Unit load advanced control method based on battery energy storage state

By constructing a unit load advance control method based on battery energy storage status and optimizing load distribution using a unit heat storage coefficient prediction model, the flexibility and lifespan issues of thermal power units and battery energy storage systems during frequency regulation are solved, thereby improving frequency regulation effect and unit load regulation performance.

CN118971179APending Publication Date: 2024-11-15NORTHWEST BRANCH OF CHINA DATANG CORP SCI & TECH RES INST +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410950127.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In existing technologies, when thermal power units participate in grid frequency regulation, they cannot flexibly determine whether the output can meet the load distribution command based on the actual operating conditions of the unit. Furthermore, the response capability of battery energy storage systems is limited under overcharging or over-discharging conditions, which affects the frequency regulation effect and service life.

Method used

The method for advance control of unit load based on battery energy storage status is to construct a predictive model of unit heat storage coefficient, and use convolutional neural network, bidirectional gated cyclic unit and multi-head self-attention mechanism to adjust the advance control amount of unit load setting in real time. Combined with the state of charge of energy storage system and unit operating conditions, the load distribution is optimized.

Benefits of technology

It enables optimized load allocation based on the real-time status of generating units and energy storage systems, extends the service life of energy storage systems, improves the load regulation performance of generating units, and enhances the quality of frequency regulation services.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118971179A_ABST
    Figure CN118971179A_ABST
Patent Text Reader

Abstract

The invention discloses a unit load advanced control method based on a battery energy storage state. The unit load advanced control method is specifically implemented according to the following steps that 1, the unit state is judged according to an AGC instruction and the current load of a unit; 2, judging whether the energy storage output meets an AGC instruction or not; step 3, establishing a unit heat storage coefficient and predicting a unit heat storage data set; 4, constructing a unit heat storage coefficient prediction model, and training the unit heat storage coefficient prediction model based on the predicted unit heat storage data set in the step 3; and step 5, inputting data during operation of the unit into the unit heat storage coefficient prediction model trained in the step 4, generating a heat storage coefficient, constructing a corresponding relation between the heat storage coefficient and a load instruction advance quantity, generating a load instruction based on the generated heat storage coefficient, and executing the load instruction by the unit. The problem that in the prior art, the frequency modulation task is distributed to the energy storage system and the conventional thermal power generating unit according to the fixed proportionality coefficient, and consequently the unit load is too high is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of thermal power energy storage automation control, and in particular relates to a method for advanced load control of a unit based on battery energy storage status. Background Art

[0002] With the expansion of wind power and photovoltaic grid-connected scale, the role of thermal power units in the power grid has gradually changed to providing more flexible and efficient auxiliary services. However, when participating in the frequency regulation of the power grid, thermal power units may suffer severe wear and tear due to equipment characteristics, resulting in a decrease in short-term power throughput capacity, affecting the load regulation potential, and thus prolonging the system response time and reducing the frequency regulation capability, making it difficult to achieve the expected frequency regulation target. The development of battery energy storage technology provides important technical support for new energy power systems. Its rapid response characteristics can effectively reduce the frequency regulation burden of conventional thermal power units and improve the quality of frequency regulation services.

[0003] At present, when receiving automatic generation control (AGC) instructions from the power grid, units equipped with battery energy storage usually allocate frequency regulation tasks to energy storage systems and conventional thermal power units according to a fixed proportional coefficient. However, this method has some limitations: for thermal power units, it cannot flexibly judge whether its output can meet the load distribution instructions based on the actual operating conditions of the units, nor can it fully utilize the thermal storage capacity of the units, thus failing to fully utilize its frequency regulation potential. For battery energy storage systems, when the energy storage system is overcharged or over-discharged, its response ability will be limited and it will not be able to effectively execute the assigned instructions. In addition, the number of charge and discharge times of the energy storage system will change with the fluctuation of the frequency modulation signal, which may have an adverse effect on the service life of the energy storage system. Summary of the invention

[0004] The purpose of the present invention is to provide a method for advanced control of unit load based on battery energy storage status, which solves the problem in the prior art that the frequency regulation task is allocated to the energy storage system and the conventional thermal power unit according to a fixed proportional coefficient, resulting in excessive unit load.

[0005] The technical solution adopted by the present invention is a method for advanced load control of a unit based on the battery energy storage state, which is specifically implemented according to the following steps:

[0006] Step 1, judging the unit status according to the AGC instruction and the current load of the unit;

[0007] Step 2, determine whether the energy storage output meets the AGC instruction;

[0008] Step 3, establishing the unit heat storage coefficient and predicting the unit heat storage data set;

[0009] Step 4, constructing a unit heat storage coefficient prediction model and training the unit heat storage coefficient prediction model based on the predicted unit heat storage data set in step 3;

[0010] Step 5, input the data of the unit during operation into the unit heat storage coefficient prediction model trained in step 4, generate the heat storage coefficient, build the corresponding relationship between the heat storage coefficient and the load instruction lead amount, generate the load instruction based on the generated heat storage coefficient and execute it by the unit.

[0011] The present invention is also characterized in that

[0012] In step 1, the unit status includes load increase, load decrease, and stable state. The load increase means that the AGC instruction is greater than the current load of the unit, the load decrease means that the AGC instruction is less than the current load of the unit, and the stable state means that the AGC instruction is equal to the current load of the unit.

[0013] In step 2, determining whether the energy storage output meets the AGC instruction is specifically that the adjustment range of the energy storage system is 0.2 to 0.9. When the energy storage charge state of the energy storage system is not within the adjustment range, the energy storage capacity cannot meet the AGC instruction.

[0014] In step 3, the heat storage coefficient of the unit is the ratio of the change in the steam flow stored in the water-cooled wall, the drum and the superheater to the rate of change of the drum pressure. The ratio C is shown in formula (1):

[0015]

[0016] In the formula, C is the heat storage coefficient of the unit; m represents the main steam flow rate; p is the drum pressure; and t is time.

[0017] The prediction unit heat storage data set in step 3 includes the input set {X i} and the output set {Y i}, the input set {X i} is the unit parameters 300 seconds before time t0 and the drum pressure and main steam flow rate of the unit at time t1, and the unit parameters include load, main steam flow rate, feed water flow rate, main steam pressure set value, main steam pressure actual value, total air volume, fuel volume, main steam temperature, furnace negative pressure, comprehensive valve position, load change rate, and drum pressure; construct i input samples at different times to form an input set {X i};

[0018] The output set {Y i} is the heat storage coefficient of the unit at time t0, and i output samples are constructed according to different times to form an output set {Y i}.

[0019] The unit heat storage coefficient prediction model includes an input layer, a convolutional neural network layer, a bidirectional gated recurrent unit layer, a multi-head self-attention mechanism layer, and an output layer.

[0020] The convolutional neural network layer includes two convolutional layers and one pooling layer; the bidirectional gated recurrent unit layer includes two GRU networks; the multi-head self-attention mechanism layer and the output layer are connected through a fully connected layer.

[0021] In step 4, the unit heat storage coefficient prediction model is trained based on the predicted unit heat storage data set in step 3, and is specifically implemented in the following steps:

[0022] Step 4.1: Input set {X i} Input the input layer, the input feature vector is 13×16;

[0023] Step 4.2, extract the feature vector of the data set through the convolution layer, and reduce the extracted feature vector to one-dimensional data through the pooling layer;

[0024] Step 4.3: The bidirectional gated recurrent unit layer performs bidirectional training on the data output from step 4.2 to learn the deeper temporal features between the unit operation data and the heat storage coefficient, and then uses all its output data as the input data of the multi-head self-attention mechanism layer;

[0025] Step 4.4, the hidden state h of the output of the bidirectional gated recurrent unit layer by the multi-head self-attention mechanism layer t Assign different weight values;

[0026] Step 4.5, the output layer uses Tanh as the activation function for the data processed in step 4.4, and maps the output to the interval (-1,1) through the activation function to obtain the predicted unit heat storage coefficient;

[0027] Step 4.6, compare the predicted unit heat storage coefficient with the output set {Y i} for comparison, repeat steps 4.1 to 4.5, and perform multiple iterative training until the model converges.

[0028] In step 4.4, the hidden state h t Different weight values ​​are assigned, as shown in formula (2) and formula (3):

[0029] a t =f(q,h t ) (2);

[0030]

[0031] In the formula, f is the GRU network, q is the prediction vector of the GRU network, and h t is the GRU hidden state at time step t, αt is the weight of time step t;

[0032] The step 4.5 is specifically shown in formula (4):

[0033]

[0034] Where x is the input value and e is the natural logarithm base.

[0035] The corresponding relationship between the heat storage coefficient and the load command lead amount is: when the heat storage coefficient is -1, the load command lead amount is -10; when the heat storage coefficient is -0.6, the load command lead amount is -5; when the heat storage coefficient is -0.4, the load command lead amount is -2; when the heat storage coefficient is -0.1, the load command lead amount is -0; when the heat storage coefficient is 0, the load command lead amount is 0; when the heat storage coefficient is 0.1, the load command lead amount is 0; when the heat storage coefficient is 0.4, the load command lead amount is 2; when the heat storage coefficient is 0.6, the load command lead amount is 5; when the heat storage coefficient is 1, the load command lead amount is 10.

[0036] In step 5, the load instruction based on the generated heat storage coefficient is generated specifically as follows: when the unit is in a load increase state and the energy storage charge state is less than the minimum output limit, the load instruction lead amount corresponding to the heat storage coefficient is increased in the generated load instruction; when the unit is in a load reduction state and the energy storage charge state is greater than the energy storage charging limit, the load instruction lead amount corresponding to the heat storage coefficient is reduced in the generated load instruction.

[0037] The beneficial effects of the present invention are:

[0038] The present invention is a method for advanced load control of a unit based on the battery energy storage state. By real-time reading of the energy storage battery and the relevant parameters of the unit, the unit load increase, load reduction, and steady state are determined according to the unit AGC instruction; according to the energy storage state of charge (SOC) and output parameters of the energy storage battery, it is determined whether it can meet the load change requirements of the unit. When the energy storage battery cannot meet the requirements, the unit heat storage coefficient is predicted in combination with the real-time operating conditions of the unit, and the advance control amount of the unit load setting is automatically adjusted. It can make full use of the unit heat storage while avoiding overcharging or over-discharging of the energy storage system, that is, it prolongs the service life of the energy storage system and comprehensively considers the energy storage state and the unit conditions, and utilizes the unit heat storage according to the operating conditions to improve the unit load regulation performance index. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of the method for advanced load control of a generator set based on the battery energy storage state of the present invention;

[0040] Figure 2 It is a structural schematic diagram of a unit heat storage coefficient prediction model in a unit load advance control method based on battery energy storage status of the present invention;

[0041] Figure 3 It is a schematic diagram of load change of the unit in response to AGC instruction in Example 3 of the present invention. DETAILED DESCRIPTION

[0042] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] Example 1

[0044] The present invention provides a method for controlling load advance of a generator set based on the battery energy storage state. Figure 1 As shown, the specific implementation steps are as follows:

[0045] Step 1, judging the unit status according to the AGC instruction and the current load of the unit;

[0046] In step 1, the unit status includes load increase, load decrease, and stable state. The load increase means that the AGC instruction is greater than the current load of the unit, the load decrease means that the AGC instruction is less than the current load of the unit, and the stable state means that the AGC instruction is equal to the current load of the unit.

[0047] Step 2, determine whether the energy storage output meets the AGC instruction;

[0048] When the AGC instruction of the unit grid changes, the energy storage system combines with the thermal power unit to output power. Taking into account the life of the energy storage system, the energy storage state of charge (SOC) is generally maintained at 0.2-0.9, which is the adjustment range of the energy storage system. When the SOC of the energy storage system is not within the adjustment range, the energy storage system does not respond to the AGC instruction. When the AGC instruction is greater than the dead zone of the energy storage adjustment, and the SOC is within the allowable range, the energy storage responds to the AGC instruction at the rated maximum power. Step 3, establish the unit heat storage coefficient and the predicted unit heat storage data set;

[0049] When the energy storage state of charge (SOC) is not within the adjustment range, the time it takes for the unit to reach the AGC command will be extended. In order to ensure that the unit's responsiveness does not decrease while fully utilizing the unit's output, the computer unit is ahead of the load.

[0050] The amount of lead load of the unit mainly depends on the heat storage of the unit at the time when the AGC command changes. When the unit has sufficient heat storage, the unit can provide more lead load, and the superimposed energy storage output can respond to the AGC command faster and more accurately, better meeting the grid regulation needs. When the unit has insufficient heat storage, the actual output of the unit in a short period of time is much smaller than that of the unit in a normal state, and the output of the energy storage system is small, and it cannot quickly make up for the insufficient actual output of the unit, resulting in a lack of output of the thermal power energy storage system, and the AGC comprehensive index cannot be met. The unit heat storage coefficient is defined as the ratio of the change in the steam flow stored in the water-cooled wall, drum and superheater to the rate of change of the drum pressure. The ratio C is shown in formula (1):

[0051]

[0052] In the formula, C is the heat storage coefficient of the unit; m represents the main steam flow rate; p is the drum pressure; and t is time.

[0053] Furthermore, the heat storage coefficient C0 when the unit AGC command changes t0 is determined by the unit combustion conditions. Due to the inertia of boiler combustion, the heat storage coefficient can be predicted based on the unit operation data.

[0054] The prediction unit heat storage data set in step 3 includes the input set {X i} and the output set {Y i}, the input set {X i} is the unit parameters 300 seconds before time t0 and the drum pressure and main steam flow rate of the unit at time t1, and the unit parameters include load, main steam flow rate, feed water flow rate, main steam pressure set value, main steam pressure actual value, total air volume, fuel volume, main steam temperature, furnace negative pressure, comprehensive valve position, load change rate, and drum pressure; construct i input samples at different times to form an input set {X i};

[0055] The output set {Y i} is the unit heat storage coefficient at time t0, and i output samples are constructed according to different times to form an output set {Y i}.

[0056] Step 4, constructing a unit heat storage coefficient prediction model and training the unit heat storage coefficient prediction model based on the predicted unit heat storage data set in step 3;

[0057] Step 5, input the data of the unit operation into the unit heat storage coefficient prediction model trained in step 4, generate the heat storage coefficient, and build the corresponding relationship between the heat storage coefficient and the load command advance amount. The specific corresponding relationship is as follows:

[0058]

[0059] Finally, through the correspondence between the above heat storage coefficient and the load instruction advance amount, a load instruction is generated based on the generated heat storage coefficient and executed by the unit.

[0060] Specifically, when the unit is in a load-increasing state and the energy storage charge state is less than the minimum output limit, the load instruction lead amount corresponding to the heat storage coefficient is increased in the generated load instruction; when the unit is in a load-reducing state and the energy storage charge state is greater than the energy storage charging limit, the load instruction lead amount corresponding to the heat storage coefficient is reduced in the generated load instruction.

[0061] Example 2

[0062] On the basis of Example 1, Figure 2 As shown, the unit heat storage coefficient prediction model in the unit load advance control method based on the battery energy storage state of the present invention includes a convolutional neural network (CNN) layer, a bidirectional gated recurrent unit (BiGRU) layer, a multihead self-attention mechanism (MSA) layer, and an output layer.

[0063] The convolutional neural network layer includes two convolutional layers and one pooling layer; the multi-head self-attention mechanism layer and the output layer are connected through a fully connected layer.

[0064] In step 4, the unit heat storage coefficient prediction model is trained based on the predicted unit heat storage data set in step 3, and is specifically implemented in the following steps:

[0065] Step 4.1: Input set {X i} Input the input layer, the input feature vector is 13×16;

[0066] Step 4.2, extract the feature vector of the data set through the convolution layer, and reduce the extracted feature vector to one-dimensional data through the pooling layer;

[0067] Step 4.3, the bidirectional gated recurrent unit layer consists of two GRU networks, one processes the time series data from front to back, and the other processes the time series data from back to front. The data output from step 4.2 is trained bidirectionally to learn the deeper time series features between the unit operation data and the heat storage coefficient, and then all its output data is used as the input data of the multi-head self-attention mechanism layer;

[0068] Step 4.4, the hidden state h of the output of the bidirectional gated recurrent unit layer by the multi-head self-attention mechanism layer tDifferent weight values ​​are assigned to highlight the impact of key features on the prediction results, so that the prediction model can more effectively focus on the time points that have a greater impact on the prediction results. Specifically, as shown in formula (2) and formula (3):

[0069] a t =f(q,h t ) (2);

[0070]

[0071] In the formula, f is the GRU network, q is the prediction vector of the GRU network, and h t is the GRU hidden state at time step t, α t is the weight at time step t.

[0072] Step 4.5: The output layer uses Tanh as the activation function for the data processed in step 4.4. The output is mapped to the interval (-1, 1) through the activation function to obtain the predicted unit heat storage coefficient, as shown in formula (4):

[0073]

[0074] Where x is the input value and e is the natural logarithm base.

[0075] Step 4.6, compare the predicted unit heat storage coefficient with the output set {Y i} for comparison, repeat steps 4.1 to 4.5, and perform multiple iterative training until the model converges.

[0076] Example 3

[0077] The load advance control method of the unit based on the battery energy storage state in this embodiment is implemented according to the following steps:

[0078] Step 1: Determine the unit status according to the AGC instruction and the current load of the unit;

[0079] According to the automatic generation control (AGC) command of the unit and the current load of the unit, it is judged whether the unit is in the load increase, load reduction or stable state. If the AGC command is greater than the current load of the unit, it is load increase; if the AGC command is less than the current load of the unit, it is load reduction; and if the AGC command is equal to the current load of the unit, it is stable.

[0080] Step 2: Analyze whether the unit and energy storage output meet the AGC instructions;

[0081] like Figure 3 As shown in the figure, when the AGC command of the unit changes at time t0, the AGC command forms a load command after the set rate limit, and the unit changes the load according to the load command. When the energy storage SOC is within the allowable range, the energy storage output P e, then the load of the unit at this time is the load P1 with superimposed energy storage.

[0082] Step 3, establishing the unit heat storage coefficient and predicting the unit heat storage data set;

[0083] like Figure 3 As shown, when the unit responds to the AGC command, the heat storage is maintained for a short time, that is, the unit receives the AGC command change at time t0, and the unit's heat storage is consumed at time t1, and the unit relies on the energy of the fuel system to respond to the load. Then the heat storage coefficient C0 of the unit when the AGC command changes t0 is shown in formula (5):

[0084]

[0085] The time interval between t0 and t1 is selected according to the unit characteristics, and 120s is selected for the implementation unit.

[0086] The heat storage coefficient C0 when the unit AGC instruction changes to t0 is determined by the unit combustion conditions. Due to the inertia of boiler combustion, the heat storage coefficient can be predicted based on the unit operation data. From the distributed control system (DCS) system of the thermal power unit, select the unit parameters 300 seconds before t0, including load, main steam flow, feed water flow, main steam pressure set value, main steam pressure actual value, total air volume, fuel volume, main steam temperature, furnace negative pressure, comprehensive valve position, load change rate, drum pressure, and drum pressure and main steam flow at t1. The above unit operation data are combined into an input sample. According to the needs of model training, i input samples are constructed, and the constructed samples are combined into an input set {X i According to formula (2), the heat storage coefficient of the unit at time t0 is calculated as the output sample, and i output samples are constructed. The constructed samples form the output set {Y i}.

[0087] Step 4, constructing a unit heat storage coefficient prediction model and training the unit heat storage coefficient prediction model based on the predicted unit heat storage data set in step 3;

[0088] Step 5, input the data of the unit operation into the unit heat storage coefficient prediction model trained in step 4 to generate the heat storage coefficient, and through the correspondence between the heat storage coefficient and the load instruction lead amount, generate the load instruction based on the generated heat storage coefficient and execute it by the unit.

[0089] Specifically, when the unit is in a load-raising state and the energy storage SOC is less than the minimum output limit, the unit load command increases by the calculated load command advance amount, such as Figure 3As shown in the figure, the superimposed unit lead amount enables the unit to respond to the AGC load requirement faster and more accurately. When the unit is in a load reduction state and the energy storage SOC is greater than the energy storage charging limit, the unit reduces the corresponding load according to the calculated load command lead amount to respond to the AGC command faster.

Claims

1. A method for advanced load control of a unit based on battery energy storage status, characterized in that: Follow the steps below to implement it: Step 1, judging the unit status according to the AGC instruction and the current load of the unit; Step 2, determine whether the energy storage output meets the AGC instruction; Step 3, establishing the unit heat storage coefficient and predicting the unit heat storage data set; Step 4, constructing a unit heat storage coefficient prediction model and training the unit heat storage coefficient prediction model based on the predicted unit heat storage data set in step 3; Step 5, input the data of the unit during operation into the unit heat storage coefficient prediction model trained in step 4, generate the heat storage coefficient, build the corresponding relationship between the heat storage coefficient and the load instruction lead amount, generate the load instruction based on the generated heat storage coefficient and execute it by the unit.

2. The method for advanced load control of a generator set based on battery energy storage status according to claim 1, characterized in that: The unit status in step 1 includes load increase, load decrease, and stable state. The load increase means that the AGC instruction is greater than the current load of the unit, the load decrease means that the AGC instruction is less than the current load of the unit, and the stable state means that the AGC instruction is equal to the current load of the unit.

3. The method for advanced load control of a generator set based on battery energy storage status according to claim 1, characterized in that: In the step 2, determining whether the energy storage output meets the AGC instruction is specifically that the adjustment range of the energy storage system is 0.2 to 0.

9. When the energy storage charge state of the energy storage system is not within the adjustment range, the energy storage capacity cannot meet the AGC instruction.

4. The method for advanced load control of a generator set based on battery energy storage status according to claim 1, characterized in that: The heat storage coefficient of the unit in step 3 is the ratio of the change in the steam flow rate stored in the water-cooled wall, the drum and the superheater to the rate of change of the drum pressure. The ratio C is shown in formula (1): In the formula, C is the heat storage coefficient of the unit; m represents the main steam flow rate; p is the drum pressure; and t is time.

5. The method for advanced load control of a generator set based on battery energy storage status according to claim 1, characterized in that: The prediction unit heat storage data set in step 3 includes an input set {X i } and the output set {Y i }, the input set {X i } is the unit parameters 300 seconds before time t0 and the drum pressure and main steam flow rate of the unit at time t1, and the unit parameters include load, main steam flow rate, feed water flow rate, main steam pressure set value, main steam pressure actual value, total air volume, fuel volume, main steam temperature, furnace negative pressure, comprehensive valve position, load change rate, and drum pressure; construct i input samples at different times to form an input set {X i }; The output set {Y i } is the heat storage coefficient of the unit at time t0, and i output samples are constructed according to different times to form an output set {Y i }.

6. The method for advanced load control of a generator set based on battery energy storage status according to claim 5, characterized in that: The unit heat storage coefficient prediction model includes an input layer, a convolutional neural network layer, a bidirectional gated recurrent unit layer, a multi-head self-attention mechanism layer and an output layer. The convolutional neural network layer includes two convolutional layers and one pooling layer; the bidirectional gated recurrent unit layer includes two GRU networks; the multi-head self-attention mechanism layer and the output layer are connected through a fully connected layer.

7. The method for advanced load control of a generator set based on battery energy storage status according to claim 6, characterized in that: In step 4, the unit heat storage coefficient prediction model is trained based on the predicted unit heat storage data set in step 3, and is specifically implemented according to the following steps: Step 4.1: Input set {X i } Input the input layer, the input feature vector is 13×16; Step 4.2, extract the feature vector of the data set through the convolution layer, and reduce the extracted feature vector to one-dimensional data through the pooling layer; Step 4.3: The bidirectional gated recurrent unit layer performs bidirectional training on the data output from step 4.2 to learn the deeper temporal features between the unit operation data and the heat storage coefficient, and then uses all its output data as the input data of the multi-head self-attention mechanism layer; Step 4.4, the hidden state h of the output of the bidirectional gated recurrent unit layer by the multi-head self-attention mechanism layer t Assign different weight values; Step 4.5, the output layer uses Tanh as the activation function for the data processed in step 4.4, and maps the output to the interval (-1,1) through the activation function to obtain the predicted unit heat storage coefficient; Step 4.6, compare the predicted unit heat storage coefficient with the output set {Y i } for comparison, repeat steps 4.1 to 4.5, and perform multiple iterative training until the model converges.

8. The method for advanced load control of a generator set based on battery energy storage status according to claim 7, characterized in that: In step 4.4, the hidden state h t Different weight values ​​are assigned, as shown in formula (2) and formula (3): a t =f(q,h t ) (2); In the formula, f is the GRU network, q is the prediction vector of the GRU network, and h t is the GRU hidden state at time step t, α t is the weight of time step t; The step 4.5 is specifically shown in formula (4): Where x is the input value and e is the natural logarithm base.

9. The method for advanced load control of a generator set based on battery energy storage status according to claim 1, characterized in that: The corresponding relationship between the heat storage coefficient and the load command lead amount is: when the heat storage coefficient is -1, the load command lead amount is -10; when the heat storage coefficient is -0.6, the load command lead amount is -5; when the heat storage coefficient is -0.4, the load command lead amount is -2; when the heat storage coefficient is -0.1, the load command lead amount is -0; when the heat storage coefficient is 0, the load command lead amount is 0; when the heat storage coefficient is 0.1, the load command lead amount is 0; when the heat storage coefficient is 0.4, the load command lead amount is 2; when the heat storage coefficient is 0.6, the load command lead amount is 5; when the heat storage coefficient is 1, the load command lead amount is 10.

10. The method for advanced load control of a generator set based on battery energy storage status according to claim 1, characterized in that: The load instruction generated based on the generated heat storage coefficient in step 5 is specifically as follows: when the unit is in a load increase state and the energy storage charge state is less than the minimum output limit, the load instruction lead amount corresponding to the heat storage coefficient is increased in the generated load instruction; when the unit is in a load reduction state and the energy storage charge state is greater than the energy storage charging limit, the load instruction lead amount corresponding to the heat storage coefficient is reduced in the generated load instruction.

Citation Information

Cited By

  • Unit load advanced control method based on battery energy storage state

    WO2026016873A1