Intelligent operation energy accounting method for coal-fired power plant

Abstract

The invention discloses an intelligent operation energy accounting method for a coal-fired power plant. Comprising the following steps: collecting data information in real time through a sensor network; the collected data information is preprocessed; converting the preprocessed data information into a data set; establishing a data visualization dynamic graph; analysis and processing are performed through an artificial intelligence model, so that the load demand, coal consumption and equipment maintenance demand of the power plant can be predicted conveniently, and resource allocation and operation plans are optimized; parameters are collected through the sensor network, data information is preprocessed, the accuracy of the data information is improved, and noise interference is eliminated; a data visual dynamic graph is established according to the data set, so that visual monitoring can be realized conveniently; analysis processing is carried out through the artificial intelligence model, the load demand, coal consumption and equipment maintenance demand of the power plant can be predicted conveniently, resource allocation and operation plan processing are optimized through particle swarm optimization, and the resource utilization rate and production efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of energy accounting, and more specifically, it relates to a coal-fired power plant intelligent operation energy accounting method. BACKGROUND

[0002] Thermal power plant, also known as thermal power plant, is a plant that uses combustible materials (such as coal) as fuel to produce electric energy. Its basic production process is: when the fuel is burned, it heats water to generate steam, converts the chemical energy of the fuel into heat energy, the steam pressure drives the steam turbine to rotate, and the heat energy is converted into mechanical energy, then the steam turbine drives the generator to rotate, and the mechanical energy is converted into electric energy.

[0003] The prime mover is usually a steam engine or a gas turbine, and in some smaller power stations, it may also use an internal combustion engine. They all generate electricity by using the pressure drop in the process of converting high-temperature, high-pressure steam or gas into low-pressure air or condensed water through a turbine.

[0004] Coal-fired power plants play an important role in global energy production, but traditional coal-fired power plants face some challenges in operation: data recording and monitoring: traditional coal-fired power plants mainly rely on basic sensors and manual recording systems to monitor equipment operating conditions. The data collection frequency is low, usually in the form of manual recording and periodic summary. This method not only takes time, but is also susceptible to human factors, resulting in low data accuracy and timeliness. Efficiency analysis: In traditional power plants, efficiency analysis often relies on periodic checks and maintenance. The operating efficiency of the equipment is obtained through periodic manual calculation and historical data analysis. This method often cannot reflect the real-time state of the equipment, and it is difficult to adjust the operation strategy in time. Optimization difficulty: Due to the lack of real-time data support, the optimization work of traditional power plants mainly relies on the experience and subjective judgment of operators. Optimization adjustment usually needs a long time to verify, which makes the energy management process more complex and inefficient.

[0005] The introduction of intelligent technology has brought revolutionary changes to coal-fired power plants, making their operations more efficient, precise, and intelligent. The main advantages and applications of intelligent technology are as follows: Real-time data monitoring and collection: By introducing high-precision sensors and advanced data collection systems, power plants can monitor the operating status of key equipment such as boilers and generators in real time. Real-time data acquisition enables the immediate recording and analysis of equipment operating conditions and energy usage, improving data accuracy and timeliness. Data analysis and intelligent algorithms: Using modern data processing platforms and machine learning algorithms, power plants can conduct in-depth analysis of massive data. These algorithms can identify patterns and trends in operation, predict equipment failures, and optimize energy use. For example, regression analysis can help predict energy demand changes, classification models can identify potential equipment problems, and time series prediction can provide suggestions for future operations. Automated control systems: Intelligent technology introduces automated control systems that can automatically adjust equipment operating parameters based on real-time data and analysis results. These systems can respond to data changes in real time and automatically adjust coal consumption, boiler temperature, and other parameters to optimize equipment performance and energy use efficiency. Intelligent optimization and energy saving: Intelligent technology enables power plants to identify and implement optimization measures in real time. For example, the system can automatically adjust operations when it detects a decrease in equipment efficiency, reducing energy waste and improving energy use efficiency. This closed-loop feedback mechanism ensures that power plants can continuously optimize operations, reduce operating costs, and improve economic efficiency.

[0006] The introduction of intelligent technology not only improves the operational efficiency and precision of coal-fired power plants, but also lays the foundation for future energy management. With continuous technological progress, coal-fired power plants will be able to further realize intelligent and digitalized operations, thereby meeting global energy demand while reducing environmental impact and achieving sustainable development. The application of intelligent technology will become the core of future coal-fired power plant management, driving the energy industry towards a more efficient and intelligent future.

[0007] However, there are some problems in the existing technology: The existing coal-fired power plants still have difficulties in collecting data, and the introduction of intelligent technology involves a large amount of data that cannot be effectively fused and processed. Moreover, the results are only given after calculation, and the analysis of data cannot be visualized, and optimization adjustment is not convenient. Therefore, we propose a method for intelligent management of energy accounting in coal-fired power plants. SUMMARY

[0008] In view of the problems existing in the prior art, the purpose of the present application is to provide a coal-fired power plant intelligent management energy accounting method, which collects various data information of the power plant, realizes data processing, and gives an optimization scheme after energy accounting, so as to realize visual display of data information, improve fusion and specification of a large amount of data, and facilitate analysis of energy accounting analysis and optimization solution.

[0009] To achieve the above purpose, the present application provides the following technical scheme: a coal-fired power plant intelligent management energy accounting method, comprising the following steps:

[0010] Real-time tracking of coal consumption data X m , heat data X r and power generation data X d , and collecting historical marketing data X x and historical monthly expenditure costs X z in a large database, and collecting coal purchase unit price X d , electricity sales unit price X q heat sales unit price X v ;

[0011] Pretreatment of collected coal consumption data X m , heat data X r , power generation data X d , historical marketing data X x and historical monthly expenditure costs X z , eliminating noise in data information, completing data information cleaning, and improving data information accuracy;

[0012] Fusion processing of pretreated data information, unifying all data information into a unified format, and respectively establishing coal consumption data information set [X m ], heat information set [X r ], power generation data information set [X d ], historical marketing data set [X x ] and historical monthly expenditure data set [X z ];

[0013] Respectively establishing data visualization dynamic diagrams for coal consumption data information set [X m ], heat information set [X r ], power generation data information set [X d ], emission data information set [X p ], historical marketing data set [X x ] and historical monthly expenditure data set [X z ];

[0014] The coal consumption data information set [X m ], the heat information set [X r ], the power generation data information set [X d ], the emission data information set [X p ], the historical marketing data set [X x ] and the historical monthly expenditure data set [X z ] are stored and analyzed and processed by an artificial intelligence model to optimize resource allocation and operation plans.

[0015] Preferably, the preprocessing processing steps are as follows:

[0016] The data X m , the heat data X r and the power generation data X d of the coal consumption collected by the sensor network, and the historical marketing data X x and the historical monthly expenditure data X z in the large database are received.

[0017] The received X m , X r , X d , X x and X z are denoised, calculated and processed by a signal processing method, converted to a frequency domain by Fourier transform, high-frequency noise is removed by filtering, the data is converted back to a time domain, and the data is reconstructed by wavelet transform: the data is wavelet transformed, the noise coefficients are removed, and the data is reconstructed.

[0018] The denoised data information is then cleaned, the abnormal data in the data information is obtained, the abnormal data information is removed, the removed positions are filled, and the integrity of the data information is maintained.

[0019] The data information is standardized to ensure consistent data format and unit for subsequent analysis.

[0020] Preferably, the Fourier transform is used to respectively transform the collected coal consumption data X m , the heat data X r and the power generation data X d , and the historical marketing data X x and the historical monthly expenditure data X z in the large database, and the calculation formula of the transformation is as follows:

[0021]

[0022] Where x(t) is the time-domain signal, representing the signal to be analyzed, w(t) is the window function, used to locally weight the signal, t is the time variable, the center position of the window, f is the frequency variable, representing the frequency-domain information, X(t, f) is the result of STFT, representing the amplitude and phase information of the signal at time t and frequency f, τ is the integral variable, representing the position in time, used to calculate the spectral information of the signal at a specific time point t, τ is used to describe the overlapping part of the window function and the signal, helping to determine the frequency characteristics of the signal at time t, and ξ is the introduced window adaptive adjustment factor, where the calculation formula of ξ is as follows:

[0023] Where rect is the rectangular window function, and Δt(τ) is the window width dynamically adjusted according to the instantaneous frequency or characteristics of the signal.

[0024] Preferably, the wavelet transform is used to perform a second transformation on the Fourier-transformed data information, and the calculation formula is as follows:

[0025] Where ψ a,b (t) is the mother wavelet, a is the scale parameter, controlling the frequency band of the analysis, and b is the translation parameter, controlling the position of the analysis.

[0026] The wavelet-transformed data is then decomposed:

[0027] And each DWT i represents the wavelet coefficient at different scales.

[0028] The data information is then denoised:

[0029] Where c i is the coefficient of the i-th layer, and λ is the threshold parameter.

[0030] Finally, the processed data information is reconstructed:

[0031] Where IDWT represents the inverse wavelet transform process.

[0032] Preferably, the standardization of the data information is as follows:

[0033] First, the denoised data information is processed by weighted mean:

[0034] Where w i is the weight of the data point x i ,

[0035] Then, the median is introduced in the weighted standardization process, and the calculation formula is as follows:

[0036] Wherein, median(X) is the median of the data information X, and MAD(X) is the absolute deviation of the data.

[0037] Preferably, the data visualization dynamic diagram is established as follows:

[0038] The data information of the processing process is stored, and the data information is fused and processed to unify the format of all the data information, so that the data format is suitable for the visualization requirements.

[0039] The polyline visualization diagram is designed to enable the processed data information to realize dynamic display effect, facilitate observation of the update state of the data information, and realize the trend of the data information.

[0040] The selected tool is used to write code or configure the chart, and the dynamic update logic is added.

[0041] The effect of the dynamic diagram is checked, and the parameters and design are adjusted to ensure clarity and accuracy.

[0042] Preferably, the data X of coal consumption m , the data X of heat r , and the data X of power generation d The output of coal is calculated, and the ratio of coal combustion is optimized:

[0043] X r +X d =X m ·μ·ζ·ψ,

[0044] Wherein, μ, ζ and ψ are oxygen supply ratio coefficient, coal blending ratio coefficient and combustion output ratio coefficient, respectively.

[0045] The output value is calculated by unit price:

[0046] (X r ·X v +X d ·X q )-X j =X m ·X d ·μ·ζ·ψ+X f ,

[0047] Wherein, X f is the other consumption amount generated in the coal-fired power generation process, and X j is the profit amount of the calculated coal output.

[0048] Preferably, the artificial intelligence model adopts a feedforward neural network, and each layer of neurons is fully connected with the next layer of neurons.

[0049] Each layer of neurons is fully connected to the next layer of neurons using linear connections, and the connection formula is as follows:

[0050] Wherein, x i is the input data information, mainly including historical marketing data set [X x ] and historical monthly expenditure data set [X z ], w ij is the weight from the i-th input to the j-th neuron, b j is the bias;

[0051] The sales amount and the expenditure amount of the month are predicted and calculated by the feedforward neural network to obtain the predicted profit amount of the month;

[0052] The predicted profit amount of the month is calculated and processed with the profit amount X j of the coal production to obtain the difference between the prediction and the actual value;

[0053] D = |z j -X j |, wherein D is the difference between the prediction and the actual value, and is the sum of the specific profit and the monthly expenditure;

[0054] And the calculated data is displayed through a visualization tool.

[0055] Preferably, the feedforward neural network uses a Sigmoid function as an activation function, and the calculation formula is as follows:

[0056] Wherein, z is the weighted sum of the neuron input, and the calculation formula is:

[0057]

[0058] Wherein: w i is the weight, which determines the influence degree of each input x i on the output, and b is the bias, which adjusts the output range of the function;

[0059] And the calculation result of the feedforward neural network is evaluated by a loss function to evaluate the difference between the network prediction result and the true value:

[0060] Wherein, y k is the true value, is the predicted value, and n is the sample number;

[0061] The weight and the bias are updated by gradient descent method or its variants;

[0062] Gradient descent method:

[0063]

[0064] where η is the learning rate, is the gradient of the loss function with respect to the weights, is the gradient of the loss function with respect to the biases.

[0065] Preferably, the optimization of resource allocation and operation plan is allocated by adjusting the particle swarm optimization, and the particle swarm optimization combines the difference value D to optimize the resource allocation and operation plan processing;

[0066] And in order to improve the efficiency of the algorithm, the hierarchical local search is carried out according to the neural cloud of the feedforward neural network, and the particle swarm is divided into multiple levels, each level uses different parameter settings for optimization, and then the results are integrated:

[0067] v i (t+1) = w k ·v i (t) + c 1,k ·r1(pBest i -x i (t)) + c 2,k ·r2(gBest i -x i (t)) + LocalSearch(x i (t)),

[0068] where v i (t) is the speed of particle i at time t, x i (t) is the position of particle i at time t, w is the inertia weight, which controls the inertia of the particle, c1 and c2 are learning factors, which control the learning degree of the particle to the individual best position and the group best position respectively, r1 and r2 are random numbers between [0, 1], and k represents the level, LocalSearch(x i (t)) is a local search algorithm, pBest i is the best position found by particle i so far, gBest i is the best position found by the entire particle swarm so far;

[0069] In order to realize the dynamic adjustment of the inertia weight and the dynamic adjustment of the learning factor for the particle swarm optimization:

[0070]

[0071] where w max is the maximum value of the inertia weight, w minis the minimum value of the inertia weight, T is the maximum number of iterations, and t is the current iteration number;

[0072]

[0073] wherein c 1,max and c 1,min are the maximum and minimum values of the individual learning factor, respectively, and c 2,max and c 2,min are the maximum and minimum values of the social learning factor, respectively.

[0074] Technical effects and advantages of the present application:

[0075] The present application realizes the collection of the data X m X r X d of coal consumption, heat and power generation through a sensor network, as well as the collection of historical marketing data X x and historical monthly expenditure costs X z in a large database, and the collection of the coal purchase unit price X d , electricity sales unit price X q and heat sales unit price X v , so as to facilitate the collection of parameters, in order to improve the accuracy of the collected parameters, and to realize the denoising, fusion and unified format of each parameter, to realize the preprocessing of data information, and to improve the accuracy of data information and eliminate the interference of noise, by filtering the data information through Fourier transform and wavelet transform, and by adaptively adjusting the Fourier transform window factor, so as to automatically adjust the window according to the fluctuation of data information, and to improve the smoothness when performing frequency domain conversion; and when standardizing the data information, weighted and median and absolute deviation are introduced, so as to improve the accuracy of data information processing;

[0076] After preprocessing the data information, the data information is established in a set processing, and a data visualization dynamic diagram is established according to the data set, so as to realize the separate visualization monitoring of coal consumption, power generation and profit margin, improve the visualization degree of data information, facilitate the direct observation of data information, and establish a visualization dynamic diagram, so as to observe the change of each parameter, observe the update state of data information and the trend of data information, so that the operator can obtain the intermediate information of data, and facilitate the observation and processing of the base number of energy accounting;

[0077] The artificial intelligence model is used for analyzing and processing the processed data information set, so as to predict the profit amount, compare the predicted profit amount, optimize resource allocation and operation plan, and when the feedforward neural network is analyzed and processed, the artificial intelligence model is convenient for calculating energy efficiency, power generation efficiency, energy efficiency analysis, realizing calculation and analysis of energy accounting, and after the energy accounting, the resource allocation and operation plan are optimized by particle swarm optimization, so that the utilization rate and production efficiency of resources can be improved.

[0078] Other features and advantages of the present application will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in connection with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0079] Fig. 1 is a step flow chart of the coal-fired power plant intelligent management energy accounting method provided by the present application;

[0080] Fig. 2 is a pre-processing step flow chart provided by the present application;

[0081] Fig. 3 is a step flow chart of the establishment of the data visualization dynamic diagram provided by the present application. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0083] As shown in Figs. 1 to 3 , the coal-fired power plant intelligent management energy accounting method provided by the present application includes the following steps:

[0084] Step one: real-time tracking of coal consumption data X m , heat data X r and power generation data X d , and collecting historical marketing data X x and historical monthly expenditure X z in the large database, and collecting coal purchase unit price X d , electricity sales unit price X q heat sales unit price X v ;

[0085] It should be noted that a plurality of sensors are provided in the sensor network, which are respectively used for coal consumption, heat, power generation and emission, mainly including weight sensors, temperature sensors, intelligent power meters, gas sensors, flow sensors and pressure sensors, wherein the weight sensors, temperature sensors, intelligent power meters, gas sensors, flow sensors and pressure sensors are mainly used for detecting coal consumption, heat, power generation and emission;

[0086] Step two: preprocessing the collected coal consumption data X m , heat data X r , power generation data X d , historical marketing data X x and historical monthly expenditure X z , eliminating noise in data information, completing data information cleaning, and improving data information accuracy;

[0087] The preprocessing processing steps are as follows:

[0088] The sensor network collects coal consumption data X m , heat data X r and power generation data X d , and collects historical marketing data X x and historical monthly expenditure X z in the large database;

[0089] Then, the received X m , X r , X d , X x and X z are denoised, calculated and processed by signal processing method, first converted to frequency domain by Fourier transform, high frequency noise is removed by filtering, then converted back to time domain, then wavelet transform is performed: wavelet transform is performed on the data, noise coefficient is removed, and the data is reconstructed;

[0090] Then, the denoised data information is cleaned, the abnormal data in the data information is obtained, and the abnormal data information is removed, and the removed position is filled, so as to maintain the integrity of the data information;

[0091] The data information is standardized to ensure that the data format and unit are consistent, which is convenient for subsequent analysis;

[0092] The Fourier transform is performed on the collected coal consumption data X m , heat data X r and power generation data X d , and historical marketing data X xand historical monthly expenditure cost X z Respective transformation processing is performed, and the calculation formula of transformation is as follows:

[0093]

[0094] Wherein, x(t) is a time domain signal, indicating a signal to be analyzed, w(t) is a window function, used for local weighting of the signal, t is a time variable, the center position of the window, f is a frequency variable, indicating frequency domain information, X(t,f) is the result of STFT, indicating the amplitude and phase information of the signal at time t and frequency f, τ is an integral variable, indicating the position in time, used to calculate the spectral information of the signal at a specific time point t, τ is used to describe the overlapping part of the window function and the signal, helping to determine the frequency characteristics of the signal at time t, ξ is a window self-adaptive adjustment factor introduced, wherein the calculation formula of ξ is as follows:

[0095] Wherein, rect is a rectangular window function, and Δt(τ) is a window width dynamically adjusted according to the instantaneous frequency or characteristics of the signal;

[0096] The wavelet transform is used for retransforming the data information after Fourier transform, and the calculation formula is as follows:

[0097] Wherein, ψ a,b (t) is a mother wavelet, a is a scale parameter, controlling the frequency band of analysis, and b is a translation parameter, controlling the position of analysis;

[0098] The data after wavelet transform is further decomposed:

[0099] And each DWT i Indicates the wavelet coefficient at different scales;

[0100] The data information is further denoised:

[0101] Wherein, c i Is the coefficient of the i-th layer, and λ is a threshold parameter;

[0102] Finally, the processed data information is reconstructed:

[0103] Wherein, IDWT indicates the inverse wavelet transform process;

[0104] It should be noted that in order to improve the accuracy of data information, and make the data information convenient for subsequent calculation processing, including denoising, cleaning and format conversion of data information, and for real-time data information through Fourier transform, the data is converted to frequency domain, the high frequency noise is removed through filtering, and then the data is converted back to time domain, and then through wavelet transform: wavelet transform of data, removal of noise coefficient, and then data reconstruction, completion of denoising processing of data information, and realization of conversion processing between time domain and frequency domain, and in the Fourier transform processing, the window self-adaptive adjustment factor is introduced, the window width is adjusted dynamically according to the instantaneous frequency or characteristics of the signal, which is convenient for optimizing the real-time collected data information, improving the smoothness of the processed data information, maintaining the accuracy of the data information, and through wavelet transform, the noise can be removed, and the position of the eliminated data information can be filled, maintaining the integrity of the data information, preventing abnormal situations during subsequent data processing.

[0105] Step three: fusion processing of the preprocessed data information, all data information is unified in format, and coal consumption data information set [X m ], heat information set [X r ], power generation data information set [X d ], historical marketing data set [X x ] and historical monthly expenditure data set [X z ];

[0106] The standardization processing of the data information is as follows:

[0107] First, the weighted mean processing of the denoising processed data information is carried out:

[0108] Wherein, w i is the weight of data point x i ,

[0109] Then introduce the median in the weighted standardization processing, the calculation formula is as follows:

[0110] Wherein, median(X) is the median of data information X, and MAD(X) is the absolute deviation of data;

[0111] It should be noted that in order to realize the conversion of the collected data information into data set, realize the fusion and unified format of different types of data information, and improve the uniformity of data, the data information is standardized to convert the data information into corresponding data information set;

[0112] Step four: coal consumption data information set [Xm ] the heat information set [X r ] the power generation data information set [X d ] the emission data information set [X p ] the historical marketing data set [X x ] and the historical monthly expenditure data set [X z ] respectively to establish a data visualization dynamic diagram;

[0113] The data visualization dynamic diagram is established as follows:

[0114] The data information of the processing process is stored, and the data information is fused and processed, and all the data information is unified in format to ensure that the data format is suitable for visualization needs;

[0115] The broken line visualization diagram is designed to enable the processed data information to achieve a dynamic display effect, facilitate observation of the update state of the data information, and the trend of the data information;

[0116] Code or chart configuration is written using selected tools, and dynamic update logic is added;

[0117] The effect of the dynamic diagram is checked, and parameters and designs are adjusted to ensure clarity and accuracy;

[0118] It should be noted that in order to visualize the display of the data information set, the processed data information set is processed, code or chart configuration is written using selected tools, and dynamic update logic is added, so that the processed data information can achieve a dynamic display effect, facilitate observation of the update state of the data information, and the trend of the data information;

[0119] Step five: the coal consumption data information set [X m ], the heat information set [X r ], the power generation data information set [X d ], the emission data information set [X p ], the historical marketing data set [X x ] and the historical monthly expenditure data set [X z ] are stored, and analyzed and processed through an artificial intelligence model to optimize resource allocation and operation plans.

[0120] The coal consumption data X m , the heat data X r and the power generation data X d calculate the output of coal, and optimize the proportioning combustion of coal:

[0121] X r +X d =X m ·μ·ζ·ψ,

[0122] Wherein, μ, ζ, ψ are oxygen supply ratio coefficient, coal blending ratio coefficient and combustion output ratio coefficient, respectively;

[0123] In the calculation of output value by unit price:

[0124] (X r ·X v +X d ·X q )-X j =X m ·X d ·μ·ζ·ψ+X f ,

[0125] Wherein, X f is other consumption amount generated in the process of coal-fired power generation, and X j is the calculated profit margin of coal-fired output.

[0126] The artificial intelligence model adopts a feedforward neural network, and each layer of neurons is fully connected with the next layer of neurons;

[0127] Each layer of neurons is fully connected with the next layer of neurons using linear connection, and the connection formula is as follows:

[0128] Wherein, x i is the input data information, mainly including historical marketing data set [X x ] and historical monthly expenditure data set [X z ], w ij is the weight from the ith input to the jth neuron, and b j is the bias;

[0129] The sales amount and expenditure amount of the month are calculated by the feedforward neural network to obtain the predicted profit margin of the month;

[0130] The predicted profit margin of the month is calculated with the profit margin X j of the coal-fired output to obtain the difference between the prediction and the actual value;

[0131] D=|z j -X j |, wherein D is the difference between the prediction and the actual value, and is the sum of the specific profit and monthly expenditure;

[0132] And the calculated data is displayed through a visualization tool.

[0133] The feedforward neural network adopts Sigmoid function as the activation function, and the calculation formula is as follows:

[0134] where z is the weighted sum of neuron inputs, and the calculation formula is:

[0135]

[0136] where w i is the weight, which determines the influence of each input x i on the output, and b is the bias, which adjusts the output range of the function;

[0137] And the calculation results of the feedforward neural network are evaluated by the loss function, which evaluates the gap between the network prediction results and the true value:

[0138] where y k is the true value, is the predicted value, and n is the number of samples;

[0139] The weights and biases are updated by gradient descent method or its variants;

[0140] Gradient descent method:

[0141]

[0142] where η is the learning rate, is the gradient of the loss function with respect to the weight, is the gradient of the loss function with respect to the bias.

[0143] The optimization of resource allocation and operation plan is adjusted by particle swarm optimization, and the particle swarm optimization combines the calculated difference C to optimize the resource allocation and operation plan processing;

[0144] And in order to improve the efficiency of the algorithm, hierarchical local search is performed according to the neural cloud of the feedforward neural network, and the particle swarm is divided into multiple levels, each level uses different parameter settings for optimization, and then the results are integrated:

[0145] v i (t+1)=w k ·v i (t)+c 1,k ·r1(pBest i -x i (t))+c 2,k ·r2(gBest i -x i (t))+LocalSearch(x i (t)),

[0146] where v i(t) is the velocity of particle i at time t, x i (t) is the position of particle i at time t, w is the inertia weight, which controls the inertia of the particle, c1 and c2 are learning factors, which control the learning degree of the particle to the individual best position and the group best position respectively, r1 and r2 are random numbers between [0, 1], and k represents the level, LocalSearch(x i (t)) is a local search algorithm, pBest i is the best position found by particle i so far, gBest i is the best position found by the entire particle group so far.

[0147] In order to realize the dynamic adjustment of the inertia weight and the dynamic adjustment of the learning factor for particle swarm optimization:

[0148]

[0149] where w max is the maximum value of the inertia weight, w min is the minimum value of the inertia weight, T is the maximum number of iterations, and t is the current number of iterations.

[0150]

[0151] where c 1,max and c 1,min are the maximum and minimum values of the individual learning factor respectively, c 2,max and c 2,min are the maximum and minimum values of the social learning factor respectively.

[0152] It should be noted that in order to realize the energy accounting, the artificial intelligence model is used for analysis and processing, realizing the prediction of the load demand, coal consumption and profit margin of the power plant, the input of data information is completed through the neurons of the feedforward neural network, which is convenient for realizing the phased calculation and processing of energy accounting. In order to realize the automatic update of the weight and bias, a loss function is introduced for evaluation, which evaluates the gap between the network prediction result and the true value, which is convenient for adjusting the calculation process of the feedforward neural network. And through the phased calculation and analysis of multiple neurons, it is convenient to calculate the energy efficiency, power generation efficiency, energy efficiency analysis and profit margin prediction, realize the accurate calculation and processing of the energy accounting of the power plant, and realize the distribution adjustment through particle swarm optimization. And the particle swarm optimization combines the calculation of energy efficiency, power generation efficiency, load prediction, energy efficiency analysis and profit margin difference to optimize resource allocation and operation plan processing. In order to realize the improvement of the efficiency and accuracy of the algorithm, hierarchical local search is carried out according to the neural cloud of the feedforward neural network, and the particle swarm is divided into multiple levels, each level uses different parameter settings for optimization, thereby realizing the optimization of resource allocation and operation plan.

[0153] Finally, it should be noted that the above only for the preferred embodiments of the present application, and is not intended to limit the present application, although the foregoing embodiments of the present application has been described in detail, for those skilled in the art, it still can be modified, or part of the technical features of the equivalent replacement, within the spirit and principles of the present application, any modification, equivalent replacement, improvement, etc., should be included within the scope of the present application.

Claims

1. A method for accounting for energy in the intelligent management of a coal-fired power plant, characterized by, Comprise the following steps: Tracking data X of coal consumption in real time through sensor network m Tracking data X of heat r Tracking data X of power generation d Collecting historical marketing data X in large database x Collecting historical monthly expenditure cost X z Collecting coal purchase unit price X d Collecting power sale unit price X q Collecting heat sale unit price X v ; Data X on collected coal consumption m Data X on heat r Data X on power generation d Historical marketing data X x and historical monthly expenditure costs X z Preprocessing is performed to eliminate noise in the data information, complete cleaning of the data information, and improve the accuracy of the data information. The preprocessed data information is fused, all data information is unified in format, and a coal consumption data information set [X m ], a heat information set [X r ], a power generation data information set [X d ], a historical marketing data set [X x ] and a historical monthly expenditure data set [X z ] are respectively established; A data visualization dynamic graph is established for each of a coal consumption data information set [X m ], a heat information set [X r ], a power generation data information set [X d ], an emission data information set [X p ], a historical marketing data set [X x ], and a historical monthly expenditure data set [X z ]. The coal consumption data information set [X m ], the heat information set [X r ], the power generation data information set [X d ], the emission data information set [X p ], the historical marketing data set [X x ] and the historical monthly expenditure data set [X z ] are stored and analyzed and processed by an artificial intelligence model to optimize resource allocation and operation plans.

2. The method for energy accounting of intelligent management of coal-fired power plants according to claim 1, characterized in that: The preprocessing processing steps are as follows: data X of coal consumption collected by a sensor network m data X of heat r and data X of power generation, and d data of historical marketing data X x and historical monthly expenditure cost X z in a large database are received; The received X m , X r , X d , X x and X z are denoised, and are calculated and processed through a signal processing method, are first converted to a frequency domain through Fourier transform, high-frequency noise is removed through filtering, and then the data is converted back to a time domain, and are then processed through wavelet transform: wavelet transform is performed on the data, noise coefficients are removed, and the data is reconstructed; Then the denoised data information is cleaned, the abnormal data in the data information is obtained, and the abnormal data information is eliminated, and the position of the eliminated data is filled to maintain the integrity of the data information; Standardize the data information to ensure consistent data format and unit for subsequent analysis.

3. The method for energy accounting of intelligent management of coal-fired power plants according to claim 2, characterized in that: The Fourier transform is applied to the collected data X of coal consumption m , data X of heat r , and data X of power generation d , and historical marketing data X x and historical monthly expenditure costs X z in a large database are respectively transformed, and the calculation formula of the transformation is as follows: Wherein, x(t) is a time domain signal, indicating the signal to be analyzed, w(t) is a window function, used for local weighting of the signal, t is a time variable, the center position of the window, f is a frequency variable, indicating the frequency domain information, X(t,f) is the result of STFT, indicating the amplitude and phase information of the signal at time t and frequency f, tau is the integral variable, indicating the position in time, used to calculate the frequency spectrum information of the signal at a specific time point t, tau is used to describe the overlapping part of the window function and the signal, which helps to determine the frequency characteristics of the signal at time t, and the calculation formula of the introduced window self-adaptive adjustment factor is as follows: where rect is a rectangular window function, and At(τ) is a window width that is dynamically adjusted according to the instantaneous frequency or characteristics of the signal.

4. The method for energy accounting of intelligent management of coal-fired power plants according to claim 2, characterized in that: The wavelet transform is used for re-transforming the data information after Fourier transform, and the calculation formula is as follows: where ψ a,b (t) is a mother wavelet, a is a scale parameter that controls the frequency band of the analysis, and b is a translation parameter that controls the location of the analysis. The data after wavelet transform is decomposed again: and each DWT i denotes wavelet coefficients at different scales; The data information is denoised again: where c i is the coefficient of the i-th layer, and λ is a threshold parameter. Finally, the processed data information is reconstructed: where IDWT denotes the inverse wavelet transform process.

5. The method for energy accounting of intelligent management of coal-fired power plants according to claim 2, characterized in that: The standardization processing of the data information is as follows: First, the weighted mean value processing is performed on the data information after the de-noising processing: where w i is the weight of data point x i , Then introduce the median when performing weighted standardization processing, and the calculation formula is as follows: where median(X) is the median of the data information X and MAD(X) is the absolute deviation of the data.

6. The method for energy accounting of intelligent management of coal-fired power plants according to claim 1, characterized in that: The establishment steps of the data visualization dynamic diagram are as follows: Store the data information of the processing process, and fuse the data information to unify the format of all data information, so as to ensure that the data format is suitable for visualization requirements; Design polyline visualization diagram to realize dynamic display effect of processed data information, so as to observe the update state of data information and the trend of data information; Use the selected tool to write code or configure chart, add dynamic update logic; Check the effect of dynamic diagram, adjust parameters and design to ensure clarity and accuracy.

7. The method for energy accounting of intelligent management of coal-fired power plants according to claim 1, characterized in that: Data X of the coal consumption m Data X of the heat r Data X of the power generation d The output of the coal is calculated, and the coal blending combustion is optimized: X r +X d = X m · μ · ζ · ψ, Wherein, mu, zeta and psi are oxygen supply ratio coefficient, coal blending ratio coefficient and combustion output ratio coefficient respectively; The output value is calculated by unit price: (X r ·X v +X d ·X q )-X j =X m ·X d ·μ·ζ·ψ+X f , where X f is the other consumption amount generated in the coal-fired power generation process, X j is the calculated profit amount of the coal-fired power generation.

8. The method for energy accounting of intelligent management of coal-fired power plants according to claim 7, characterized in that: The artificial intelligence model adopts feedforward neural network, and each layer of neurons is fully connected with the next layer of neurons; Each layer of neurons is fully connected with the next layer of neurons by linear connection, and the connection formula is as follows: where x i is the input data information, mainly including a historical marketing data set [X x ] and a historical monthly expenditure data set [X z ], w ij is the weight from the i-th input to the j-th neuron, and b j is the bias; The feedforward neural network is used to realize the prediction and calculation of the sales amount and the expenditure amount of this month, and the predicted profit amount of this month is obtained; The predicted amount of profit for the current month is compared with the amount of profit X from the coal production j A calculation process is performed to obtain the difference between the prediction and the actual value; D = |z j X j where D is the difference between the calculated prediction and the actual, and z is the sum of the specific profit and monthly expenses. And the calculated data is displayed through the visualization tool. 9.The energy accounting method for intelligent management of a coal-fired power plant according to claim 8, characterized in that: The feedforward neural network adopts Sigmoid function as the activation function, and the calculation formula is as follows: where z is the weighted sum of the neuron inputs, calculated as: where: w i is the weight, which determines the degree of influence of each input x i on the output, and b is the bias, which adjusts the output range of the function; And the calculation result of the feedforward neural network is evaluated by the loss function to evaluate the gap between the network prediction result and the true value: where y k is the true value, is the predicted value, n is the number of samples; The weights and biases are updated by gradient descent method or its variants; Gradient descent method: where η is the learning rate, is the gradient of the loss function with respect to the weights, is the gradient of the loss function with respect to the bias. 10.The energy accounting method for intelligent management of a coal-fired power plant according to claim 8, characterized in that: The optimized resource allocation and operation plan are allocated and adjusted by particle swarm optimization, and the particle swarm optimization combines the calculated difference D to optimize the resource allocation and operation plan processing; And in order to improve the efficiency of the algorithm, according to the neural cloud of feedforward neural network, hierarchical local search is carried out, and combined with local search algorithm, the particle swarm is divided into multiple levels, each level uses different parameter settings for optimization, and then the results are integrated: v i (t+1) = w k ·v i (t) + c 1,k ·r1(pBest i -x i (t)) + c 2,k ·r2(gBest i -x i (t)) + LocalSearch(x i (t)), where v i (t) is the velocity of particle i at time t, x i (t) is the position of particle i at time t, w is the inertia weight that controls the inertia of particles, c1 and c2 are learning factors that control the learning degree of particles to the individual best position and the global best position respectively, r1 and r2 are random numbers between [0, 1], and k represents the level, LocalSearch(x i (t)) is a local search algorithm, pBest i is the best position found so far by particle i, gBest i is the best position found so far by the entire particle swarm. In order to realize the dynamic adjustment of inertia weight and dynamic adjustment of learning factor in particle swarm optimization: where w max is the maximum value of the inertia weight, w min is the minimum value of the inertia weight, T is the maximum number of iterations, and t is the current iteration number. where c 1,max and c 1,min are the maximum and minimum values of the individual learning factor, respectively, and c 2,max and c 2,min are the maximum and minimum values of the social learning factor, respectively.

Images