Establishment and configuration method of engineering design and planning model

Abstract

The invention relates to an establishment and configuration method of an engineering design and planning model, in particular to application of a stochastic programming model considering real-time constraints in optimal scheduling of a reservoir hydropower station. According to the method, a time period is divided into a current time period and a future time period, discretization description is carried out on water flow in the future time period by adopting a random water flow mode network, and a model objective function is established with the goal of maximizing the expected generating capacity of a reservoir group system in a dispatching period; and configuring model constraint conditions including a water balance equation, upstream and downstream hydraulic connection, generating capacity, reservoir capacity limitation, output limitation and initial and boundary conditions. Meanwhile, the method further provides a step of cutting the model mode network, the water flow mode network is simplified in a systematic and operable mode, and therefore the calculation efficiency and the real-time performance of the model are improved.

Description

Technical Field

[0001] This invention relates to the field of engineering design and planning, and more specifically to a method for establishing and configuring an engineering design and planning model. Background Technology

[0002] Optimal scheduling of reservoirs and hydropower stations is crucial for achieving efficient water resource utilization and stable power system operation. However, traditional deterministic programming models struggle to address the challenges of low accuracy in flow forecasting and uncertainty in future inflow information. Stochastic programming models, by incorporating stochastic flow pattern networks, can better describe the uncertainty of future flow, thereby improving the robustness of scheduling decisions. However, existing stochastic programming models still suffer from the following problems in practical applications: High model complexity: Stochastic flow pattern networks typically contain a large number of flow pattern sequences, resulting in high computational costs and making it difficult to meet the requirements of real-time scheduling.

[0003] Poor model interpretability: The complex network of water flow patterns makes the model's decision-making process difficult to understand, which is not conducive to decision-makers making real-time adjustments and optimizations.

[0004] Poor model adaptability: Existing stochastic programming models are usually based on fixed flow pattern networks, which are difficult to adapt to the dynamic changes in flow forecasts. Summary of the Invention

[0005] The purpose of this invention is to provide a method for establishing and configuring engineering design and planning models to solve the problems existing in the prior art. Specifically, this invention provides an application of a stochastic programming model considering real-time constraints in the optimal scheduling of reservoir hydropower stations, including the following steps: Step 1: Establishing the model objective function. The objective function is established with the goal of maximizing the expected power generation of the reservoir system during the scheduling period. This objective function includes the deterministic power generation in the current period and the expected power generation in future periods, where the expected power generation in future periods is calculated using a stochastic flow pattern network. Step 2: Configuring model constraints. Model constraints are configured, including water balance equations, upstream and downstream hydraulic connections, power generation, reservoir capacity limits, output limits, and initial and boundary conditions, to ensure the feasibility and rationality of scheduling decisions. Step 3: Pruning the model pattern network. A systematic and operable process for pruning the model pattern network is provided, including initialization and data preparation, application probability threshold pruning, application scenario clustering pruning, and real-time integration and updating. This step simplifies the flow pattern network, thereby improving the computational efficiency and real-time performance of the model. The method for establishing and configuring an engineering design and planning model provided by this invention has the following beneficial effects: Improved computational efficiency: By pruning the model pattern network, the number of flow patterns can be significantly reduced, thereby lowering the computational complexity of the model and improving computational efficiency. Enhanced interpretability: The simplified flow pattern network makes the model's decision-making process clearer, facilitating real-time adjustments and optimizations by decision-makers. Improved adaptability: Through real-time integration and updating steps, the flow pattern network can be dynamically adjusted to adapt to dynamic changes in flow forecasts, improving the model's adaptability. Attached Figure Description

[0006] Figure 1 This is a flowchart illustrating a method for establishing and configuring an engineering design and planning model according to the present invention. Detailed Implementation

[0007] In multi-period stochastic programming models considering real-time constraints, time periods are divided into current and future periods. When applying this model to solve the optimal scheduling problem of a reservoir hydropower station, the time periods are used as time periods. The accuracy of water flow forecasts for the current period is generally high and can be directly used to formulate scheduling strategies. However, the accuracy of water flow forecasts for future periods is difficult to guarantee and is generally assumed to be a stochastic process, using a stochastic water flow pattern network for discretization. In the current period, decision-makers formulate a water release strategy based on the predicted inflow for this period and the stochastic inflow pattern for future periods. In real-time scheduling, the water release strategy for the current period, the reservoir storage at the end of the period, and the corresponding period benefits are all deterministic and unique values. Because the inflow information for future periods is uncertain, decision-makers can formulate different scheduling strategies based on different water flow patterns to correct the impact of insufficient water flow information on scheduling decisions. Therefore, the water release strategy for future periods is uncertain but related to specific water flow patterns. Decision-makers only use the water release strategy for the current period to guide real-time scheduling; as time progresses, decision-makers will adjust the scheduling strategy in real-time based on the latest information. Figure 1 As shown: 1. Establish the objective function of the model. The objective is to maximize the expected power generation of the reservoir system during the scheduling period. ; In the formula: For reservoir During the period Water flow mode The amount of electricity generated; Number of reservoirs; This represents the expected power generation of the reservoir system in the future time period (from time period 2 to time period T, per hour); Water flow mode The probability of; For water flow pattern network The objective function value corresponding to the stochastic programming model.

[0008] Power generation during the current period It is a deterministic and unique value; however, due to the uncertainty of future water release strategies, the corresponding power generation also has a randomness. Corresponding to water flow pattern The amount of electricity generated reflects the impact of the randomness of water flow on the efficiency of dispatching.

[0009] in, i This is the index for the water flow pattern, with values ​​ranging from 1, 2, ..., I (where I is the total number of patterns). For the current time period ( t =1) Reservoir m Deterministic power generation (unit: MW·h); For future time periods ( t ≥2) Reservoir m In water flow mode i Electricity generation (unit: MW·h).

[0010] The objective function is to maximize the expected power generation of the reservoir system during the scheduling period. This expected power generation is equal to the sum of the power generation of all reservoirs in the current period, plus the expected power generation in each future period (from the second period to the last period). The expected power generation in the future period is the sum of the products obtained by multiplying the probability of each flow pattern by the sum of the power generation of all reservoirs in the future period under that pattern.

[0011] like Figure 1 As shown: 2. Configure model constraints: (1) Water balance equation.

[0012] ; In the formula: and Reservoirs In water flow mode Down, Reservoir water storage at the beginning and end of the time period as well as Reservoirs During the period Water flow mode The following are the inflow, power generation, and wastewater discharge rates, in units of: ; For reservoir During the period Water flow mode Evaporation rate of water surface below, unit ; For time period The time period is long, and the unit is Example: 2.592×106s.

[0013] The reservoir's water storage at the beginning of the next period is equal to the reservoir's water storage at the beginning of the current period plus the product of the inflow rate of that period minus the power generation flow, the abandoned water flow, and the water surface evaporation rate, and the duration of that period.

[0014] (2) Upstream and downstream hydraulic connections.

[0015] ; Let the number of the upstream reservoir be 1, and the number of the downstream reservoir be... . For reservoir During the period Water flow mode The inflow interval below, i.e., the flow vector The Each element.

[0016] When the reservoir is the upstream reservoir (number 1), the inflow of the reservoir is equal to the interval inflow of the reservoir in the corresponding time period and flow pattern.

[0017] When a reservoir is not the upstream reservoir (serial number greater than or equal to 2), the inflow of the reservoir is equal to the interval inflow of the reservoir in the corresponding time period and flow pattern, plus the sum of the power generation flow and water discharge flow of its upstream reservoir in the same time period and flow pattern.

[0018] (3) Electricity generation.

[0019] ; ; In the formula: For reservoir In water flow mode Time period Average gross head, unit and These are the calculation functions for the reservoir water level and the tailrace water level, respectively. For reservoir power generation efficiency, unit .

[0020] Wherein, the reservoir water level calculation function is f1(x)=a1x 2 +a2x+a3, where a1, a2, a3 are the fitting coefficients of the reservoir's capacity-water level relationship curve. Example: a1=−2.1×10 −8 a2=0.015, a3=100; Tailwater level calculation function f2(x2)=b1x 2 +b2, where b1 and b2 are the fitting coefficients for the tailwater level-discharge relationship curve. Example: b1 = 3.5 × 10⁻⁶ −5 b2=50.2.

[0021] Power generation efficiency function =e1e2 ,e1 e2 is the unit conversion factor (usually 9.81×10⁻³), and e2 is the overall unit efficiency (range 0.85–0.95).

[0022] Power generation efficiency is equal to the product of the unit conversion factor, the overall unit efficiency, and the average gross head.

[0023] The power generation of a reservoir is equal to the product of the power generation flow rate during the corresponding period, the power generation efficiency of the corresponding average gross head, and the duration of that period.

[0024] The average gross head of the reservoir is equal to the reservoir water level (the input for calculating the reservoir water level is the average of the reservoir water storage at the beginning and end of the corresponding time period) minus the tailwater level (the input for calculating the tailwater level is the sum of the power generation flow and the water discharge flow in the corresponding time period).

[0025] (4) Storage capacity limitations.

[0026] ; In the formula: and Reservoirs exist Upper and lower limits of storage capacity at the end of the period, in units Example: 1×10 6 m³ and 5×10 6 m³.

[0027] The water storage capacity of a reservoir must be between the lower and upper limits of the reservoir's capacity at the end of the corresponding period.

[0028] (5) Output restrictions.

[0029] ; In the formula: and These are the lower and upper limits of the unit's output, respectively, in units. Examples: 10MW and 100MW.

[0030] The power generation per unit time of the reservoir (i.e., the power generation of the corresponding period divided by the duration of that period) must be between the lower and upper limits of the unit's output.

[0031] (6) Initial and boundary conditions.

[0032] ; In the formula: as well as Reservoirs Initial and ending storage capacity, in units The ending storage capacity refers to the year-end storage capacity.

[0033] The water storage volume of a reservoir in the first period is equal to its initial capacity; the water storage volume of a reservoir in the last period is equal to its final capacity.

[0034] 3. Pruning Model Pattern Network: For the model, the following steps are presented in a systematic and operational manner. Assume the input is a complete network of flow patterns (containing the number of time periods T, all flow pattern sequences and their probabilities p_g), and the output is a pruned, simplified network. Threshold parameters (such as probability threshold ε, number of clusters k) can be adjusted according to the specific problem (recommended initial values: ε=0.02, k=5).

[0035] Step 1: Initialization and Data Preparation Obtain the complete flow pattern network: Construct the network structure based on historical data or forecast models. The root node is the current time period (t=1, flow determinism), and each branch path corresponds to a future flow pattern sequence (e.g., the sequence is represented as a vector Q_g=[q_{t=2},q_{t=3},...,q_{t=T}], where q_t is the water flow rate at time t, in m³ / s).

[0036] Extracting pattern probability: Record the probability p_g of each path (or child node) (satisfying Σp_g=1).

[0037] Set the cropping parameters: Probability threshold ε (recommended ε=0.02, remove branches with probabilities less than this value).

[0038] Maximum number of branches N (optional, e.g., N=50, to control network size).

[0039] Similarity distance metric (for scene clustering): In non-current time periods (time periods t=2 to T), use Euclidean distance to calculate the distance between two water flow pattern sequences Q_g and Q_j: d(Q_g,Q_j).

[0040] Output: A list of the complete network structure and probabilities.

[0041] Step 2: Apply probability threshold pruning (remove low probability branches) Traverse all child nodes of the water flow pattern network (each child node represents a complete path): Calculate the total probability p_g for each path.

[0042] If p_g < ε, mark the path as "to be removed".

[0043] Remove all marked paths: Delete these low-probability branches and their parent nodes (if the parent node has no other child nodes).

[0044] Update probability: The probability of renormalizing the remaining path: p_j'=p_j / Σp_j (for all remaining j), ensuring Σp_j'=1.

[0045] After removing the low-probability branch, the probability of the remaining path is updated as follows: the original probability of the path divided by the sum of the original probabilities of all remaining paths.

[0046] Check the network size: If the number of remaining branches is ≤ N, go to step 4; otherwise, go to step 3 for further pruning.

[0047] Low-probability events contribute little to the expected power generation, and their removal has a negligible impact on the objective function. Decision-makers rely solely on the current period's water release strategy, and future uncertainties are handled through a model network; therefore, removing low-probability branches does not disrupt the real-time decision-making framework.

[0048] Step 3: Application scenario clustering and pruning (merging similar patterns) If the network size is still too large after step 2 (number of branches > N), use a clustering algorithm to merge similar flow patterns: Representing all paths as vectors: Treat the water flow sequence Q_g of each path as a vector of a point in the R{T-1} space (time period t=2 to T).

[0049] Applying k-means clustering: Input: All Q_g vectors and their probabilities p_g.

[0050] Set the number of clusters k (it is recommended to start with k=5, and adjust according to computing resources).

[0051] Perform clustering to obtain k clusters C_1, C_2, ..., C_k.

[0052] Select a representative pattern for each cluster: Calculate the cluster center Q_c = Σ_{g∈C}(p_g / p_C)*Q_g, where p_C = Σ_{g∈C}p_g is the total probability of clustering.

[0053] The cluster center is equal to the sum of the probabilities of each path within the cluster multiplied by the ratio of the probability of that path to the total probability of the cluster; where the total probability of the cluster is the sum of the probabilities of all paths within the cluster.

[0054] Replace all original patterns within the cluster with Q_c to form a new branch.

[0055] Update probability: For each new branch, the probability is set to p_C.

[0056] Reconstructing the network structure: The cluster centers are used as new child nodes, while maintaining the time-segment structure of the network (nodes in time segment t still correspond to time segment t).

[0057] Similar flow patterns (small distance) have similar impacts on scheduling strategies (e.g., power generation calculation depends on head, and head is affected by flow). Merging these patterns simplifies the network but does not significantly change the expected value. Power generation calculation depends on flow patterns, but similar flows lead to similar heads f1(x) and f2(x), therefore clustering is reasonable.

[0058] Step 4: Real-time integration and updates Initial pruning: Steps 1-3 are completed during the offline planning period to generate a simplified network for optimizing the scheduling model.

[0059] Real-time updates: During scheduling execution, when new water flow information arrives (such as "the decision-maker adjusts the scheduling strategy in real time based on the latest information"), a portion of the network is regenerated based on the new forecast (usually updating the pattern for future time periods).

[0060] Apply the same cropping steps (steps 2-3), but use a dynamic threshold (e.g., ε is adjusted based on forecast uncertainty).

[0061] Ensure rapid updates: Clustering algorithms (such as k-means) can be updated incrementally to support real-time performance.

[0062] Output: The dynamically adjusted network, used for strategy formulation in the next time period.

[0063] The method for establishing and configuring an engineering design and planning model provided by this invention has the following beneficial effects: Improved computational efficiency: By pruning the model pattern network, the number of flow patterns can be significantly reduced, thereby reducing the computational complexity of the model and improving computational efficiency. Enhanced interpretability: The simplified flow pattern network makes the model's decision-making process clearer, facilitating real-time adjustments and optimizations by decision-makers. Improved adaptability: Through real-time integration and updating steps, the flow pattern network can be dynamically adjusted to adapt to dynamic changes in flow forecasts, improving the model's adaptability. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of this invention should be determined by the preceding claims.

Claims

1. A method for establishing and configuring an engineering design and planning model, characterized in that, Includes the following steps: Establish a model objective function with the goal of maximizing the expected power generation of the reservoir system during the scheduling period; Configure model constraints, including water balance equations, upstream and downstream hydraulic connections, power generation, reservoir capacity limits, power output limits, and initial and boundary conditions; The model pattern network is pruned, including initialization and data preparation, application probability threshold pruning, application scenario clustering pruning, and real-time integration and updating.

2. The method according to claim 1, characterized in that, The objective function of the model includes the deterministic power generation in the current period and the expected power generation in the future period, wherein the expected power generation in the future period is calculated using a stochastic hydrodynamic model network.

3. The method according to claim 1, characterized in that, The water balance equation is used to calculate the reservoir's water storage volume at the beginning of the next period. The upstream and downstream hydraulic connection is used to determine the reservoir's inflow. The power generation is used to calculate the reservoir's power generation. The reservoir capacity limit is used to ensure that the reservoir's water storage volume is within the allowable range. The power output limit is used to ensure that the reservoir's power generation per unit time is within the allowable range. The initial and boundary conditions are used to determine the reservoir's initial and final storage capacity.

4. The method according to claim 1, characterized in that, The steps of pruning the model pattern network include: Initialization and data preparation: Obtain the complete flow pattern network and set the trimming parameters; Apply probability threshold pruning, remove low-probability branches and update the probabilities; Application scenario clustering and pruning, merging similar water flow patterns and updating probabilities; Real-time integration and updates allow for dynamic adjustment of the water flow pattern network during scheduling and execution.

5. The method according to claim 4, characterized in that, The pruning parameters include a probability threshold and a maximum number of allowed branches.

6. The method according to claim 4, characterized in that, The steps of applying probability threshold clipping include: Traverse all child nodes of the water flow pattern network and calculate the total probability of each path; If the total probability is less than the probability threshold, mark the path as "to be removed"; The probability of removing all marked paths and updating the remaining paths.

7. The method according to claim 4, characterized in that, The steps for clustering and cropping in the application scenario include: Treat the water flow sequence of each path as a vector of a point in space; Clustering algorithms are used to merge similar water flow patterns; Select a representative pattern for each cluster and update the probability.

8. The method according to claim 4, characterized in that, The real-time integration and updating steps include: Initial trimming is completed during the offline planning period; During scheduling execution, when new water flow information arrives, a partial water flow pattern network is regenerated; Apply the same cropping steps and use a dynamic threshold; Ensure rapid updates and output of dynamically adjusted water flow pattern networks.

Images