Hydroelectric photovoltaic combined peak shaving method and terminal

Abstract

The application discloses a kind of water electricity photovoltaic combined peak shaving method and terminal, obtain the water and electricity parameters of water and electricity system to be adjusted, and the photovoltaic parameter of photovoltaic system corresponding to the water and electricity system to be adjusted;According to the photovoltaic parameter and the water and electricity parameter, with the minimum of power grid residual load variance as target and with the opening and closing gate time of upstream hydropower station in the water and electricity system to be adjusted and the water release as undetermined parameter, construct water electricity photovoltaic complementary combined peak shaving model;Solving the water electricity photovoltaic complementary combined peak shaving model obtains the optimal opening and closing gate time of the upstream hydropower station and optimal discharge flow;The present application considers the time lag influence of upstream hydropower station to current hydropower station, and the opening and closing gate time of upstream hydropower station and water release are used as undetermined parameter, so as to determine the controllable part in water electricity photovoltaic combined system, realize peak shaving, reduce the negative influence of water flow time lag on power generation plan, realize more efficient, better water-light system peak shaving.

Description

Technical Field

[0001] This invention relates to the field of power grid peak shaving, and in particular to a method and terminal for combined hydropower and photovoltaic peak shaving. Background Technology

[0002] Fully leveraging the regulating role of cascade hydropower in river basins to achieve complementary and joint power generation of cascade hydropower and photovoltaic (PV) systems is an important way to promote the consumption of clean energy. However, PV output is greatly affected by weather, exhibiting strong volatility and uncertainty. When the fluctuation in the power output of the complementary system exceeds the regulation capacity of the power system, it can lead to the risk of power curtailment, which is detrimental to the stable operation of the power system. Furthermore, many cascade hydropower stations are run-of-river stations. The arrival of water from the upstream adjustable reservoirs at the downstream reservoirs involves a certain flow lag, resulting in a time delay in the power output of the hydropower stations. This presents certain challenges for joint hydropower-PV peak regulation.

[0003] Existing hydro-solar joint peak shaving methods still have some shortcomings in solving the problems of calculating water flow lag time and peak shaving scheduling. First, they do not take into account the impact of water flow lag time on the output time of downstream run-of-river hydropower stations; second, hydro-solar joint peak shaving is a nonlinear, multi-dimensional problem, and its solution involves a considerable amount of computation and is prone to getting trapped in local optima. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a method and terminal for combined hydropower and photovoltaic peak shaving, so as to achieve higher efficiency and better effect of hydropower and photovoltaic system peak shaving.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] A method for combined hydropower and photovoltaic peak shaving includes the following steps:

[0007] Obtain the hydropower parameters of the hydropower system to be regulated, and the photovoltaic parameters of the photovoltaic system corresponding to the hydropower system to be regulated;

[0008] Based on the photovoltaic parameters and the hydropower parameters, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing times and water discharge of the upstream hydropower station in the hydropower system to be regulated as undetermined parameters.

[0009] Solving the hydropower-photovoltaic complementary peak-shaving model yields the optimal gate opening and closing time and the optimal outflow from the upstream hydropower station.

[0010] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows:

[0011] A hydropower-photovoltaic combined peak-shaving terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps:

[0012] Obtain the hydropower parameters of the hydropower system to be regulated, and the photovoltaic parameters of the photovoltaic system corresponding to the hydropower system to be regulated;

[0013] Based on the photovoltaic parameters and the hydropower parameters, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing times and water discharge of the upstream hydropower station in the hydropower system to be regulated as undetermined parameters.

[0014] Solving the hydropower-photovoltaic complementary peak-shaving model yields the optimal gate opening and closing time and the optimal outflow from the upstream hydropower station.

[0015] The beneficial effects of this invention are as follows: With the goal of minimizing the variance of the grid surplus load, a hydropower-photovoltaic complementary joint peak-shaving model is established by introducing the hydropower parameters of the hydropower system to be regulated and the photovoltaic parameters of the photovoltaic system. Considering the impact of the time lag in the outflow of the upstream hydropower station in the hydropower system, and since the photovoltaic system is greatly affected by weather, which is mostly an uncontrollable factor, the time lag of the upstream hydropower station on the current hydropower station is taken into account. The gate opening and closing time and water release of the upstream hydropower station are taken as undetermined parameters, thereby determining the controllable part of the hydropower-photovoltaic joint system. Peak shaving is achieved while reducing the negative impact of water flow time lag on the power generation plan, thus achieving more efficient and better peak shaving of the hydropower-photovoltaic system. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of a combined hydropower and photovoltaic peak-shaving method according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the reinforcement learning process for solving the hydropower-photovoltaic complementary joint peak-shaving model in an embodiment of the present invention.

[0018] Figure 3 This is a flowchart illustrating the steps of solving the hydropower-photovoltaic complementary joint peak-shaving model according to an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of the structure of a hydropower-photovoltaic combined peak-shaving terminal according to an embodiment of the present invention;

[0020] Label Explanation:

[0021] 1. A hydropower-photovoltaic combined peak-shaving terminal; 2. A processor; 3. A memory. Detailed Implementation

[0022] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0023] Please refer to Figure 1 A method for combined hydropower and photovoltaic peak shaving includes the following steps:

[0024] Obtain the hydropower parameters of the hydropower system to be regulated, and the photovoltaic parameters of the photovoltaic system corresponding to the hydropower system to be regulated;

[0025] Based on the photovoltaic parameters and the hydropower parameters, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing times and water discharge of the upstream hydropower station in the hydropower system to be regulated as undetermined parameters.

[0026] Solving the hydropower-photovoltaic complementary peak-shaving model yields the optimal gate opening and closing time and the optimal outflow from the upstream hydropower station.

[0027] As can be seen from the above description, the beneficial effects of the present invention are as follows: taking the minimum variance of the grid surplus load as the objective, a hydropower-photovoltaic complementary joint peak-shaving model is established by introducing the hydropower parameters of the hydropower system to be regulated and the photovoltaic parameters of the photovoltaic system. Considering the impact of the time lag in the outflow of the upstream hydropower station in the hydropower system, and since the photovoltaic system is greatly affected by the weather, which is mostly an uncontrollable factor, the time lag of the upstream hydropower station on the current hydropower station is taken into account. The gate opening and closing time and water release of the upstream hydropower station are taken as undetermined parameters, thereby determining the controllable part of the hydropower-photovoltaic joint system. Peak shaving is achieved while reducing the negative impact of the water flow time lag on the power generation plan, and achieving higher efficiency and better effect of hydropower-photovoltaic system peak shaving.

[0028] Furthermore, the process of constructing the hydropower-photovoltaic complementary joint peak-shaving model includes:

[0029] Obtain the photovoltaic installed capacity of the photovoltaic system described in the historical data and the target photovoltaic output data under the same weather scenario;

[0030] The photovoltaic installed capacity and the weather scenario are used to obtain the photovoltaic output data prediction value through the initial prediction network. The initial prediction network is then optimized based on the preset number of iterations and the comparison results between the photovoltaic output data prediction value and the target photovoltaic output data to obtain the target prediction network.

[0031] The acquisition of photovoltaic parameters of the photovoltaic system includes:

[0032] The photovoltaic parameters include predicted photovoltaic output data;

[0033] The current weather scenario of the photovoltaic system is obtained, and the predicted photovoltaic output data is obtained through the target prediction network based on the current weather scenario and the photovoltaic installed capacity.

[0034] As described above, the target prediction network is obtained by training the initial prediction network based on the photovoltaic power output data corresponding to different weather scenarios in historical data. Each weather scenario corresponds to a target prediction network, which is more targeted to different weather scenarios and improves the prediction accuracy under different weather scenarios.

[0035] Furthermore, it also includes:

[0036] Set constraints on predicted photovoltaic output data:

[0037]

[0038]

[0039] in, The predicted photovoltaic power output data for time period t; The photovoltaic prediction error between the predicted photovoltaic output data for time period t and the target photovoltaic output data is denoted as t. This represents the maximum downward deviation of the photovoltaic power output from the prediction error during time period t. This represents the maximum value of the upward output deviation of the photovoltaic prediction error during time period t.

[0040] As described above, setting constraints on the predicted photovoltaic processing data helps prevent the final predicted photovoltaic output data from exceeding normal values, thus affecting the accuracy of subsequent calculations and ensuring that the predicted photovoltaic output data is close to the actual data.

[0041] Furthermore, the process of constructing the hydropower-photovoltaic complementary joint peak-shaving model includes:

[0042] Based on historical data, the inflow, interval inflow, power generation flow, water discharge, outflow, reservoir capacity at the end of the time period, the outflow of the upstream hydropower station corresponding to the hydropower station in each time period, and the sum of the delayed outflow reaching the hydropower station within the time period, the water balance equation of the hydropower station is constructed.

[0043] Based on the dam water level, lower limit, upper limit, initial and final water levels, lower limit of reservoir capacity, upper limit of reservoir capacity, and reservoir capacity function of the hydropower station for each time period, construct the water level-reservoir capacity constraint equation.

[0044] Based on the lower limit of power generation flow, upper limit of power generation flow, tailwater level, and tailwater level discharge flow function of the hydropower station for each time period, a tailwater level-discharge flow constraint equation is constructed.

[0045] A head constraint is constructed based on the net head and head loss of the hydropower station for each time period.

[0046] Based on the total inflow and interval flow of the hydropower station in each time period, and the time it takes for the outflow from the upstream hydropower station to reach the hydropower station in each time period, the hydraulic connection constraint equations of the cascade system are constructed.

[0047] Based on the water balance equation, water level-storage capacity constraint equation, tailwater level-discharge flow constraint equation, head constraint, and cascade system hydraulic connection constraint equation, the predicted power generation of the hydropower station and the relationship function between the power generation of the hydropower station and the power generation of the upstream hydropower station are predicted.

[0048] Construct a function relating the power generation of the upstream hydropower station to the gate opening and closing time and the water release volume.

[0049] As described above, in order to obtain more realistic optimization results when solving the hydropower-photovoltaic complementary joint peak-shaving model, corresponding constraints are constructed based on the objective conditions in the actual environment, so that the final optimization results can be achieved without exceeding the actual limitations.

[0050] Furthermore, the construction of the water balance equation for the hydropower station based on historical data of the inflow, interval inflow, power generation flow, wastewater discharge, outflow, reservoir capacity at the end of the time period, the outflow from the upstream hydropower station corresponding to the hydropower station for each time period, and the sum of the delayed outflow reaching the hydropower station within the time period includes:

[0051]

[0052]

[0053]

[0054] Where Δt is the length of time interval t; Q i,t S i,t , These represent the inflow, interval inflow, power generation flow, wastewater discharge, and outflow of hydropower station i during time period t; V i,t Let i be the reservoir capacity of hydropower station i at the end of time period t; Ω represents the outflow from the upstream hydropower station k during time period n; i Let i be the set of upstream hydropower stations k of hydropower station i; Let be the sum of the delayed outflow from power station k, which is directly upstream of power station i, during time period n, reaching downstream power station i during time period t.

[0055] As described above, constraining the water balance means that for a fixed hydropower station, its inflow, outflow, and reservoir capacity are interconnected and mutually influential. Constraining these relationships can simulate more realistic flow data.

[0056] Furthermore, the construction of head constraints based on the net head and head loss of the hydropower station for each time period includes:

[0057]

[0058]

[0059] Among them, H i,t The net head of power station i during time period t; The head loss of hydropower station i during time period t; Let Q be the head loss function of power station i; i,t Let be the power generation flow of hydropower station i during time period t.

[0060] As can be seen from the above description, constraining the head of the hydropower station allows for better simulation.

[0061] Furthermore, the construction of the hydropower-photovoltaic complementary joint peak-shaving model includes:

[0062] Based on the predicted photovoltaic output data and the predicted hydropower generation, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing times and water release volumes of the upstream hydropower stations in the hydropower system to be regulated as undetermined parameters.

[0063] As described above, it is difficult to monitor photovoltaic data and hydropower generation in real time. Furthermore, if real-time data is obtained before peak shaving calculations are started, there will be a lag. Therefore, making appropriate advance predictions based on historical data can enable timely peak shaving and further ensure the stability of the power grid.

[0064] Furthermore, the process of solving the hydropower-photovoltaic complementary joint peak-shaving model to obtain the optimal gate opening and closing time and the optimal outflow from the upstream hydropower station includes:

[0065] Building a reinforcement learning environment

[0066]

[0067]

[0068] in, This is the policy state, which is influenced by the action policy. For the environmental state, these feature variables are independent of the action policy; Nwt N represents the set of hydropower stations in the hydropower system to be regulated; Q N represents the set of water flow rates in the hydropower system to be regulated; ev Represents a collection of photovoltaic units; The predicted load on the power grid during time period t. This refers to the predicted photovoltaic output data of the photovoltaic system during time period t. The predicted power generation of the hydropower station during time period t;

[0069] Constructing Action Space Let be the outflow from the reservoir of hydropower station i at time t;

[0070] Using Actor neural network μ i (s i,t ;θ i The strategy function π is approximately continuous and deterministic. i The operating status s of system i, consisting of the photovoltaic system and the hydropower system to be regulated, during time period t. i,t As input, output action a i,t Acting on the learning environment, wherein a i,t ∈A t θ i The parameters of the Actor neural network are represented;

[0071] Using Critic Neural Network Evaluate the quality of the action and obtain the new operating state s after the action is executed. i,t+1 ;

[0072]

[0073] in, Indicates in s t ,a t Evaluation value of the Critic neural network under the following conditions The parameters of the Critic neural network are: K represents the preset number of training rounds, k represents the current number of training rounds; t represents the current time period, T represents the preset total number of time periods; and r represents the reward function.

[0074] Will Store in the experience pool;

[0075] Determine if t is greater than T. If so, determine if k is greater than K. If so, obtain the trained Actor neural network and use the trained Actor neural network to obtain the optimal gate opening and closing time and the optimal outbound flow rate for the environmental state to be predicted.

[0076] As described above, the introduction of Critic neural network and Actor neural network in the solution process, which work together to determine the optimal action by evaluating the action space, solves the problem of large computational load and easy getting trapped in local optima in the peak shaving model optimization process, and improves the accuracy of peak shaving strategy.

[0077] Furthermore, the reward function

[0078] Among them, P max P represents the peak value of the initial load curve. min P' is the trough of the initial load curve. max P' represents the peak value of the load curve after peak shaving. min This represents the valley value of the load curve after peak shaving.

[0079] As described above, the reward function measures the result of peak shaving by the difference between the peak and trough values ​​of the conformity curve before and after adjustment, which can intuitively reflect the effect of peak shaving.

[0080] Please refer to Figure 4 A hydropower-photovoltaic combined peak-shaving terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the aforementioned hydropower-photovoltaic combined peak-shaving method.

[0081] The above-mentioned hydropower-photovoltaic combined peak shaving method and terminal of the present invention can be applied to peak shaving in hydropower systems, especially in hydropower systems with upstream and downstream relationships or in hydropower-photovoltaic combined systems. The following is a description through specific embodiments.

[0082] Please refer to Figure 1 Embodiment 1 of the present invention is as follows:

[0083] A method for combined hydropower and photovoltaic peak shaving includes the following steps:

[0084] S011. Obtain the photovoltaic installed capacity of the photovoltaic system and the target photovoltaic output data under the same weather scenario from the historical data.

[0085] S012. The photovoltaic installed capacity and the weather scenario are used to obtain the photovoltaic output data prediction value through the initial prediction network. The initial prediction network is then optimized based on the preset number of iterations and the comparison results between the photovoltaic output data prediction value and the target photovoltaic output data to obtain the target prediction network.

[0086] In one optional implementation, the initial prediction network is an LSTM network, and the optimization process is as follows:

[0087] 1) Forgetting stage: Using the sigmoid function, by reading h t-1 and x t To determine which useless information to discard, namely:

[0088] f t =σ(W f [h t-1 ,x t ]+b f );

[0089] In the formula: f t The forget gate state at time t; σ() represents the sigmoid function; W f b f Represents the weight matrix and bias vector; h t-1 Output the short-term memory information of the sequence at time t-1; x t The input sequence, in this embodiment, is a time series of photovoltaic power output data, which can be obtained from photovoltaic installed capacity and weather scenarios; [h t-1 ,x t This implements the concatenation of two vectors;

[0090] Furthermore, since the sigmoid function compresses any input into the (0,1) interval, if a component of the vector becomes 0 after passing through the sigmoid layer, the corresponding component of the unit state after the bitwise multiplication will also become 0, thus discarding useless information.

[0091] 2) Input determination stage: Use the sigmoid function to obtain the new information that needs to be determined. And input c t The ratio, that is:

[0092]

[0093] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0094]

[0095] In the formula: i t Importance of extracted effective information; c t Output the long-term memory information of the sequence at time t; W represents the temporary information stored in the memory unit at time t. i and b i The weight matrix and bias vector of the sigmoid layer are represented by tanh; tanh represents the hyperbolic tangent function; W cand b c This represents the weight matrix and bias vector of the tanh layer; that is, c t This refers to model information that has been saved before new data is input, and the new information obtained after the new data is input. i t The extracted new information is filtered, and a score of (0,1) is given; finally, c is updated. t ;

[0096] 3) Output stage: The output will be processed using the tanh function. And the data classified by the sigmoid function o t The hidden layer data value h passed to the next time step t Generate the output results.

[0097] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0098] h t =o t ·tanh(c t );

[0099] Among them, W o and b o These represent the weight matrix and bias vector of the sigmoid layer, respectively.

[0100] Specifically, in one optional implementation, the training process includes:

[0101] S0121. Set the number of steps to 10;

[0102] S0122. Divide the data into training and testing sets, with the training set accounting for 70% of the total data and the testing set accounting for 30% of the total data.

[0103] S0123. Change the data dimension to three-dimensional data (n,1,1), where n represents the length of the data;

[0104] S0124. Initialize the LSTM model, set the number of neural network layers to 3, the number of iterations to 100, select the Adam optimizer as the optimizer, use the mean squared error (MSE) loss function, and set the optimization coefficient to 0.1.

[0105] S0125. Draw the loss function;

[0106] S0126. Use the test set to make predictions using the model;

[0107] S0127. Construct evaluation indicators to evaluate the regression:

[0108] Mean square error:

[0109] Root mean square error:

[0110] Mean absolute error:

[0111] Among them, y i It is the actual value. This is a predicted value;

[0112] S0128. Determine whether the evaluation result meets the preset evaluation index. If yes, stop training and obtain the target LSTM network, i.e., the target prediction network. Otherwise, if the number of iterations has not reached the preset number of iterations, increase the number of iterations by 1 and return to step S0126. When the number of iterations reaches the preset number of iterations, obtain the target LSTM network.

[0113] S013. Set constraints for predicted photovoltaic output data:

[0114]

[0115]

[0116] in, The predicted photovoltaic power output data for time period t; The photovoltaic prediction error between the predicted photovoltaic output data for time period t and the target photovoltaic output data is denoted as t. This represents the maximum downward deviation of the photovoltaic power output from the prediction error during time period t. This represents the maximum value of the upward output deviation of the photovoltaic prediction error during time period t;

[0117] S021. Based on historical data of the inflow, interval inflow, power generation flow, wastewater discharge, outflow, reservoir capacity at the end of the time period, the outflow of the upstream hydropower station corresponding to the hydropower station in each time period, and the sum of the delayed outflow reaching the hydropower station within the time period, construct the water balance equation of the hydropower station:

[0118]

[0119]

[0120]

[0121] Where Δt is the length of time period t. For example, if it is 1 day, the corresponding data needs to be multiplied by 3600 seconds of a day. Other lengths can be adjusted according to the determined length. Q i,t S i,t, These represent the inflow, interval inflow, power generation flow, wastewater discharge, and outflow of hydropower station i during time period t, respectively, in cubic meters per second; V i,t This represents the reservoir capacity of hydropower station i at the end of time period t, for example, in cubic meters. Ω represents the outflow from the upstream hydropower station k during time period n, for example, in cubic meters per second; i Let i be the set of upstream hydropower stations k of hydropower station i; Let be the sum of the delayed outflow from the reservoir of the direct upstream power station k during time period n, which arrives at the downstream power station i in time period t; where, the upstream hydropower station refers to the direct upstream hydropower station, that is, if there are hydropower stations A→B→C→D from upstream to downstream, the direct upstream hydropower station of C is B, and the direct upstream hydropower station of D is C.

[0122] S022. Based on the dam water level, lower limit water level, upper limit water level, initial and final water levels, lower limit reservoir capacity, upper limit reservoir capacity, and reservoir capacity function of the hydropower station for each time period, construct the water level-reservoir capacity constraint equation:

[0123]

[0124]

[0125]

[0126] V i,min ≤V i,t ≤V i,max

[0127] in, Let i be the water level above the dam of hydropower station i during time period t, for example, in meters; Z is the water level-capacity function of reservoir i; i,min and Z i,max These represent the lower and upper limits of the water level at hydropower station i during time period t, respectively, in meters, for example. and V represents the initial and final water levels given for hydropower station i, for example, in meters; i,min and V i,max These represent the lower and upper limits of the reservoir capacity of hydropower station i during time period t, respectively, in cubic meters.

[0128] S023. Construct the tailwater level-discharge flow constraint equation based on the lower limit of power generation flow, upper limit of power generation flow, tailwater level, and tailwater level discharge flow function for each time period of the hydropower station:

[0129] Q i,min ≤Q i,t ≤Q i,max ;

[0130]

[0131] In the formula: Q i,min and Q i,max These represent the lower and upper limits of the power generation flow of hydropower station i during time period t, respectively, in cubic meters per second. The tailwater level of hydropower station i during time period t, for example, in meters; The tailrace discharge flow function represents the flow rate of hydropower station i.

[0132] S024. Construct head constraints based on the net head and head loss of the hydropower station for each time period:

[0133]

[0134] Among them, H i,t The net head of hydropower station i during time period t, for example, in meters; This represents the head loss of hydropower station i during time period t, for example, in meters. Let Q be the head loss function of hydropower station i; i,t Let i be the power generation flow of hydropower station i during time period t;

[0135] S025 Constructs the hydraulic connection constraint equations for the cascade system based on the inflow and interval flow of the hydropower station at each time period, the outflow of the upstream hydropower station at each time period, and the time it takes for the outflow to reach the hydropower station:

[0136]

[0137] In the formula, TI d,t and NI d,t These are the total inflow and interval flow of the hydropower station d during time period t; U d,t τ is the outflow from the hydropower station d during time period t; d It is the time delay, that is, the time required for the outflow from the upstream hydropower station d to reach the hydropower station d at a certain time within the time period t;

[0138] S026. Based on the water balance equation, water level-storage capacity constraint equation, tailwater level-discharge flow constraint equation, head constraint, and cascade system hydraulic connection constraint equation, the predicted power generation of the hydropower station and the relationship function between the power generation of the hydropower station and the power generation of the upstream hydropower station are predicted.

[0139] The predicted power generation of the hydropower station includes:

[0140] S0260. Construct water balance constraints and total flow constraints during the scheduling period;

[0141] The water balance constraint is:

[0142]

[0143] In the formula, The outflow from upstream hydropower station k to hydropower station i during time period t is expressed in cubic meters per second. This is the number of time delays for the outflow from hydropower station i to hydropower station i during time period n. For example, if the unit is days, then a time delay of 2 means a delay of 2 days.

[0144] Furthermore, t represents the inflow period of the downstream hydropower station, and n represents the outflow period of the upstream hydropower station. The question is whether the outflow period plus the water flow stagnation time falls within the t-period of the downstream hydropower station's inflow. This represents the outflow from the reservoir of hydropower station k during time period n;

[0145] The total flow constraint is:

[0146]

[0147] The flow transferred to downstream hydropower station i in time period t is mainly divided into three parts: the first part is the remaining flow from the previous scheduling cycle, the second part is the scheduling flow for the current scheduling cycle, and the third part is the remaining flow to be transferred to the next scheduling cycle. This can be expressed by the formula:

[0148]

[0149] in, The outflow from the upstream power station k to the power station i during time period 0 is the delayed flow of the previous major scheduling cycle, for example, in cubic meters per second. To convert the hourly time period into a daily scheduling step length, This refers to the outflow from the upstream power plant during this scheduling cycle. The outflow from the reservoir of the direct upstream power station k to the power station i during time period T is expressed in cubic meters per second.

[0150] Furthermore, T represents the total number of time periods t (one scheduling cycle), i.e., one large scheduling cycle; t < 1 represents the portion of the previous large scheduling cycle, so the traffic within time period t < 1 is the remaining traffic (i.e., delayed traffic) of the previous large scheduling cycle; 1 < t ≤ T represents the intermediate scheduling cycle, which can receive the remaining traffic (i.e., delayed traffic) of the previous scheduling cycle and the scheduled traffic of the current scheduling cycle; t > T means that the scheduling of the T time periods t in the current period is completed, and the next large scheduling cycle begins, which is equivalent to t < 1 in the next round.

[0151] S0261. Calculation of power impact outside the scheduling cycle:

[0152]

[0153]

[0154] in, The outflow from the upstream power station at the end of the previous major scheduling period is a known value; This refers to the outflow from the upstream power station at the end of this scheduling period;

[0155] In an optional implementation, the method further includes: constructing a time-delay piecewise function for the flow rate of the cascade hydropower project.

[0156] This refers to dividing the incoming flow into three parts under the condition of satisfying the water balance constraints within the scheduling period. The first part is the remaining flow from the previous scheduling cycle, the second part is the scheduling flow for the current scheduling cycle, and the third part is the remaining flow to be transferred to the next scheduling cycle. The flow information is converted into power information using the power plant's average water consumption. The aforementioned incoming flow can be transformed into three parts: the power E transferred from the previous scheduling period. Ⅰ The power generation E generated by the water flow during this scheduling period Ⅱ And the transfer out volume E in the next scheduling period Ⅲ If the effect of water flow lag time is taken into account, the power generation during this scheduling period is E. Ⅰ +E Ⅱ If the effect of water flow lag is not considered, the power generation during this scheduling period is E. Ⅱ +E Ⅲ ;

[0157] S0262, External head during scheduling cycle Determination: The average water consumption rate is related to power generation water consumption and power generation. The head-water consumption rate curve is used to measure the impact of water consumption during the lag period on power generation and the maximum power generation. Specifically, the average tailwater level for each time period is obtained from S023, and the head outside the dispatch period is determined through S024.

[0158] S0263, Based on head Calculate the average water consumption rate r of power plant i in part u. i,u :

[0159]

[0160] Where u∈{I,III}; Let the head-water consumption rate function of power station i be...

[0161] S0264. Calculation of water volume impact outside the scheduling cycle:

[0162]

[0163]

[0164] Among them, E Ⅰ E represents the amount transferred in from the previous scheduling period. Ⅲ This represents the outflow for the next scheduling period; N is the total number of hydropower stations i. Since the data is in hours, it needs to be multiplied by the 24 hours included in Δt, which can be modified according to the length of Δt; I, II, and III refer to the inflow from the previous scheduling cycle, the flow in the current scheduling period, and the flow transferred to the next scheduling period, respectively.

[0165] S0265. Net power generation generated by water flow during this scheduling cycle:

[0166]

[0167] Among them, E II This represents the net power generation for this dispatch cycle; The outflow from upstream power station k to downstream power station i during time period t.

[0168] S0266. Power generation during this dispatch period:

[0169]

[0170] Where E Ⅱ E represents the net power generation generated by the water flow during this scheduling period. I E II E III All of these can be calculated using the formulas above, thus enabling the prediction of power generation during the current scheduling period. It can be seen that power generation is directly related to flow rate, i.e., the predicted power generation E of the hydropower station during time period t is calculated accordingly. a Directly related to the flow rates of I, II, and III;

[0171] S027. Construct a function relating the power generation of the hydropower station to the gate opening and closing times and water release volume of the upstream hydropower station, including:

[0172] S0271. Input initialization data and set constraints;

[0173] S0272. Calculate the inflow of hydropower station i in time period t according to step S0260.

[0174] S0273. Calculate the power generation during the dispatch period according to steps S0261 to S0266;

[0175] In an optional implementation, prior to executing S011-S027, the method further includes: preprocessing the historical data.

[0176] Data transformation. To avoid the influence of variable units, the cleaned data is normalized. The formula for min-max normalization is:

[0177]

[0178] Where V is the value of the original data, min(A) is the minimum value of the original data, max(A) is the maximum value of the original data, and V' is the value of the original data after min-max normalization. Pre-cleaning the original data can eliminate outliers and avoid errors in subsequent predictions and calculations.

[0179] S1. Obtain the hydropower parameters of the hydropower system to be regulated, and the photovoltaic parameters of the photovoltaic system corresponding to the hydropower system to be regulated, including:

[0180] S11. Obtain the predicted power generation of the hydropower station in the hydropower system to be regulated;

[0181] S12. The photovoltaic parameters include predicted photovoltaic output data; the current weather scenario of the photovoltaic system is obtained, and the predicted photovoltaic output data is obtained through the target prediction network based on the current weather scenario and the photovoltaic installed capacity;

[0182] S2. Based on the photovoltaic parameters and the hydropower parameters, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the objective of minimizing the grid surplus load variance and the gate opening and closing times and water release volumes of the upstream hydropower stations in the hydropower system to be regulated as undetermined parameters, including:

[0183] S21. Based on the predicted photovoltaic output data and the predicted hydropower generation, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing time and water release of the upstream hydropower station in the hydropower system to be regulated as undetermined parameters.

[0184] The objective function is:

[0185]

[0186] Where E represents the mean square error of the grid surplus load, T is the total number of scheduling periods t, and R... t The surplus load of the power grid during time period t is obtained by subtracting the power generation of the photovoltaic system and the hydropower system to be regulated during time period t from the power grid load during time period t; time period t is a dispatching period.

[0187] S22, and constrain the hydropower-photovoltaic complementary joint peak-shaving model by the water balance equation, water level-storage capacity constraint equation, tailwater level-discharge flow constraint equation, head constraint, and cascade system hydraulic connection constraint equation.

[0188] Please refer to Figure 2-3 S3. Solve the hydropower-photovoltaic complementary joint peak-shaving model to obtain the optimal gate opening and closing time and the optimal outflow from the upstream hydropower station, including:

[0189] S30. Initialize network parameters, set the number of training epochs N, learning rate α, and Critic network parameters. Actor network parameters θ i Initialize the current training round k = 1;

[0190] Build goals

[0191] S31. Constructing a reinforcement learning environment

[0192]

[0193]

[0194] in, This is the policy state, which is influenced by the action policy. For the environmental state, these feature variables are independent of the action policy; N wt N represents the set of hydropower stations in the hydropower system to be regulated; Q N represents the set of water flow rates in the hydropower system to be regulated; ev Represents a collection of photovoltaic units; The predicted load on the power grid during time period t. This refers to the predicted photovoltaic output data of the photovoltaic system during time period t. The predicted power generation of the hydropower station during time period t; The predicted water flow of the hydropower station in time period t; it can be seen from steps S0-S2 that, as well as It is related to the gate opening and closing times of the upstream hydropower station and the outflow from the reservoir;

[0195] S32, Constructing Action Space Let be the outflow from the reservoir of hydropower station i at time t;

[0196] S33. Initialize the system running state S1 at t=1;

[0197] S34. Utilizing the Actor neural network μ i (s i,t ;θ i The strategy function π is approximately continuous and deterministic. i The operating status s of system i, consisting of the photovoltaic system and the hydropower system to be regulated, during time period t. i,t As input, output action a i,t Acting on the learning environment, wherein a i,t ∈A t θi The parameters of the Actor neural network are represented;

[0198] In step S3, i is defined as system i, which combines scattered photovoltaic power stations and hydropower stations in a certain area into a power generation system, making it convenient to calculate the regional power output and to adjust the regional load.

[0199] S35. Utilizing Critic Neural Network Evaluate the quality of the action and obtain the new operating state s after the action is executed. i,t+1 ;

[0200]

[0201] in, Indicates in s t ,a t Evaluation value of the Critic neural network under the following conditions The parameters of the Critic neural network are: K represents the preset number of training rounds, k represents the current number of training rounds; t represents the current time period, T represents the preset total number of time periods; and r represents the reward function.

[0202]

[0203] Among them, P max P represents the peak value of the initial load curve. min P' is the trough of the initial load curve. max P' represents the peak value of the load curve after peak shaving. min This represents the trough value of the load curve after peak shaving.

[0204] S36, will Store in the experience pool;

[0205] S371. Determine if t is greater than T. If yes, execute S362; otherwise, execute S364.

[0206] S372. Determine if k is greater than K. If yes, execute S363; otherwise, execute S365.

[0207] S373. Obtain the trained Actor neural network, and use the trained Actor neural network to obtain the optimal gate opening and closing time and the optimal outbound flow rate, i.e. the optimal action, of the environmental state to be predicted.

[0208] S374, Return to step S34 and increment the value of t by 1;

[0209] S375. Return to step S33 and increment the value of k by 1.

[0210] Please refer to Figure 2Embodiment two of the present invention is as follows:

[0211] A hydropower-photovoltaic combined peak-shaving terminal 1 includes a processor 2, a memory 3, and a computer program stored in the memory 3 and executable on the processor 2. When the processor 2 executes the computer program, it implements the steps in Embodiment 1.

[0212] In summary, this invention provides a combined hydropower and photovoltaic peak-shaving method and terminal. Addressing the issue of water flow lag directly affecting the water balance between upstream and downstream power stations, leading to water wastage or under-generation, it proposes a hydropower output prediction method considering water flow lag. By employing a segmented water flow lag method and the correlation between hydropower station output and flow rate, it reduces the potential negative impact of water flow lag on power generation plans and feasibility. Furthermore, it constructs a cascade hydropower-photovoltaic complementary combined peak-shaving model considering water flow lag. By predicting grid load and photovoltaic power generation, it considers the impact of water flow lag, improving the precise regulation capability of hydropower, promoting the absorption of photovoltaic output, and resolving the impact of the randomness and instability of photovoltaic power generation on peak shaving. Finally, it proposes a reinforcement learning-based peak-shaving strategy optimization method, solving the problem of high computational cost and susceptibility to local optima in the peak-shaving model optimization process, improving the accuracy of the peak-shaving strategy, and better addressing practical peak-shaving problems.

[0213] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for combined hydropower and photovoltaic peak shaving, characterized in that, Including the following steps: Obtain the hydropower parameters of the hydropower system to be regulated, and the photovoltaic parameters of the photovoltaic system corresponding to the hydropower system to be regulated; Based on the photovoltaic parameters and the hydropower parameters, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing times and water discharge of the upstream hydropower station in the hydropower system to be regulated as undetermined parameters. Solving the hydropower-photovoltaic complementary joint peak-shaving model yields the optimal gate opening and closing times and the optimal outflow from the upstream hydropower station, including: Constructing a reinforcement learning environment S[ , ]; ； ； in, This is the policy state. Environmental state; This represents the collection of hydropower stations in the hydropower system to be regulated; This represents the set of water flow rates in the hydropower system to be regulated; Represents a collection of photovoltaic units; The predicted load on the power grid during time period t. This refers to the predicted photovoltaic output data of the photovoltaic system during time period t. The predicted power generation of the hydropower station during time period t; Constructing Action Space ; Let be the outflow from the reservoir of hydropower station i at time t; Using Actor Neural Network Approximately continuous deterministic policy function The operating status of system i, consisting of the photovoltaic system and the hydropower system to be regulated, during time period t. As input, output action It acts on the learning environment, wherein, , The parameters of the Actor neural network are represented; Using Critic Neural Network Evaluate the quality of the action and obtain the new operating state after the action is executed. s i,t+1 ; ； in, Indicates in Evaluation value of the Critic neural network under the following conditions The parameters of the Critic neural network are: K represents the preset number of training rounds, k represents the current number of training rounds; t represents the current time period, T represents the preset total number of time periods; and r represents the reward function. Will Store in the experience pool; Determine if t is greater than T. If so, determine if k is greater than K. If so, obtain the trained Actor neural network and use the trained Actor neural network to obtain the optimal gate opening and closing time and the optimal outbound flow rate of the environment to be predicted. The reward function ; in, This represents the peak value of the initial load curve. This represents the trough of the initial load curve. This represents the peak value of the load curve after peak shaving. This represents the valley value of the load curve after peak shaving.

2. The method for combined hydropower and photovoltaic peak shaving according to claim 1, characterized in that, Before constructing the hydropower-photovoltaic complementary joint peak-shaving model, the following steps are included: Obtain the photovoltaic installed capacity of the photovoltaic system described in the historical data and the target photovoltaic output data under the same weather scenario; The photovoltaic installed capacity and the weather scenario are used to obtain the photovoltaic output data prediction value through the initial prediction network. The initial prediction network is then optimized based on the preset number of iterations and the comparison results between the photovoltaic output data prediction value and the target photovoltaic output data to obtain the target prediction network. Obtaining the photovoltaic parameters of a photovoltaic system includes: The photovoltaic parameters include predicted photovoltaic output data; The current weather scenario of the photovoltaic system is obtained, and the predicted photovoltaic output data is obtained through the target prediction network based on the current weather scenario and the photovoltaic installed capacity.

3. The method for combined hydropower and photovoltaic peak shaving according to claim 2, characterized in that, Also includes: Set constraints for current photovoltaic output data: ； ； in, The predicted photovoltaic power output data for time period t; The photovoltaic prediction error between the predicted photovoltaic output data for time period t and the target photovoltaic output data is denoted as t. This represents the maximum downward deviation of the photovoltaic power output from the prediction error during time period t. This represents the maximum value of the upward output deviation of the photovoltaic prediction error during time period t.

4. The method for combined hydropower and photovoltaic peak shaving according to claim 2, characterized in that, Before constructing the hydropower-photovoltaic complementary joint peak-shaving model, the following steps are included: The water balance equation of the hydropower station is constructed based on the historical data of the inflow, interval inflow, power generation flow, water discharge, outflow, reservoir capacity at the end of the time period, the outflow of the upstream hydropower station corresponding to the hydropower station in each time period, and the sum of the delayed outflow reaching the hydropower station within the time period. Based on the dam water level, lower limit, upper limit, initial and final water levels, lower limit of reservoir capacity, upper limit of reservoir capacity, and reservoir capacity function of the hydropower station for each time period, construct the water level-reservoir capacity constraint equation; Based on the lower limit of power generation flow, upper limit of power generation flow, tailwater level, and tailwater level discharge flow function of the hydropower station for each time period, a tailwater level-discharge flow constraint equation is constructed. A head constraint is constructed based on the net head and head loss of the hydropower station for each time period. Based on the total inflow and interval flow of the hydropower station in each time period, and the time it takes for the outflow from the upstream hydropower station to reach the hydropower station in each time period, the hydraulic connection constraint equations of the cascade system are constructed. The predicted power generation of the hydropower station is obtained based on the water balance equation, water level-storage capacity constraint equation, tailwater level-discharge flow constraint equation, head constraint, and hydraulic connection constraint equation of the cascade system.

5. The method for combined hydropower and photovoltaic peak shaving according to claim 4, characterized in that, The water balance equation for the hydropower station is constructed based on historical data, including the inflow, inter-regional inflow, power generation flow, wastewater discharge, outflow, reservoir capacity at the end of the period, the outflow from the upstream hydropower station corresponding to the hydropower station for each period, and the sum of the delayed outflow reaching the hydropower station within the period. ； ； ； in, The length of time interval t; , , , , Hydropower stations During the period Inflow, inter-regional inflow, power generation flow, wastewater discharge, and outflow; For hydroelectric power station During the period The final storage capacity; For upstream hydropower station During the period Outbound flow; For hydroelectric power station upstream hydropower station A set; For power station direct upstream power station Time period Delayed outbound flow during the time period Reaching downstream power station The sum of .

6. The method for combined hydropower and photovoltaic peak shaving according to claim 4, characterized in that, The head constraint construction based on the net head and head loss of the hydropower station for each time period includes: ； in, For power station During the period The internal water purification head; For hydroelectric power station During the period The water level above the dam; For hydroelectric power station During the period The tailwater level inside; For hydroelectric power station During the period Internal head loss; For power station The head loss function; Let be the power generation flow of hydropower station i during time period t.

7. The method for combined hydropower and photovoltaic peak shaving according to claim 4, characterized in that, The construction of the hydropower-photovoltaic complementary joint peak-shaving model includes: Based on the predicted photovoltaic output data and the predicted hydropower generation, a hydropower-photovoltaic complementary joint peak-shaving model is constructed with the goal of minimizing the grid surplus load variance and the gate opening and closing times and water release volumes of the upstream hydropower stations in the hydropower system to be regulated as undetermined parameters.

8. A hydropower-photovoltaic combined peak-shaving terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the hydropower-photovoltaic combined peak-shaving method according to any one of claims 1-7.

Images