Method for measuring and regulating adjustment capacity of regional integrated energy system participating in power market

Abstract

The application discloses a method for measuring and regulating the adjustment capacity of a regional comprehensive energy system participating in the power market, and belongs to the technical field of power systems, which comprises collecting response potential power data and state of charge data, taking the load response rate and the sustainable adjustment time length as inputs, measuring the total up-regulation capacity and the total down-regulation capacity through an up-regulation capacity measurement model and a down-regulation capacity measurement model, outputting the up-regulation capacity value and the down-regulation capacity value by using an iterative optimization gate cycle unit, comparing the up-regulation capacity value and the down-regulation capacity value with the constraint conditions of the power market, judging whether to participate in the power market, giving a risk warning on the out-of-limit operation state of the regional comprehensive energy system, and presenting the adjustment capacity boundary of the regional comprehensive energy system in a visual way.The application can comprehensively improve the adaptability, safety and profitability of the regional comprehensive energy system participating in the power market.

Description

Technical Field

[0001] This invention belongs to the field of power system technology, specifically relating to a method for calculating and regulating the adjustment capacity of a regional integrated energy system participating in the electricity market. Background Technology

[0002] A regional integrated energy system is a distributed energy network system built on the principle of multi-energy synergy and complementarity. It integrates heterogeneous energy carriers such as electricity, gas, heat, and cooling to achieve cascaded utilization of energy and dynamic matching of supply and demand within the region. Based on a physical interconnection network, the regional integrated energy system includes energy production units, conversion devices, energy storage units, and an intelligent management and control platform. Through multi-energy flow coupling modeling and optimization scheduling algorithms, it breaks the independent operation mode of traditional single energy systems. Its core feature is the construction of a two-way interaction mechanism between the energy supply side and the demand side. Relying on digital technology, it achieves synergistic optimization of the entire process of energy production, transmission, storage, and consumption, significantly improving comprehensive energy efficiency and reducing carbon emission intensity, providing a replicable technical solution for regional energy transformation.

[0003] To address the challenge of accurately assessing the overall regulation capacity and effectively participating in the electricity market for regional integrated energy systems due to the variability in the states and uncertainties in the responses of their diverse and heterogeneous resources, existing technologies primarily employ simple superposition of the rated regulation power of each adjustable resource unit and estimation of overall regulation potential using a single static prediction model. However, these methods neglect dynamic temporal characteristics, including energy storage state of charge and load response rate, fail to reveal the complex nonlinear coupling relationships between resources, and lack a mechanism for closed-loop correction of the assessment model based on actual market participation results. Consequently, the accuracy and reliability of regulation capacity calculations are low, potentially leading regional integrated energy systems to make incorrect bidding decisions when participating in the electricity market. This not only limits their market profitability but also risks economic penalties and operational risks due to unreliable contract fulfillment. Summary of the Invention

[0004] The purpose of this invention is to provide a method for measuring and regulating the adjustment capacity of a regional integrated energy system participating in the electricity market. By accurately measuring the total upward and downward adjustment capacity of the regional integrated energy system, iteratively optimizing the gated cycle unit based on actual market participation results, and implementing real-time risk warning, the adaptability, security, and profitability of the regional integrated energy system participating in the electricity market can be comprehensively improved.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for calculating and regulating the regulation capacity of a regional integrated energy system participating in the electricity market, comprising the following steps:

[0006] Collect and preprocess the response potential power data of adjustable loads and the state of charge data of each energy storage unit in the regional integrated energy system, and obtain the load response rate and sustainable adjustment duration from the preprocessed response potential power data and state of charge data.

[0007] Using load response rate and sustainable adjustment duration as inputs, establish upward and downward adjustment capacity calculation models to calculate the total upward and downward adjustment capacity of the regional integrated energy system.

[0008] The total upward and downward adjustment capabilities are input into the gated loop unit, and the gated loop unit is optimized by the adaptive moment estimation optimization algorithm. The upward and downward adjustment capability values ​​are output.

[0009] The constraints of the electricity market are compared with the upward and downward adjustment capacity values ​​to determine whether the regional integrated energy system should participate in the electricity market.

[0010] The system compares the state of charge data, up-adjustment capacity values, and down-adjustment capacity values ​​with preset safety thresholds in real time to provide risk warnings for exceeding the limits of the regional integrated energy system's operating status.

[0011] Based on actual participation results in the electricity market, the gated loop unit is iteratively revised and optimized;

[0012] Dynamically plot the upward and downward adjustment capacity values ​​to visually represent the adjustment capacity boundary of the regional integrated energy system.

[0013] Preferably, the process of collecting and preprocessing the response potential power data of adjustable loads and the state-of-charge data of each energy storage unit in the regional integrated energy system includes:

[0014] At each adjustable load unit within the regional integrated energy system, smart energy terminals are deployed. Each smart energy terminal has a built-in computing unit, storage unit, communication unit, and control interface that connects to the adjustable load unit itself.

[0015] The regional energy management system broadcasts data collection instructions to smart energy terminals within the region according to a preset time cycle. The smart energy terminals call the load operation constraint parameter set pre-stored in the storage unit, quantify the adjustable power into response potential power data, and then add a timestamp and device identification code to form a data frame.

[0016] A battery management system is deployed on the energy storage unit within the regional integrated energy system. The battery management system is connected to a data acquisition gateway deployed near the energy storage unit via an external communication interface.

[0017] The regional energy management system broadcasts data query commands to the data acquisition gateways within the region according to a time cycle. The data acquisition gateways, based on the data query commands, initiate data read requests to the battery management system. After receiving the state of charge data, they attach a timestamp and device identification code to form a data frame.

[0018] The outlier removal algorithm and the min-max normalization algorithm are sequentially applied to the response potential power data and the state of charge data, followed by cleaning and dimensionless processing, to obtain the preprocessed response potential power data and state of charge data.

[0019] Preferably, the process of obtaining the load response rate and sustainable regulation duration from the preprocessed response potential power data and state of charge data includes:

[0020] Calculate the difference between the current value and the previous value in the response potential power data, divide the difference by the time period, and obtain the load response rate at the current time.

[0021] Retrieve the pre-configured static parameters for the energy storage unit, including the rated capacity, rated regulating power, upper limit of safe state of charge, and lower limit of safe state of charge of the energy storage unit;

[0022] Based on static parameters and the current state of charge data, the sustainable discharge duration and sustainable charging duration are calculated respectively. The sustainable discharge duration and sustainable charging duration together constitute the sustainable adjustment duration.

[0023] The difference between the current state of charge data and the safe lower limit of the state of charge is multiplied by the rated capacity of the energy storage unit to obtain the current total discharge capacity. The sustainable discharge duration is calculated by dividing the total discharge capacity by the rated regulating power.

[0024] The difference between the safe upper limit of the state of charge and the current state of charge data is multiplied by the rated capacity of the energy storage unit to obtain the current total chargeable capacity. The continuous charging time is calculated by dividing the total chargeable capacity by the rated regulating power.

[0025] Preferably, the process of establishing an upward adjustment capacity calculation model using load response rate and sustainable adjustment duration as inputs, and calculating the total upward adjustment capacity of the regional integrated energy system, includes:

[0026] Using load response rate and sustainable discharge duration as input features, an up-adjustment capability measurement model based on gradient boosting decision tree model architecture is established;

[0027] Construct a historical training dataset, which includes input features and corresponding output labels. The output label is the historical total up-adjustment capacity of the regional integrated energy system obtained through actual measurements.

[0028] Establish an initial baseline model, which outputs the average value of the output labels in the historical training dataset.

[0029] Based on the historical training dataset and the initial benchmark model, the gradient boosting algorithm is used to iteratively optimize the initial benchmark model.

[0030] In the first iteration, the residual between the average value and each true output label in the historical training dataset is calculated. Using the input features as input, the residual generated in the first round is used as the learning target to train the first decision tree regression model. The first decision tree regression model is multiplied by the preset learning rate and then accumulated with the initial baseline model to form a new decision tree regression model.

[0031] Repeat the first round of iterations until the preset number of iterations is reached. Then, weight the decision tree regression models generated in each iteration round, solidify them into an upregulation capacity measurement model, and output the total upregulation capacity.

[0032] Preferably, the process of establishing a down-regulation capacity calculation model using load response rate and sustainable adjustment duration as inputs, and calculating the total down-regulation capacity of the regional integrated energy system, includes:

[0033] Using load response rate and sustainable charging duration as input sequences, a down-regulation capability measurement model based on a long short-term memory network model architecture is established.

[0034] Construct a historical sequence training dataset, which includes input sequences and corresponding measurement values. The measurement values ​​are the historical total downward adjustment capacity of the regional integrated energy system obtained through actual measurements.

[0035] Based on the historical sequence training dataset, a long short-term memory network model architecture is defined to obtain an initial downregulation capacity measurement model. The long short-term memory network model architecture includes the number of network layers, the number of memory units in the network layers, and fully connected layers.

[0036] The historical sequence training dataset is input into the initial downregulation capability measurement model. The initial downregulation capability measurement model performs forward propagation calculation for each input sequence. The forward propagation calculation process includes processing the input sequence data at each time point through the forget gate, input gate and output gate inside the memory unit, updating and passing the memory unit state.

[0037] At the end of the input sequence, the fully connected layer of the initial downregulation capability measurement model outputs a predicted value based on the hidden state, and compares the predicted value with the actual measured value to calculate the prediction error;

[0038] Based on the prediction error, an adaptive moment estimation optimization algorithm is adopted. The weight matrix and bias vector in the initial down-adjustment capability measurement model are adjusted through the backpropagation mechanism. The backpropagation process is repeated on the historical sequence training dataset until the preset training termination condition is reached, and a trained down-adjustment capability measurement model is obtained. The trained down-adjustment capability measurement model outputs the predicted value as the total down-adjustment capability at the current time.

[0039] Preferably, the process of optimizing the gated recurrent unit using an adaptive moment estimation optimization algorithm and outputting the up-adjustment capability value and the down-adjustment capability value includes:

[0040] The total up-adjustment capability and total down-adjustment capability are constructed as two-dimensional input sequences with a preset sequence length. The historical true up-adjustment capability and historical true down-adjustment capability obtained through actual measurement at the last time point of the two-dimensional input sequence are constructed as two-dimensional output label vectors. The two-dimensional input sequence and the two-dimensional output label vector constitute the total training dataset.

[0041] Based on the total training dataset, the gated recurrent unit model architecture is defined to obtain the initial gated recurrent unit. The gated recurrent unit model architecture includes the number of network layers, the number of gated recurrent units in the network layers, and the number of fully connected layers.

[0042] The total training dataset is input into the initial gated recurrent unit. The initial gated recurrent unit performs forward propagation computation on each two-dimensional input sequence. The forward propagation computation process includes processing the two-dimensional input sequence at each time point through the reset gate and update gate inside the initial gated recurrent unit to update and pass the hidden state.

[0043] At the end of the two-dimensional input sequence, the fully connected layer of the initial gated recurrent unit outputs a two-dimensional prediction vector containing up-adjustment and down-adjustment capabilities based on the hidden state. The two-dimensional prediction vector is compared with the true two-dimensional output label vector to calculate the capability prediction error.

[0044] Based on the capability prediction error, an adaptive moment estimation optimization algorithm is adopted. The weight matrix and bias vector in the initial gated recurrent unit are adjusted through the backpropagation mechanism. The backpropagation process is repeated on the total training dataset until the preset training termination condition is reached, and a trained gated recurrent unit is obtained. The trained gated recurrent unit is deconstructed into a two-dimensional prediction vector to obtain the up-adjusted capability value and the down-adjusted capability value.

[0045] Preferably, the process of comparing the constraints of the electricity market with the upward and downward adjustment capacity values ​​to determine whether the regional integrated energy system should participate in the electricity market includes:

[0046] From the market constraint database obtained through communication with the electricity market trading system, retrieve the set of market constraint parameters applicable to the current market cycle, including the minimum upward adjustment application capacity, the minimum downward adjustment application capacity, and the required minimum duration.

[0047] The increased capacity value is compared with the minimum increase application capacity, and the sustainable discharge duration is compared with the shortest duration. When the increased capacity value exceeds the minimum increase application capacity and the sustainable discharge duration exceeds the shortest duration, the regional integrated energy system is deemed to meet the conditions for participation in the increase.

[0048] The reduced capacity value is compared with the minimum reduced capacity declared, and the sustainable charging time is compared with the shortest continuous time. When the reduced capacity value exceeds the minimum reduced capacity declared and the sustainable charging time exceeds the shortest continuous time, the regional integrated energy system is deemed to meet the conditions for participating in the reduction.

[0049] The judgment results are logically combined. If the judgment results satisfy both the upward and downward participation conditions, a participation decision instruction is generated. If neither the upward nor downward participation conditions are satisfied, a non-participation decision instruction is generated.

[0050] Preferably, the process of comparing the state of charge data, the up-adjustment capacity value, and the down-adjustment capacity value with preset safety thresholds in real time to provide risk warnings for exceeding the limits of the regional integrated energy system's operating status includes:

[0051] A set of safety threshold parameters is pre-configured, which includes thresholds for excessively high state of charge, excessively low state of charge, and excessively low regulation capability.

[0052] The real-time acquired state of charge data is continuously compared with the over-high and under-high thresholds of state of charge. When the state of charge data is greater than the over-high threshold or less than the under-high threshold, the preset state of charge over-limit judgment condition is triggered.

[0053] The real-time values ​​of the upward and downward adjustment capabilities are continuously compared with the threshold of the adjustment capability being too low. When the value of the upward adjustment capability is less than the threshold of the adjustment capability being too low or the value of the downward adjustment capability is less than the threshold of the adjustment capability being too low, the preset condition for insufficient adjustment capability is triggered.

[0054] When the conditions for determining the state of charge exceeding the limit and the conditions for determining insufficient regulation capacity are triggered, risk warning information is generated, which includes the warning trigger time, warning type, specific value of the warning trigger, and corresponding safety threshold parameters. The warning types include state of charge exceeding the limit and insufficient regulation capacity.

[0055] Preferably, the process of iteratively correcting and optimizing the gated cyclic unit based on actual participation results in the electricity market includes:

[0056] After the regional integrated energy system completes an actual regulation service response in the electricity market according to the participation decision-making instructions, it obtains the actual execution data of the actual regulation service from the electricity market settlement system. The actual execution data includes the execution time of the actual regulation service and the actual upward and downward power provided at the execution time. The upward and downward power constitute the true capacity label for the corresponding execution time.

[0057] The actual capability labels are timestamped and associated with the two-dimensional input sequence corresponding to the execution time, so as to construct an incremental training sample consisting of a two-dimensional input sequence and actual capability labels from the actual participation results of a power market.

[0058] According to the preset period, the incremental training samples generated within the period will be added to the total training dataset, thereby forming an expanded training dataset;

[0059] Starting with the weight matrix and bias vector of the pre-trained gated recurrent unit, the expanded training dataset is input into the pre-trained gated recurrent unit. An adaptive moment estimation optimization algorithm is used to perform incremental training on the pre-trained gated recurrent unit for a specified number of rounds. During the incremental training process, the pre-trained gated recurrent unit fine-tunes the weight matrix and bias vector based on the prediction error revealed by the incremental training samples.

[0060] After incremental training is completed, the original weight matrix and bias vector of the gated recurrent unit are replaced with the finely tuned weight matrix and bias vector to obtain the iteratively optimized gated recurrent unit.

[0061] Preferably, the process of dynamically plotting upward and downward adjustment capacity values ​​to visually represent the adjustment capacity boundary of the regional integrated energy system includes:

[0062] On the monitoring interface of the regional energy management system, a two-dimensional rectangular coordinate system is established. The horizontal axis of the two-dimensional rectangular coordinate system is defined as the time axis, and the range of the time axis extends from the current moment to the end of the preset duration. The vertical axis of the two-dimensional rectangular coordinate system is defined as the power axis, with zero as the center. Positive values ​​upward represent upward adjustment capacity, and negative values ​​downward represent downward adjustment capacity.

[0063] The sequence of output upward adjustment capability values ​​is plotted as the upper boundary curve on the positive half-axis of a two-dimensional rectangular coordinate system in chronological order. The sequence of output downward adjustment capability values ​​is plotted as the lower boundary curve on the negative half-axis of a two-dimensional rectangular coordinate system after taking the opposite value in chronological order.

[0064] The drawing process is executed cyclically according to the time cycle, generating new upper and lower boundary curves. These new upper and lower boundary curves then cover the original upper and lower boundary curves on the monitoring interface, thereby visually presenting the regulation capacity boundary of the regional integrated energy system.

[0065] The beneficial effects of this invention are as follows:

[0066] This invention deeply mines the dynamic temporal characteristics of adjustable resources and uses a nonlinear model to reveal the coupling relationship between adjustment capacity and adjustment capacity, solving the problem of inaccurate evaluation by traditional methods and significantly improving the accuracy and reliability of adjustment capacity measurement results.

[0067] Based on accurate calculations of regulation capacity, this invention automatically makes decisions regarding participation in the electricity market, effectively avoiding performance risks caused by overestimation of capacity and profit losses caused by underestimation of capacity, thus greatly enhancing the market competitiveness and economic benefits of the regional integrated energy system.

[0068] This invention possesses closed-loop feedback and iterative optimization capabilities, which can continuously improve the accuracy of the model based on actual participation results in the electricity market. Combined with a real-time risk warning mechanism, it comprehensively ensures the adaptability and security of the long-term operation of the regional integrated energy system. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0070] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0071] Example 1: As Figure 1 As shown, the method for calculating and regulating the regional integrated energy system's ability to participate in the electricity market includes steps 1-7:

[0072] Step 1: Collect and preprocess the response potential power data of adjustable loads and the state of charge data of each energy storage unit in the regional integrated energy system to obtain the preprocessed response potential power data and state of charge data.

[0073] At each adjustable load unit within the regional integrated energy system, smart energy terminals are deployed. Each smart energy terminal has a built-in computing unit, storage unit, communication unit, and control interface that connects to the adjustable load unit itself, and establishes a communication connection with the regional energy management system wirelessly.

[0074] The regional energy management system broadcasts data collection instructions to smart energy terminals within the region according to a preset time cycle. The smart energy terminals call the load operation constraint parameter set pre-stored in the storage unit, quantify the adjustable power into response potential power data, and then attach a timestamp and device identification code to form a data frame. The load operation constraint parameter set includes the upper and lower temperature limits set for users for temperature-controlled loads, the minimum interruption duration and maximum number of interruptions set for users for interruptible loads, and the expected charging time and target power set for users for charging loads. The adjustable power is calculated by the calculation unit based on the actual operating status of the current load and the load operation constraint parameter set, without violating the load operation constraints.

[0075] A battery management system is deployed on the energy storage unit within the regional integrated energy system. The battery management system is connected to a data acquisition gateway deployed near the energy storage unit through an external communication interface. The data acquisition gateway establishes a communication connection with the regional energy management system.

[0076] The regional energy management system broadcasts data query commands to the data acquisition gateways within the region according to a time cycle. The data acquisition gateways initiate data read requests to the battery management system based on the data query commands. After receiving the state of charge data, they attach a timestamp and device identification code to form a data frame. The state of charge data is calculated using the ampere-hour integration method combined with the open-circuit voltage correction method and is returned to the data acquisition gateway by the battery management system.

[0077] The outlier removal algorithm and the min-max normalization algorithm are sequentially applied to the response potential power data and the state of charge data, followed by cleaning and dimensionless processing, to obtain the preprocessed response potential power data and state of charge data.

[0078] In step 1, it is also necessary to obtain the load response rate and sustainable regulation duration from the preprocessed response potential power data and state of charge data.

[0079] The load response rate at the current moment is obtained by calculating the difference between the current value and the previous value in the response potential power data, and then dividing the difference by the time period.

[0080] Retrieve the pre-configured static parameters for the energy storage unit. The static parameters include the rated capacity, rated regulating power, upper limit of safe state of charge, and lower limit of safe state of charge of the energy storage unit.

[0081] Based on static parameters and the current state of charge data, the sustainable discharge duration and sustainable charging duration are calculated separately. The sustainable discharge duration and sustainable charging duration together constitute the sustainable adjustment duration.

[0082] The difference between the current state of charge (SOC) data and the safe lower limit of SOC is multiplied by the rated capacity of the energy storage unit to obtain the current total discharge capacity. The sustainable discharge duration is calculated by dividing the total discharge capacity by the rated regulating power.

[0083] The difference between the safe upper limit of the state of charge and the current state of charge data is multiplied by the rated capacity of the energy storage unit to obtain the current total chargeable capacity. The continuous charging time is calculated by dividing the total chargeable capacity by the rated regulating power.

[0084] Step 2: Using load response rate and sustainable adjustment duration as inputs, establish upward adjustment capacity calculation models and downward adjustment capacity calculation models to calculate the total upward adjustment capacity and total downward adjustment capacity of the regional integrated energy system.

[0085] Using load response rate and duration of continuous discharge as input characteristics We established an upscaling capability measurement model based on a gradient boosting decision tree model architecture.

[0086] Construct a historical training dataset, which includes input features x and corresponding output labels. The output labels are the historical total up-adjustment capacity of the regional integrated energy system obtained through actual measurements.

[0087] Establish an initial benchmark model Initial baseline model The output is the average of the output labels in the historical training dataset.

[0088] Based on historical training datasets and initial benchmark models The gradient boosting algorithm is used to iteratively optimize the initial baseline model.

[0089] In the first iteration, the residual between the average value and each true output label in the historical training dataset is calculated. Using the input feature x as input, the residual generated in the first round is used as the learning objective to train the first decision tree regression model. The first decision tree regression model is then multiplied by a preset learning rate. Then, compared with the initial baseline model The results are accumulated to form a new decision tree regression model. .

[0090] Repeat the first iteration process until the preset number of iterations M is reached, and then weight the decision tree regression models generated in each iteration. This is solidified into an upward adjustment capacity calculation model, which outputs the total upward adjustment capacity. , where m is the index of the iteration number.

[0091] Using load response rate and sustainable charging duration as input sequences A downregulation capability measurement model based on a long short-term memory network model architecture was established.

[0092] Construct a historical sequence training dataset, which includes the input sequences. The corresponding measured value is the historical total downward adjustment capacity of the regional integrated energy system obtained through actual measurement.

[0093] Based on the historical sequence training dataset, a long short-term memory network model architecture is defined to obtain an initial downregulation capacity measurement model. The long short-term memory network model architecture includes the number of network layers, the number of memory units in the network layers, and fully connected layers.

[0094] The historical sequence training dataset is input into the initial downregulation capability measurement model. The initial downregulation capability measurement model calculates the downregulation capability for each input sequence. The forward propagation computation is performed. The forward propagation computation process includes processing the input sequence data at each time point through the forget gate, input gate, and output gate inside the memory unit, updating and passing the state of the memory unit. The specific calculation formula is as follows:

[0095] ;

[0096] in, The output represents the hidden state from the previous moment. Represents the state of the memory unit at the previous moment. , and These represent the outputs of the forget gate, input gate, and output gate, respectively. Indicates the state of candidate memory cells. Represents the updated state of the memory cell at the current moment. The output represents the hidden state at the current moment. , , and The weight matrix represents the forget gate, input gate, memory unit, and output gate. , , and The bias vectors representing the forget gate, input gate, memory unit, and output gate. It is a logistic sigmoid activation function. The hyperbolic tangent activation function is used. This represents element-wise multiplication.

[0097] In the input sequence At the end, the fully connected layer of the initial downscaling capability measurement model is based on the hidden state. It outputs the predicted value and compares the predicted value with the actual measured value to calculate the prediction error.

[0098] Based on the prediction error, an adaptive moment estimation optimization algorithm is adopted. The weight matrix and bias vector in the initial down-adjustment capability measurement model are adjusted through the backpropagation mechanism. The backpropagation process is repeated on the historical sequence training dataset until the preset training termination condition is reached, and a trained down-adjustment capability measurement model is obtained. The trained down-adjustment capability measurement model outputs the predicted value as the total down-adjustment capability at the current time.

[0099] Step 3: Input the total upward adjustment capability and the total downward adjustment capability into the gated loop unit, optimize the gated loop unit using the adaptive moment estimation optimization algorithm, and output the upward adjustment capability value and the downward adjustment capability value.

[0100] The total up-adjustment and total down-adjustment capabilities are constructed as a two-dimensional input sequence with a preset sequence length. , two-dimensional input sequence The historical true upward and downward capabilities obtained through actual measurements at the last time point are used to construct a two-dimensional output label vector and a two-dimensional input sequence. Together with the two-dimensional output label vector, they form the total training dataset.

[0101] Based on the total training dataset, a gated recurrent unit model architecture is defined to obtain the initial gated recurrent unit. The gated recurrent unit model architecture includes the number of network layers, the number of gated recurrent units in the network layers, and the number of fully connected layers.

[0102] The total training dataset is input into the initial gated recurrent unit, which processes each two-dimensional input sequence. Perform forward propagation computation, the forward propagation computation process includes processing the two-dimensional input sequence... At each time point, the hidden state is updated and passed on through the reset and update gates within the initial gating loop unit. The specific calculation formula is as follows:

[0103] ;

[0104] in, and This represents the output of the reset and update doors. Represents the candidate hidden state. , and The weight matrix represents the reset gate, update gate, and hidden state. , and The bias vectors represent the reset gate, update gate, and hidden state.

[0105] In two-dimensional input sequence At the end, the fully connected layer of the initial gated loop unit is based on the hidden state. The output includes a two-dimensional prediction vector containing both up-adjusted and down-adjusted capabilities. The two-dimensional prediction vector is compared with the true two-dimensional output label vector to calculate the capability prediction error.

[0106] Based on the capability prediction error, an adaptive moment estimation optimization algorithm is adopted. The weight matrix and bias vector in the initial gated recurrent unit are adjusted through the backpropagation mechanism. The backpropagation process is repeated on the total training dataset until the preset training termination condition is reached, and a trained gated recurrent unit is obtained. The trained gated recurrent unit is deconstructed into a two-dimensional prediction vector to obtain the up-adjusted capability value and the down-adjusted capability value.

[0107] Step 4: Compare the constraints of the electricity market with the upward and downward adjustment capacity values ​​respectively to determine whether the regional integrated energy system should participate in the electricity market.

[0108] The market constraint parameter set applicable to the current market cycle is retrieved from the market constraint database obtained through communication with the electricity market trading system. The market constraint parameter set includes the minimum upward adjustment application capacity, the minimum downward adjustment application capacity, and the required minimum duration.

[0109] The increased capacity value is compared with the minimum increase application capacity, and the sustainable discharge duration is compared with the shortest duration. When the increased capacity value exceeds the minimum increase application capacity and the sustainable discharge duration exceeds the shortest duration, the regional integrated energy system is deemed to meet the conditions for participation in the increase.

[0110] The reduced capacity value is compared with the minimum reduced capacity declared, and the sustainable charging time is compared with the shortest continuous time. When the reduced capacity value exceeds the minimum reduced capacity declared and the sustainable charging time exceeds the shortest continuous time, the regional integrated energy system is determined to meet the conditions for participating in the reduction.

[0111] The judgment results are logically combined. If the judgment results satisfy both the upward and downward participation conditions, a participation decision instruction is generated. If neither the upward nor downward participation conditions are satisfied, a non-participation decision instruction is generated.

[0112] Step 5: Compare the state of charge data, up-adjustment capacity values, and down-adjustment capacity values ​​with preset safety thresholds in real time to issue risk warnings for exceeding the limits of the regional integrated energy system's operating status.

[0113] A set of safety threshold parameters is pre-configured, which includes thresholds for excessively high state of charge, excessively low state of charge, and excessively low regulation capability.

[0114] The real-time acquired state of charge data is continuously compared with the over-high and under-high thresholds of state of charge. When the state of charge data is greater than the over-high threshold or less than the under-high threshold, the preset over-limit judgment condition of state of charge is triggered.

[0115] The real-time values ​​of the upward and downward adjustment capabilities are continuously compared with the threshold of insufficient adjustment capability. When the value of the upward adjustment capability is less than the threshold of insufficient adjustment capability or the value of the downward adjustment capability is less than the threshold of insufficient adjustment capability, the preset condition for insufficient adjustment capability is triggered.

[0116] When the conditions for determining the state of charge exceeding the limit and the conditions for determining insufficient regulation capacity are triggered, risk warning information is generated, which includes the warning trigger time, warning type, specific value of the warning trigger, and corresponding safety threshold parameters. The warning types include state of charge exceeding the limit and insufficient regulation capacity.

[0117] Step 6: Based on the actual participation results in the electricity market, iteratively revise and optimize the gated loop unit.

[0118] After the regional integrated energy system completes an actual regulation service response in the electricity market according to the decision-making instructions, it obtains the actual execution data of the actual regulation service from the electricity market settlement system. The actual execution data includes the execution time of the actual regulation service and the actual upward and downward power provided at the execution time. The upward and downward power constitute the true capacity label for the corresponding execution time.

[0119] The actual capability labels are timestamped and associated with the two-dimensional input sequence corresponding to the execution time, so as to construct an incremental training sample consisting of a two-dimensional input sequence and actual capability labels for the actual participation results in a power market.

[0120] According to the preset period, the incremental training samples generated within the period will be added to the total training dataset, thereby forming an expanded training dataset.

[0121] Starting with the weight matrix and bias vector of the pre-trained gated recurrent unit, the expanded training dataset is input into the pre-trained gated recurrent unit. An adaptive moment estimation optimization algorithm is used to perform incremental training on the pre-trained gated recurrent unit for a specified number of rounds. During the incremental training process, the pre-trained gated recurrent unit fine-tunes its weight matrix and bias vector based on the prediction error revealed by the incremental training samples.

[0122] After incremental training is completed, the original weight matrix and bias vector of the gated recurrent unit are replaced with the finely tuned weight matrix and bias vector to obtain the iteratively optimized gated recurrent unit.

[0123] Step 7: Dynamically plot the upward and downward adjustment capacity values ​​to visually represent the adjustment capacity boundary of the regional integrated energy system.

[0124] On the monitoring interface of the regional energy management system, a two-dimensional rectangular coordinate system is established. The horizontal axis of the two-dimensional rectangular coordinate system is defined as the time axis, which extends from the current moment to the end of the preset duration. The vertical axis of the two-dimensional rectangular coordinate system is defined as the power axis, with zero as the center. Positive values ​​upward represent upward adjustment capacity, and negative values ​​downward represent downward adjustment capacity.

[0125] The sequence of output upward adjustment capability values ​​is plotted as the upper boundary curve on the positive half-axis of a two-dimensional rectangular coordinate system in chronological order. The sequence of output downward adjustment capability values ​​is plotted as the lower boundary curve on the negative half-axis of a two-dimensional rectangular coordinate system after taking the opposite value in chronological order.

[0126] The drawing process is executed cyclically according to the time cycle, generating new upper and lower boundary curves. These new upper and lower boundary curves then cover the original upper and lower boundary curves on the monitoring interface, thereby visually presenting the regulation capacity boundary of the regional integrated energy system.

[0127] Example 2: Based on Example 1, in step 1, when the regional energy management system broadcasts data acquisition commands to smart energy-consuming terminals and data query commands to the data acquisition gateway, a dynamic periodic adjustment mechanism is adopted. The specific process includes:

[0128] A pre-configured set of periodic adjustment trigger thresholds is provided, which includes thresholds for state of charge fluctuation amplitude, load response rate change rate, and minimum and maximum thresholds for the data acquisition period.

[0129] The regional energy management system calculates in real time the fluctuation range of the state of charge data collected between two consecutive data collections and the rate of change of the load response rate between two consecutive data collections.

[0130] When the state of charge fluctuation exceeds the state of charge fluctuation threshold, or the load response rate change rate exceeds the load response rate change rate threshold, the current data acquisition cycle will be shortened by a preset step size, and the shortened cycle will not be lower than the minimum cycle threshold.

[0131] When the state of charge fluctuation amplitude is lower than the state of charge fluctuation amplitude threshold for a preset duration and the load response rate change rate is lower than the load response rate change rate threshold for a preset duration, the current data acquisition cycle will be extended by a preset step size, and the extended cycle will not be higher than the maximum cycle threshold.

[0132] The regional energy management system will synchronize the adjusted dynamic cycle to the corresponding smart energy terminals and data acquisition gateways, and the smart energy terminals and data acquisition gateways will perform data acquisition and reporting according to the adjusted dynamic cycle.

[0133] Example 3: Based on Example 1, in step 5, after generating risk warning information, a tiered linkage control mechanism is added. The specific process includes:

[0134] The system pre-configures early warning classification standards and corresponding control strategy libraries. The early warning classification standards divide over-limit charging state into Level 1 over-limit (charging state exceeds the safety threshold by 10% or more) and Level 2 over-limit (charging state exceeds the safety threshold by 5%-10%). The system also divides insufficient regulation capacity into Level 1 insufficient regulation capacity (regulation capacity is 20% or more below the safety threshold) and Level 2 insufficient regulation capacity (regulation capacity is 10%-20% below the safety threshold). In the control strategy library, Level 1 early warning corresponds to emergency control strategies, and Level 2 early warning corresponds to conventional control strategies.

[0135] The regional energy management system determines the warning level by matching the specific value that triggers the warning with the safety threshold and the warning classification standard.

[0136] If it is a Level 1 state of charge exceeding the limit warning, the emergency control strategy is as follows: immediately send a charging and discharging power adjustment command to the battery management system of the corresponding energy storage unit, control the energy storage unit to perform charging and discharging operations at 80%-100% of the rated regulation power, and at the same time suspend the application of the energy storage unit to participate in the power market regulation.

[0137] If the warning is for insufficient first-level regulation capacity, the emergency regulation strategy is as follows: send a load response priority instruction to the smart energy terminal of the adjustable load unit in the area, and prioritize calling the top 30% of adjustable loads in terms of response potential power data to participate in regulation to supplement the regulation capacity gap.

[0138] If it is a Level II warning, the conventional control strategy is to send step-by-step control commands to the corresponding equipment according to the preset resource scheduling priority (energy storage units take priority over adjustable loads) to gradually restore the state of charge or regulation capacity to the safe threshold range.

[0139] After the control is executed, the regional energy management system monitors the changing trends of the state of charge data and the value of the regulation capacity in real time, records the parameter changes before and after the execution of the control command, and stores the changes in the value of the changes in the corresponding control strategy for the warning level. This information is then used to optimize the parameters in the control strategy library (such as the charging and discharging power ratio and the load dispatch ratio).

[0140] Example 4: A device for measuring and controlling the regulation capacity of a regional integrated energy system participating in the electricity market, comprising:

[0141] One or more processors;

[0142] Memory, used to store one or more computer programs;

[0143] When one or more programs are executed by one or more processors, the one or more processors perform the methods in Embodiment 1, Embodiment 2, or Embodiment 3.

[0144] Example 5: A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, implement the methods of Example 1, Example 2 or Example 3.

Claims

1. A method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market, characterized in that, Includes the following steps: Collect and preprocess the response potential power data of adjustable loads and the state of charge data of each energy storage unit in the regional integrated energy system, and obtain the load response rate and sustainable adjustment duration from the preprocessed response potential power data and state of charge data. Using load response rate and sustainable adjustment duration as inputs, establish upward and downward adjustment capacity calculation models to calculate the total upward and downward adjustment capacity of the regional integrated energy system. The total upward and downward adjustment capabilities are input into the gated loop unit, and the gated loop unit is optimized by the adaptive moment estimation optimization algorithm. The upward and downward adjustment capability values ​​are output. The constraints of the electricity market are compared with the upward and downward adjustment capacity values ​​to determine whether the regional integrated energy system should participate in the electricity market. The system compares the state of charge data, up-adjustment capacity values, and down-adjustment capacity values ​​with preset safety thresholds in real time to provide risk warnings for exceeding the limits of the regional integrated energy system's operating status. Based on actual participation results in the electricity market, the gated loop unit is iteratively revised and optimized; Dynamically plot the upward and downward adjustment capacity values ​​to visually represent the adjustment capacity boundary of the regional integrated energy system.

2. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 1, characterized in that, The process of collecting and preprocessing the response potential power data of adjustable loads and the state-of-charge data of each energy storage unit in the regional integrated energy system includes: At each adjustable load unit within the regional integrated energy system, smart energy terminals are deployed. Each smart energy terminal has a built-in computing unit, storage unit, communication unit, and control interface that connects to the adjustable load unit itself. The regional energy management system broadcasts data collection instructions to smart energy terminals within the region according to a preset time cycle. The smart energy terminals call the load operation constraint parameter set pre-stored in the storage unit, quantify the adjustable power into response potential power data, and then add a timestamp and device identification code to form a data frame. A battery management system is deployed on the energy storage unit within the regional integrated energy system. The battery management system is connected to a data acquisition gateway deployed near the energy storage unit via an external communication interface. The regional energy management system broadcasts data query commands to the data acquisition gateways within the region according to a time cycle. The data acquisition gateways, based on the data query commands, initiate data read requests to the battery management system. After receiving the state of charge data, they attach a timestamp and device identification code to form a data frame. The outlier removal algorithm and the min-max normalization algorithm are sequentially applied to the response potential power data and the state of charge data, followed by cleaning and dimensionless processing, to obtain the preprocessed response potential power data and state of charge data.

3. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 1, characterized in that, The process of obtaining load response rate and sustainable regulation duration from preprocessed potential power data and state of charge data includes: Calculate the difference between the current value and the previous value in the response potential power data, divide the difference by the time period, and obtain the load response rate at the current time. Retrieve the pre-configured static parameters for the energy storage unit, including the rated capacity, rated regulating power, upper limit of safe state of charge, and lower limit of safe state of charge of the energy storage unit; Based on static parameters and the current state of charge data, the sustainable discharge duration and sustainable charging duration are calculated respectively. The sustainable discharge duration and sustainable charging duration together constitute the sustainable adjustment duration. The difference between the current state of charge data and the safe lower limit of the state of charge is multiplied by the rated capacity of the energy storage unit to obtain the current total discharge capacity. The sustainable discharge duration is calculated by dividing the total discharge capacity by the rated regulating power. The difference between the safe upper limit of the state of charge and the current state of charge data is multiplied by the rated capacity of the energy storage unit to obtain the current total chargeable capacity. The continuous charging time is calculated by dividing the total chargeable capacity by the rated regulating power.

4. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 3, characterized in that, Using load response rate and sustainable adjustment duration as inputs, a model for calculating upward adjustment capacity is established. The process of calculating the total upward adjustment capacity of the regional integrated energy system includes: Using load response rate and sustainable discharge duration as input features, an up-adjustment capability measurement model based on gradient boosting decision tree model architecture is established; Construct a historical training dataset, which includes input features and corresponding output labels. The output label is the historical total up-adjustment capacity of the regional integrated energy system obtained through actual measurements. Establish an initial baseline model, which outputs the average value of the output labels in the historical training dataset. Based on the historical training dataset and the initial benchmark model, the gradient boosting algorithm is used to iteratively optimize the initial benchmark model. In the first iteration, the residual between the average value and each true output label in the historical training dataset is calculated. Using the input features as input, the residual generated in the first round is used as the learning target to train the first decision tree regression model. The first decision tree regression model is multiplied by the preset learning rate and then accumulated with the initial baseline model to form a new decision tree regression model. Repeat the first round of iterations until the preset number of iterations is reached. Then, weight the decision tree regression models generated in each iteration round, solidify them into an upregulation capacity measurement model, and output the total upregulation capacity.

5. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 3, characterized in that, Using load response rate and sustainable adjustment duration as inputs, a model for calculating down-regulation capacity is established. The process of calculating the total down-regulation capacity of the regional integrated energy system includes: Using load response rate and sustainable charging duration as input sequences, a down-regulation capability measurement model based on a long short-term memory network model architecture is established. Construct a historical sequence training dataset, which includes input sequences and corresponding measurement values. The measurement values ​​are the historical total downward adjustment capacity of the regional integrated energy system obtained through actual measurements. Based on the historical sequence training dataset, a long short-term memory network model architecture is defined to obtain an initial downregulation capacity measurement model. The long short-term memory network model architecture includes the number of network layers, the number of memory units in the network layers, and fully connected layers. The historical sequence training dataset is input into the initial downregulation capability measurement model. The initial downregulation capability measurement model performs forward propagation calculation for each input sequence. The forward propagation calculation process includes processing the input sequence data at each time point through the forget gate, input gate and output gate inside the memory unit, updating and passing the memory unit state. At the end of the input sequence, the fully connected layer of the initial downregulation capability measurement model outputs a predicted value based on the hidden state, and compares the predicted value with the actual measured value to calculate the prediction error; Based on the prediction error, an adaptive moment estimation optimization algorithm is adopted. The weight matrix and bias vector in the initial down-adjustment capability measurement model are adjusted through the backpropagation mechanism. The backpropagation process is repeated on the historical sequence training dataset until the preset training termination condition is reached, and a trained down-adjustment capability measurement model is obtained. The trained down-adjustment capability measurement model outputs the predicted value as the total down-adjustment capability at the current time.

6. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 1, characterized in that, The process of optimizing the gated recurrent unit using an adaptive moment estimation optimization algorithm and outputting the up-adjustment and down-adjustment capability values ​​includes: The total up-adjustment capability and total down-adjustment capability are constructed as two-dimensional input sequences with a preset sequence length. The historical true up-adjustment capability and historical true down-adjustment capability obtained through actual measurement at the last time point of the two-dimensional input sequence are constructed as two-dimensional output label vectors. The two-dimensional input sequence and the two-dimensional output label vector constitute the total training dataset. Based on the total training dataset, the gated recurrent unit model architecture is defined to obtain the initial gated recurrent unit. The gated recurrent unit model architecture includes the number of network layers, the number of gated recurrent units in the network layers, and the number of fully connected layers. The total training dataset is input into the initial gated recurrent unit. The initial gated recurrent unit performs forward propagation computation on each two-dimensional input sequence. The forward propagation computation process includes processing the two-dimensional input sequence at each time point through the reset gate and update gate inside the initial gated recurrent unit to update and pass the hidden state. At the end of the two-dimensional input sequence, the fully connected layer of the initial gated recurrent unit outputs a two-dimensional prediction vector containing up-adjustment and down-adjustment capabilities based on the hidden state. The two-dimensional prediction vector is compared with the true two-dimensional output label vector to calculate the capability prediction error. Based on the capability prediction error, an adaptive moment estimation optimization algorithm is adopted. The weight matrix and bias vector in the initial gated recurrent unit are adjusted through the backpropagation mechanism. The backpropagation process is repeated on the total training dataset until the preset training termination condition is reached, and a trained gated recurrent unit is obtained. The trained gated recurrent unit is deconstructed into a two-dimensional prediction vector to obtain the up-adjusted capability value and the down-adjusted capability value.

7. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 3, characterized in that, The process of comparing the constraints of the electricity market with the upward and downward adjustment capacity values ​​to determine whether a regional integrated energy system should participate in the electricity market includes: From the market constraint database obtained through communication with the electricity market trading system, retrieve the set of market constraint parameters applicable to the current market cycle, including the minimum upward adjustment application capacity, the minimum downward adjustment application capacity, and the required minimum duration. The increased capacity value is compared with the minimum increase application capacity, and the sustainable discharge duration is compared with the shortest duration. When the increased capacity value exceeds the minimum increase application capacity and the sustainable discharge duration exceeds the shortest duration, the regional integrated energy system is deemed to meet the conditions for participation in the increase. The reduced capacity value is compared with the minimum reduced capacity declared, and the sustainable charging time is compared with the shortest continuous time. When the reduced capacity value exceeds the minimum reduced capacity declared and the sustainable charging time exceeds the shortest continuous time, the regional integrated energy system is deemed to meet the conditions for participating in the reduction. The judgment results are logically combined. If the judgment results satisfy both the upward and downward participation conditions, a participation decision instruction is generated. If neither the upward nor downward participation conditions are satisfied, a non-participation decision instruction is generated.

8. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 1, characterized in that, The process of comparing state of charge data, up-adjustment capacity values, and down-adjustment capacity values ​​with preset safety thresholds in real time to provide risk warnings for exceeding the limits of the regional integrated energy system's operating status includes: A set of safety threshold parameters is pre-configured, which includes thresholds for excessively high state of charge, excessively low state of charge, and excessively low regulation capability. The real-time acquired state of charge data is continuously compared with the over-high and under-high thresholds of state of charge. When the state of charge data is greater than the over-high threshold or less than the under-high threshold, the preset state of charge over-limit judgment condition is triggered. The real-time values ​​of the upward and downward adjustment capabilities are continuously compared with the threshold of the adjustment capability being too low. When the value of the upward adjustment capability is less than the threshold of the adjustment capability being too low or the value of the downward adjustment capability is less than the threshold of the adjustment capability being too low, the preset condition for insufficient adjustment capability is triggered. When the conditions for determining the state of charge exceeding the limit and the conditions for determining insufficient regulation capacity are triggered, risk warning information is generated, which includes the warning trigger time, warning type, specific value of the warning trigger, and corresponding safety threshold parameters. The warning types include state of charge exceeding the limit and insufficient regulation capacity.

9. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 6, characterized in that, Based on actual participation results in the electricity market, the process of iteratively revising and optimizing the gated loop unit includes: After the regional integrated energy system completes an actual regulation service response in the electricity market according to the participation decision-making instructions, it obtains the actual execution data of the actual regulation service from the electricity market settlement system. The actual execution data includes the execution time of the actual regulation service and the actual upward and downward power provided at the execution time. The upward and downward power constitute the true capacity label for the corresponding execution time. The actual capability labels are timestamped and associated with the two-dimensional input sequence corresponding to the execution time, so as to construct an incremental training sample consisting of a two-dimensional input sequence and actual capability labels from the actual participation results of a power market. According to the preset period, the incremental training samples generated within the period will be added to the total training dataset, thereby forming an expanded training dataset; Starting with the weight matrix and bias vector of the pre-trained gated recurrent unit, the expanded training dataset is input into the pre-trained gated recurrent unit. An adaptive moment estimation optimization algorithm is used to perform incremental training on the pre-trained gated recurrent unit for a specified number of rounds. During the incremental training process, the pre-trained gated recurrent unit fine-tunes the weight matrix and bias vector based on the prediction error revealed by the incremental training samples. After incremental training is completed, the original weight matrix and bias vector of the gated recurrent unit are replaced with the finely tuned weight matrix and bias vector to obtain the iteratively optimized gated recurrent unit.

10. The method for calculating and regulating the regulatory capacity of a regional integrated energy system participating in the electricity market according to claim 1, characterized in that, The process of dynamically plotting upward and downward adjustment capacity values ​​to visually represent the adjustment capacity boundary of a regional integrated energy system includes: On the monitoring interface of the regional energy management system, a two-dimensional rectangular coordinate system is established. The horizontal axis of the two-dimensional rectangular coordinate system is defined as the time axis, and the range of the time axis extends from the current moment to the end of the preset duration. The vertical axis of the two-dimensional rectangular coordinate system is defined as the power axis, with zero as the center. Positive values ​​upward represent upward adjustment capacity, and negative values ​​downward represent downward adjustment capacity. The sequence of output upward adjustment capability values ​​is plotted as the upper boundary curve on the positive half-axis of a two-dimensional rectangular coordinate system in chronological order. The sequence of output downward adjustment capability values ​​is plotted as the lower boundary curve on the negative half-axis of a two-dimensional rectangular coordinate system after taking the opposite value in chronological order. The drawing process is executed cyclically according to the time cycle, generating new upper and lower boundary curves. These new upper and lower boundary curves then cover the original upper and lower boundary curves on the monitoring interface, thereby visually presenting the regulation capacity boundary of the regional integrated energy system.

Images