CN-122026506-A - Thermal power depth peak regulation-hydrogen production coupling collaborative optimization method and system

CN122026506ACN 122026506 ACN122026506 ACN 122026506ACN-122026506-A

Abstract

The invention belongs to the technical field of comprehensive energy systems, and provides a thermal power deep peak regulation-hydrogen production coupling collaborative optimization method and a thermal power deep peak regulation-hydrogen production coupling collaborative optimization system, wherein the technical scheme is that a power generation prediction result is obtained based on a historical power generation side input feature set and a power generation power prediction sub-model, and a hydrogen production yield prediction result is obtained based on a historical hydrogen production side input feature set and a hydrogen production yield prediction sub-model; and constructing a collaborative prediction model by taking the load power as a coupling variable, establishing a linkage relation between the power generation power prediction sub-model and the hydrogen production yield prediction sub-model based on the coupling variable, taking a power generation power prediction result as constraint of hydrogen production side reasoning, and obtaining a power generation power prediction result and a hydrogen production yield prediction result based on power generation side and hydrogen production side data acquired in real time and the trained collaborative prediction model. And the prediction precision and response consistency of the cooperative operation of the thermal power and the hydrogen production are improved.

Inventors

ZHANG LIANGCHEN
CHEN GAOLIANG
WANG HUI
LV YANHONG
CHEN CHEN
Dan Shaopeng
LI YUFENG
WANG LUYUAN
HU YIGONG
CAO HONGZHEN
LIU JIANMIN
LI XIN
YU HAIMENG
ZHANG JIBING
LV YUJUAN

Assignees

山东电力工程咨询院有限公司

Dates

Publication Date: 20260512
Application Date: 20251219

Claims (10)

1. A thermal power depth peak regulation-hydrogen production coupling collaborative optimization method is characterized by comprising the following steps: Acquiring an input characteristic set of a historical power generation side and an input characteristic set of a hydrogen production side; training the constructed collaborative prediction model based on the input feature set of the historical power generation side and the input feature set of the hydrogen production side to obtain a trained collaborative prediction model, wherein the construction process of the collaborative prediction model comprises the following steps: obtaining a power generation prediction result based on the historical power generation side input feature set and the power generation prediction sub-model, and obtaining a hydrogen generation yield prediction result based on the historical hydrogen generation side input feature set and the hydrogen generation yield prediction sub-model; Taking load power as a coupling variable, establishing a linkage relation between the power generation power prediction sub-model and the hydrogen production yield prediction sub-model based on the coupling variable, and constructing a collaborative prediction model by taking a power generation power prediction result as constraint of hydrogen production side reasoning; And obtaining a power generation power prediction result and a hydrogen production yield prediction result based on the data of the power generation side and the hydrogen production side obtained in real time and the trained collaborative prediction model.
2. The thermal power depth peak shaving-hydrogen production coupling collaborative optimization method according to claim 1, wherein a power generation power prediction sub-model and a hydrogen production yield prediction sub-model adopt symmetrical parallel three-branch and fusion layer structures, and the unified preprocessing data set is used as common input, and the parallel structure is composed of a multi-layer perceptron branch, a one-dimensional convolution branch and a multi-head attention branch, and the two models execute the following steps in parallel: The input corresponding data set is branched by a multi-layer perceptron, the global nonlinear relation is extracted, and a first feature vector is output; Extracting local change modes and trends from the input corresponding data sets through one-dimensional convolution branches, and outputting second feature vectors; The input corresponding data sets are branched through multiple attention branches, a dynamic weighting relation between input features is calculated, and attention features are obtained through a linear layer after the input of the flat; And splicing the first feature vector, the second feature vector and the attention feature, fusing the first feature vector, the second feature vector and the attention feature by two layers of fully-connected networks, and finally outputting a prediction result by a single-node linear regression layer.
3. The thermal power depth peak shaving-hydrogen production coupling collaborative optimization method according to claim 1, wherein in the power generation side prediction step, a first feature vector of branch output of the multi-layer perceptron is used for representing a global working condition affecting power generation, a second feature vector of branch output of the one-dimensional convolution is used for representing a local dynamic feature of rapid load fluctuation, short-time climbing or falling, and the attention feature of the multi-head attention branch output reflects contribution weights of input variables to prediction under the current working condition by calculating an attention weight matrix; In the hydrogen production side prediction step, a first feature vector of the multi-layer perceptron branch output is used for representing a steady-state nonlinear mapping relation of an electrolysis system under the constraints of electrolysis power, coal consumption rate and coupling load, a second feature vector of the one-dimensional convolution branch output is used for representing a short-time yield fluctuation trend caused by rapid adjustment of the electrolysis power, and the attention feature of the multi-head attention branch output is used for representing dominant dynamic migration of the electrolysis power, the coupling load and the coal consumption rate under different working conditions.
4. The thermal power depth peaking-hydrogen production coupling collaborative optimization method according to claim 1, wherein the collaborative prediction model is expressed as: , , , Wherein, the In order to predict the generated power, the generated power is predicted, Representing the generated power predictor model, In order to be able to determine the temperature, In order to be a degree of humidity, In order to achieve a concentration of the particulate matter, Is the heat value of the fire coal, Is the intensity of solar radiation; For coupling the load power, alpha, beta are coupling coefficients, epsilon is a systematic error term, The predicted result of the hydrogen production yield is shown, In order to produce a hydrogen production yield predictor model, In order to achieve the coal consumption rate, Is electrolytic power.
5. The thermal power depth peak shaving-hydrogen production coupling collaborative optimization method according to claim 1, wherein when the collaborative prediction model is trained, a historical actual measurement result is used as a supervision quantity, and training and parameter updating are carried out on the collaborative prediction model to obtain a converged collaborative prediction model, comprising: the historical data of the power generation side and the hydrogen production side are subjected to unified cleaning, alignment and normalization treatment, so that the data quality and consistency are ensured; Independent pre-training is carried out on the power generation power predictor model and the hydrogen production yield predictor model respectively to obtain a trained power generation power predictor model and a trained hydrogen production yield predictor model, and basic prediction capacity is obtained; Establishing a coupling relation model between the generated power and the load power based on historical data, and determining a mapping relation between the generated power and the load power; And simultaneously optimizing the prediction precision of the generated power, the prediction precision of the hydrogen production yield and the fitting degree of the coupling relation based on the constructed multi-objective loss function, and stopping training when the total loss function changes tend to be stable and reach the preset precision requirement to obtain a converged collaborative prediction model.
6. The thermal power depth peak shaving-hydrogen production coupling collaborative optimization method according to claim 1, wherein the multi-objective loss function is: Wherein, the The number of training samples; Indexing for the samples; , the actual measurement and the forecast of the power generation power (or the unified caliber power generation capacity) are respectively carried out; , The actual measurement and the predicted hydrogen production yield are respectively; , Load powers (coupling consistency constraints) predicted for the measured/coupled models, respectively; , , , a weight coefficient; is a regular term used to suppress overfitting, and Θ is a full model parameter set.
7. The thermal power depth peak regulation-hydrogen production coupling collaborative optimization system is characterized by comprising: the historical data acquisition module is used for acquiring an input characteristic set of the historical power generation side and an input characteristic set of the hydrogen production side; The collaborative optimization module is used for training the constructed collaborative prediction model based on the input feature set of the historical power generation side and the input feature set of the hydrogen production side to obtain a trained collaborative prediction model, wherein the construction process of the collaborative prediction model comprises the following steps: obtaining a power generation prediction result based on the historical power generation side input feature set and the power generation prediction sub-model, and obtaining a hydrogen generation yield prediction result based on the historical hydrogen generation side input feature set and the hydrogen generation yield prediction sub-model; Taking load power as a coupling variable, establishing a linkage relation between the power generation power prediction sub-model and the hydrogen production yield prediction sub-model based on the coupling variable, and constructing a collaborative prediction model by taking a power generation power prediction result as constraint of hydrogen production side reasoning; And the prediction module is used for obtaining a power generation power prediction result and a hydrogen production yield prediction result based on the data of the power generation side and the hydrogen production side obtained in real time and the trained collaborative prediction model.
8. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps of a thermal power depth peaking-hydrogen production coupling co-optimization method as claimed in any one of claims 1-6.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor performs the steps of a thermal power depth peaking-hydrogen production coupling co-optimization method as claimed in any one of claims 1-6 when the program is executed.
10. A program product, which is a computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of a thermal power depth peaking-hydrogen production coupling co-optimization method according to any one of claims 1-6.

Description

Thermal power depth peak regulation-hydrogen production coupling collaborative optimization method and system Technical Field The invention belongs to the technical field of comprehensive energy systems, and particularly relates to a thermal power depth peak regulation-hydrogen production coupling collaborative optimization method and system. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. Renewable energy sources such as wind power, photovoltaic and the like are rapidly increased in installed capacity, and the electric power system is pushed to display high fluctuation and strong uncertainty. The traditional thermal power generating unit is used as a peak regulation ballast, and the operation mode is changed from the original stable base load working condition into a deep peak regulation state of frequent start-stop and rapid climbing. Under the high-proportion new energy grid-connected background, the peak regulation frequency of the thermal power generating unit is increased, the lowest running load is continuously detected, a series of problems of unstable thermodynamic system, limited heat supply, increased coal consumption, shortened equipment service life and the like are brought, and the depth and breadth of thermal power flexibility release are seriously restricted. On the other hand, the water electrolysis hydrogen production technology is used as an important adjustable flexible load form, has the operation characteristics of quick response and large-scale adjustment, can actively absorb electric energy in a grid load valley period or in a renewable energy surplus period, converts the electric energy into hydrogen in a chemical energy form, and realizes electric-gas energy form conversion. Especially in the low-load operation stage of the thermal power, if the hydrogen production device is embedded into the thermal power plant or the power grid side as a virtual load, the peak regulation pressure of the unit can be effectively relieved, the thermal efficiency is improved, the localized preparation and storage of hydrogen energy resources can be realized, and the form of a multi-energy complementary energy system is promoted. However, in the existing research, most of the thermal power system and the hydrogen production system are independently modeled and distributed and scheduled, and a systematic modeling mechanism and prediction capability considering the coupling dynamics of the thermal power system and the hydrogen production system are lacking. On one hand, the thermal power peak regulation process has the operation characteristics of nonlinearity, multiple time lags and strong physical constraint, and on the other hand, the load response of the electrolytic hydrogen production system is also limited by the multi-parameter dynamic characteristics of the temperature, voltage, current and the like of the electrolytic tank, and a complex 'electric-thermal-gas' coupling relation exists between the two. Therefore, the thermal power output or the hydrogen output is predicted independently, and the mutual feedback effect of the thermal power output and the hydrogen output in the dynamic operation process is difficult to reflect, so that the scheduling strategy is lagged or redundant, and the overall economy and the energy utilization efficiency of the system are affected. Disclosure of Invention In order to solve at least one technical problem in the background art, the invention provides a thermal power deep peak regulation-hydrogen production coupling collaborative optimization method and a thermal power deep peak regulation-hydrogen production coupling collaborative optimization system, which realize unified modeling, collaborative prediction and bidirectional optimization control of a thermal power and hydrogen production system by introducing a load power coupling variable and a multi-branch deep neural network structure, and remarkably improve prediction precision, system stability and energy management intelligence level under complex working conditions. In order to achieve the above purpose, the present invention adopts the following technical scheme: The first aspect of the invention provides a thermal power depth peak shaving-hydrogen production coupling collaborative optimization method, which comprises the following steps: Acquiring an input characteristic set of a historical power generation side and an input characteristic set of a hydrogen production side; training the constructed collaborative prediction model based on the input feature set of the historical power generation side and the input feature set of the hydrogen production side to obtain a trained collaborative prediction model, wherein the construction process of the collaborative prediction model comprises the following steps: obtaining a power generation prediction result based on the historical power generati