the iteration number threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action;

[0045] Integrate several execution actions to obtain real-time power plant operation optimization strategy.

[0046] A power coal purchase quantity prediction system is used to implement a power coal purchase quantity prediction method, the system includes a cloud data center, several power plants, several data acquisition devices and several power plant servers;

[0047] Each power plant is equipped with power generation equipment, control equipment and monitoring equipment, and the cloud data center is connected to several power plant servers in communication;

[0048] A plurality of power plant servers are arranged in a plurality of power plants in a one-to-one correspondence, each power plant server is communicatively connected to a control device of a corresponding power plant, and each power plant server includes a model reconstruction unit, a power coal purchase quantity prediction unit, and a power plant operation optimization unit which are connected in sequence;

[0049] A plurality of data acquisition devices are arranged in a plurality of power plants in a one-to-one correspondence, and each data acquisition device is communicatively connected with a power plant server, power generation equipment and monitoring equipment of a corresponding power plant.

[0050] The beneficial effects of the present invention are:

[0051] The invention discloses a method and system for predicting the purchase volume of electric coal. Through a prediction model of the purchase volume of electric coal constructed based on deep learning technology, the nonlinear relationship between the purchase volume of electric coal and various complex factors can be fully captured by utilizing its powerful feature extraction and pattern recognition capabilities, so as to achieve high-precision prediction. The prediction model of the purchase volume of electric coal based on deep learning technology has a data memory function. When facing a new data environment or changes in application scenarios, the generalization ability of the model is strong and the accuracy of prediction can be maintained. Dimensionality reduction processing is performed on real-time electric coal purchase related data to remove noise and redundancy in the data, so that the data is easier to understand and process, and the data dimension is effectively reduced, the calculation efficiency and prediction speed of the prediction model are improved, and high-quality, low-dimensional input data are provided for the prediction model of the purchase volume of electric coal, thereby enhancing the prediction ability of the model. Through a power plant operation optimization model constructed based on reinforcement learning technology, the power plant operation optimization strategy can be dynamically generated according to the prediction result of the purchase volume of electric coal, and the change of the purchase volume of electric coal and the operation status of the power plant can be realized in real time. It has strong adaptability and flexibility, can significantly improve the economic benefits of the power plant, reduce the operating cost, and improve the energy utilization rate and environmental protection level at the same time.

[0052] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flowchart of the method for predicting the power coal purchase quantity in the present invention.

[0054] Figure 2 It is a structural block diagram of the power coal purchase quantity prediction system in the present invention. DETAILED DESCRIPTION

[0055] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0056] Embodiment 1:

[0057] like Figure 1 As shown, this embodiment provides a method for predicting the purchase amount of electric coal, comprising the following steps:

[0058] S1: Based on the cloud data center, a coal purchase quantity prediction model and a power plant operation optimization model are constructed and deployed to all power plant servers connected to the cloud data center, including the following steps:

[0059] S1-1: Based on the cloud data center, collect historical coal purchase related data of several historical power plants, and pre-process them to obtain several pre-processed historical coal purchase related data;

[0060] The historical coal purchase-related data include historical coal purchase volume data, historical power plant operation data, historical climate and environmental data, historical market supply and demand data, historical alternative energy data, historical coal inventory data, and historical power plant transportation data;

[0061] The purchasing trend and cyclical changes in the historical power coal purchasing volume data affect the prediction of the next purchasing volume. The historical power plant operation data include historical steam pressure and temperature data, historical thermal pipeline efficiency data, historical plant power consumption rate, historical boiler thermal efficiency, historical turbine heat consumption rate and historical unit load, which are closely related to the power coal consumption of the power plant. The historical climate and environmental data are used to characterize seasonal demand changes. For example, the demand for electricity in summer and winter is different, which affects the purchase of power coal, and weather factors, such as extreme weather events, affect electricity demand and power coal purchase. The historical market supply and demand data include historical power coal market supply data, that is, the total amount of power coal available in the historical market, which affects the purchasing strategy, and historical power coal market demand data, that is, the demand for power coal in the historical market, which is directly related to the purchasing volume. The historical alternative energy data include alternative energy data such as natural gas, oil, renewable energy, etc., the purchase volume, market supply and demand changes may affect the demand for power coal. The power coal inventory level of the power plant in the historical power coal inventory data affects the purchasing strategy and purchase volume. The changes in transportation costs of historical power plant transportation data affect the purchasing decision, and transportation capacity restrictions may affect the purchase volume and timing of power coal.

[0062] Through the various indicator data involved in the power coal procurement related data, all factors related to the power coal procurement volume can be covered, improving the comprehensiveness and accuracy of the power coal procurement volume forecast;

[0063] For indicator data from different processing systems, preprocessing includes data format conversion, converting heterogeneous indicator data into a data format that can be recognized by the power plant server, and then cleaning and filtering the data to remove duplicate, erroneous and noisy data to provide data support for subsequent model training;

[0064] S1-2: Using the principal component analysis (PCA) method, a number of pre-processed historical coal purchase related data are subjected to data dimension reduction to obtain a principal component set and a number of reduced-dimensional historical coal purchase related data, including the following steps:

[0065] S1-2-1: Perform matrix conversion on some pre-processed historical coal purchase related data to obtain the corresponding historical coal purchase related data matrix X = [x 1 ,x 2 ,...x p ,...,x n ] T , where x p is the p-th preprocessed historical coal purchase related data row vector, p is the row vector indicator, and n is the total number of preprocessed historical coal purchase related data;

[0066] The initial row vector of the historical coal purchase related data matrix is ​​the preprocessed historical coal purchase related data, and the initial column vector of the historical coal purchase related data matrix is ​​the historical coal purchase related index data, coal purchase quantity index data, power plant operation index data, climate environment index data, market supply and demand index data, alternative energy index data, coal inventory index data and power plant transportation index data, each index data corresponds to the data value of each column in the historical coal purchase related data matrix;

[0067] S1-2-2: Standardize the historical coal purchase related data matrix to obtain the corresponding standardized historical coal purchase related data matrix;

[0068] The formula is:

[0069]

[0070] Where X' is the historical coal purchase related data matrix after standardization; μ is the mean of the historical coal purchase related data matrix; σ is the variance of the historical coal purchase related data matrix;

[0071] S1-2-3: Obtain the covariance matrix of the standardized historical coal purchase related data matrix, and obtain the corresponding candidate principal component matrix according to the standardized historical coal purchase related data matrix and the covariance matrix; the candidate row vectors of the candidate principal component matrix are the preprocessed historical coal purchase related data, and the candidate column vectors of the candidate principal component matrix are the historical coal purchase related index data;

[0072] The formula is:

[0073]

[0074] Where D is the covariance matrix of the historical coal purchase related data matrix after standardization; Y is the candidate principal component matrix; P is the conversion matrix; E is the unit eigenvector matrix; n is the total number of historical coal purchase related data after preprocessing;

[0075] Y=PX'

[0076] Where, Y is the candidate principal component matrix; P is the conversion matrix; X' is the historical coal purchase correlation data matrix after standardization;

[0077] S1-2-4: Set the candidate principal component matrix Y = [y 1 ,y 2 ,...y l ,...,y L ] Several candidate column vectors y' of the top 90% of the cumulative contribution rate of variance l As the corresponding principal component column vectors, the reduced dimension historical coal purchase correlation data matrix Y'=[y' 1 ,y' 2 ,...y' l ,...,y K ] and the principal component set;

[0078] The key row vector of the historical coal purchase related data matrix after dimensionality reduction is the historical coal purchase related data after dimensionality reduction, and the data index corresponding to the key column vector of the historical coal purchase related data matrix after dimensionality reduction is the principal component, and the principal component set is obtained;

[0079] The formula is:

[0080]

[0081] In the formula, G is the cumulative contribution rate of variance; l The lth candidate principal component y l The variance of ; l is the indicator of the candidate principal component; L is the total number of candidate principal components; K is the total number of principal components;

[0082] S1-2-5: Perform an inverse matrix transformation on the historical coal purchase related data matrix after dimension reduction to obtain a number of historical coal purchase related data after dimension reduction;

[0083] The PCA method transforms the original data into a set of linearly independent representations in each dimension through linear transformation, which is used to remove noise and redundancy in the data, making the data easier to understand and process;

[0084] S1-3: Based on several dimensionally reduced historical coal purchase related data, a coal purchase volume prediction model is constructed using a deep learning algorithm, and several historical coal purchase volume prediction results are generated;

[0085] The electric coal purchase volume prediction model is built based on the Random Forest (RF)-Attention-Bidirectional Long Short-Term Memory (BiLSTM) algorithm, and the electric coal purchase volume prediction model includes a key feature extraction module built based on the RF algorithm, an attention weight module built based on the Attention mechanism, and an electric coal purchase volume prediction module built based on the BiLSTM algorithm;

[0086] The key feature extraction module uses the internal Classification And Regression Tree (CART) to screen the key features of the input data and extract several key features related to the data prediction label; the attention weight generation module uses the preset attention weights to realize the splicing of key features and integrate the scattered features; enhance the influence of key features on label prediction, and the coal purchase volume prediction module uses a multi-layer hidden layer structure to explore the potential connection between key features and data prediction labels, train its own network structure through a large number of key features, predict new data features, and realize fast, efficient and accurate prediction of data labels;

[0087] S1-4: Based on several historical coal purchase volume forecast results, a power plant operation optimization model is constructed using the Deep Q Network (DQN) algorithm, and several historical power plant operation optimization experiences are generated, including the following steps:

[0088] A-1: Generate the power plant operation optimization problem as the simulation environment of the DQN algorithm, and build the intelligent agent and experience replay pool;

[0089] A-2: Based on the historical coal purchase volume forecast results for each power plant operating state type, define the state space of the DQN algorithm, and the parameters of the state space correspond to each power plant operating state;

[0090] A-3: Define the action space of the DQN algorithm according to the actions that need to be output for power plant operation optimization; the action space includes various actions to control power generation equipment, such as expanding or contracting unit load, increasing turbine power, increasing thermal system drain equipment power, etc.

[0091] A-4: Based on the possible impact of each action in the action space, define the reward function of the DQN algorithm to evaluate the quality or impact of the action;

[0092] A-5: Construct the input layer, several hidden layers, and output layer of the deep Q network, connect the input layer to the state space, and connect the output layer to the action space;

[0093] A-6: Based on the state space, action space and reward function, according to several historical coal purchase volume forecast results, the deep Q network and intelligent agent are optimized and trained to build a power plant operation optimization model, and several historical power plant operation optimization experiences are generated;

[0094] A-7: storing several historical power plant operation optimization experiences into the experience playback pool of the power plant operation optimization model;

[0095] The power plant operation optimization model is built based on the DQN algorithm, and the power plant operation optimization model includes a deep Q network, an intelligent agent, and an experience replay pool;

[0096] S1-5: extracting first model metadata of the power coal purchase quantity prediction model and second model metadata of the power plant operation optimization model;

[0097] S1-6: Sending the first model metadata and the second model metadata to all power plant servers connected to the cloud data center;

[0098] S1-7: Based on the power plant server, reconstruct the model according to the first model metadata to obtain a reconstructed power coal procurement quantity prediction model;

[0099] S1-8: Based on the power plant server, reconstruct the model according to the second model metadata to obtain a reconstructed power plant operation optimization model;

[0100] S2: Based on the data collection device, the real-time coal purchase related data of the current power plant is collected, and the data dimension is reduced to obtain the real-time coal purchase related data after dimension reduction, and the data is sent to the power plant server, including the following steps:

[0101] S2-1: Based on the data collection device, the real-time coal purchase related data of the current power plant is collected and preprocessed to obtain the preprocessed real-time coal purchase related data;

[0102] The data acquisition device extracts real-time coal inventory data, real-time power plant transportation data, and historical coal purchase data of the previous forecast period from the database, extracts real-time power plant operation data from the monitoring equipment of the power plant's power generation equipment, extracts real-time climate and environment data from the monitoring equipment of the power plant's climate and environment, and downloads real-time market supply and demand data and real-time alternative energy data from the Internet, and performs data format conversion, data denoising and normalization processing on the above-mentioned real-time coal purchase related indicator data to improve data quality;

[0103] Real-time coal purchase-related data include historical coal purchase volume data, real-time power plant operation data, real-time climate and environmental data, real-time market supply and demand data, real-time alternative energy data, real-time coal inventory data, and real-time power plant transportation data;

[0104] S2-2: According to the principal component set, the pre-processed real-time coal purchase related data is subjected to data dimension reduction to obtain the reduced-dimensional real-time coal purchase related data;

[0105] After preprocessing, the data dimension of real-time coal purchase correlation data is very high. Using PCA method for preliminary dimensionality reduction helps to reduce the amount of calculation, remove noise and redundant features, and improve the performance of RF module of coal purchase volume prediction model;

[0106] S2-3: Send the real-time coal purchase related data after dimension reduction to the power plant server;

[0107] S3: Based on the power plant server, according to the real-time coal purchase related data after dimensionality reduction, using the coal purchase quantity prediction model, the coal purchase quantity is predicted, and the real-time coal purchase quantity prediction result is obtained, including the following steps;

[0108] S3-1: Based on the power plant server, the real-time coal purchase related data after dimension reduction is input into the coal purchase quantity prediction model;

[0109] S3-2: Based on the real-time coal purchase related data after dimensionality reduction, use the key feature extraction module of the coal purchase volume prediction model to extract M real-time key features, including the following steps:

[0110] S3-2-1: Use the RF structure trained in the key feature extraction module to extract the feature contribution of M candidate features in the real-time coal procurement related data after dimensionality reduction, where M is the total number of key features;

[0111] The formula is:

[0112]

[0113] In the formula, is the feature contribution of the j'th candidate feature; is the feature contribution of the j'th candidate feature in the i'th tree of the random forest; i' is the CART tree indicator; j' is the candidate feature indicator; n' is the total number of CARTs;

[0114]

[0115] In the formula, GI m GI l' GI r" is the Gini index of the CART tree node m, node l' and node r" of the random forest; pmk" is the proportion of category k" in CART tree node m; K' is the total number of categories; m, l', r" are node indicators; k" is the category indicator;

[0116] S3-2-2: normalizing the feature contributions of several candidate features to obtain corresponding normalized feature contributions;

[0117] The formula is:

[0118]

[0119] In the formula, VIM j' is the contribution of the feature after normalization; J is the total number of candidate features;

[0120] S3-2-3: Generate feature selection standard values ​​for several candidate features based on the normalized feature contribution;

[0121] The formula is:

[0122]

[0123] In the formula, CFC j' The feature selection standard value for the j'th candidate feature; is the normalized feature contribution of the j*th candidate feature; j* is the indicator of the candidate feature;

[0124] S3-2-4: According to the feature selection standard value, the candidate features are sorted in descending order, and the first M candidate features are selected as real-time key features to obtain M real-time key features;

[0125] S3-3: according to the attention weight preset by the attention weight module of the electric coal purchase volume prediction model, feature splicing is performed on the M real-time key features to obtain a real-time splicing feature;

[0126] S3-4: According to the real-time splicing features, the electric coal purchase volume prediction module of the electric coal purchase volume prediction model is used to predict the electric coal purchase volume, and a real-time electric coal purchase volume prediction result is obtained;

[0127] S4: Based on the power plant server, according to the real-time coal purchase quantity forecast result, using the power plant operation optimization model, the power plant operation optimization strategy is generated to obtain the real-time power plant operation optimization strategy, including the following steps;

[0128] S4-1: Based on the power plant server, the real-time coal purchase quantity forecast result is input into the power plant operation optimization model;

[0129] S4-2: Based on the real-time coal purchase forecast results, the state space of the power plant operation optimization model is updated to obtain the updated state space S' = [s' 1 ,...,s' i" ,...,s' I' ], where s' i" is the updated i'th state value, i' is the state indicator, and I' is the total number of state space dimensions;

[0130] S4-3: According to the real-time power coal purchase volume forecast result, the corresponding historical power plant operation optimization experience is retrieved from the experience playback pool of the power plant operation optimization model to obtain several matching historical power plant operation optimization experiences;

[0131] S4-4: Based on some matching historical power plant operation optimization experiences, the action space of the power plant operation optimization model is updated to obtain the updated action space A'=[a' 1 ,...,a' j" ,...,a' I ], where a' j" is the updated j-th "action value", j" is the action indicator, and I is the total number of action space dimensions;

[0132] S4-5: Use the updated state space as the input of the deep Q network and use the agent to generate the Q value of each possible action in the updated action space;

[0133] S4-6: Use the reward function to obtain the reward value of each possible action in the updated action space, and update the Q value of the possible action according to the reward value to obtain the updated Q value of each possible action;

[0134] The formula is:

[0135] Q(s' p' ,a' p' )=(1-α")·Q(s p' ,a p' )+α"·(R(s p' ,a p' ,s' p' )+γ·Q max (s p' ,a p' ))

[0136] In the formula, Q(s' p' ,a' p' ) is the updated state value s' p' and the updated action value a' p' The corresponding updated Q value; Q(s p' ,a p' ) is the state value sp' and action value a p' The Q value of the corresponding possible action; α" is the learning rate; Q max (s p' ,a p' ) is the highest possible Q value of the action; p' is the comprehensive indicator; γ is the update parameter;

[0137] S4-7: Repeat the above Q-value updating steps until the number of iterations reaches the iteration number threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action;

[0138] S4-8: Integrate several execution actions to obtain a real-time power plant operation optimization strategy, and send it to the control equipment of the power plant.

[0139] Embodiment 2:

[0140] like Figure 2 As shown, this embodiment provides a power coal purchase quantity prediction system for implementing a power coal purchase quantity prediction method, the system includes a cloud data center, several power plants, several data acquisition devices and several power plant servers;

[0141] Each power plant is equipped with power generation equipment, control equipment and monitoring equipment, and the cloud data center is connected to several power plant servers in communication;

[0142] Power generation equipment, used to perform power generation tasks in power plants;

[0143] Control equipment, used to execute real-time power plant operation optimization strategies and control corresponding power generation equipment, including industrial computers or main control servers of steam turbines, thermal systems, etc.;

[0144] Monitoring equipment, used to collect real-time power plant operation data and real-time climate and environmental data of power generation equipment;

[0145] A plurality of power plant servers are arranged in a plurality of power plants in a one-to-one correspondence, each power plant server is communicatively connected to a control device of a corresponding power plant, and each power plant server includes a model reconstruction unit, a power coal purchase quantity prediction unit, and a power plant operation optimization unit which are connected in sequence;

[0146] A model reconstruction unit is used to reconstruct the model according to the first model metadata to obtain a reconstructed coal purchase quantity prediction model; and to reconstruct the model according to the second model metadata to obtain a reconstructed power plant operation optimization model;

[0147] The electric coal purchase volume prediction unit is used to predict the electric coal purchase volume based on the real-time electric coal purchase related data after dimensionality reduction and use the electric coal purchase volume prediction model to obtain the real-time electric coal purchase volume prediction result;

[0148] The power plant operation optimization unit is used to generate the power plant operation optimization strategy based on the real-time coal purchase quantity forecast result and the power plant operation optimization model to obtain the real-time power plant operation optimization strategy;

[0149] A plurality of data acquisition devices are arranged in a plurality of power plants in a one-to-one correspondence, and each data acquisition device is communicatively connected with a power plant server, power generation equipment and monitoring equipment of a corresponding power plant;

[0150] The data collection device is used to collect the real-time coal purchase related data of the current power plant, perform data dimension reduction, obtain the real-time coal purchase related data after dimension reduction, and send it to the power plant server.

[0151] The invention discloses a method and system for predicting the purchase volume of electric coal. Through a prediction model of the purchase volume of electric coal constructed based on deep learning technology, the nonlinear relationship between the purchase volume of electric coal and various complex factors can be fully captured by utilizing its powerful feature extraction and pattern recognition capabilities, so as to achieve high-precision prediction. The prediction model of the purchase volume of electric coal based on deep learning technology has a data memory function. When facing a new data environment or changes in application scenarios, the generalization ability of the model is strong and the accuracy of prediction can be maintained. Dimensionality reduction processing is performed on real-time electric coal purchase related data to remove noise and redundancy in the data, so that the data is easier to understand and process, and the data dimension is effectively reduced, the calculation efficiency and prediction speed of the prediction model are improved, and high-quality, low-dimensional input data are provided for the prediction model of the purchase volume of electric coal, thereby enhancing the prediction ability of the model. Through a power plant operation optimization model constructed based on reinforcement learning technology, the power plant operation optimization strategy can be dynamically generated according to the prediction result of the purchase volume of electric coal, and the change of the purchase volume of electric coal and the operation status of the power plant can be realized in real time. It has strong adaptability and flexibility, can significantly improve the economic benefits of the power plant, reduce the operating cost, and improve the energy utilization rate and environmental protection level at the same time.

[0152] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.

Claims

1. A method for predicting the purchase volume of electric coal, characterized in that: The steps include: Based on the cloud data center, a coal purchase forecasting model and a power plant operation optimization model are built and deployed to all power plant servers connected to the cloud data center; Based on the data collection device, the real-time coal purchase related data of the current power plant is collected, and the data dimension is reduced to obtain the real-time coal purchase related data after dimension reduction, and the data is sent to the power plant server; Based on the power plant server, according to the real-time coal purchase related data after dimensionality reduction, the coal purchase volume prediction model is used to predict the coal purchase volume, and the real-time coal purchase volume prediction result is obtained; Based on the power plant server, according to the real-time coal purchase forecast results, the power plant operation optimization model is used to generate the power plant operation optimization strategy, and the real-time power plant operation optimization strategy is obtained.

2. A method for predicting the purchase volume of electric coal according to claim 1, characterized in that: The real-time coal purchase-related data include historical coal purchase volume data, real-time power plant operation data, real-time climate and environment data, real-time market supply and demand data, real-time alternative energy data, real-time coal inventory data and real-time power plant transportation data.

3. A method for predicting the purchase volume of electric coal according to claim 1, characterized in that: Based on the cloud data center, a coal purchase quantity prediction model and a power plant operation optimization model are constructed and deployed to all power plant servers connected to the cloud data center, including the following steps: Based on the cloud data center, historical coal purchase related data of several historical power plants are collected and preprocessed to obtain several preprocessed historical coal purchase related data; Performing data dimensionality reduction on a number of pre-processed historical coal purchase related data to obtain a principal component set and a number of dimensionality-reduced historical coal purchase related data; Based on several historical coal purchase related data after dimensionality reduction, a coal purchase volume prediction model is constructed using a deep learning algorithm, and several historical coal purchase volume prediction results are generated; Based on several historical coal purchase forecast results, a reinforcement learning algorithm is used to build a power plant operation optimization model and generate several historical power plant operation optimization experiences; Extracting first model metadata of a power coal purchase quantity prediction model and second model metadata of a power plant operation optimization model; sending the first model metadata and the second model metadata to all power plant servers connected to the cloud data center; Based on the power plant server, the model is reconstructed according to the first model metadata to obtain a reconstructed power coal purchase quantity prediction model; Based on the power plant server, the model is reconstructed according to the second model metadata to obtain a reconstructed power plant operation optimization model.

4. A method for predicting the purchase volume of electric coal according to claim 3, characterized in that: The PCA method is used to reduce the dimension of several preprocessed historical coal purchase related data to obtain the principal component set and several reduced dimension historical coal purchase related data.

5. A method for predicting the purchase volume of electric coal according to claim 3, characterized in that: The electric coal procurement volume prediction model is constructed based on the RF-Attention-BiLSTM algorithm, and the electric coal procurement volume prediction model includes a key feature extraction module constructed based on the RF algorithm, an attention weight module constructed based on the Attention mechanism, and an electric coal procurement volume prediction module constructed based on the BiLSTM algorithm.

6. A method for predicting the purchase volume of electric coal according to claim 3, characterized in that: The power plant operation optimization model is constructed based on the DQN algorithm, and the power plant operation optimization model includes a deep Q network, an intelligent agent and an experience replay pool.

7. A method for predicting the purchase volume of electric coal according to claim 4, characterized in that: Based on the data collection device, the real-time coal purchase related data of the current power plant is collected, and the data dimension is reduced to obtain the real-time coal purchase related data after dimension reduction, and the data is sent to the power plant server, including the following steps: Based on the data collection device, the real-time coal purchase related data of the current power plant is collected and preprocessed to obtain the real-time coal purchase related data after preprocessing; According to the principal component set, the pre-processed real-time coal purchase related data is subjected to data dimension reduction to obtain the reduced-dimensional real-time coal purchase related data; The real-time coal procurement related data after dimension reduction is sent to the power plant server.

8. A method for predicting the purchase volume of electric coal according to claim 5, characterized in that: Based on the power plant server, according to the real-time coal purchase related data after dimensionality reduction, the coal purchase quantity prediction model is used to predict the coal purchase quantity, and the real-time coal purchase quantity prediction result is obtained, including the following steps; Based on the power plant server, the real-time coal purchase related data after dimension reduction is input into the coal purchase quantity prediction model; According to the real-time coal purchase related data after dimensionality reduction, the key feature extraction module of the coal purchase quantity prediction model is used to extract M real-time key features; According to the attention weights preset by the attention weight module of the electric coal purchase volume prediction model, feature splicing is performed on the M real-time key features to obtain real-time splicing features; According to the real-time splicing features, the electric coal procurement volume prediction module of the electric coal procurement volume prediction model is used to predict the electric coal procurement volume and obtain the real-time electric coal procurement volume prediction result.

9. A method for predicting the purchase volume of electric coal according to claim 6, characterized in that: Based on the power plant server, according to the real-time coal purchase quantity forecast result, the power plant operation optimization model is used to generate the power plant operation optimization strategy, and the real-time power plant operation optimization strategy is obtained, including the following steps; Based on the power plant server, the real-time coal purchase forecast results are input into the power plant operation optimization model; According to the real-time prediction result of the power coal purchase volume, the state space of the power plant operation optimization model is updated to obtain an updated state space; According to the real-time power coal purchase volume forecast result, the corresponding historical power plant operation optimization experience is retrieved from the experience playback pool of the power plant operation optimization model to obtain several matching historical power plant operation optimization experiences; According to a number of matching historical power plant operation optimization experiences, the action space of the power plant operation optimization model is updated to obtain an updated action space; Use the updated state space as the input of the deep Q network and use the agent to generate the Q value of each possible action in the updated action space; Use the reward function to obtain the reward value of each possible action in the updated action space, and update the Q value of the possible action according to the reward value to obtain the updated Q value of each possible action; Repeat the above Q-value update steps until the number of iterations reaches the iteration number threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action; Integrate several execution actions to obtain real-time power plant operation optimization strategy.

10. A power coal purchase volume forecasting system, used to implement the power coal purchase volume forecasting method according to any one of claims 1 to 9, characterized in that: The system includes a cloud data center, several power plants, several data acquisition devices, and several power plant servers; Each of the power plants is provided with power generation equipment, control equipment and monitoring equipment, and the cloud data center is respectively connected to the servers of several power plants in communication; The plurality of power plant servers are arranged in a plurality of power plants in a one-to-one correspondence, each power plant server is communicatively connected to a control device of a corresponding power plant, and each power plant server comprises a model reconstruction unit, a power coal purchase quantity prediction unit, and a power plant operation optimization unit which are connected in sequence; The data acquisition devices are arranged in a corresponding manner in a plurality of power plants, and each data acquisition device is communicatively connected with a power plant server, power generation equipment and monitoring equipment of the corresponding power plant.

Images