SOFC cogeneration system operation regulation method based on short-term load prediction

Abstract

This invention provides a method for operating and controlling SOFC (SOFC) combined heat and power (CHP) systems based on short-term load forecasting, aimed at improving the source-load mismatch problem in SOFC CHP systems. The method mainly includes: testing the system's heat and electricity output power under different combinations of operating conditions to obtain the correspondence between them; acquiring and preprocessing historical load data and meteorological data, training the model using a pre-set machine learning algorithm to generate a load forecasting model; obtaining hourly heat and electricity load forecasts from hourly weather forecasts issued by the meteorological bureau; determining corresponding operating conditions based on the load forecasts, and controlling the operation of the SOFC CHP system. This invention's control method, by introducing a reasonable machine learning algorithm, identifies the intrinsic relationship between building load and meteorological parameters, thereby predicting the load at a relatively low cost and achieving efficient control of the SOFC CHP system.

Description

Technical Field

[0001] This invention belongs to the field of distributed energy systems, and particularly relates to an operation method for a solid oxide fuel cell cogeneration system based on short-term load forecasting. Background Technology

[0002] Solid oxide fuel cells (SOFCs), as a new generation of fuel cells, can convert the chemical energy stored in fuel and oxidant into electrical energy at high temperatures (500-1000℃). They have advantages such as high power generation efficiency, wide range of fuel selection, high waste heat temperature, modular structure and flexible installation. The energy recovery efficiency through cogeneration can reach more than 80%, and they are considered to be the best prime mover choice for future cogeneration systems.

[0003] Because SOFC operation involves complex heat and mass transfer processes, strong thermoelectric coupling, and high hysteresis, SOFC combined heat and power systems typically exhibit slow dynamic response during actual operation. This makes it difficult to track load changes in real time and to quickly meet user heat load demands. Given the significant uncertainty of the heat and electricity loads required by users, achieving stable system operation under conditions of large load fluctuations is crucial. Furthermore, the mismatch between the output of the SOFC combined heat and power system and the energy load can lead to substantial energy waste or insufficient energy supply. Summary of the Invention

[0004] In view of this, the purpose of this invention is to overcome the slow response speed of SOFC (SO2-FC) cogeneration systems and provide a method for operating and controlling SOFC cogeneration systems based on short-term load forecasting. This method employs a feedback control strategy, using short-term load forecasting to adjust system operation in advance, thereby maximizing the supply-demand matching between the source and load sides while ensuring stable system operation. By acquiring time-varying building environmental variables and historical load data, the prediction model can learn load variation patterns from historical data. Simultaneously, by introducing appropriate machine learning algorithms, the inherent relationships between the above information are identified, thereby achieving load forecasting at a relatively low cost and realizing efficient control of the SOFC cogeneration system.

[0005] To address the aforementioned technical problems, this invention proposes a method for operation and control of SOFC combined heat and power systems based on short-term load forecasting, comprising the following steps:

[0006] Step S1: Test the system state and system output power of SOFC cogeneration system under different combinations of operating conditions to obtain the correspondence between the system output power and operating conditions of SOFC cogeneration system in the full load range. By changing the operating conditions, measure the system output power in the full load range to obtain the functional relationship between operating conditions and system output power. The system output power includes power generation and power supply.

[0007] Step S2: Obtain historical load data and meteorological data corresponding to the load data, and perform preprocessing. The preprocessing method includes deviation data processing and missing data processing. Construct a training sample set using the preprocessed historical load data and corresponding meteorological data. Train the training sample set using a machine learning algorithm to generate a load prediction model.

[0008] Step S3: The hourly weather forecast parameters issued by the meteorological bureau are used to form the predicted sample features of the load forecasting model. The load forecasting model obtains the predicted heat and electricity load values ​​for a specified target by reading the hourly weather forecast values ​​issued by the meteorological bureau. The predicted heat and electricity load values ​​include the predicted values ​​of daily average heat load and hourly heat load and electricity load, which are used to reflect the load changes in the next day. The load forecasting model outputs the predicted load at all time points, which is the predicted short-term load value.

[0009] Step S4: Based on the short-term load value predicted by the load prediction model, the functional relationship between the operating conditions and the system output power is obtained using step S1. The operating conditions under the target load are searched based on the pattern search algorithm, and a time control instruction table is generated. The input parameters of the SOFC cogeneration system, namely fuel flow rate, air flow rate, feedwater flow rate and operating current, are controlled in advance so that the power generation and heating power of the system reach the predicted short-term load value within a specified time, thereby realizing the output of the corresponding heat and electricity loads.

[0010] Furthermore, in the SOFC combined heat and power system operation and control method of the present invention, wherein:

[0011] In step S1, the operating conditions include: fuel flow rate n fuel airflow n air Water flow rate n water And the operating current I; the functional relationship between operating conditions and system output power is obtained by following the steps below:

[0012] S1-1) Maintain airflow n air Fuel flow rate n fuel and water supply flow rate n water With the current constant, the power generation Q of the SOFC combined heat and power system was tested under different operating currents I.ele,load and heating power Q heat,load ;

[0013] S1-2) Maintain airflow n air and fuel flow rate n fuel With constant flow rate n, water The test described in step S1-1) will be performed below;

[0014] S1-3) Maintain airflow n air With constant fuel flow rate n fuel The following steps (S1-2) will be performed;

[0015] S1-4) At different airflow rates n air The tests described in steps S1-3) will be performed below;

[0016] After the above four steps of testing, the power generation Q of the SOFC cogeneration system under any operating condition is obtained. ele,load and heating power Q heat,load The system power generation Q of the SOFC combined heat and power system was obtained across the entire load range. ele,load and heating power Q heat,load The functional relationship that corresponds one-to-one with the operation conditions, i.e.:

[0017] [Q ele,load Q heat,load ]=f(I,n air ,n fuel ,n water ).

[0018] In step S2, the load data includes electricity load data and heat load data. The electricity load data is collected by the difference between the meter readings at two adjacent moments. The heat load data is collected by monitoring the supply and return water temperatures and volumetric flow rates of the heating water supply. The meteorological data includes temperature, humidity, rainfall, and wind speed. The deviation data refers to values ​​that show a sudden increase or decrease in load values ​​at adjacent moments. The missing data refers to data that shows a single missing data point or consecutive zero values ​​over a period of time. The deviation data and the missing data are processed by interpolating the load values ​​using the Langran interpolation method.

[0019] In step S2, the process of constructing the training sample set is as follows: statistical methods are used to analyze the preprocessed data to obtain the statistical characteristics of the data and form a sample set; then the sample set is divided according to the statistical characteristics of the data; the state of the historical load data at each moment constitutes a training sample, thus forming the training sample set.

[0020] In step S2, the machine learning algorithm includes the Gradient Progressive Regression Tree (GBRT) algorithm, the Artificial Neural Network (ANN) algorithm, the Multiple Linear Regression algorithm, the Support Vector Regression (SVR) algorithm, the Autoregressive Differential Integrated Moving Average (ARIMA) algorithm, and any combination of the above algorithms.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] The SOFC (Solar-Fuel Combined Heat and Power) system operation and control method described in this invention introduces a reasonable machine learning algorithm to predict the energy load in the short term. Based on the load forecast, the corresponding SOFC cogeneration system operating parameters are obtained, allowing for advance control of various parameter variables. This method has good applicability and can be used for energy load forecasting in different regions and usage scenarios. Based on short-term load forecasting, feedforward control of the SOFC cogeneration system is achieved, avoiding the control time lag caused by the hysteresis and inertia of the SOFC cogeneration system. It also ensures, to a certain extent, the matching between system energy supply and load energy consumption, keeping the SOFC cogeneration system in a safe and energy-saving operating state, while improving user satisfaction. However, due to the system's own capacity limitations, it cannot fully meet the user's energy needs. Attached Figure Description

[0023] Figure 1 This is a flowchart of the SOFC combined heat and power system operation and control method of the present invention;

[0024] Figure 2 This is a schematic diagram illustrating the relationship between system operating conditions and output power in an embodiment.

[0025] Figure 3 This is a block diagram illustrating the architecture for controlling the operation of an SOFC combined heat and power system in this embodiment. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the following embodiments are by no means intended to limit the present invention.

[0027] like Figure 1 The present invention proposes an operation and control method for SOFC combined heat and power systems based on short-term load forecasting, which mainly includes the following steps:

[0028] Step S1: Test the system state and system output power of the SOFC cogeneration system under different combinations of operating conditions to obtain the correspondence between the system output power and operating conditions of the SOFC cogeneration system in the full load range. By changing the operating conditions, measure the system output power in the full load range to obtain the functional relationship between the operating conditions and the system output power. The system output power includes power generation and power supply.

[0029] The premise of the operation control method described in this invention is to determine the functional relationship between system operating conditions and output power. The operating conditions include: fuel flow rate n fuel airflow n air Water flow rate n water And the operating current I; by changing the operating conditions of the SOFC combined heat and power system, the output power of the system is measured across the entire load range, thereby determining the functional relationship between the system operating conditions and the output power, which can be obtained through actual testing under laboratory conditions. Figure 2 As shown, the specific process is as follows:

[0030] S1-1) Maintain airflow n air Fuel flow rate n fuel and water supply flow rate n water With the current constant, the power generation Q of the SOFC combined heat and power system was tested under different operating currents I. ele,load and heating power Q heat,load ;

[0031] S1-2) Maintain airflow n air and fuel flow rate n fuel With constant flow rate n, water The test described in step S1-1) will be performed below;

[0032] S1-3) Maintain airflow n air With constant fuel flow rate n fuel The following steps (S1-2) will be performed;

[0033] S1-4) At different airflow rates n air The tests described in steps S1-3) will be performed below;

[0034] After the above four steps of testing, the power generation Q of the SOFC cogeneration system under any operating condition is obtained. ele,load and heating power Q heat,load Finally, the system power generation Q of the SOFC cogeneration system was obtained across the entire load range. ele,load and heating power Q heat,load A one-to-one functional relationship with the operating conditions.

[0035] An alternative implementation method could be: fixing the operating current I and the air flow rate n. air The fuel flow rate n is obtained. fuel The x-axis represents the water flow rate n. water The vertical axis represents the output performance spectrum; for the performance spectrum, with a fixed operating current I, different airflow rates n are obtained. air The following is based on fuel flow rate n fuel The x-axis represents the water flow rate n. water The output performance spectrum is plotted on the vertical axis; the system performance output parameters at any operating point are obtained by iterating through different operating currents I; and the correspondence between the system output parameters and input parameters is obtained within the full load range of the system.

[0036] [Q ele,load Q heat,load ]=f(I,n air ,n fuel ,n water )

[0037] Step S2: An accurate machine learning load forecasting model typically requires a large number of samples for training. It can learn the impact of weather changes on electricity and heat loads from historical load data and corresponding historical meteorological data, thereby training a model capable of predicting future energy loads. The specific process is as follows: acquire historical load data and corresponding meteorological data, and preprocess them. The preprocessing methods include bias data processing and missing data processing. Construct a training sample set using the preprocessed historical load data and corresponding meteorological data. Train the training sample set using a machine learning algorithm to generate a load forecasting model.

[0038] The load data includes electricity load data and heat load data; the electricity load data is determined by the difference between the meter readings at two adjacent times; the heat load data is obtained by monitoring the supply and return water temperatures and volumetric flow rates of the heating water supply.

[0039] For any load, determine the system operating conditions corresponding to that load. As an optional implementation, the collected historical load data includes: historical electricity load data and historical heat load data; the collected historical meteorological data is the meteorological data corresponding to the historical load, including: temperature, humidity, rainfall, and wind speed.

[0040] The historical load data and the historical meteorological data are preprocessed. In the preprocessing method, the deviation data refers to the values ​​that show a sudden increase or decrease in load values ​​at adjacent times, and the missing data refers to the data that shows a single point of missing data or a continuous zero value over a period of time. The deviation data and the missing data are processed by interpolating the load values ​​using the Langron interpolation method.

[0041] The process of constructing a training sample set using the preprocessed historical data is as follows: statistical methods are used to analyze the preprocessed data to obtain the statistical characteristics of the data and form a sample set; then the sample set is divided according to the statistical characteristics of the data; the state of the historical load data at each moment constitutes a training sample, thus forming a training sample set.

[0042] The preset machine learning algorithm is used to train the training sample set to generate a load prediction model. It is worth noting that the machine learning algorithm includes, but is not limited to, Gradient Progressive Regression Tree (GBRT) algorithm; Artificial Neural Network (ANN) algorithm; Multiple Linear Regression algorithm; Support Vector Regression (SVR) algorithm; Differential Integrated Moving Average Autoregressive (ARIMA) algorithm; and any combination of the above algorithms.

[0043] Step S3: Using the hourly weather forecast values ​​issued by the meteorological bureau as the feature variables of the predicted sample, perform feature extraction processing on the predicted sample data to obtain a predicted sample feature set; input the predicted sample features into the load prediction model for prediction to obtain the predicted heat and electricity load values ​​for the specified object. That is: the hourly weather forecast parameter values ​​issued by the meteorological bureau are used to construct the predicted sample features of the load prediction model. The load prediction model obtains the predicted heat and electricity load values ​​for the specified target by reading the hourly weather forecast values ​​issued by the meteorological bureau. The predicted heat and electricity load values ​​include the predicted values ​​of daily average heat load and hourly heat load and electricity load, which are used to reflect the load changes in the next day; the load predicted by the load prediction model at all time points is the predicted short-term load value.

[0044] Step S4: Based on the short-term load value predicted by the load prediction model, and using the functional relationship between the operating conditions and system output power obtained in step S1, the operating conditions under the target load are searched using the pattern search algorithm to determine the corresponding operating conditions, and a time control instruction table is generated. The input parameters of the SOFC combined heat and power system, namely fuel flow rate, air flow rate, feedwater flow rate and operating current, are controlled in advance so that the power generation and heating power of the system reach the predicted short-term load value within a specified time, thereby realizing the output of the corresponding heat and electricity loads.

[0045] The framework diagram for implementing the above-mentioned SOFC cogeneration system operation and control method is as follows: Figure 3 As shown. It mainly includes a data sampling module, a prediction model generation module, a prediction load output module, and a control operation module. The functions of each module are as follows:

[0046] 1) The data sampling module is used to acquire historical data on electricity and heat loads, as well as historical weather information corresponding to these load data. The collected historical meteorological data includes temperature, humidity, rainfall, and wind speed. In addition, data needs to be read from the system's real-time operating data to update the historical load data and corresponding meteorological data in real time. For electricity load data, the difference between meter readings at two adjacent times can be used to determine the load; for heat load data, it can be obtained by monitoring the supply and return water temperatures and volumetric flow rates of the heating water supply.

[0047] 2) The prediction model generation module is used to preprocess historical load data, construct a training sample set, and generate a load prediction model.

[0048] Preprocessing: The historical load data and the historical gas phase data are preprocessed. The preprocessing method includes deviation data processing and missing data processing. Both deviation data processing and missing data processing use the Langron interpolation method to interpolate the load values.

[0049] Constructing a training sample set: After the preprocessing step, sample analysis can be performed on the preprocessed data to form a sample set. This sample analysis refers to using statistical methods, such as calculating the expected value and variance, to understand the statistical characteristics of the data, such as uniform distribution, Gaussian distribution, etc. Based on these statistical characteristics, the sample set is then divided. The state of the energy load at each time step can constitute a training sample.

[0050] Generate a load forecasting model: The preset machine learning algorithm is used to train the training sample set to obtain a load forecasting model.

[0051] 3) The predicted load output module uses the hourly weather forecast values ​​issued by the meteorological bureau as the feature variables of the predicted sample, performs feature extraction processing on the predicted sample data, and obtains the predicted sample feature set; the predicted sample features are input into the load prediction model for prediction to obtain the predicted heat and electricity load values ​​for the specified object. The predicted load at all time points is output, and the predicted load at all time points is the predicted short-term load.

[0052] 4) The control operation module mainly consists of a controller, which receives the short-term load value predicted by the predicted load output module, searches for the operating conditions under the target load, and issues control commands to the system to control the SOFC cogeneration system to achieve the specified heat and electricity load output. After obtaining the hourly heat and electricity load prediction values, the controller, based on a pattern search algorithm, searches for the system operating conditions corresponding to the predicted hourly heat and electricity load values, generates a time control command table, adjusts the control strategy of the SOFC cogeneration system, and controls the various input parameters of the system in advance so that the various adjustment parameters of the system reach the preset parameters at the specified time. As an optional implementation, the controller in this invention is specifically used for: obtaining the hourly heat and electricity load prediction values; searching for the system operating conditions corresponding to the predicted hourly heat and electricity load values; and sending control commands containing the system operating conditions to the SOFC cogeneration system to control the SOFC cogeneration system to provide the corresponding electric and heat power to the user.

[0053] Although the present invention has been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many modifications under the guidance of the present invention without departing from the spirit of the present invention, and these modifications are all within the protection scope of the present invention.

Claims

1. A method for operation and control of SOFC combined heat and power system based on short-term load forecasting, characterized in that, Includes the following steps: Step S1: Test the system state and system output power of SOFC cogeneration system under different combinations of operating conditions to obtain the correspondence between the system output power and operating conditions of SOFC cogeneration system in the full load range. By changing the operating conditions, measure the system output power in the full load range to obtain the functional relationship between operating conditions and system output power. The system output power includes power generation and power supply. Step S2: Obtain historical load data and meteorological data corresponding to the load data, and perform preprocessing. The preprocessing method includes deviation data processing and missing data processing. A training sample set is constructed using preprocessed historical load data and corresponding meteorological data; a load prediction model is generated by training the training sample set using machine learning algorithms. Step S3: The hourly weather forecast parameters issued by the meteorological bureau are used to form the predicted sample features of the load forecasting model. The load forecasting model obtains the predicted heat and electricity load values ​​for a specified target by reading the hourly weather forecast values ​​issued by the meteorological bureau. The predicted heat and electricity load values ​​include the predicted values ​​of daily average heat load and hourly heat load and electricity load, which are used to reflect the load changes in the next day. The load forecasting model outputs the predicted load at all time points, which is the predicted short-term load value. Step S4: Based on the short-term load value predicted by the load prediction model, the functional relationship between the operating conditions and the system output power is obtained using step S1. The operating conditions under the target load are searched based on the pattern search algorithm, and a time control instruction table is generated. The input parameters of the SOFC cogeneration system, namely fuel flow rate, air flow rate, feedwater flow rate and operating current, are controlled in advance so that the power generation and heating power of the system reach the predicted short-term load value within a specified time, thereby realizing the output of the corresponding heat and electricity loads.

2. The SOFC combined heat and power system operation and control method according to claim 1, characterized in that, In step S1, the operating conditions include: fuel flow rate n fuel , air flow rate n air , feed water flow rate n water , and working current I; and a functional relationship between the operating conditions and the system output power is obtained according to the following steps: S1-1) Maintain airflow n air Fuel flow rate n fuel and water supply flow rate n water With the current constant, the power generation Q of the SOFC combined heat and power system was tested under different operating currents I. ele,load and heating power Q heat,load ; S1-2) Maintain airflow n air and fuel flow rate n fuel With constant flow rate n, water The test described in step S1-1) will be performed below; S1-3) Maintain airflow n air With constant flow rate n, at different fuel flow rates fuel The following steps (S1-2) will be performed; S1-4) At different airflow rates n air The tests described in steps S1-3) will be performed below; After the above four steps of testing, the power generation Q of the SOFC cogeneration system under any operating condition is obtained. ele,load and heating power Q heat,load The system power generation Q of the SOFC combined heat and power system was obtained across the entire load range. ele,load and heating power Q heat,load The functional relationship that corresponds one-to-one with the operation conditions, i.e.: [Q ele,load ,Q heat,load ]=f(I,n air ,n fuel ,n water )。 3. The SOFC combined heat and power system operation and control method according to claim 1, characterized in that, In step S2, the load data includes electrical load data and heat load data; the electrical load data is collected by the difference between the meter readings at two adjacent times; the heat load data is collected by monitoring the supply and return water temperatures and volumetric flow rates of the heating water. The meteorological data includes temperature, humidity, rainfall, and wind speed; The deviation data refers to the values ​​that show a sudden increase or decrease in load values ​​at adjacent time points. The missing data refers to the data that shows a single point of missing data or consecutive zero values ​​over a period of time. The deviation data and the missing data are processed by interpolating the load values ​​using the Langron interpolation method.

4. The SOFC combined heat and power system operation and control method according to claim 1, characterized in that, In step S2, the process of constructing the training sample set is as follows: statistical methods are used to analyze the preprocessed data to obtain the statistical characteristics of the data and form a sample set; then the sample set is divided according to the statistical characteristics of the data; the state of the historical load data at each moment constitutes a training sample, thus forming the training sample set.

5. The SOFC combined heat and power system operation and control method according to claim 1, characterized in that, In step S2, the machine learning algorithm includes the Gradient Progressive Regression Tree (GBRT) algorithm, the Artificial Neural Network (ANN) algorithm, the Multiple Linear Regression algorithm, the Support Vector Regression (SVR) algorithm, the Autoregressive Differential Integrated Moving Average (ARIMA) algorithm, and any combination of the above algorithms.

Images