Heat supply unit heat storage peak regulation strategy construction method and device and storage medium

Abstract

The invention relates to the technical field of heat supply unit peak regulation capacity optimization, in particular to a heat supply unit heat storage peak regulation strategy construction method and device and a storage medium, and the method comprises the steps: carrying out the hourly prediction of a day-ahead electrical load and a day-ahead thermal load through a target BO-ResNet-AM neural network prediction model, obtaining a day-ahead hourly electric load predicted value and a day-ahead hourly thermal load predicted value; based on a power supply balance constraint, a heat supply balance constraint, a unit ramp rate constraint and a target function, taking pipe network heat storage / release, actual power supply power and actual heat supply as decision variables, and constructing an initial source network load storage collaborative interaction mode optimization scheduling model; and converging the initial source network load storage collaborative interaction mode optimization scheduling model to a globally optimal solution through an adaptive weight learning factor particle swarm optimization algorithm. By accurately predicting the electric heating load and quantifying the heat storage capacity of the pipe network, peak clipping and valley filling of the electric load and space-time translation of the thermal load are achieved, and the deep adjustment and peak capacity of the unit is improved.

Description

Technical Field

[0001] The present application relates to the technical field of optimizing the peak-shaving capacity of heating units, and in particular to a method, device, and storage medium for constructing a heat storage peak-shaving strategy for heating units. Background Art

[0002] Against the backdrop of new power system construction, thermal power units have gradually transformed from the main power source to a regulating power source. The annual utilization hours have significantly decreased, and the demand for peak-shaving has become increasingly urgent, especially in the northern region, where the proportion of cogeneration units is high. Their "heat-to-electricity" operation mode has severely limited the peak-shaving capacity during the heating period. However, traditional heating unit scheduling strategies usually ignore the thermal inertia characteristics of the heating pipeline network, and only allocate electricity and heat loads based on static heat loads, failing to effectively tap the potential of pipeline network heat storage to enhance peak-shaving capacity. In existing technologies, thermal load forecasting mostly adopts a single time series model (such as LSTM or traditional BP neural network), which does not fully integrate multi-dimensional features such as meteorological and day types, resulting in limited prediction accuracy. In addition, thermal network modeling is mostly simplified into a centralized parameter model, ignoring the delay characteristics and temperature attenuation characteristics of thermal energy transmission, resulting in large deviations in the calculation of thermal storage characteristics. In addition, existing electric thermal optimization scheduling models often take single economic or environmental protection as the goal, lack of coordinated optimization of the dynamic characteristics of pipeline network thermal storage, and difficult to achieve flexible scheduling under the source-grid-load-storage interactive mode. As a result, the peak-shaving potential of heating units has not been fully released, and the contradiction between economy and grid regulation needs is prominent. Summary of the Invention

[0003] In order to solve the above-mentioned problems in the existing technology, the present application provides a method, equipment and storage medium for constructing a heat storage peak-shaving strategy for a heating unit. By accurately predicting the electric and thermal loads and quantifying the heat storage capacity of the pipeline network, a source-grid-load-storage collaborative scheduling model is constructed to achieve peak shaving and valley filling of the electric load and temporal and spatial shifting of the thermal load, thereby improving the deep regulation and peak capacity of the unit.

[0004] In a first aspect, an embodiment of the present application provides a method for constructing a heat storage peak-shaving strategy for a heating unit, the method comprising:

[0005] Obtaining a historical operating data set of the heating unit, a library of external influencing factors of the heating system, pipe network geometry and fluid dynamics parameters, pipe network thermal operating parameters, and a key element data set of an optimal scheduling model for the cogeneration system, wherein the key element data set of the optimal scheduling model for the cogeneration system includes electricity price, heat price, actual power supply of the heating unit, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient, heating operation and maintenance cost coefficient, steam inlet volume, and ramp rate upper limit;

[0006] Preprocessing the historical operation data set of the heating unit and the external influencing factor library of the heating system respectively to obtain unit operation characteristic data and external characteristic data;

[0007] Constructing a target BO-ResNet-AM neural network prediction model, and inputting the unit operation characteristic data and the external characteristic data into the target BO-ResNet-AM neural network prediction model, so as to perform hourly prediction of the day-ahead electric load and the day-ahead thermal load through the target BO-ResNet-AM neural network prediction model, thereby obtaining a day-ahead hourly electric load prediction value and a day-ahead hourly thermal load prediction value;

[0008] Determine the heat energy transfer delay time based on the pipe network geometry and fluid mechanics parameters;

[0009] According to the heat energy transmission delay time and the thermal operation parameters of the pipe network, a heat energy exponential decay relationship model is obtained, and according to the heat energy exponential decay relationship model, the heat load side supply water temperature, the heat load side return water temperature, the heat source side return water temperature, and the heat source side supply water temperature are determined;

[0010] Calculating the heat dissipation of the pipe network, the heat supply on the heat source side, and the heat supply on the heat load side according to the heat load side water supply temperature, the heat load side return water temperature, the heat source side return water temperature, and the heat source side water supply temperature;

[0011] Calculating the heat storage / release of the pipe network based on the heat dissipation of the pipe network, the heat supply on the heat source side, and the heat supply on the heat load side;

[0012] The objective function is constructed by using the heat price, the actual power supply, the actual heat supply, the fuel price, the coal consumption function, the power generation operation and maintenance cost coefficient, and the heating operation and maintenance cost coefficient;

[0013] Determine the power supply balance constraint, the heat supply balance constraint, and the unit ramp rate constraint based on the pipe network heat storage / release, the day-ahead hourly electricity load forecast value, the day-ahead hourly heat load forecast value, the actual power supply, the actual heat supply, the steam intake, and the ramp rate upper limit;

[0014] Based on the power supply balance constraint, the heat supply balance constraint, the unit ramp rate constraint, and the objective function, and with the pipe network heat storage / release, the actual power supply power, and the actual heat supply as decision variables, an initial source-grid-load-storage collaborative interaction mode optimization scheduling model is constructed;

[0015] The initial source-grid-load-storage collaborative interaction mode optimization scheduling model is converged to the global optimal solution through the adaptive weight learning factor particle swarm optimization algorithm to obtain the target source-grid-load-storage collaborative interaction mode optimization scheduling model.

[0016] According to some embodiments of the first aspect of the present application, the historical operation data of the heating unit includes historical power supply load data and historical heating load data, the heating system external influencing factor library includes historical gas phase information and day type, and the heating unit historical operation data set and the heating system external influencing factor library are respectively preprocessed to obtain unit operation characteristic data and external characteristic data, including:

[0017] By using a box plot method, abnormal data screening and elimination processing are performed on the historical power supply load data, the historical heating load data, the historical gas phase information, and the day type, respectively, to obtain a first historical operation data set and a first external influencing factor data;

[0018] The first historical operation data set and the first external influencing factor data are smoothed in sequence based on cubic spline difference, and abnormal data reconstruction is performed on the first historical operation data set and the first external influencing factor data respectively to obtain unit operation characteristic data and external characteristic data.

[0019] According to some embodiments of the first aspect of the present application, inputting the unit operation characteristic data and the external characteristic data into the target BO-ResNet-AM neural network prediction model includes:

[0020] A correlation analysis is performed on the unit operation characteristic data and the external characteristic data, and the unit operation characteristic data and the external characteristic data after the correlation analysis are input into the target BO-ResNet-AM neural network prediction model.

[0021] According to some embodiments of the first aspect of the present application, constructing a target BO-ResNet-AM neural network prediction model includes:

[0022] Get the training dataset;

[0023] The hyperparameters in the ResNet network are automatically optimized using the Bayesian optimization algorithm, and the optimized hyperparameters are substituted into the ResNet-AM network structure to obtain the initial BO-ResNet-AM neural network prediction model;

[0024] Inputting the training data set into the initial BO-ResNet-AM neural network prediction model for model training, and monitoring the regression error index;

[0025] An error judgment is performed on the regression error index until the regression error index reaches a preset regression error threshold, thereby obtaining a target BO-ResNet-AM neural network prediction model.

[0026] According to some embodiments of the first aspect of the present application, automatically optimizing hyperparameters in the ResNet network using a Bayesian optimization algorithm includes:

[0027] Get the number of convolutional layers, convolution kernels, and neurons in the fully connected layer in the ResNet network;

[0028] Initializing the number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer respectively;

[0029] Establishing a Gaussian process regression model, inputting the initialized number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer into the Gaussian process regression model to construct a mapping relationship with hyperparameters and search for optimal parameters through the expected improvement method;

[0030] The number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer are iteratively updated until a maximum number of iterations is reached, and the optimal parameter search is stopped to obtain the optimized hyperparameters.

[0031] According to some embodiments of the first aspect of the present application, the pipe network geometry and fluid dynamics parameters include the pipe number, pipe heating medium density, pipe heating medium flow rate, pipe length, and pipe inner diameter. The calculation formula for the heat energy transfer delay time is as follows:

[0032]

[0033] Wherein, k represents the pipeline number, τ represents the heat energy transmission delay time, ρ represents the density of the heat supply medium in the pipeline, L k Denotes the pipe length, d k Indicates the inner diameter of the pipe, m k Indicates the flow rate of the heating medium in the pipeline.

[0034] According to some embodiments of the first aspect of the present application, the thermal operation parameters of the pipeline network include: pipeline heat dissipation coefficient, pipeline starting node temperature, ambient temperature and constant pressure specific heat, and the calculation formula of the thermal energy exponential decay relationship model is as follows:

[0035]

[0036] Wherein, λ represents the heat dissipation coefficient of the pipeline, τ represents the heat energy transmission delay time, T start (t) represents the temperature of the node at the beginning of the pipeline at time t, T end (t+τ k ) represents the delay of heat energy transfer (t+τ k ) time, the temperature of the pipe end node, c p represents the specific heat at constant pressure, m kIndicates the heat supply medium flow rate of the pipeline, T a (t) is the ambient temperature at time t, L k Indicates the length of the pipeline.

[0037] According to some embodiments of the first aspect of the present application, the calculation formula of the objective function is as follows:

[0038]

[0039] Among them, F represents net income, P e，j (t) represents the actual power supply power of the j-th heating unit in time period t, Q j (t) represents the actual heat supply of the j-th heating unit in period t, c e (t) represents the electricity price during period t, c h (t) represents the heat price during the period t, c coal represents the fuel price, f j (·) represents the coal consumption function of the j-th heating unit, represents the power generation operation and maintenance cost coefficient, Represents the heating operation and maintenance cost coefficient.

[0040] In a second aspect, an embodiment of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements: the method for constructing a heat storage and peak-shaving strategy for a heating unit as described in the first aspect above.

[0041] In a third aspect, an embodiment of the present application further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the method for constructing a heat storage peak-shaving strategy for a heating unit as described in the first aspect above.

[0042] The beneficial effects of the present application are embodied in that, by obtaining the historical operation data set of the heating unit, the external influencing factor library of the heating system, the geometric and fluid mechanics parameters of the pipe network, the thermal operation parameters of the pipe network and the key element data set of the optimization scheduling model of the cogeneration system, the key element data set of the optimization scheduling model of the cogeneration system includes electricity price, heat price, actual power supply power of the heating unit, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient, heating operation and maintenance cost coefficient, steam intake and climbing rate upper limit; the historical operation data set of the heating unit and the external influencing factor library of the heating system are respectively preprocessed to obtain the unit operation Characteristic data and external characteristic data; construct a target BO-ResNet-AM neural network prediction model, and input the unit operation characteristic data and external characteristic data into the target BO-ResNet-AM neural network prediction model to make hourly predictions on the day-ahead electric load and the day-ahead thermal load through the target BO-ResNet-AM neural network prediction model, and obtain the day-ahead hourly electric load prediction value and the day-ahead hourly thermal load prediction value; confirm the heat energy transmission delay time according to the pipe network geometry and fluid mechanics parameters; obtain the heat energy exponential attenuation relationship model according to the heat energy transmission delay time and the pipe network thermal operation parameters and calculate the heat energy exponential attenuation relationship model according to the heat energy transmission delay time and the pipe network thermal operation parameters. The exponential decay relationship model confirms the heat load side water supply temperature, heat load side return water temperature, heat source side return water temperature and heat source side water supply temperature; according to the heat load side water supply temperature, heat load side return water temperature, heat source side return water temperature and heat source side water supply temperature, the heat dissipation of the pipeline network, the heat supply of the heat source side and the heat supply of the heat load side are calculated; according to the heat dissipation of the pipeline network, the heat supply of the heat source side and the heat supply of the heat load side, the heat storage / release of the pipeline network is calculated; according to the electricity price, heat price, actual power supply, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient and heating operation and maintenance cost coefficient, the objective function is constructed; according to the heat storage / release of the pipeline network, hourly The power load forecast value, the hourly heat load forecast value of the day before, the actual power supply power, the actual heat supply, the steam intake and the upper limit of the ramp rate are used to confirm the power supply balance constraint, the heat supply balance constraint and the unit ramp rate constraint; based on the power supply balance constraint, the heat supply balance constraint, the unit ramp rate constraint and the objective function, and with the pipe network heat storage / release, the actual power supply power and the actual heat supply as decision variables, an initial source-grid-load-storage collaborative interaction mode optimization scheduling model is constructed; the initial source-grid-load-storage collaborative interaction mode optimization scheduling model is converged to the global optimal solution through the adaptive weight learning factor particle swarm optimization algorithm to obtain the target source-grid-load-storage collaborative interaction mode optimization scheduling model. This application uses this method to accurately predict the electric and thermal loads and quantify the heat storage capacity of the pipe network to construct a source-grid-load-storage collaborative scheduling model, realize peak shaving and valley filling of the electric load and spatiotemporal shift of the heat load, and improve the deep adjustment and peak capacity of the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1A flow chart of a method for constructing a heat storage peak-shaving strategy for a heating unit provided in an embodiment of the first aspect of the present application;

[0044] Figure 2 This is a schematic diagram of the process of obtaining heat storage / release of the official website provided in the embodiment of the first aspect of the present application;

[0045] Figure 3 1 is a flow chart of constructing a target BO-ResNet-AM neural network prediction model according to an embodiment of the first aspect of the present application;

[0046] Figure 4 This is a flow chart of unit operation characteristic data and external characteristic data provided by the embodiment of the first aspect of the present application;

[0047] Figure 5 This is a schematic diagram of a process for automatically optimizing hyperparameters in a ResNet network using a Bayesian optimization algorithm, as provided in an embodiment of the first aspect of the present application;

[0048] Figure 6 This is a schematic diagram of a prediction process for hourly electric load prediction values ​​and hourly thermal load prediction values ​​provided by an embodiment of the first aspect of the present application;

[0049] Figure 7 Schematic diagram of the structure of the target BO-ResNet-AM neural network prediction model provided by the embodiment of the first aspect of the present application;

[0050] Figure 8 Schematic diagram of the attention mechanism provided by the embodiment of the first aspect of the present application;

[0051] Figure 9 It is a schematic diagram of a box diagram provided by an embodiment of the first aspect of the present application;

[0052] Figure 10 Schematic diagram of the heat energy transmission delay characteristics provided by the embodiment of the first aspect of the present application;

[0053] Figure 11 This is a structural diagram of another U-shaped Transformer network provided in an embodiment of the second aspect of the present application. DETAILED DESCRIPTION

[0054] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and are not to be construed as limiting the present application.

[0055] In the description of this application, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on this application.

[0056] In the description of this application, if there is a description of first or second, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0057] In the description of this application, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in this application based on the specific content of the technical solution.

[0058] Against the backdrop of the "dual carbon" goals and the construction of a new power system, thermal power units have gradually transformed from the main power source to a regulating power source, with annual utilization hours significantly reduced and the demand for peak regulation becoming increasingly urgent, especially in the northern region, where cogeneration units account for a high proportion. Their "heat-based electricity" operation mode has severely limited peak regulation capacity during the heating period; however, traditional heating unit scheduling strategies usually ignore the thermal inertia characteristics of the heating pipeline network, and only allocate electricity and heat loads based on static heat loads, failing to effectively tap the potential of pipeline network heat storage to enhance peak regulation capacity. In existing technologies, thermal load forecasting mostly adopts a single time series model (such as LSTM or traditional BP neural network), which does not fully integrate multi-dimensional features such as meteorological and day types, resulting in limited prediction accuracy. In addition, thermal network modeling is mostly simplified into a centralized parameter model, ignoring the delay characteristics and temperature attenuation characteristics of thermal energy transmission, resulting in large deviations in the calculation of thermal storage characteristics. In addition, existing electric thermal optimization scheduling models often take single economic or environmental protection as the goal, lack of coordinated optimization of the dynamic characteristics of pipeline network thermal storage, and difficult to achieve flexible scheduling under the source-grid-load-storage interactive mode. As a result, the peak-shaving potential of heating units has not been fully released, and the contradiction between economy and grid regulation needs is prominent.

[0059] In order to solve the above problems, the present application proposes a method for constructing a heat storage peak-shaving strategy for a heating unit. The embodiments of the present application are further explained below in conjunction with the accompanying drawings.

[0060] Reference Figures 1 to 2 , Figures 1 to 2 A method for constructing a heat storage peak-shaving strategy for a heating unit provided in an embodiment of the first aspect of the present application is shown. This method is also applied to and executed by a device. In other words, the method can be executed by software or hardware installed on the device. The method includes the following steps:

[0061] Step S110, obtaining a historical operation data set of the heating unit, a library of external influencing factors of the heating system, a pipe network geometry and fluid mechanics parameters, a pipe network thermal operation parameters and a key element data set of the cogeneration system optimization scheduling model.

[0062] In this step, the key element data set of the cogeneration system optimization scheduling model includes electricity price, heat price, actual power supply power of the heating unit, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient, heating operation and maintenance cost coefficient, steam input volume and climbing rate upper limit.

[0063] Step S120 , preprocessing the historical operation data set of the heating unit and the external influencing factor library of the heating system to obtain unit operation characteristic data and external characteristic data.

[0064] In this step, a dual-channel input architecture of unit operation characteristic data + external characteristic data is adopted to break through the limitations of traditional single time series prediction.

[0065] Step S130: construct a target BO-ResNet-AM neural network prediction model, and input the unit operation characteristic data and external characteristic data into the target BO-ResNet-AM neural network prediction model to perform hourly prediction of the day-ahead electric load and the day-ahead thermal load through the target BO-ResNet-AM neural network prediction model to obtain the day-ahead hourly electric load prediction value and the day-ahead hourly thermal load prediction value.

[0066] Specifically, a short-term prediction model for electric load and thermal load is established based on the BO-ResNet-AM neural network. The BO-ResNet-AM neural network prediction model mainly consists of two parts: a three-layer convolutional deep residual network and an attention mechanism. Figure 7 The electric load and thermal load prediction model based on BO-ResNet-AM neural network is constructed. Each convolution layer contains multiple convolution kernels. Perform convolution calculations and add a pooling layer after the convolution layer to reduce the dimension of the feature data, where x represents the feature input, h represents the output value after convolution, and f(·) represents the nonlinear feature operation. Represents the convolution operation, W represents the weight of the convolution kernel, and b represents the bias of the convolution kernel; in the deep residual network, the output of a residual block is: x i+1 =F(x i , w i ), where x i 、x i+1 Represent the input and output of the i-th residual block in the residual network, F represents the residual mapping to be learned and trained, and w iis the weight parameter to be learned and trained; the deep residual network is composed of multiple stacked residual blocks, and the forward propagation of the k residual block structure can be expressed as: In the formula, x0 represents the input of the residual network, x k represents the output of the kth residual block, W j ={w j,t |1≤j≤m} represents the weight set associated with the jth residual block, m represents the number of layers in the residual block; the back propagation of the error is: Among them, loss represents the loss function of the neural network. Figure 8 As shown in Figure 2, the attention mechanism can adaptively assign greater weights to targets with higher relevance. By emphasizing the contribution of the most informative and influential parts of the input to the output, the attention module can reduce the loss of historical information and extract more relevant information. Figure 8 In the formula, X = (X1, X2, …, Xt) is the input vector; q is the query vector; the relevance of each input vector to the query vector is calculated by the attention scoring function. The most commonly used scoring function is the dot product function, which is: Where S i (i∈[1,t]) is the attention score, and the softmax function is used to calculate the attention distribution weight α i : According to the attention distribution weight α i , perform weighted averaging on the input data to get the final output:

[0067] Step S131: Determine the heat energy transfer delay time based on the pipe network geometry and fluid mechanics parameters.

[0068] In one possible implementation, the pipe network geometry and fluid dynamics parameters include the pipe number, pipe heating medium density, pipe heating medium flow rate, pipe length, and pipe inner diameter. The heat energy transfer delay time is calculated as follows:

[0069]

[0070] Where k represents the pipe number, and the unit is kg / m 3 , τ represents the heat energy transmission delay time, the unit is h, ρ represents the density of the pipeline heating medium, L k Indicates the length of the pipeline in m, d k Indicates the inner diameter of the pipe, m k Indicates the flow rate of pipeline heating medium, in kg / h.

[0071] Step S132: obtaining a heat energy exponential decay relationship model based on the heat energy transmission delay time and the thermal operation parameters of the pipe network, and confirming the heat load side supply water temperature, the heat load side return water temperature, the heat source side return water temperature, and the heat source side supply water temperature based on the heat energy exponential decay relationship model.

[0072] In one possible implementation, the thermal operation parameters of the pipeline network include: pipeline heat dissipation coefficient, pipeline starting node temperature, ambient temperature, and constant pressure specific heat. The calculation formula of the thermal energy exponential decay relationship model is as follows:

[0073]

[0074] Where λ represents the heat dissipation coefficient of the pipe, in kW / (m·K), and τ represents the heat transfer delay time, T start (t) represents the temperature of the node at the beginning of the pipeline at time t, in °C, T end (t+τ k ) represents the delay of heat energy transfer (t+τ k ) time, the temperature of the pipe end node, in °C, c p represents the specific heat at constant pressure, m k Indicates the flow rate of heating medium in the pipeline, T a (t) is the ambient temperature at time t, in °C, L k Indicates the length of the pipe.

[0075] Reference Figure 10 , the relationship between the supply water temperature on the heat load side and the return water temperature on the heat load side is: Among them, T 1, Load Indicates the water supply temperature on the heat load side, T 1,CHP represents the return water temperature on the heat load side, k represents the pipe number, and τ represents the heat energy transmission delay time; the relationship between the supply water temperature on the heat source side and the return water temperature on the heat source side is: Among them, T 2, Load Indicates the water supply temperature on the heat source side, T 2,CHP represents the return water temperature on the heat source side, k represents the pipe number, and τ represents the heat energy transfer delay time.

[0076] Step S133, according to the heat load side water supply temperature, the heat load side return water temperature, the heat source side return water temperature and the heat source side water supply temperature, calculate the heat dissipation of the pipe network, the heat source side heat supply and the heat load side heat supply.

[0077] In one possible implementation, the heat dissipation of the pipe network is calculated as follows:

[0078] Q loss (t) = mk ×4.187×(T 1,CHP (t-τ k )-T 1,Load (t)) / 1000+m k ×4.187×(T 2,Load (t-τ k )-T 2,CHP (t)) / 1000, where Q loss (t) represents the heat dissipation of the pipe network at time t; the heat supply on the heat source side is calculated as follows:

[0079] Q CHP (t) = m k ×4.187×(T 1,CHP (t)-T 2,CHP (t)) / 1000, where Q CHP (t) represents the heat supply on the heat source side at time t, in GJ / h; the heat supply on the heat load side is calculated as follows:

[0080] Q Load (t) = m k ×4.187×(T 1,Load (t)-T 2,Load (t)) / 1000, where Q Load (t) represents the heat supply on the heat source side at time t, in GJ / h.

[0081] Step S134: Calculate the heat storage / release amount of the pipe network based on the heat dissipation of the pipe network, the heat supply on the heat source side, and the heat supply on the heat load side.

[0082] In one possible implementation, the calculation formula for the heat storage / release of the pipe network is as follows:

[0083] Q s (t) = Q CHP (t)-Q Load (t)-Q loss (t)), where Q s (t) represents the heat storage / release of the pipe network at time t.

[0084] Step S140 , constructing an objective function based on electricity price, heat price, actual power supply, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient and heat supply operation and maintenance cost coefficient.

[0085] In one possible implementation, the objective function is calculated as follows:

[0086]

[0087] Among them, F represents net income, the unit is yuan / h, Pe，j (t) represents the actual power supply of the jth heating unit in period t, in MW, Q j (t) represents the actual heat supply of the jth heating unit in period t, in MW, c e (t) represents the electricity price in period t, in RMB / MWh, c h (t) represents the heat price during period t, in RMB / MWh, c coal Indicates fuel price, in yuan / t, f j (·) represents the coal consumption function of the jth heating unit, φ e represents the power generation operation and maintenance cost coefficient, φ h represents the heating operation and maintenance cost coefficient; the total revenue from electricity and heat sales of the j-th heating unit in period t is: I j (t) = P e，j (t)c e (t)+Q j (t)c h (t), I j (t) represents the total revenue from electricity and heat sales of the j-th heating unit in period t. The fuel cost of the j-th heating unit in period t is calculated as follows: C f，j (t) = c coal ·f j (P e，j (t),Q j (t), C f，j (t) represents the fuel cost of the jth heating unit in period t. The calculation formula for the operation and maintenance cost of each unit in period t is as follows: C OM，j (t) = φ e P e，j (t)+φ h Q j (t), C OM，j (t) represents the operation and maintenance cost of each unit in period t, that is,

[0088] Step S150: Confirm the power supply balance constraint, heat supply balance constraint, and unit ramp rate constraint based on the network's stored / released heat, the day's hourly electricity load forecast, the day's hourly heat load forecast, the actual power supply, the actual heat supply, the steam intake, and the ramp rate upper limit.

[0089] It should be noted that the power balance constraint is as follows: The relationship between the heat balance constraint is as follows: Q D(t) represents the hourly heat load forecast value of the day before. The inequality constraints of the cogeneration unit are the feasible region composed of the upper and lower limits of the electricity and heat loads. Mathematically, it presents an irregular polygonal region of the electricity and heat loads. The inequality constraints of the heat load and electricity load are: The power balance constraint is as follows: Among them, D 0，j The relationship between the steam inlet of the jth heating unit and the ramp rate constraint of the cogeneration unit is as follows: |P e，j (t)-P e，j (t-1)|≤λ reg,j , where λ reg,j It represents the upper limit of the ramp rate of heating unit j, calculated as 2% / min of the rated output.

[0090] Step S160: Based on the power supply balance constraint, heat supply balance constraint, unit ramp rate constraint and objective function, and with the network heat storage / release, actual power supply power and actual heat supply as decision variables, an initial source-grid-load-storage collaborative interaction mode optimization scheduling model is constructed.

[0091] Step S170 , the initial source-grid-load-storage collaborative interaction mode optimization scheduling model is converged to a global optimal solution through an adaptive weight learning factor particle swarm optimization algorithm to obtain a target source-grid-load-storage collaborative interaction mode optimization scheduling model.

[0092] It should be noted that this application innovatively constructs a BO-ResNet-AM neural network prediction model, which effectively solves the problem of "LSTM / BP neural network not integrating multidimensional features" pointed out in the background technology through the integration of Bayesian optimization (BO) and attention mechanism (AM); adopts a dual-channel input architecture of unit operation feature data + external feature data to break through the limitations of traditional single time series prediction; innovates the pipeline storage / release heat calculation system, and quantifies the spatiotemporal distribution characteristics of thermal inertia through the three-dimensional coupling of heat supply on the heat source side / load side and heat dissipation on the pipeline network; constructs a multi-dimensional constraint system including power supply / heat supply balance constraints + unit ramp rate constraints to achieve dynamic coupling of "source-grid-load-storage".

[0093] It should be noted that the adaptive weight learning factor particle swarm optimization algorithm is an improved particle swarm optimization algorithm. The specific algorithm steps of the improved particle swarm optimization algorithm are: start → particle swarm initialization → calculate the objective function of each particle → determine the individual extreme value and the overall extreme value → change the inertia weight and learning factor, and update the speed and position → calculate the objective function of the updated particle → determine the updated individual extreme value → determine whether the number of iterations has been reached → output the optimal value after the number of iterations has been reached → end.

[0094] In some embodiments, in the improved particle swarm optimization algorithm, "particle swarm initialization" refers to generating a set of initial candidate solutions for the optimization problem. Each particle represents a possible scheduling strategy, and its position is composed of all decision variables, including: the amount of heat stored / released by the pipeline network, the actual power supply power, and the actual heat supply. During initialization, the algorithm will randomly generate multiple particles, and the position of each particle must meet the physical constraints of the variables (such as upper and lower limits of electric load, climbing rate limit, etc.). Each particle corresponds to a complete scheduling plan, and its objective function calculates the economic index of the plan. The significance of the particle objective function is: by calculating the F of each particle, it is evaluated whether the corresponding scheduling plan is economically optimal, thereby guiding the algorithm to search in the direction of high returns and low costs, and obtaining the objective function with the optimal economy. The specific process of the improved particle swarm algorithm in this application is as follows: Step 1: Particle swarm initialization: Particle structure: The position vector of each particle contains all decision variables. The initial position is randomly generated, but it must meet basic constraints (such as upper and lower limits of electric load) or be processed by subsequent penalty functions; Step 2: Calculate the objective function of each particle. For each particle, the electric load, heating supply, and heat storage are analyzed according to its position, and substituted into the objective function F to calculate the net benefit. Constraint violation penalty: For particles that do not meet the constraints (such as heat balance, climbing rate), the penalty is deducted from F; Step 3: Update of individual extreme values ​​and group extreme values, individual extreme value: record the historical optimal position of each particle, group extreme value: record the global optimal position of all particles; Step 4: Dynamically adjust parameters and change inertia weight: adopt a linear decrease strategy, enhance global search in the early stage, and improve local convergence in the later stage; change learning factor: adaptively adjust cognitive and social weights, for example, focus on individual experience in the early stage and focus on group experience in the later stage; Step 5: Update speed and position; Step 6: Termination condition judgment: reach the maximum number of iterations, or the group extreme value has no significant improvement for multiple consecutive generations.

[0095] It should be noted that this application proposes a short-term prediction model based on the BO-ResNet-AM neural network, which automatically searches for residual network hyperparameters (number of convolution layers, number of convolution kernels, etc.) through Bayesian optimization, and dynamically allocates input feature weights in combination with the attention mechanism, thereby solving the problem of insufficient integration of multi-dimensional heterogeneous data (meteorological, historical load, day type) of traditional models, and effectively reducing prediction errors. For the first time, this application collaboratively considers the delay characteristics of heat energy transmission (dynamic calculation of transmission delay time based on pipeline geometry and flow parameters) and attenuation characteristics (quantification of temperature loss along the way through an exponential attenuation model), constructs a dynamic equation for heat storage / release in the pipeline network, improves the calculation accuracy of heat storage characteristics, and provides a reliable basis for load temporal and spatial translation. ; Construct a source-grid-load-storage interactive scheduling model with the goal of maximizing net profit, take the heat storage / release of the pipeline network as an independent decision variable, and solve the global optimal solution through the improved particle swarm algorithm (adaptive adjustment of inertia weight and learning factor) to achieve coordinated optimization of peak shaving and valley filling of electric load and equivalent translation of thermal load; introduce thermal storage characteristic constraints (all-day heat storage / release balance) and climbing rate constraints (based on dynamic limit of steam inlet volume) to ensure that the scheduling strategy maximizes peak-shaving flexibility within the heating safety boundary, and solves the conservative problem of optimization results caused by static constraints of traditional models. This application provides a systematic solution to resolve the contradiction between thermal and electric coupling, and has important engineering value in promoting the transformation of heating units from "passive peak regulation" to "active response".

[0096] The beneficial effects of the present application are embodied in that, by obtaining the historical operation data set of the heating unit, the external influencing factor library of the heating system, the geometric and fluid mechanics parameters of the pipe network, the thermal operation parameters of the pipe network and the key element data set of the optimization scheduling model of the cogeneration system, the key element data set of the optimization scheduling model of the cogeneration system includes electricity price, heat price, actual power supply power of the heating unit, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient, heating operation and maintenance cost coefficient, steam intake and climbing rate upper limit; the historical operation data set of the heating unit and the external influencing factor library of the heating system are respectively preprocessed to obtain the unit operation Characteristic data and external characteristic data; construct a target BO-ResNet-AM neural network prediction model, and input the unit operation characteristic data and external characteristic data into the target BO-ResNet-AM neural network prediction model to make hourly predictions on the day-ahead electric load and the day-ahead thermal load through the target BO-ResNet-AM neural network prediction model, and obtain the day-ahead hourly electric load prediction value and the day-ahead hourly thermal load prediction value; confirm the heat energy transmission delay time according to the pipe network geometry and fluid mechanics parameters; obtain the heat energy exponential attenuation relationship model according to the heat energy transmission delay time and the pipe network thermal operation parameters and calculate the heat energy exponential attenuation relationship model according to the heat energy transmission delay time and the pipe network thermal operation parameters. The exponential decay relationship model confirms the heat load side water supply temperature, heat load side return water temperature, heat source side return water temperature and heat source side water supply temperature; according to the heat load side water supply temperature, heat load side return water temperature, heat source side return water temperature and heat source side water supply temperature, the heat dissipation of the pipeline network, the heat supply of the heat source side and the heat supply of the heat load side are calculated; according to the heat dissipation of the pipeline network, the heat supply of the heat source side and the heat supply of the heat load side, the heat storage / release of the pipeline network is calculated; according to the electricity price, heat price, actual power supply, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient and heating operation and maintenance cost coefficient, the objective function is constructed; according to the heat storage / release of the pipeline network, hourly The power load forecast value, the hourly heat load forecast value of the day before, the actual power supply power, the actual heat supply, the steam intake and the upper limit of the ramp rate are used to confirm the power supply balance constraint, the heat supply balance constraint and the unit ramp rate constraint; based on the power supply balance constraint, the heat supply balance constraint, the unit ramp rate constraint and the objective function, and with the pipe network heat storage / release, the actual power supply power and the actual heat supply as decision variables, an initial source-grid-load-storage collaborative interaction mode optimization scheduling model is constructed; the initial source-grid-load-storage collaborative interaction mode optimization scheduling model is converged to the global optimal solution through the adaptive weight learning factor particle swarm optimization algorithm to obtain the target source-grid-load-storage collaborative interaction mode optimization scheduling model. This application uses this method to accurately predict the electric and thermal loads and quantify the heat storage capacity of the pipe network to construct a source-grid-load-storage collaborative scheduling model, realize peak shaving and valley filling of the electric load and spatiotemporal shift of the heat load, and improve the deep adjustment and peak capacity of the unit.

[0097] It is understandable that, referring to Figure 4The historical operation data of the heating unit includes historical power supply load data and historical heating load data. The external influencing factor library of the heating system includes historical gas phase information and day type. Step S120 includes but is not limited to the following steps:

[0098] Step S121 , using a box plot method, abnormal data screening and elimination processing are performed on historical power supply load data, historical heating load data, historical gas phase information, and day type, respectively, to obtain a first historical operation data set and first external influencing factor data.

[0099] Step S122 : smoothing the first historical operation data set and the first external influencing factor data in sequence based on the cubic spline difference, and reconstructing abnormal data on the first historical operation data set and the first external influencing factor data respectively to obtain unit operation characteristic data and external characteristic data.

[0100] Specifically, refer to Figure 9 , the box plot method is used to screen and eliminate abnormal values ​​from the original data of the unit. Figure 3 As shown, the upper quartile is defined as Q1, the median is Q2, the lower quartile is Q3, and the interquartile range IQR is: IQR=Q1-Q3. If any value in the data set satisfies the formula x≥Q1+1.5IQR or x≤Q3-1.5IQR, it is considered an outlier. In order to maintain the continuity of the data, the abnormal data is smoothed based on cubic spline interpolation and the abnormal data is reconstructed. The cubic spline interpolation method divides the known data set into several intervals, and fits a cubic interpolation function to each data segment to calculate the function value of the missing data point and obtain a complete data sequence; the known data set has n+1 data points, and the data set is divided into n intervals: [f(x1),f(x2)], [f(x2),f(x3)],…, [f( x n),f( x n+1)]. Fitting a cubic function to each of the above intervals, we get n cubic functions in total: G(x1), G(x2),…, G(x n-1 ). The cubic function G(x) of each interval is defined as follows: G(x)

[0101] Among them, x i Denotes the i-th data point in the dataset, where i = 1, 2, …, n. Under the condition that each piecewise function smoothly connects, the coefficients of the equations are solved and the missing values ​​in the dataset are reconstructed.

[0102] It is understandable that, referring to Figure 6 Inputting the unit operation characteristic data and external characteristic data into the target BO-ResNet-AM neural network prediction model in step S130 includes but is not limited to the following steps:

[0103] Step S1311: perform correlation analysis on the unit operation characteristic data and the external characteristic data, and input the unit operation characteristic data and the external characteristic data after the correlation analysis into the target BO-ResNet-AM neural network prediction model.

[0104] It is understandable that, referring to Figure 3 The step S130 of constructing the target BO-ResNet-AM neural network prediction model includes but is not limited to the following steps:

[0105] Step S1311: Obtain a training data set.

[0106] In step S1312, the hyperparameters in the ResNet network are automatically optimized using the Bayesian optimization algorithm, and the optimized hyperparameters are substituted into the ResNet-AM network structure to obtain the initial BO-ResNet-AM neural network prediction model.

[0107] In step S1313, the training data set is input into the initial BO-ResNet-AM neural network prediction model for model training, and the regression error index is monitored.

[0108] In step S1314, an error judgment is performed on the regression error index until the regression error index reaches a preset regression error threshold, thereby obtaining a target BO-ResNet-AM neural network prediction model.

[0109] It should be noted that the Bayesian optimization algorithm in this application is abbreviated as BO algorithm, and its full name is Bayesian Optimization; ResNet is the abbreviation of Residual Network, which is a classic neural network that serves as the backbone of many computer vision tasks; AM is the full name of Attention Mechanism, which is translated as attention mechanism.

[0110] In this embodiment, Bayesian optimization is a key technology to improve load forecasting accuracy and solve the problem of parameter adjustment in complex neural networks; correlation analysis provides high-quality input for the model, enhancing the physical rationality and computational efficiency of the prediction; the electric and thermal optimization scheduling model achieves a dual breakthrough in economy and peak-shaving capacity through multi-energy coordination and heat storage regulation.

[0111] In some embodiments, according to Figure 6 The process in obtains the hourly electricity load forecast value and the hourly heat load forecast value of the day before.

[0112] It is understandable that, referring to Figure 5In step S1312, the hyperparameters of the ResNet network are automatically optimized using the Bayesian optimization algorithm, including but not limited to the following steps:

[0113] Step S13121, obtaining the number of convolutional layers, the number of convolution kernels, and the number of neurons in the fully connected layer in the ResNet network;

[0114] Step S13122, initializing the number of convolutional layers, the number of convolution kernels, and the number of neurons in the fully connected layer respectively;

[0115] Step S13123: Establish a Gaussian process regression model, input the initialized number of convolution layers, number of convolution kernels, and number of neurons in the fully connected layer into the Gaussian process regression model to build a mapping relationship with hyperparameters and search for optimal parameters through the expected improvement method;

[0116] In step S13124, the number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer are iteratively updated until the maximum number of iterations is reached, the optimal parameter search is stopped, and the optimized hyperparameters are obtained.

[0117] Alternatively, as Figure 11 As shown, the second embodiment of the present application further provides an electronic device 700, including a processor 710 and a memory 720, and the memory 720 stores a program or instruction that can be run on the processor 710. When the program or instruction is executed by the processor 710, the various processes of the embodiment of the method for constructing a heat storage and peak-shaving strategy for a heating unit in the first aspect mentioned above are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0118] It should be noted that the devices in the embodiments of the present application include: servers, terminals, or other devices other than terminals.

[0119] The above device structure does not constitute a limitation of the device. The device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, the input unit may include a graphics processing unit (GPU) and a microphone, and the display unit may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. to configure the display panel. The user input unit includes a touch panel and at least one of other input devices. The touch panel is also called a touch screen. Other input devices may include but are not limited to a physical keyboard, function keys (such as volume control buttons, switch buttons, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.

[0120] The memory can be used to store software programs and various data. The memory may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, applications or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory may include a volatile memory or a non-volatile memory, or the memory may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus random access memory (DRRAM).

[0121] The processor may include one or more processing units; optionally, the processor may integrate an application processor and a modem processor, wherein the application processor primarily handles operations related to the operating system, user interface, and application programs, and the modem processor primarily processes wireless communication signals, such as a baseband processor. It is understood that the modem processor may not be integrated into the processor.

[0122] The present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the embodiment of the method for constructing a heat storage peak-shaving strategy for a heating unit in the first aspect is implemented, and the same technical effect is achieved. To avoid repetition, it is not described here. The processor is the processor in the device in the above embodiment. The readable storage medium includes a computer-readable storage medium such as ROM, RAM, a magnetic disk or an optical disk. It should be noted that, in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such a process, method, article or device. In the absence of further restrictions, an element defined by the sentence "including one..." does not exclude the presence of other identical elements in the process, method, article or device including the element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of the present application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in reverse order depending on the functions involved. For example, the described methods may be performed in an order different from that described. In addition, features described with reference to certain examples may be combined in other examples.

[0123] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods in each embodiment of the present application.

[0124] In the description of the embodiments of the present application, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Thus, a feature defined as "first," "second," "third," or "fourth" may explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, "plurality" means two or more.

[0125] In the description of the embodiments of the present application, specific features, structures, materials or characteristics may be combined in an appropriate manner in any one or more embodiments or examples.

[0126] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a heat storage peak-shaving strategy for a heating unit, characterized in that: include: Obtaining a historical operating data set of the heating unit, a library of external influencing factors of the heating system, pipe network geometry and fluid dynamics parameters, pipe network thermal operating parameters, and a key element data set of an optimal scheduling model for the cogeneration system, wherein the key element data set of the optimal scheduling model for the cogeneration system includes electricity price, heat price, actual power supply of the heating unit, actual heat supply, fuel price, coal consumption function, power generation operation and maintenance cost coefficient, heating operation and maintenance cost coefficient, steam inlet volume, and ramp rate upper limit; Preprocessing the historical operation data set of the heating unit and the external influencing factor library of the heating system respectively to obtain unit operation characteristic data and external characteristic data; Constructing a target BO-ResNet-AM neural network prediction model, and inputting the unit operation characteristic data and the external characteristic data into the target BO-ResNet-AM neural network prediction model, so as to perform hourly prediction of the day-ahead electric load and the day-ahead thermal load through the target BO-ResNet-AM neural network prediction model, thereby obtaining a day-ahead hourly electric load prediction value and a day-ahead hourly thermal load prediction value; Determine the heat energy transfer delay time based on the pipe network geometry and fluid mechanics parameters; According to the heat energy transmission delay time and the thermal operation parameters of the pipe network, a heat energy exponential decay relationship model is obtained, and according to the heat energy exponential decay relationship model, the heat load side supply water temperature, the heat load side return water temperature, the heat source side return water temperature, and the heat source side supply water temperature are determined; Calculating the heat dissipation of the pipe network, the heat supply on the heat source side, and the heat supply on the heat load side according to the heat load side water supply temperature, the heat load side return water temperature, the heat source side return water temperature, and the heat source side water supply temperature; Calculating the heat storage / release of the pipe network based on the heat dissipation of the pipe network, the heat supply on the heat source side, and the heat supply on the heat load side; constructing an objective function based on the electricity price, the heat price, the actual power supply, the actual heat supply, the fuel price, the coal consumption function, the power generation operation and maintenance cost coefficient, and the heat supply operation and maintenance cost coefficient; Determine the power supply balance constraint, the heat supply balance constraint, and the unit ramp rate constraint based on the pipe network heat storage / release, the day-ahead hourly electricity load forecast value, the day-ahead hourly heat load forecast value, the actual power supply, the actual heat supply, the steam intake, and the ramp rate upper limit; Based on the power supply balance constraint, the heat supply balance constraint, the unit ramp rate constraint, and the objective function, and with the pipe network heat storage / release, the actual power supply power, and the actual heat supply as decision variables, an initial source-grid-load-storage collaborative interaction mode optimization scheduling model is constructed; The initial source-grid-load-storage collaborative interaction mode optimization scheduling model is converged to the global optimal solution through the adaptive weight learning factor particle swarm optimization algorithm to obtain the target source-grid-load-storage collaborative interaction mode optimization scheduling model.

2. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 1, characterized in that: The historical operation data of the heating unit includes historical power supply load data and historical heating load data, and the external influencing factor library of the heating system includes historical gas phase information and day type. The historical operation data set of the heating unit and the external influencing factor library of the heating system are respectively preprocessed to obtain unit operation characteristic data and external characteristic data, including: By using a box plot method, abnormal data screening and elimination processing are performed on the historical power supply load data, the historical heating load data, the historical gas phase information, and the day type, respectively, to obtain a first historical operation data set and a first external influencing factor data; The first historical operation data set and the first external influencing factor data are smoothed in sequence based on cubic spline difference, and abnormal data reconstruction is performed on the first historical operation data set and the first external influencing factor data respectively to obtain unit operation characteristic data and external characteristic data.

3. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 1, characterized in that: The step of inputting the unit operation characteristic data and the external characteristic data into the target BO-ResNet-AM neural network prediction model includes: A correlation analysis is performed on the unit operation characteristic data and the external characteristic data, and the unit operation characteristic data and the external characteristic data after the correlation analysis are input into the target BO-ResNet-AM neural network prediction model.

4. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 1, characterized in that: The target BO-ResNet-AM neural network prediction model is constructed, including: Get the training dataset; The hyperparameters in the ResNet network are automatically optimized using the Bayesian optimization algorithm, and the optimized hyperparameters are substituted into the ResNet-AM network structure to obtain the initial BO-ResNet-AM neural network prediction model; Inputting the training data set into the initial BO-ResNet-AM neural network prediction model for model training, and monitoring the regression error index; An error judgment is performed on the regression error index until the regression error index reaches a preset regression error threshold, thereby obtaining a target BO-ResNet-AM neural network prediction model.

5. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 4, characterized in that: The automatic optimization of the hyperparameters in the ResNet network by the Bayesian optimization algorithm includes: Get the number of convolutional layers, convolution kernels, and neurons in the fully connected layer in the ResNet network; Initializing the number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer respectively; Establishing a Gaussian process regression model, inputting the initialized number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer into the Gaussian process regression model to construct a mapping relationship with hyperparameters and search for optimal parameters through the expected improvement method; The number of convolution layers, the number of convolution kernels, and the number of neurons in the fully connected layer are iteratively updated until a maximum number of iterations is reached, and the optimal parameter search is stopped to obtain the optimized hyperparameters.

6. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 1, characterized in that: The pipe network geometry and fluid mechanics parameters include pipe number, pipe heating medium density, pipe heating medium flow rate, pipe length, and pipe inner diameter. The heat energy transfer delay time is calculated as follows: Wherein, k represents the pipeline number, τ represents the heat energy transmission delay time, ρ represents the density of the heat supply medium in the pipeline, L k Denotes the pipe length, d k Indicates the inner diameter of the pipe, m k Indicates the flow rate of the heating medium in the pipeline.

7. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 6, characterized in that: The thermal operation parameters of the pipeline network include: pipeline heat dissipation coefficient, pipeline starting node temperature, ambient temperature and constant pressure specific heat. The calculation formula of the thermal energy exponential decay relationship model is as follows: Wherein, λ represents the heat dissipation coefficient of the pipeline, τ represents the heat energy transmission delay time, T start (t) represents the temperature of the node at the beginning of the pipeline at time t, T end (t+τ k ) represents the delay of heat energy transfer (t+τ k ) time, the temperature of the pipe end node, c p represents the specific heat at constant pressure, m k Indicates the heat supply medium flow rate of the pipeline, T a (t) is the ambient temperature at time t, L k Indicates the length of the pipeline.

8. The method for constructing a heat storage peak-shaving strategy for a heating unit according to claim 1, characterized in that: The calculation formula of the objective function is as follows: Among them, F represents net income, P e，j (t) represents the actual power supply power of the j-th heating unit in time period t, Q j (t) represents the actual heat supply of the j-th heating unit in period t, c e (t) represents the electricity price during period t, c h (t) represents the heat price during the period t, c coal represents the fuel price, f j (·) represents the coal consumption function of the j-th heating unit, represents the power generation operation and maintenance cost coefficient, Represents the heating operation and maintenance cost coefficient.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for constructing a heat storage peak-shaving strategy for a heating unit according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used for: constructing a heat storage peak-shaving strategy for a heating unit according to any one of claims 1 to 8.

Images