Auxiliary peak regulation method and device for cogeneration unit, storage medium and electronic equipment

Abstract

The invention discloses a cogeneration unit auxiliary peak regulation method and device, a storage medium and electronic equipment. The method comprises the following steps: processing and analyzing meteorological data to obtain a meteorological prediction result; obtaining a new energy output prediction result based on the weather prediction result and the new energy operation data; determining an electric heating load demand prediction result according to the electric heating load data, the weather prediction result and the operation state of the thermodynamic system; according to the new energy output prediction result and the electric heating load demand prediction result, peak regulation or peak demand parameters are generated; according to the real-time operation parameters, the configuration parameters of the multiple thermoelectric decoupling modes and the application scene, the peak regulation and peak potential of the cogeneration unit is determined; and according to the weather prediction result, the new energy output prediction result, the electric heating load demand prediction result, the peak regulation or peak related demand parameters and the peak regulation and peak potential of the cogeneration unit, an auxiliary peak regulation scheme of the cogeneration unit is determined in combination with the supply and demand change of the electric power spot market and the electric heating load fluctuation condition.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to an auxiliary peak-shaving method, device, storage medium and electronic equipment for a combined heat and power unit. Background Technology

[0002] In the process of transforming the power energy structure towards cleaner energy, the installed capacity of new energy sources (wind power, photovoltaic, etc.) has increased significantly. Their output is significantly affected by meteorological conditions and fluctuates. In order to ensure the stability of the power grid frequency and the balance of power supply and demand, cogeneration units, as core power sources that have both power supply and heating functions, need to undertake more critical peak-shaving tasks to absorb the power surplus caused by the fluctuation of new energy sources or fill the power gap during peak load periods.

[0003] Currently, combined heat and power (CHP) units need to meet multiple demands simultaneously during operation: on the one hand, they must ensure a stable supply of electricity and heat loads for different scenarios such as residential life, industrial production, and commercial services, especially during peak energy consumption periods such as the winter heating season, prioritizing residential heating needs; on the other hand, they must dynamically adapt to real-time changes in renewable energy output and adjust their own power generation in a timely manner to maintain grid supply and demand balance. However, dynamic changes in meteorological conditions directly cause fluctuations in renewable energy output, and differences in electricity and heat demand in different scenarios lead to continuous adjustments in electricity and heat loads. In addition, the operating status of the thermal system constantly changes with supply and demand. These intertwined factors make the grid's peak-shaving demand complex and dynamic. CHP units operating under the current mode cannot accurately match this dynamic peak-shaving demand, thus affecting the overall peak-shaving efficiency of the grid. Summary of the Invention

[0004] In view of the above problems, this application provides an auxiliary peak shaving method, device, storage medium and electronic equipment for cogeneration units.

[0005] To solve the above-mentioned technical problems, this application proposes the following solution: Firstly, this application provides an auxiliary peak-shaving method for cogeneration units. The method includes: acquiring meteorological data, processing and analyzing the meteorological data to obtain meteorological forecast results; acquiring renewable energy operation data, and determining renewable energy output forecast results based on the meteorological forecast results and renewable energy operation data; acquiring electricity and heat load data within the power supply and heating coverage area corresponding to the cogeneration unit, classifying and analyzing the electricity and heat load data, and processing it in conjunction with the meteorological forecast results and the operating status of the thermal system to obtain electricity and heat load demand forecast results; generating peak-shaving or peak-load demand parameters based on the renewable energy output forecast results and electricity and heat load demand forecast results; acquiring real-time operating parameters of the cogeneration unit, and determining the peak-shaving and peak-load potential of the cogeneration unit based on the real-time operating parameters, configuration parameters of various heat and power decoupling methods, and application scenarios; and determining an auxiliary peak-shaving scheme for the cogeneration unit based on the meteorological forecast results, renewable energy output forecast results, electricity and heat load demand forecast results, peak-shaving or peak-load related demand parameters, and the peak-shaving and peak-load potential of the cogeneration unit, combined with changes in electricity spot market supply and demand and fluctuations in electricity and heat load.

[0006] Secondly, this application provides an auxiliary peak-shaving device for a combined heat and power (CHP) unit, which includes: The weather forecasting module is used to acquire weather data, process and analyze the weather data, and obtain weather forecast results. The new energy output prediction module is used to acquire new energy operation data and determine the new energy output prediction results based on meteorological forecast results and new energy operation data; The electric heating load forecasting module is used to acquire electric heating load data within the power supply and heating coverage area corresponding to the cogeneration unit, classify and analyze the characteristics of the electric heating load data, and process it in combination with meteorological forecast results and the operating status of the thermal system to obtain the electric heating load demand forecast results. The parameter determination module is used to generate peak-shaving or peak demand parameters based on the forecast results of new energy output and the forecast results of electric and heat load demand. The thermoelectric decoupling strategy module is used to obtain the real-time operating parameters of the cogeneration unit and determine the peak shaving and peak load potential of the cogeneration unit based on the real-time operating parameters, the configuration parameters of various thermoelectric decoupling methods and application scenarios. The scheme generation module is used to determine the auxiliary peak-shaving scheme for cogeneration units based on meteorological forecasts, new energy output forecasts, electricity and heat load demand forecasts, peak-shaving or peak-climbing related demand parameters, and the peak-shaving and peak-climbing potential of cogeneration units, combined with changes in electricity spot market supply and demand and fluctuations in electricity and heat load.

[0007] To achieve the above objectives, according to a third aspect of this application, a storage medium is provided, the storage medium including a stored program, wherein, when the program is executed, the device where the storage medium is located is controlled to perform the auxiliary peak shaving method for a cogeneration unit described in the first aspect.

[0008] To achieve the above objectives, according to a third aspect of this application, an electronic device is provided, the device including at least one processor, and at least one memory and bus connected to the processor; wherein the processor and memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the auxiliary peak shaving method for cogeneration units described in the first aspect.

[0009] By employing the above-described technical solution, the technical solution provided in this application has at least the following advantages: This application first acquires and processes meteorological data to obtain meteorological forecast results, which can capture changes in meteorological elements such as temperature and wind speed in advance, providing a basis for predicting fluctuations in renewable energy output and adjusting electricity and heat loads, and avoiding delays in peak-shaving response due to sudden meteorological changes. Next, it combines meteorological forecast results with renewable energy operation data to determine renewable energy output forecast results, and simultaneously analyzes electricity and heat load data and combines meteorological forecast results with the operating status of the thermal system to obtain electricity and heat load demand forecast results. Through these two types of forecasts, the real-time supply and demand relationship between renewable energy supply and electricity and heat demand can be accurately grasped. Then, based on the difference between the two, peak-shaving or peak-load demand parameters are generated, clarifying whether peak-shaving or peak-load is needed at different times, and specifying the power and duration requirements, thus solving the problem of ambiguous judgment of supply and demand differences. Subsequently, by combining the configuration parameters and application scenarios of various heat-power decoupling methods, peak-shaving and peak-load potential are determined, clearly defining the boundary of the unit's adjustable capacity under the current state, avoiding blindly formulating peak-shaving measures. Finally, by comprehensively considering meteorological forecasts, renewable energy output forecasts, electricity and heat load demand forecasts, demand parameters, and unit potential, and combining the supply and demand in the electricity spot market with fluctuations in electricity and heat load, an auxiliary peak-shaving scheme is determined. This ensures that peak-shaving measures not only adapt to actual supply and demand changes but also match the unit's regulation capacity, effectively addressing the complex peak-shaving demands brought about by the interplay of meteorological, renewable energy, load, and thermal system conditions, and improving the grid's peak-shaving efficiency.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0011] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic flowchart of an auxiliary peak-shaving method for a combined heat and power unit provided in an embodiment of this application is shown; Figure 2 This illustration shows a structural schematic diagram of an auxiliary peak-shaving device for a combined heat and power unit provided in an embodiment of this application; Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0012] The auxiliary peak-shaving method for combined heat and power units will be explained in detail below with reference to the accompanying drawings. Figure 1 This application provides a flowchart illustrating an auxiliary peak-shaving method for a combined heat and power (CHP) unit. The method includes the following steps.

[0013] Step 110: Obtain meteorological data, process and analyze the meteorological data, and obtain meteorological forecast results.

[0014] First, multi-dimensional meteorological data was acquired for the area covered by the combined heat and power (CHP) unit and its surrounding regions. Data sources included real-time monitoring data from ground meteorological observation stations, meteorological satellite remote sensing data, upper-air sounding data, and data collected from regional automatic weather stations. This included core meteorological elements such as temperature, humidity, wind speed, wind direction, precipitation, sunshine duration, and air pressure. Historical meteorological data from the same period over the past five years was also collected as the foundation for model training and feature analysis. The acquired meteorological data underwent preprocessing, removing outliers (such as data exceeding reasonable physical ranges or abrupt changes caused by sensor malfunctions). Linear interpolation was used to fill in missing data, and data standardization was applied to convert meteorological element data of different dimensions to the same numerical range, ensuring the accuracy and effectiveness of subsequent feature extraction.

[0015] After completing the meteorological data preprocessing, the processed meteorological data is decomposed into multiple scales using wavelet transform. The db4 wavelet basis function is selected as the decomposition core, and the original meteorological data is decomposed into one approximate component (reflecting the overall trend of the data) and four detail components (reflecting the fluctuation characteristics at different scales). Statistical domain features such as mean, variance, extreme values, and peak factor are extracted from the approximate component, and time-frequency domain features such as energy entropy, wavelet coefficient modulus maxima, instantaneous frequency, and amplitude spectrum are extracted from each detail component to form a preliminary feature set.

[0016] The extracted statistical domain features and time-frequency domain features are then input into a multilayer perceptron-based attention model. This model consists of an input layer, three hidden layers (with 256, 128, and 64 neurons respectively), and an output layer. Through fully connected operations and the ReLU activation function in the hidden layers, each feature is non-linearly mapped. A scaled dot product attention mechanism is introduced into the model to calculate the contribution weight of each feature relative to the weather forecast result. Specifically, matrix operations are performed on the feature vector, query vector, and key vector to obtain an attention score, which is then normalized using a softmax function to generate the contribution weight of each feature. A preset threshold of 0.05 is set for the contribution weight. Features with contribution weights greater than this threshold are selected to form a weather feature set, while redundant features with low weights are removed to improve the computational efficiency of subsequent models.

[0017] A multimodal meteorological prediction model is constructed based on the selected meteorological feature set. This model includes a spatial correlation feature extraction branch, a temporal correlation feature extraction branch, and a feature fusion layer. The spatial correlation feature extraction branch employs a convolutional neural network (CNN) structure, mapping spatially distributed related features from the meteorological feature set through three convolutional layers (kernel sizes of 3×3, 5×5, and 3×3, with a stride of 1 for each layer). This is combined with a max-pooling layer (pooling kernel size of 2×2, stride of 2) to achieve feature dimensionality reduction, ultimately obtaining spatial correlation features that reflect the correlation between meteorological elements in different regions. The temporal correlation feature extraction branch employs a long short-term memory (LSTM) network structure, with two LSTM layers (128 neurons per layer). Through gating mechanisms (input gate, forget gate, output gate), it mines the dependencies of meteorological data in the temporal dimension, capturing hourly, daily, and weekly temporal variation patterns, thereby obtaining temporal correlation features. The feature fusion layer adopts an attention-weighted fusion strategy. First, it calculates the importance weights of spatial correlation features and temporal correlation features (adaptively learned through the training process). Then, it adds and normalizes the two types of features element-wise according to their weights to complete feature integration and form a unified fusion feature vector.

[0018] After the model was built, preprocessed meteorological data from the same period in the past 5 years were used as the training set, divided into training and validation samples in a 7:3 ratio. The mean squared error (MSE) between the predicted and actual values ​​was used as the loss function. The Adam optimizer (learning rate 0.001, decay coefficient 0.9) was used to iteratively update the model parameters, with 100 iterations. The model was considered stable after 5 consecutive iterations (the change was less than 10⁻⁻⁶). 4 Training is stopped when the time is right, and the optimized multimodal weather prediction model is obtained.

[0019] Finally, the meteorological data (after preprocessing) for the period to be predicted is input into the trained multimodal meteorological prediction model. The spatial correlation feature extraction branch obtains the spatial correlation information of meteorological elements for that period, and the temporal correlation feature extraction branch captures its temporal evolution trend. After integration by the feature fusion layer, the model output layer outputs the meteorological prediction results for the period to be predicted, including the predicted values ​​of meteorological elements such as hourly temperature, humidity, wind speed, and precipitation, as well as the prediction confidence interval (95% confidence level) of each element. This provides accurate meteorological data support for subsequent prediction of new energy output and electricity and heat load demand.

[0020] Step 120: Obtain new energy operation data, and determine the new energy output forecast result based on meteorological forecast results and new energy operation data.

[0021] Data on the operation of new energy sources is acquired, including real-time monitoring data from the SCADA (Supervisory Control and Data Acquisition) system of new energy power plants, historical operation databases, and records from equipment management systems. Specifically, this includes data on photovoltaic module temperature, irradiance, inverter output power, and array tilt angle for photovoltaic power plants, and wind turbine speed, pitch angle, wind speed at hub height, nacelle temperature, and grid-connected power for wind farms. Simultaneously, output data, equipment start-up and shutdown records, and fault maintenance logs from new energy power plants over the same period for the past three years are collected to form a complete dataset on new energy operation. The acquired new energy operation data undergoes preprocessing. Abnormal output data caused by equipment failures or communication interruptions (such as sudden output drops to 0 or exceeding 120% of rated power) are removed. Short-term fluctuations are smoothed using a moving average method, and different types of operating parameters are converted to the [0,1] range through data normalization to ensure consistency between the data format and the characteristic data of meteorological forecasts.

[0022] After data preparation, two types of features were extracted. The first type of features directly adopted the statistical and time-frequency domain features corresponding to the meteorological forecast results, including mean temperature, humidity variance, wind speed energy entropy, and instantaneous frequency of sunshine duration, resulting in 28 core features. The second type of features was extracted from the preprocessed new energy operation data. Operational status features included 15 parameters such as photovoltaic module temperature fluctuation range, wind turbine pitch angle adjustment amplitude, inverter efficiency, and nacelle temperature stability. Output response features included 12 parameters such as output change rate per unit time, output ramp-up rate, maximum output duration, and output trough recovery time. Subsequently, the coupling correlation between the two types of features was determined through feature cross-operation. Specifically, a combination of polynomial cross-operation and Hadamard product was used: first, polynomial cross-operation was performed on the wind speed time-frequency domain features in the first type of features and the wind turbine output response features in the second type of features, generating cross-terms such as "wind speed-output change rate" and "wind speed fluctuation amplitude-ramp-up rate". Then, a Hadamard product operation is performed on the temperature statistical domain features and the photovoltaic module temperature operating status features to obtain a feature vector reflecting the coupling relationship between temperature and module thermal state. Finally, a feature set containing 42 coupled and correlated features is formed, covering the core correlation dimensions between wind-solar new energy and meteorological elements.

[0023] Based on the temporal evolution patterns of meteorological forecast data and new energy operation data, we extract the in-depth features of coupled correlation features in the time dimension as temporal collaborative features. First, we construct a time series window with a length of 24 hours (corresponding to a day's meteorological and power output cycle) and a sliding step of 1 hour. We then perform temporal analysis on the coupled correlation features within each time window. Specifically, we use difference operations to capture the short-term trends of features, autocorrelation analysis to uncover the periodic patterns of features (such as the photovoltaic characteristic cycle caused by intraday sunshine variation and the wind power characteristic cycle caused by diurnal wind speed variation), and cross-correlation analysis to quantify the time lag relationships between different coupled correlation features (such as the lag time between wind speed characteristic changes and wind turbine power output characteristic changes). The above analysis results were then input into a bidirectional long short-term memory network (Bi-LSTM), which contains a forward LSTM layer and a backward LSTM layer (each layer has 128 neurons). The forward layer extracts temporal features from historical data to the future, while the backward layer traces dependencies from future data to the historical data. By splicing and fusing the outputs of the two layers, temporal co-evolutionary features that reflect the long-term evolution trend and short-term fluctuation correlation of coupled features are obtained, generating a total of 36 temporal co-evolutionary feature parameters.

[0024] Based on temporal collaborative features, feature fusion is achieved through bidirectional interaction between forward and reverse temporal collaborative features, extracting long-term evolutionary trend features of new energy output from the fused features. Specifically, a bidirectional interaction channel is constructed through an attention mechanism: forward temporal collaborative features are used as query vectors, and reverse temporal collaborative features are used as key and value vectors, calculating the attention weight between them. The weight value reflects the supplementary explanatory power of the reverse feature to the forward feature. Simultaneously, reverse temporal collaborative features are used as query vectors, and forward temporal collaborative features are used as key and value vectors, calculating another set of attention weights to form a bidirectional interactive attention matrix. Based on this matrix, forward and reverse temporal collaborative features are weighted and fused to obtain bidirectional interactive fused features. Subsequently, trend extraction is performed on the fused features. Wavelet packet decomposition is used to decompose the fused features into components of different frequencies, filtering out low-frequency components with frequencies below 0.04Hz (corresponding to a period greater than 25 hours). Through linear fitting and trend term extraction algorithms, the long-term evolutionary trend features of new energy output are obtained, including eight core trend parameters such as the daily peak output shift trend, the weekly output fluctuation decay trend, and the monthly average output change trend.

[0025] Based on the coupling correlation characteristics, multi-scale modal decomposition is performed to separate the output fluctuation components at different frequencies. Variational Mode Decomposition (VMD) is used to decompose the output sequence corresponding to the coupling correlation characteristics. Initially, the number of modes is set to 5 (corresponding to different frequency fluctuation types), the penalty factor is 2000, and the noise tolerance is 10⁻. 7 By iteratively updating the center frequency and bandwidth of each mode, the original power output sequence is decomposed into five modal components with different frequencies, corresponding to: high-frequency fluctuation components (frequency 0.5-1Hz, corresponding to minute-level power output fluctuations, such as wind power fluctuations caused by instantaneous gusts), second-high frequency fluctuation components (frequency 0.1-0.5Hz, corresponding to hourly power output fluctuations, such as photovoltaic fluctuations caused by cloud cover), mid-frequency fluctuation components (frequency 0.02-0.1Hz, corresponding to daytime power output fluctuations, such as photovoltaic power output changes caused by changes in solar radiation intensity), second-low frequency fluctuation components (frequency 0.008-0.02Hz, corresponding to cross-day power output fluctuations, such as wind power output changes caused by the movement of weather systems), and low-frequency fluctuation components (frequency below 0.008Hz, corresponding to cross-week power output fluctuations, such as power output trend changes caused by seasonal changes). Each modal component corresponds to a set of power output fluctuation characteristic parameters, including fluctuation amplitude, fluctuation frequency, fluctuation duration, and fluctuation peak value.

[0026] The correlation strength between each power output fluctuation component and the first type of feature is established through quantitative analysis. This correlation strength is used as the first-level weight to perform preliminary weighted adjustment of each power output fluctuation component. The correlation strength is calculated using a combination of Pearson correlation coefficient and grey relational analysis: for high-frequency fluctuation components, the Pearson correlation coefficients with the instantaneous frequency of wind speed and the abrupt change in precipitation among the first type of features are calculated; the larger the absolute value of the correlation coefficient, the higher the correlation strength. For low-frequency fluctuation components, the grey relational degree with the mean temperature and the mean sunshine duration is calculated; the larger the relational degree value, the higher the correlation strength. The calculated correlation strength values ​​are normalized (mapped to the [0,1] interval) and used as the first-level weight. Subsequently, the feature parameters of each power output fluctuation component are multiplied by their corresponding weights to complete the preliminary weighted adjustment. For example, if the correlation strength between a high-frequency fluctuation component and the instantaneous frequency of wind speed is 0.85, its fluctuation amplitude parameter is multiplied by 0.85 to achieve optimization of the fluctuation components based on the correlation degree of meteorological features.

[0027] Based on the correlation between the impact of each output fluctuation component and the long-term evolution trend characteristics on the prediction of new energy output after preliminary weighting adjustment, the cross-category fusion weight parameters of the two are determined, and the collaborative fusion operation is completed to form a new energy output prediction model. First, the correlation is calculated using the random forest algorithm. Using the pre-weighted output fluctuation components and the long-term evolution trend characteristics as input variables, and the actual output value of new energy as the output variable, a random forest model (100 decision trees, maximum depth 15) is constructed. The correlation between the impact of each fluctuation component feature and the trend feature on output prediction is determined by scoring the feature importance of each input variable during model training. The correlation of the impact of similar features (fluctuation component class, trend feature class) is summed and normalized to obtain the class weight; then, the correlation of the impact of each feature within the same class is normalized to obtain the intra-class feature weight. Finally, a two-level fusion weight parameter of "class weight × intra-class feature weight" is formed. Based on this weight parameter, the collaborative fusion of each output fluctuation component and the long-term evolution trend characteristics is completed by weighted summation to generate the final fusion feature vector. The fused feature vector is input into a fully connected neural network (containing two hidden layers with 64 and 32 neurons respectively, and the activation function is ReLU). The output layer uses a linear activation function to obtain the predicted value of new energy power output, thereby constructing a complete new energy power output prediction model.

[0028] The meteorological forecast results and new energy operation data are input into the new energy output prediction model constructed above. The model first automatically extracts the first type of features and the second type of features, and completes the calculation and fusion of coupled correlation features, time series synergy features and long-term evolution trend features. Then, through multi-scale modal decomposition and weighted fusion operation, it outputs the hourly output prediction value of the new energy power station in the period to be predicted. At the same time, it outputs the confidence interval (confidence level 90%) of the output prediction at each time and the error analysis report (including the historical prediction error mean, maximum error, error distribution standard deviation, etc.), which provides accurate new energy output data support for the subsequent generation of peak shaving or peak demand parameters.

[0029] Step 130: Obtain the power and heat load data within the power supply and heating coverage area corresponding to the cogeneration unit, classify and analyze the characteristics of the power and heat load data, and process it in combination with meteorological forecast results and the operating status of the thermal system to obtain the power and heat load demand forecast results.

[0030] The electricity and heat load data within the power supply and heating coverage area of ​​the combined heat and power (CHP) units come from sources including the load monitoring system of the regional power grid dispatch center, the heating metering platform of the heating company, and data collected by smart meters and heat meters on the user side. Data types include hourly total electricity load, total heat load, and sub-load data broken down by user type (e.g., residential electricity load, industrial enterprise heat load, commercial electricity and heat load). Historical electricity and heat load data for the same period over the past three years, holiday load data, and load data under special weather conditions (e.g., extreme high temperatures, cold waves) are also collected to form a complete electricity and heat load dataset. The acquired electricity and heat load data undergoes preprocessing to remove outliers caused by metering equipment failures or data transmission interruptions (e.g., load values ​​suddenly dropping to 0 or exceeding 150% of the historical maximum for the same period). A weighted average method is used to fill in short-term missing data, and data smoothing processing (e.g., 5-point moving average) is used to eliminate high-frequency interference signals, ensuring the continuity and reliability of the load data.

[0031] The preprocessed electricity and heat load data were divided into three categories based on electricity consumption scenarios: residential load data, industrial production load data, and commercial service load data. Residential load data covers electricity and heat consumption data for lighting, appliances, and heating (cooling) of residential users, filtered based on the user's address area (e.g., residential community, urban village) and load curve characteristics (e.g., lower daytime load, peak load between 18:00-22:00). Industrial production load data includes electricity consumption of production equipment and heat consumption for process heating in factory workshops, categorized according to user industry type (e.g., manufacturing, heavy industry) and load characteristics (e.g., stable load, fluctuation with production shifts, no significant day-night peak-valley difference). Commercial service load data includes air conditioning, lighting, and equipment operation load data for shopping malls, office buildings, hotels, etc., differentiated according to user business type and load characteristics (e.g., higher daytime load between 9:00-21:00, higher weekend load than weekday load). After classification, based on historical load data from the same period over the past three years, the K-means clustering algorithm was used to identify typical daily curves for each scenario. For residential load, clustering yields typical daily curves for weekdays (double peaks from 6:00-8:00 AM and 6:00-10:00 PM), typical daily curves for weekends (double peaks from 10:00-12:00 PM and 7:00-11:00 PM), and typical daily curves for holidays (extended peak load periods). For industrial production load, clustering yields typical daily curves for three-shift production (stable load with fluctuations less than 5%) and typical daily curves for two-shift production (double peaks from 8:00-4:00 PM and 8:00-4:00 AM the next day). For commercial service load, clustering yields typical daily curves for weekdays (double peaks from 12:00-2:00 PM and 6:00-8:00 PM) and typical daily curves for weekends (high load throughout the day with gentle fluctuations), forming a library of typical daily curves for each scenario.

[0032] Meteorological influencing factors were extracted for each scenario based on different electricity demands. For residential load, core meteorological influencing factors included average daily temperature (heating / cooling load increases significantly when temperature is below 10℃ or above 28℃), daily temperature range (load fluctuation increases when temperature range is greater than 10℃), sunshine duration (lighting load increases when sunshine is less than 4 hours), and wind speed (building heat dissipation increases, leading to increased heating load when wind speed is greater than 5m / s). Pearson correlation analysis was used to calculate the correlation coefficient between each factor and residential load, and factors with an absolute correlation coefficient greater than 0.6 were included. For industrial production load, meteorological influencing factors mainly included extreme temperature (equipment heat dissipation / insulation causes load changes when temperature is below 0℃ or above 35℃), precipitation (outdoor production process adjustments lead to load decrease when daily precipitation is greater than 20mm), and humidity (some process loads increase when humidity is greater than 85%). Grey relational analysis was used to determine the correlation degree between each factor and industrial load, and factors with a correlation degree greater than 0.7 were selected as core influencing factors. For commercial service load, meteorological influencing factors include daytime average temperature (air conditioning load is lowest when the temperature is between 20℃ and 25℃; the load increases when it deviates from this range), solar radiation intensity (the cooling load of shopping malls increases when the solar radiation intensity is greater than 800W / ㎡), and wind speed (the building air conditioning load decreases when the wind speed is greater than 3m / s). The 3-5 factors with the highest explanatory power for commercial load are extracted through partial least squares regression analysis to form a set of meteorological influencing factors specific to each scenario.

[0033] A thermal system operation status matrix was constructed based on parameters such as the supply and return water temperature difference in the pipeline network, the power of the circulating water pump, and the energy charging and discharging status of the thermal storage device. First, the matrix dimensions were determined: rows represented key monitoring nodes of the thermal system (e.g., heat source plant outlet, intermediate heat exchange stations in the pipeline network, and user inlet, totaling 12 nodes); columns represented the types of operating parameters (supply and return water temperature difference, circulating water pump power, energy charging and discharging power of the thermal storage device, and pipeline pressure loss, totaling 4 parameters); and matrix elements represented the real-time monitoring values ​​of each parameter at each node (collected every 15 minutes). Based on this operation status matrix, principal component analysis (PCA) was used to extract core operating features reflecting the system's heating capacity. After standardizing the matrix data, the covariance matrix was calculated, and the eigenvalues ​​and eigenvectors were solved. Principal components with eigenvalues ​​greater than 1 were selected (a total of 3 principal components were extracted, with a cumulative variance contribution rate exceeding 85%). The first principal component consists of the temperature difference between the supply and return water in the pipeline network and the energy charging and discharging power of the thermal storage device, reflecting the system's heat transfer capacity; the second principal component consists of the power of the circulating water pump and the pressure loss in the pipeline network, reflecting the system's power transmission efficiency; and the third principal component consists of the energy charging and discharging state of the thermal storage device (such as the heat storage rate and the duration of energy charging and discharging), reflecting the system's heat regulation capacity. The eigenvectors corresponding to these three principal components are used as the core operating characteristic parameters.

[0034] Based on typical daily curves, meteorological influencing factors, and core operational characteristics for each scenario, a three-dimensional load-meteorological system mapping model is constructed to output preliminary load forecasts for each scenario. For residential load, the model input consists of selected typical daily curves (retrieved from a curve library based on the forecast date type, such as weekday / weekend / holiday), meteorological influencing factors (predicted values ​​for daily temperature, temperature difference, and sunshine duration), and core operational characteristics (heat transport capacity parameters of the thermal system). A backpropagation (BP) neural network is used to construct the mapping relationship: the input layer has 12 nodes (load values ​​for 8 time periods of the typical daily curve + 3 meteorological influencing factors + 1 core operational characteristic), the hidden layer has 2 layers (24 and 12 neurons respectively), and the output layer has 24 nodes (corresponding to hourly load values ​​for the 24 hours of the forecast day). Using the actual residential load values ​​and input parameters from the past year as training data, the model parameters are optimized using gradient descent until the prediction error (root mean square error) is less than 5%. For industrial production load, support vector regression (SVR) is used. A three-dimensional mapping model is constructed, with the inputs being a typical daily curve (selected according to production shifts), meteorological influencing factors (extreme temperature, precipitation, etc.), and core operating characteristics (system power transmission efficiency parameters). The kernel function is RBF (radial basis function), and the penalty parameters C=10 and gamma=0.1 are determined through cross-validation. The output is the predicted 24-hour industrial load value. For commercial service load, a Long Short-Term Memory (LSTM) network is used to construct the model. The inputs are a typical daily curve, meteorological influencing factors (daytime temperature, solar radiation intensity, etc.), and core operating characteristics (system heat regulation capacity parameters). One LSTM layer (32 neurons) and one fully connected layer (24 neurons) are set. The model is trained with time-series data to capture the dynamic relationship between load and time, meteorology, and system state, and outputs a preliminary predicted value of commercial load.

[0035] The weight of each scenario is determined based on its actual electricity consumption percentage. Combined with the preliminary load forecast for each scenario, the predicted electricity and heat load demand for each scenario is obtained. First, the actual electricity consumption percentage for each scenario is calculated: the average of the total electricity and heat load within the power and heating coverage area over the past three months and the average load for each scenario category are calculated. Residential load percentage = Residential average load / Total average load; Industrial production load percentage = Industrial average load / Total average load; Commercial service load percentage = Commercial average load / Total average load. If a scenario exhibits seasonal fluctuations (e.g., a higher residential heating load percentage in winter and a lower percentage in summer), the average percentage for the same month and year over the past three years is used for correction to ensure the weights reflect seasonal characteristics. For example, in winter, the residential load percentage is 45%, industrial production load percentage is 35%, and commercial service load percentage is 20%; in summer, the residential load percentage is 30%, industrial production load percentage is 40%, and commercial service load percentage is 30%. After determining the weights, the preliminary load forecast for each scenario is multiplied by its corresponding weight to obtain the weighted load value for each scenario. Subsequently, the weighted load value is time-series calibrated, and fine-tuned based on special events of the day (such as industrial enterprise maintenance plans or commercial promotional activities) (e.g., industrial load is reduced by 10%-20% on industrial maintenance days, and commercial load is increased by 5%-15% on commercial promotion days). Finally, the electricity and heat load demand forecast results for each scenario are output, including hourly electricity and heat load forecasts for the three major scenarios of residential life, industrial production, and commercial services, as well as the load fluctuation range for each scenario (e.g., a forecast deviation range of ±8%).

[0036] Step 140: Based on the forecast results of new energy output and the forecast results of electricity and heat load demand, generate peak shaving or peak demand parameters.

[0037] Based on the predicted renewable energy output determined in step 120 and the predicted electricity and heat load demand determined in step 130, the difference between the two at the same time dimension is calculated to obtain the electricity supply and demand gap data. Specifically, the time interval is divided into hourly units. Within each hour, four time nodes (one node every 15 minutes) are selected to calculate the supply and demand difference. The average of the differences at the four nodes is then taken as the electricity supply and demand gap value for that hour. The calculation formula is: Hourly electricity supply and demand gap value = Average predicted electricity and heat load demand for that hour - Average predicted renewable energy output for that hour. If the calculation result is positive, it means that the electricity supply for that hour cannot meet the load demand, indicating a supply and demand shortage. If the result is negative, it means that the renewable energy output for that hour exceeds the load demand, indicating a supply and demand surplus. If the result is close to zero (e.g., the absolute value is less than 2% of the total load), it is considered that the supply and demand are basically balanced. Meanwhile, for specific time periods (such as peak electricity consumption from 8:00-10:00 AM and 6:00-8:00 PM, and peak photovoltaic output from 12:00-2:00 PM), it is necessary to increase the density of time nodes (one node every 5 minutes) and refine the details of changes in the supply-demand gap. For example, during the period from 6:00-7:00 PM, the superimposed effect of a rapid decline in renewable energy output and a rapid increase in residential load should be captured to accurately calculate the dynamic change process of the gap value, forming an hourly electricity supply-demand gap dataset that includes detailed data for specific time periods. Each data entry should be labeled with the corresponding timestamp, load demand value, renewable energy output value, and gap calculation result.

[0038] The obtained electricity supply and demand gap data are statistically analyzed in segments according to the time dimension to identify the supply and demand surplus or shortage status in each period. First, the day is divided into 6 periods: 0:00-6:00 AM (off-peak period), 6:00-9:00 AM (morning peak preparation period), 9:00-12:00 PM (morning flat period), 12:00-3:00 PM (midday renewable energy peak period), 3:00-6:00 PM (afternoon flat period), and 6:00-12:00 AM (evening peak period). The supply and demand gap value of all hours in each period is statistically analyzed, and the average, maximum, minimum and standard deviation of the gap value in the period are calculated to determine the overall supply and demand status of the period. For example, if the hourly supply-demand gap values ​​are all negative during the period from 12:00 to 15:00, with an average of -120MW (negative values ​​represent surplus), a maximum of -80MW, a minimum of -150MW, and a standard deviation of 25MW, then this period is determined to be a "stable surplus period"; if the hourly supply-demand gap values ​​are all positive during the period from 18:00 to 24:00, with an average of 80MW, a maximum of 120MW (occurring at 19:00), a minimum of 50MW, and a standard deviation of 30MW, then this period is determined to be a "perpetual shortage period". Meanwhile, by combining the segmented statistical results of several consecutive days, the changing trends of supply and demand are identified. For example, if the shortage value during the evening peak period increases day by day for three consecutive days (80MW on the first day, 95MW on the second day, and 110MW on the third day), it is determined that the shortage status during that period is intensifying. If the surplus value during the midday period decreases day by day for two consecutive days (-120MW on the first day and -90MW on the second day), it is determined that the surplus status is easing. This forms a supply and demand status statistical report that includes time period division, status determination, and trend analysis.

[0039] When the supply-demand gap data for a certain period shows a shortage and the gap value exceeds a preset shortage threshold, the peak demand parameter is determined by combining the electricity and heat load growth rate for that period. First, a shortage threshold is set. This threshold is determined comprehensively based on the total installed capacity of the regional power grid, the total regulation capacity of cogeneration units, and historical load peak data. It is typically set to 8%-12% of the regional maximum load. For example, if the regional maximum load is 1000MW, the shortage threshold is set to 80MW-120MW. If the average supply-demand gap for a certain period is 100MW (exceeding the 80MW threshold), and the gap value exceeds 100MW for at least 3 hours within that period, then the peak demand determination is triggered. Next, the electricity and heat load growth rate for that period is calculated. The two-hour interval with the most significant load growth within that period (e.g., 17:00-19:00) is selected, and the load growth rate is calculated as: Load growth rate = (Load value at the end of the interval - Load value at the beginning of the interval) / Duration of the interval (2 hours). If the calculation result is 15MW / h, it means that the load increases by 15MW per hour. By combining the shortfall value and the load growth rate, peak demand parameters are determined, specifically including peak power demand (the maximum power that needs to be supplemented), peak duration (the shortest time to maintain peak output), and peak response speed (the maximum allowable time to increase output from the current level to peak output). For example, if the average shortfall value for a certain period is 100MW, the maximum shortfall value is 120MW, the load growth rate is 15MW / h, and the period lasts for 4 hours, then the peak power demand is set to 120MW (covering the maximum shortfall), the peak duration is set to 4 hours (covering the entire shortage period), and the peak response speed is set to 30 minutes (ensuring that the output increase is completed before the rapid load growth). The start time (e.g., 17:30), end time (e.g., 21:30), and changes in peak power demand at each time point are also marked (e.g., 100MW at 18:00 and 120MW at 19:00), forming a complete peak demand parameter table.

[0040] When the supply-demand gap data for a certain period shows a surplus and the surplus value exceeds a preset surplus threshold, the peak-shaving demand parameters are determined by combining the fluctuation range of renewable energy output during that period. First, a surplus threshold is set, similar to the shortage threshold, based on the regional power grid absorption capacity, renewable energy curtailment rate control targets, and the load reduction capacity of cogeneration units. It is usually 10%-15% of the region's maximum renewable energy output. For example, if the region's maximum renewable energy output is 500MW, then the surplus threshold is set to 50MW-75MW (the surplus value is calculated in absolute value, such as -60MW, where the absolute value of the surplus is 60MW). If the average supply-demand gap for a certain period is -70MW (absolute value of 70MW, exceeding the 50MW threshold), and the absolute value of the surplus exceeds 70MW for at least 2 hours during that period, then the peak-shaving demand determination is triggered. Subsequently, the fluctuation range of renewable energy output during this period is calculated. The most significant fluctuation occurs within a consecutive one-hour interval, and the fluctuation range is calculated as follows: Fluctuation range = Maximum renewable energy output within this interval - Minimum renewable energy output within this interval. If the calculated result is 30MW, it indicates that the renewable energy output fluctuation within this interval reaches 30MW. Combining the surplus value and the fluctuation range, peak-shaving demand parameters are determined, specifically including peak-shaving power demand (the maximum power to be reduced or transferred), peak-shaving duration (the shortest time to maintain peak-shaving status), and peak-shaving response sensitivity (the adjustment frequency for tracking renewable energy output fluctuations). For example, if the average absolute value of surplus during a certain period is 70MW, the absolute value of maximum surplus is 90MW, the fluctuation range of renewable energy output is 30MW, and the period lasts for 3 hours, then the peak-shaving power demand is set to 90MW (covering the maximum surplus), the peak-shaving duration is set to 3 hours (covering the entire surplus period), and the peak-shaving response sensitivity is set to be adjusted once every 5 minutes (to match the high-frequency fluctuations of renewable energy output). At the same time, the changes in peak-shaving power demand for each hour during the peak-shaving period are marked (e.g., 70MW peak-shaving is required at 12:00 noon, 90MW at 13:00 noon, and 80MW at 14:00 noon) and the corresponding renewable energy output fluctuation range (e.g., the output fluctuation range at 13:00 noon is 420MW-450MW). This forms a peak-shaving demand parameter table that includes peak-shaving power, duration, sensitivity, and dynamic adjustment requirements, providing a clear basis for the subsequent formulation of peak-shaving schemes for cogeneration units.

[0041] Step 150: Obtain the real-time operating parameters of the cogeneration unit, and determine the peak shaving and peak load potential of the cogeneration unit based on the real-time operating parameters, the configuration parameters of various heat and power decoupling methods, and the application scenarios.

[0042] In step 150, three types of key data need to be acquired. The first type is the real-time operating parameters of the cogeneration unit. The data comes from the unit's DCS (Distributed Control System) and SIS (Plant-level Monitoring Information System). Specifically, it includes core operating indicators such as the unit's current power generation, heating load, main steam pressure, main steam temperature, condenser vacuum, feedwater temperature, flue gas temperature, and furnace negative pressure. The acquisition frequency is once every 10 seconds to ensure that the unit's operating status can be reflected in real time. The second category consists of configuration parameters for various thermoelectric decoupling methods. These parameters cover the rated capacity (e.g., rated power of electric boiler 20MW, rated heat storage of heat storage tank 100GJ) of the decoupling device types currently equipped in the unit (e.g., electric boiler 0.12kWh / kWh, heat storage tank 0.5% / h), operation energy consumption data (e.g., electric boiler unit heat production power consumption 0.12kWh / kWh, heat storage tank heat dissipation loss rate 0.5% / h), input response time (e.g., heat storage tank charging and discharging response time 3 minutes, low-pressure cylinder cut-off device response time 15 minutes), and operation adjustment range (e.g., heat pump heating load adjustment range 20%-100%). These parameters need to be calibrated in conjunction with the equipment manufacturer's manual and actual operation test data. The third category is application scenario-related data, including current grid dispatch requirements (such as peak-shaving response priority and peak output command threshold), heat user needs (such as minimum water supply temperature for residential heating and stable pressure range for industrial heating), and time period type (such as weekday / weekend, heating season / non-heating season), which are used to match heat and power decoupling combination schemes that meet the needs of the scenario.

[0043] Based on the real-time operating parameters, including the unit's current power generation, heating load, main steam pressure, and condenser vacuum, the current operating boundary conditions of the cogeneration unit are determined. First, a quantitative evaluation model of the operating boundary conditions is established, using the current power generation and heating load as the core dimensions, combined with the constraints of main steam pressure and condenser vacuum, to clarify the safe operating range of the unit under the current conditions. Specifically, the current power generation must match the allowable range of the main steam pressure. If the current main steam pressure is 16 MPa (rated value 18 MPa), then according to the unit's variable operating condition curve, the upper limit of the current power generation must be controlled within 90% of the rated power generation (e.g., 270 MW for a rated power of 300 MW), and the lower limit must meet the power generation corresponding to the boiler's minimum stable combustion load (e.g., 30% of the rated power, i.e., 90 MW). The current heating load needs to be determined in conjunction with the condenser vacuum. If the current condenser vacuum is 92 kPa (design value 95 kPa), a decrease in vacuum will lead to an increase in the turbine exhaust enthalpy, requiring a corresponding adjustment to the extraction steam volume for heating. In this case, the upper limit of the heating load needs to be reduced from the rated heating load of 120 GJ / h to 110 GJ / h, while the lower limit remains at 40 GJ / h to meet the minimum user demand (e.g., minimum residential heating demand). Simultaneously, other real-time parameters (such as feedwater temperature and flue gas temperature) need to be considered to correct the boundary conditions: if the feedwater temperature is 5°C lower than the design value, the boiler thermal efficiency will decrease, requiring a further 5% reduction in the upper limit of power generation (from 270 MW to 256.5 MW). The final result is a current operating boundary curve diagram with "power generation - heating load" as the coordinates, marking the four vertices of the safe operating area (maximum power generation - minimum heating load, maximum power generation - maximum heating load, minimum power generation - minimum heating load, minimum power generation - maximum heating load), as well as the constraint lines corresponding to the main steam pressure and condenser vacuum, thus clarifying the adjustable power generation and heating load range of the unit under the current state.

[0044] Based on the current operating boundary conditions and combined with the rated capacity and operating energy consumption data of the decoupling device in the configuration parameters, the power generation adjustment range of each thermoelectric decoupling method when individually and in combination is calculated. First, for a single thermoelectric decoupling method, its ability to expand the power generation adjustment range is calculated. Taking an electric boiler as an example, its rated capacity is 20MW (electric energy converted into heat energy). When the electric boiler is put into operation alone, the unit can transfer the steam originally used for heating to power generation. According to the thermoelectric conversion formula (power generation increment = steam transfer amount × turbine relative internal efficiency × generator efficiency), if the steam corresponding to 20MW of electricity consumed per hour by the electric boiler can increase the power generation by 15MW, and considering the operating energy consumption loss of the electric boiler (e.g., 90% efficiency, actually requiring 22.2MW of electricity for steam extraction), the actual power generation upper limit can be increased by 13.5MW from the current boundary upper limit (e.g., 256.5MW), reaching 270MW. The lower limit of power generation remains unchanged at 90MW, unaffected by the operation of the electric boiler. Therefore, the power generation adjustment range when the electric boiler is operated alone is 90MW-270MW. Taking a thermal storage tank as another example (rated heat storage capacity of 100GJ, heat release time of 5 hours), when the thermal storage tank is operated alone for heat release, the unit can reduce the amount of steam extracted for heating to increase power generation. Based on the transfer of steam extraction corresponding to the heat release rate of the thermal storage tank (20GJ / h), the upper limit of power generation can be increased by 12MW (considering a heat dissipation loss rate of 0.5% / h, the actual increase is 11.94MW), and the power generation adjustment range expands to 90MW-268.44MW. If the thermal storage tank is in a charging state, the amount of steam extracted for heating needs to be increased, causing the lower limit of power generation to drop to 85MW, and the adjustment range becomes 85MW-256.5MW. For combined investment scenarios, common combinations such as "electric boiler + thermal storage tank" and "low-pressure cylinder cutoff + heat pump" are selected to calculate the synergistic effect of combined investment. For example, the combined investment of "electric boiler (increases the upper limit by 13.5MW) + thermal storage tank heat release (increases the upper limit by 11.94MW)" considers the superposition of energy consumption during the operation of both (total loss rate increases by 1%), the actual upper limit of power generation is increased by 24.2MW, and the adjustment range is expanded to 90MW-280.7MW; the combined investment of "low-pressure cylinder cutoff (reduces the lower limit to 60MW) + heat pump (increases the upper limit by 8MW)" changes the power generation adjustment range to 60MW-264.5MW. In the calculation process, it is necessary to establish the correlation formula between the energy consumption loss of each decoupling method and the power generation adjustment amount to ensure that each adjustment range expansion has a clear quantitative basis. Finally, a comparison table of power generation adjustment ranges corresponding to single and combined thermoelectric decoupling methods is formed, marking the upper and lower limits of adjustment and energy consumption costs of each method.

[0045] Based on the input response time and corresponding power generation adjustment range in the configuration parameters of each thermoelectric decoupling method, feasible thermoelectric decoupling combination schemes for each thermoelectric decoupling method in corresponding scenarios are determined. First, the time periods are classified according to the application scenarios, into four categories: "Emergency peak shaving scenario of power grid" (requiring response time ≤ 10 minutes and peak shaving depth ≥ 30%), "Conventional peak shaving scenario" (response time ≤ 30 minutes and peak shaving depth 10%-30%), "Winter peak heating scenario" (requiring stable heating load and peak response time ≤ 15 minutes), and "Flexible adjustment scenario in non-heating season" (relaxed heating constraints and large adjustment range requirements). For each scenario, first, select the thermoelectric decoupling method that meets the response time requirements: For example, in the "emergency peak shaving scenario of the power grid", the decoupling methods with a response time of ≤10 minutes include thermal storage tanks (3 minutes) and electric boilers (5 minutes), excluding low-pressure cylinder cut-off devices with a response time of 15 minutes and heat pumps with a response time of 20 minutes; then, combine the power generation adjustment range requirements in this scenario (e.g., peak shaving depth of 30%, the current power generation of 200MW needs to be reduced to 140MW), and combine feasible solutions from the selected decoupling methods. "Temperature storage tank charging (single operation can reduce the lower limit to 85MW, meeting the 140MW requirement)" and "electric boiler + thermal storage tank charging (reducing the lower limit to 80MW, with a larger adjustment margin)" are both feasible solutions. For the "peak winter heating scenario," priority must be given to ensuring stable heating load (e.g., water supply temperature ≥ 80℃). Therefore, decoupling methods that can maintain heating (e.g., heat pumps, heat storage tanks releasing heat to avoid reduced steam extraction due to low-pressure cylinder cutoff) are selected. Considering the requirement of peak response time ≤ 15 minutes, a feasible combination is "heat pump (response time 8 minutes, maximum boost 8MW) + heat storage tank releasing heat (response time 3 minutes, maximum boost 11.94MW)". This solution can complete the response within 15 minutes and supplement heat through the heat pump, ensuring stable heating. When determining feasible solutions, a matching matrix between scenario requirements (response time, adjustment range, heating constraints) and decoupling method parameters needs to be established. A scoring method is used to evaluate each combination solution (response time compliance accounts for 40%, adjustment range compliance accounts for 30%, and heating stability accounts for 30%). Solutions with a score ≥ 80 points are listed as feasible solutions. At the same time, combinations that cause operational conflicts need to be excluded (such as the inability to charge and release heat in the heat storage tank at the same time, or the simultaneous operation of the low-pressure cylinder cut-off and the electric boiler, which would lead to excessive fluctuations in the main steam pressure). Finally, a list of feasible thermoelectric decoupling combination schemes for each scenario is formed, with each scheme marked with scenario type, response time, adjustment range, heating guarantee measures and reasons for conflict exclusion.

[0046] Based on the rated capacity of the decoupling device, operating energy consumption data, and the current heating load of the cogeneration unit for feasible heat and power decoupling schemes, the upper limit of peak-shaving power, the upper limit of peak power, and the duration of continuous adjustment for different schemes are determined, and the peak-shaving potential parameters and peak-shaving potential parameters of the cogeneration unit are summarized. First, the upper limit of peak-shaving power (the maximum total power that can be reduced during the peak-shaving period under a certain scheme) is calculated. The formula is "Upper limit of peak-shaving power = (Current power generation - Lower limit of power generation adjustment corresponding to the scheme) × Duration of continuous adjustment", where the duration of continuous adjustment is determined by the rated capacity of the decoupling device and the current heating load. For example, in the "heat storage tank charging" scheme, the rated heat storage tank has a heat storage capacity of 100 GJ, the current heating load is 80 GJ / h, and 20 GJ / h of heat needs to be extracted from the heating system during charging. Therefore, the duration of continuous charging = 100 GJ ÷ 20 GJ / h = 5 hours. The current power generation is 200MW, the lower limit of the scheme's regulation is 85MW, and the upper limit of peak power is (200-85)MW × 5h = 575MWh. The formula for calculating the upper limit of peak power (the maximum total power that a scheme can increase during the peak period) is "Upper limit of peak power = (upper limit of power generation regulation corresponding to the scheme - current power generation) × continuous regulation duration". For example, in the "electric boiler + thermal storage tank heat release" scheme, the regulation upper limit is 280.7MW, the current power generation is 200MW, and the continuous regulation duration is determined by the rated capacity of the electric boiler (20MW) and the heating load (80GJ / h). The electric boiler consumes 20MW of electricity per hour, which corresponds to a reduction of 20GJ / h of heating steam extraction. The thermal storage tank releases 20GJ / h of heat per hour to supplement the heating. Therefore, the heating load remains stable, and the continuous regulation duration is not constrained by the heating supply. It is determined by the continuous operating capacity of the equipment to be 8 hours. The upper limit of peak power is (280.7-200)MW × 8h = 645.6MWh. The calculation process needs to consider the impact of operating energy consumption loss on the duration of continuous regulation: for example, if the heat loss rate of the thermal storage tank is 0.5% / h, the heat storage capacity will decrease by 2.5 GJ after 5 hours of continuous operation, and the actual duration needs to be shortened to 4.8 hours. The upper limit of peak shaving power should be adjusted accordingly to (200-85)×4.8=552MWh. For each feasible scheme, three core parameters need to be calculated: the upper limit of peak shaving power, the upper limit of peak power, and the duration of continuous regulation. Then, they should be summarized by scenario: for example, under the "conventional peak shaving scenario", the average upper limit of peak shaving power for all feasible schemes is 520MWh, the maximum value is 575MWh, and the duration of continuous regulation is concentrated in 4-6 hours. Based on this, the peak shaving potential parameters for this scenario are determined to be "peak shaving power 450-575MWh, duration of continuous regulation 4-6 hours". Under the "peak heating scenario in winter", the feasible scheme has an average peak power limit of 600MWh and a minimum of 550MWh, with a continuous adjustment time of 5-8 hours. The peak potential parameters are "peak power 550-645.6MWh, continuous adjustment time 5-8 hours".

[0047] Step 160: Based on meteorological forecasts, new energy output forecasts, electricity and heat load demand forecasts, peak shaving or peak-shaving related demand parameters, and the peak shaving and peak-shaving potential of cogeneration units, combined with changes in electricity spot market supply and demand and fluctuations in electricity and heat load, determine the auxiliary peak shaving scheme for cogeneration units.

[0048] First, the six core basic data categories mentioned above are integrated. The first category is meteorological forecast results, which focuses on extracting information on extreme meteorological events every 15 minutes over the next 72 hours (such as the probability of short-term strong winds and the duration of cold waves) and the boundary values ​​of the forecast intervals with 90% confidence levels for each element. It focuses on the degree of interference of meteorological conditions on the stability of new energy output (such as the impact coefficient of wind power output when wind speed fluctuations are >5m / s) and the correlation between heat loss in the thermal system (such as the increase in pipeline heat loss when the temperature is <-5℃), avoiding duplication of basic statistical dimensions of existing meteorological data. The second category is new energy output forecast results, which only retains key parameters that are highly adaptable to peak-shaving schemes, including the hourly ramp-up threshold for photovoltaic / wind power output (such as the percentage of time periods where wind power output changes by more than 20MW every 15 minutes), the overlap duration of stable output windows (such as the number of hours overlapping with midday load troughs), and the extreme values ​​of forecast errors under the same meteorological conditions over the past month (such as the maximum negative deviation of 30MW for wind power), omitting the repetitive listing of conventional output values. The third category is the forecast results of electric heating load, highlighting the dynamic response characteristics of loads in various scenarios, such as the sensitivity curve of residential heating load to temperature changes (the load increase gradient for every 1°C drop in temperature), and the recovery time after sudden changes in industrial load (such as the time required for stabilization after a sudden increase of 10MW). Only the standard deviation of load fluctuations in the same period of the past 3 days (such as ±3% for commercial load) is retained to replace long-term historical data. The fourth category is peak-shaving or peak demand parameters, simplified to the time-specific characteristic value of the supply-demand gap (such as an average gap of 80MW during the evening peak from 18:00 to 22:00), the classification of response time levels (emergency ≤10 minutes, normal ≤30 minutes), and the minimum threshold for continuous adjustment duration (such as peak not less than 4 hours), without the need to list the gap value hourly. The fifth category is data on unit peak shaving and peak potential. Classified by the function of the heat-power decoupling combination scheme (e.g., "fast response type" and "large capacity regulation type"), core performance indicators for each scheme are extracted, such as the maximum instantaneous regulation capacity of 120MW for "heat pump + thermal storage tank," the continuous regulation limit of 5 hours for "low-pressure cylinder cutoff + electric boiler," and the energy consumption loss comparison coefficients of different schemes (e.g., the heat dissipation loss of the thermal storage tank is 1.2 times that of the electric boiler), omitting basic parameters of individual units. The sixth category is electricity spot market data, focusing on information strongly related to the economics of the schemes, including the deviation of the current marginal electricity price from the historical same period (e.g., 15% higher), the electricity price fluctuation coefficient corresponding to the supply and demand tension level (20% increase when the supply is red), and the turning point of the market trend in the next 2 hours (e.g., the electricity price is expected to decrease after 14:00), without needing to fully list past clearing cycle data. The integrated data is calibrated with minute-level timestamps to ensure that the time nodes of each data dimension are fully aligned. The 3σ criterion is used to remove extreme outliers (such as load mutation data that exceeds the normal range by 3 times). Short-term (<1 hour) missing data is filled in by linear interpolation, and finally a concise dataset focused on solution decision-making is formed.

[0049] Based on the integrated data, a decision-making matrix for peak-shaving schemes is constructed using a "time period-decision dimension" approach. The row dimension is divided into 72 time periods per hour, and the column dimension includes five core dimensions: meteorological impact, renewable energy stability, load urgency, market trends, and potential matching degree. Each dimension is quantitatively scored on a 10-point scale (higher scores indicate stronger constraints). Meteorological impact is scored based on the probability of extreme weather events; for example, a forecast of strong winds (10 m / s) causing large wind power fluctuations scores 8 points. Renewable energy stability is assessed based on output deviation and historical accuracy; for example, a photovoltaic deviation of ±5% and an accuracy rate of 90% scores 3 points. Load urgency combines peak periods and real-time fluctuations; for example, a residential evening peak with load fluctuations of ±5% scores 9 points. Market trends are scored based on marginal electricity prices and supply-demand levels; for example, rising electricity prices and a yellow-level tight supply-demand situation score 7 points. Potential matching degree assesses the degree to which the decoupled scheme meets demand; for example, a scheme that can provide 80MW of peak shaving capacity when 60MW is required scores 4 points. Calculate the comprehensive score for each time period (weight: meteorology 15%, new energy 20%, load 25%, market 20%, potential 20%), and divide them into three priority adjustment periods: high (≥8 points), medium (5-7 points), and low (<5 points) to clarify the focus of the plan formulation.

[0050] For different priority periods, suitable solutions were selected and details were optimized: For high priority periods (such as the evening peak, with a comprehensive score of 8.5), demand parameters were first matched (100MW peak, 4 hours of continuous operation, 15 minutes of response). From the potential data, two candidate solutions were selected: "heat pump + thermal storage tank" (120MW, 5 hours, 10 minutes) and "electric boiler + low-pressure cylinder cutoff" (110MW, 4.5 hours, 12 minutes). Combining market data (marginal electricity price of 1.2 yuan / kWh), the economic efficiency was calculated. The former has an hourly energy cost of 25,000 yuan, which is lower than the latter's 30,000 yuan. It also meets the requirements of real-time load fluctuations (residential heating ±2%) for stable heating. Finally, this solution was selected, with 80MW of heat pump and 40MW of thermal storage tank, and 20MW reserved to cope with sudden load changes. For medium-priority periods (e.g., midday peak shaving, overall score 6), the demand is 80MW peak shaving, 3 hours of continuous operation, and 30-minute response time. Candidate options are "heat storage tank charging" (90MW, 3.5 hours, 20 minutes) and "low-pressure cylinder cutoff" (85MW, 3 hours, 25 minutes). Considering the market electricity price of 0.8 yuan / kWh and industrial load fluctuations of ±1%, the former (18,000 yuan / hour) has a lower cost and less impact on heating, so this option is selected and implemented at 80MW charging power, with adjustments made every 30 minutes based on load. For low-priority periods (e.g., early morning off-peak, overall score 3), the demand is 30MW peak shaving, 5 hours of continuous operation. A single "electric boiler" option (40MW, 6 hours, 40 minutes) can meet this. Considering the low market price (0.6 yuan / kWh) and residential load fluctuations of ±1%, a base power of 30MW is implemented, with 10MW reserved for emergencies.

[0051] After the initial formulation of the auxiliary peak-shaving scheme is completed, to ensure the scheme is accurately adapted to the actual operating conditions, it is necessary to trace the root causes of the differences between the forecast and the actual situation through deviation analysis. The final clearing data of the electricity spot market objectively reflects the true balance of power grid supply and demand and is the core reference for deviation quantification. Therefore, it is necessary to build a benchmark framework for deviation quantification analysis based on the final clearing data of the electricity spot market. Among them, the actual supply and demand gap and marginal electricity price fluctuations are the core indicators reflecting the true operating state of the market. The degree of deviation between the forecast and the actual value can be measured by analyzing the correlation between the two and various forecast parameters. The inherent mapping relationship between supply and demand tension and deviation is first extracted from the clearing data. For example, when the marginal electricity price exceeds the forecast value by 10%, the deviation of the renewable energy output forecast is often accompanied by an average increase of 5%. Such patterns can be used as a reference benchmark for subsequent deviation quantification calculations to ensure that the deviation analysis is closely related to the actual operating characteristics of the market.

[0052] Regarding meteorological forecast results, the focus is on calculating the "deviation impact coefficient" for extreme weather events. The specific formula is (actual impact duration - predicted impact duration) / predicted impact duration × 100%. Taking short-duration strong winds as an example, if the predicted impact duration is 2 hours and the actual impact duration is 3 hours, the corresponding deviation impact coefficient is 50%. Simultaneously, the percentage of periods where this coefficient exceeds 20% is statistically analyzed to quantify the cascading impact of meteorological deviations on subsequent forecasting stages such as renewable energy output and electricity and heat load. For renewable energy output forecast results, the "output deviation contribution rate" indicator is introduced. This calculates the proportion of renewable energy output deviation in the total supply-demand gap deviation for a given period. For example, if the total supply-demand gap deviation is 30MW, and renewable energy output deviation accounts for 20MW, its contribution rate is 66.7%. Combined with the actual renewable energy absorption rate in the clearing data, if the absorption rate decreases by 3% due to this output deviation, the deviation level for that period is marked as "significant," clearly defining the actual impact of the deviation on renewable energy absorption. For the forecast results of electricity and heat load demand, the focus is on calculating the "load impact weight" based on scenario-specific deviations. This weight represents the proportion of total load deviation caused by a single scenario's load forecast deviation. For example, a 2% deviation in industrial load forecasting results in a 1.2% deviation in total load, with a weight of 60%. Particular attention is paid to situations where this weight exceeds 50% during peak periods to analyze the interference of key scenario load deviations on the overall supply and demand balance. For peak-shaving or peak-peak demand parameters, the degree of deviation is quantified using the "demand deviation rate," calculated as (predicted supply-demand gap - actual supply-demand gap) / actual supply-demand gap × 100%. If the predicted peak demand is 120MW but the actual demand is only 90MW, the deviation rate is 33.3%, reflecting the degree of alignment between demand parameters and actual market demand. For unit peak-shaving and peak-peak potential parameters, the "potential deviation rate" is calculated as (predicted regulation capacity - actual regulation capacity) / actual regulation capacity × 100%. This is further corrected by incorporating actual energy consumption data from the decoupling device. If the predicted energy loss is 5% but the actual loss reaches 8%, an energy consumption correction coefficient of 1.2 is introduced to obtain a more realistic corrected potential deviation rate. Finally, the deviation calculation results of various parameters are integrated to form a deviation quantification table, which clearly marks the deviation value of each parameter and the associated clearing supply and demand information, providing data support for subsequent deviation analysis.

[0053] Based on the quantified degree of deviation, the distribution characteristics of deviation are analyzed in depth from two dimensions: time and parameter correlation, thereby identifying the dominant deviation link. In the time dimension, the average deviation rate of various parameters is statistically analyzed for different time periods, such as morning peak, midday peak of new energy, evening peak, and trough, to identify periods of concentrated deviation. For example, it was found that the average deviation rate of new energy output reaches 15% during the midday period of 12:00-14:00, significantly higher than the 5% of other time periods. Furthermore, the proportion of light intensity prediction deviation in the total deviation during this period exceeds 70%, indicating that this period is a critical time period with high deviation incidence and dominated by meteorological factors. In the parameter correlation dimension, the deviation transmission path is identified through correlation analysis. For example, wind speed deviation in meteorological forecasts directly leads to new energy output deviation, which in turn causes supply-demand gap deviation, ultimately affecting peak-shaving demand parameters. If the correlation coefficient between wind speed prediction deviation and new energy output deviation reaches 0.85, then "meteorological forecast deviation (wind speed)" can be identified as the initial dominant link. Further investigation is conducted to pinpoint the root causes of different deviation scenarios: If the deviation rate of renewable energy output is high but the meteorological forecast deviation is small, the preprocessing process of renewable energy operation data is examined. It is found that abnormal data during equipment maintenance periods (such as low output due to wind turbine maintenance) were not removed, making "omission of maintenance records in renewable energy operation data preprocessing" the core leading factor. If the deviation rate of peak-shaving demand is high but the deviations of electric heating load and renewable energy output forecasts are small, the supply-demand gap calculation logic is analyzed. It is found that grid transmission losses (actual loss rate of 3%) were not included, making "supply-demand gap calculation not including transmission loss factors" the leading cause of deviation. If the unit potential deviation rate is high, comparing the actual operating data of the decoupling device with the forecast parameters reveals that equipment aging has led to a 10% decrease in efficiency (such as the thermal efficiency of electric boilers decreasing from 90% to 81%), making "aging of the decoupling device and failure to update parameters" the leading factor. A comprehensive deviation distribution characteristic report and a leading factor location table are generated to clarify the temporal patterns, transmission paths, and core influencing factors of the deviations.

[0054] Based on the clearly defined characteristics of the deviation distribution and the dominant deviation links, targeted optimization measures for the auxiliary peak-shaving scheme were formulated. For the dominant link of "meteorological forecast deviation (wind speed)," the meteorological forecast model was optimized by adding a terrain wind speed correction factor to the original model. For example, wind speed forecasts in mountainous areas need to be multiplied by a correction coefficient of 0.9. The wind speed data update cycle was shortened from 1 hour to 30 minutes. Furthermore, portable anemometers and other backup monitoring equipment were added during the high-deviation period at midday to improve wind speed forecast accuracy from multiple dimensions. For the link of "omitted maintenance records in new energy operation data preprocessing," the data preprocessing process was optimized by adding an equipment maintenance calendar module to automatically identify equipment maintenance periods and remove abnormal data during those periods. At the same time, the weights of the model training data were adjusted, using a weighted training of data from the past month (weight 0.7) and older data (weight 0.3) to reduce the interference of historical abnormal data on the prediction results. To address the issue of "supply-demand gap calculation not including transmission loss factors," a dynamic loss calculation module is added to the supply-demand gap calculation logic. Based on the power supply distance (e.g., 2% loss rate for 10km lines, 4% loss rate for 20km lines) and real-time load current, the transmission loss rate is updated in real time to correct the supply-demand gap calculation results, making peak-shaving or peak demand parameters more closely match the actual market supply and demand. To address the issue of "aging decoupling devices with outdated parameters," a decoupling device status monitoring model is established. By analyzing parameters such as equipment operating current and temperature (e.g., when the current of an electric boiler exceeds the rated value by 5%, efficiency is determined to have decreased), the energy consumption loss parameters of the equipment are updated in real time. At the same time, an aging correction coefficient is added to the unit potential calculation. For example, a correction coefficient of 1.1 is used for equipment that has been in operation for more than 5 years to avoid overestimation of potential due to parameter lag. Based on the characteristics of the deviation distribution, for periods with high deviations such as 12:00-14:00, the peak-shaving scheme will increase the reserve regulation capacity by 10% and adopt a rolling adjustment strategy every 15 minutes, dynamically correcting the peak-shaving instructions based on the latest actual output data. For periods with large fluctuations in marginal electricity prices, the scheme's response priority will be optimized, shortening the peak-shaving response time from 30 minutes to 15 minutes. This will be achieved by activating decoupling devices such as the thermal storage tank in advance, improving the scheme's adaptability to market changes. The optimized scheme needs to provide a detailed explanation of the optimization measures, a comparison of deviations before and after optimization (e.g., reducing the wind speed prediction deviation rate from 15% to 8%), and expected effects (e.g., improving the peak-shaving accuracy by 12%). A closed-loop feedback mechanism will also be established. After each clearing cycle, the deviation rates of various parameters will be recalculated to verify the effectiveness of the optimization measures. If there is no significant improvement in deviations, new dominant deviation factors will be further analyzed, and continuous iterative optimization will be carried out to ensure that the peak-shaving scheme is always highly matched with the actual operating conditions.

[0055] Furthermore, as a response to the above Figure 1The implementation of the method embodiment shown in this application provides an auxiliary peak-shaving device for a combined heat and power unit. The embodiment of this device corresponds to the foregoing method embodiments. For ease of reading, this embodiment will not repeat the details of the foregoing method embodiments one by one, but it should be clear that the device in this embodiment can correspondingly implement all the contents of the foregoing method embodiments. Specifically, as shown... Figure 2 As shown, the auxiliary peak-shaving device 200 of the combined heat and power unit includes: The weather forecast module 210 is used to acquire meteorological data, process and analyze the meteorological data, and obtain weather forecast results. The new energy output prediction module 220 is used to acquire new energy operation data and determine the new energy output prediction result based on meteorological forecast results and new energy operation data. The electric heating load prediction module 230 is used to acquire electric heating load data within the power supply and heating coverage area corresponding to the cogeneration unit, classify and analyze the characteristics of the electric heating load data, and process it in combination with meteorological forecast results and the operating status of the thermal system to obtain the electric heating load demand prediction results. The parameter determination module 240 is used to generate peak shaving or peak demand parameters based on the forecast results of new energy output and the forecast results of electric and heat load demand. The thermoelectric decoupling strategy module 250 is used to obtain the real-time operating parameters of the cogeneration unit and determine the peak shaving and peak load potential of the cogeneration unit based on the real-time operating parameters, the configuration parameters of various thermoelectric decoupling methods and application scenarios. The scheme generation module 260 is used to determine the auxiliary peak-shaving scheme of the cogeneration unit based on meteorological forecast results, new energy output forecast results, electric and heat load demand forecast results, peak-shaving or peak-climbing related demand parameters and the peak-shaving and peak-climbing potential of the cogeneration unit, combined with the supply and demand changes in the electricity spot market and the fluctuation of electric and heat load.

[0056] Optionally, the auxiliary peak-shaving device of the cogeneration unit can be an electronic device with data processing capabilities, or a functional module in the electronic device, without limitation.

[0057] For example, the electronic device can be a server, which can be a single server or a server cluster consisting of multiple servers. As another example, the electronic device can be a mobile phone, tablet computer, desktop computer, laptop computer, handheld computer, notebook computer, ultra-mobile personal computer (UMPC), netbook, as well as cellular phones, personal digital assistants (PDAs), augmented reality (AR) devices, virtual reality (VR) devices, and other terminal devices. As yet another example, the electronic device can also be a recording device, video surveillance equipment, etc. This application does not impose any special limitations on the specific form of the electronic device.

[0058] The following example uses electronic equipment as an auxiliary peak-shaving device in a combined heat and power (CHP) unit. Figure 3 As shown, Figure 3 The hardware structure of an electronic device 300 provided in this application.

[0059] like Figure 3 As shown, the electronic device 300 includes a processor 310, a communication line 320, and a communication interface 330.

[0060] Optionally, the electronic device 300 may also include a memory 340. The processor 310, memory 340, and communication interface 330 can be connected via a communication line 320.

[0061] The processor 310 can be a central processing unit (CPU), a general-purpose processor, a network processor (NP), a digital signal processor (DSP), a microprocessor, a microcontroller, a programmable logic device (PLD), or any combination thereof. The processor 310 can also be any other device with processing capabilities, such as a circuit, device, or software module, without limitation.

[0062] In one example, processor 310 may include one or more CPUs, for example Figure 3 CPU0 and CPU1 in the CPU.

[0063] As an optional implementation, the electronic device 300 may include multiple processors, for example, in addition to processor 310, it may also include processor 370. A communication line 320 is used to transmit information between the components included in the electronic device 300.

[0064] Communication interface 330 is used for communication with other devices or other communication networks. These other communication networks can be Ethernet, Radio Access Network (RAN), Wireless Local Area Networks (WLAN), etc. Communication interface 330 can be a module, circuit, transceiver, or any device capable of enabling communication.

[0065] The memory 340 is used to store instructions. These instructions can be computer programs.

[0066] The memory 340 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and / or instructions; it may also be a random access memory (RAM) or other type of dynamic storage device capable of storing information and / or instructions; it may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, etc., without limitation.

[0067] It should be noted that the memory 340 can exist independently of the processor 310 or can be integrated with the processor 310. The memory 340 can be used to store instructions, program code, or some data, etc. The memory 340 can be located inside or outside the electronic device 300, without restriction.

[0068] The processor 310 is configured to execute instructions stored in the memory 340 to implement the communication method provided in the following embodiments of this application. For example, when the electronic device 300 is a terminal or a chip in a terminal, the processor 310 can execute instructions stored in the memory 340 to implement the steps performed by the sending end in the following embodiments of this application.

[0069] As an optional implementation, the electronic device 300 also includes an output device 350 and an input device 360. The output device 350 can be a display screen, speaker, or other device capable of outputting data from the electronic device 300 to the user. The input device 360 ​​can be a keyboard, mouse, microphone, joystick, or other device capable of inputting data into the electronic device 300.

[0070] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the electronic device, except... Figure 3 In addition to the components shown, the electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0071] The auxiliary peak-shaving devices and application scenarios of cogeneration units described in this application are for the purpose of more clearly illustrating the technical solutions of this application, and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the evolution of auxiliary peak-shaving devices for cogeneration units and the emergence of new business scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0072] This application provides a storage medium storing a program that, when executed by a processor, implements the auxiliary peak-shaving method for combined heat and power units.

[0073] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for auxiliary peak shaving of a combined heat and power plant, characterized by, The method comprises: acquiring meteorological data, processing and analyzing the meteorological data to obtain a meteorological prediction result; acquiring new energy operation data, determining a new energy output prediction result based on the meteorological prediction result and the new energy operation data; acquiring electric-thermal load data in a power and heat supply coverage range corresponding to a cogeneration unit, classifying and analyzing the electric-thermal load data, processing in combination with the meteorological prediction result and a thermal system operation state to obtain an electric-thermal load demand prediction result; generating a peak shaving or peak demand parameter according to the new energy output prediction result and the electric-thermal load demand prediction result; acquiring real-time operation parameters of the cogeneration unit, determining a peak shaving and peak potential of the cogeneration unit according to the real-time operation parameters, configuration parameters and application scenarios of multiple thermal and electric decoupling modes; determining an auxiliary peak shaving scheme of the cogeneration unit in combination with power spot market supply and demand changes and electric-thermal load fluctuation conditions according to the meteorological prediction result, the new energy output prediction result, the electric-thermal load demand prediction result, the peak shaving or peak related demand parameter and the peak shaving and peak potential of the cogeneration unit.

2. The method of claim 1, wherein, Processing and analyzing the meteorological data to obtain a meteorological prediction result comprises: performing multi-scale decomposition on the meteorological data through wavelet transform to extract statistical domain features and time-frequency domain features; inputting the statistical domain features and the time-frequency domain features into an attention model based on a multi-layer perception machine to calculate contribution weights of each feature to the meteorological prediction result, and screening features with contribution weights meeting preset conditions to form a meteorological feature set; constructing a multi-modal meteorological prediction model comprising a spatial correlation feature extraction branch, a time sequence correlation feature extraction branch and a feature fusion layer based on the meteorological feature set, the spatial correlation feature extraction branch acquires spatial correlation features of the meteorological data through feature mapping and dimension reduction, the time sequence correlation feature extraction branch acquires time sequence correlation features by mining time sequence dependency of the meteorological data, and the feature fusion layer integrates and processes the spatial correlation features and the time sequence correlation features; inputting meteorological data of a to-be-predicted period into the multi-modal meteorological prediction model to output a meteorological prediction result.

3. The method of claim 2, wherein, Determining a new energy output prediction result based on the meteorological prediction result and the new energy operation data comprises: extracting first type features and second type features, the first type features being statistical domain features and time-frequency domain features corresponding to the meteorological prediction result, the second type features being operation state features and output response features corresponding to the new energy operation data, and determining coupling correlation features between the first type features and the second type features through feature cross operation; extracting deepening features of the coupling correlation features in a time dimension as time sequence coordination features based on time sequence evolution rules of the meteorological prediction data and the new energy operation data; extracting a long-term evolution trend feature of new energy output from the time sequence coordination features through feature fusion realized by bidirectional interaction of forward and reverse time sequence coordination features. According to the coupling correlation feature, multi-scale modal decomposition processing is performed to separate different frequency output fluctuation components; The correlation strength of each output fluctuation component with the first type of feature is established through quantitative analysis, and the correlation strength is used as the first level weight to preliminarily weight and adjust each output fluctuation component; According to the influence correlation degree of each output fluctuation component after preliminary weighting adjustment and the long-term evolution trend feature on new energy output prediction, the cross-category fusion weight parameter of each output fluctuation component after preliminary weighting adjustment and the long-term evolution trend feature is determined, and the synergistic fusion operation of each output fluctuation component and the long-term evolution trend feature is completed according to the cross-category fusion weight parameter, to form a new energy output prediction model; The meteorological prediction result and the new energy operation data are input into the new energy output prediction model to obtain a new energy output prediction result.

4. The method of claim 3, wherein, The electric heating load data is classified and analyzed, and the meteorological prediction result and the thermal system operation state are processed to obtain an electric heating load demand prediction result, including: The electric heating load data is divided into residential life load data, industrial production load data and commercial service load data according to the power consumption scene, and the typical day curve of each scene is identified according to the historical load data of each scene; According to different power consumption demands of each scene, the corresponding meteorological influence factor of each scene is extracted; The core operation features reflecting the system heating capacity are extracted based on the heat supply and return water temperature difference, circulating water pump power and heat storage device charging and discharging state parameters in the thermal system operation state matrix; The load-meteorological system three-dimensional mapping model of each scene is constructed according to the typical day curve, meteorological influence factor and core operation feature of each scene to output the load preliminary prediction value of each scene; The weight of each scene is determined according to the actual power consumption proportion of each scene, and the electric heating load demand prediction result of each scene is obtained according to the weight of each scene and the load preliminary prediction value of the corresponding scene.

5. The method of claim 4, wherein, According to the new energy output prediction result and the electric heating load demand prediction result, a peak shaving or peak demand parameter is generated, including: The difference between the new energy output prediction result and the electric heating load demand prediction result is calculated to obtain power supply and demand gap data; The power supply and demand gap data is segmented and counted according to the time dimension to identify the surplus or shortage state of each period; When the supply and demand gap data of a period is in a shortage state and the gap value exceeds a preset shortage threshold, the peak demand parameter is determined in combination with the electric heating load growth rate of the period; When the supply and demand gap data of a period is in a surplus state and the surplus value exceeds a preset surplus threshold, the peak shaving demand parameter is determined in combination with the new energy output fluctuation amplitude of the period.

6. The method of claim 5, wherein, The peak shaving and peak potential of the combined heat and power unit is determined according to the real-time operation parameters, the configuration parameters and the application scenarios of multiple heat and electricity decoupling modes, including: The current operation boundary condition of the combined heat and power unit is determined according to the current power generation power, heating load, main steam pressure and condenser vacuum data in the real-time operation parameters; Based on the current operating boundary condition, the rated capacity of the decoupling device in the configuration parameter is combined with the operating energy consumption loss data to calculate the power regulation range of each heat and power decoupling mode when the heat and power decoupling mode is independently and complexly put into operation; According to the input response time in the configuration parameter of each heat and power decoupling mode and the corresponding power regulation range, a feasible heat and power decoupling combination scheme under the corresponding scene of each heat and power decoupling mode is determined; According to the rated capacity of the decoupling device, the operating energy consumption loss data of the feasible heat and power decoupling combination scheme and the current heat supply load of the combined heat and power unit, the upper limit of the peak shaving power, the upper limit of the peak power and the duration of the sustained adjustment corresponding to different schemes are determined, and the peak shaving potential parameter and the peak potential parameter of the combined heat and power unit are obtained by summarizing.

7. The method according to any one of claims 1 to 6, characterized in that, The method further comprises: obtaining final clearing data of the electricity spot market and actual operation data of the corresponding period; combining the supply and demand correlation information in the clearing data to determine the deviation degree of the weather prediction result, the new energy output prediction result, the electric heat load demand prediction result, the peak shaving or peak demand parameter and the unit peak shaving and peak potential parameter and the actual operation data corresponding thereto; According to the deviation degree, the deviation distribution characteristics are analyzed and the dominant deviation link is located; Based on the deviation distribution characteristics and the dominant deviation link, the auxiliary peak shaving scheme of the combined heat and power unit is optimized.

8. A device for assisting peak regulation of a combined heat and power unit, characterized in that, The device comprises: a weather prediction module for obtaining weather data, processing and analyzing the weather data to obtain a weather prediction result; a new energy output prediction module for obtaining new energy operation data, determining a new energy output prediction result based on the weather prediction result and the new energy operation data; an electric heat load prediction module for obtaining electric heat load data in the power supply and heat supply coverage range corresponding to the combined heat and power unit, classifying and analyzing the characteristics of the electric heat load data, processing in combination with the weather prediction result and the heat supply system operation state to obtain an electric heat load demand prediction result; a parameter determination module for generating a peak shaving or peak demand parameter according to the new energy output prediction result and the electric heat load demand prediction result; a heat and power decoupling strategy module for obtaining real-time operation parameters of the combined heat and power unit, determining the peak shaving and peak potential of the combined heat and power unit according to the real-time operation parameters, configuration parameters and application scenes of multiple heat and power decoupling modes; a scheme generation module for determining an auxiliary peak shaving scheme of the combined heat and power unit in combination with power supply and demand changes in the electricity spot market and electric heat load fluctuation conditions according to the weather prediction result, the new energy output prediction result, the electric heat load demand prediction result, the peak shaving or peak related demand parameter and the peak shaving and peak potential of the combined heat and power unit.

9. A storage medium, characterized by The storage medium comprises a stored program, wherein when the program is running, the device where the storage medium is located executes the combined heat and power unit auxiliary peak shaving method of any one of claims 1-7. The storage medium comprises a stored program, wherein when the program is running, the device where the storage medium is located executes the combined heat and power unit auxiliary peak shaving method of any one of claims 1-7.

10. An electronic device, comprising: The device comprises at least one processor, at least one memory connected with the processor, and a bus; wherein the processor, the memory and the bus complete mutual communication; the processor is used to call program instructions in the memory, so as to execute the method for auxiliary peak regulation of the combined heat and power unit according to any one of claims 1-7.

Images