图的遍历(DFS,BFS,Topology Sort)

本文详细介绍了图的存储结构,包括稠密图的邻接矩阵和稀疏图的邻接链表。接着,文章阐述了图的深度优先搜索(DFS)和广度优先搜索(BFS)的遍历方法,提供了相应的C++代码实现。最后,讨论了有向无环图(DAG)的拓扑排序算法,通过BFS遍历寻找拓扑序列。此外,文中给出了相关例题和ACWing平台上的题目链接,帮助读者加深理解。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >


图系列专题:

图的深搜,宽搜,判断有向无环图

图最短路问题

yxc路的最短路相关算法对应场景图:



1. 图的存储结构

  • 对于稠密图,使用邻接矩阵进行存储和表示。邻接矩阵即二维数组,数组中的每个元素分别表示一有向边的边权值,在实际使用时,我们在初始化时先按照字节设置为无穷大 ∞ \infty ,实际使用memset()函数字节填充0x3f,为何使用该值见:编程中将无穷大常量的设定为0x3f3f3f3f。常用模板为:
#include <cstring> //为了使用memset()函数
#include <iostream>
using namespace std;
const int N = 100010;
int g[N][N];

...

int main() {
	memset(g, 0x3f, sizeof(g));
	...
}
  • 对于稀疏图,使用邻接链表。若是使用邻接矩阵的话会造成较大的空间浪费,这里对于邻接链表的构造,使用数组多链表的形式进行,也即将每个点与其所邻接的点构建一个链表,所有的点的链表组合起来就可以表示一个图。对于邻接链表,含有h[N](存储头节点下标索引(即对于idx),初始化将所有头节点索引设置为 − 1 -1 1),e[N](从 h [ t ] h[t] h[t] 进行遍历,则 e [ i ] e[i] e[i] 存储节点索引为 i i i 的节点的实际值,在实际使用中int j = e[i];取出节点 t t t 的邻接点),ne[N](存储链表中的下一节点,在遍历中i = ne[i]作为迭代方式),idx(记录数组中当前已用了多少索引值)四个要素,若是还要记录节点ti之间的距离,则还需w[N]( w [ i ] w[i] w[i] 表示由某个节点 t t t h [ t ] h[t] h[t] 遍历过来到节点 e [ i ] e[i] e[i],即由节点 e [ i ] e[i] e[i] 到节点 t t t 的边权值);同时最主要的操作是add(int a, int b, int c),表示的是增加一条由节点 a a a 指向节点 b b b 的边权值为 c c c 的边。常用模板为:
#include <cstring>
#include <iostream>
using namespace std;
const int N = 100010;
int h[N], e[N], ne[N], idx, w[N];

void add(int a, int b, int c) {
	e[idx] = b; //记录节点a有指向b的有向边
	ne[idx] = h[a]; //使用头插法将节点b插入到a的邻接链表当中,
	//将新的链表节点的next指针指向h[a]
	w[idx] = c; //记录节点a到节点b的距离/边权值
	h[a] = idx++; //更新节点a的邻接链表的头节点,同时完成了添加操作后idx须自加
}

...

int main() {
	memset(h, -1, sizeof(h));
	...
}


2. 图的遍历

2.1 图的DFS遍历

DFS是一个执著的人,每次搜索若不搜索到叶子节点,即遇到了死胡同绝不返回到上一步(也即是回溯)。对于DFS,我们知道一般使用的是栈这一数据结构,程序设计当中的递归就是使用递归栈来实现的,所以自然而言的我们想到使用递归方式进行图的DFS深度优先搜索遍历。同时为了记录图中的节点是否已经遍历过,所以我们需要增添一个bool st[N]数组对遍历情况加以记录。

DFS遍历的代码实现如下:

#include <cstring>
#include <iostream>
#include <algorithm>

using namespace std;
const int N = 100010;
int h[N], e[N], ne[N], idx; //数组含义见邻接链表的图表示方法
bool st[N]; //记录节点是否遍历过

void add(int a, int b) {
	e[idx] = b; ne[idx] = h[a]; h[a] = idx++;
}

void dfs(int u) {
	//将当前节点u标记为已经遍历过,然后对其子节点进行dfs遍历
	st[u] = true;
	//对节点u的的所有邻接点分别做dfs遍历
	for(int i = h[u]; i != -1; i = ne[i]) { 
		int j = e[i]; //取出与u节点相邻的节点j
		if(!st[j]) { //若是未遍历过,则进行dfs遍历
			dfs(j);
			printf("%d ", j);	
		}
	}
}

int main() {
	memset(h, -1, sizeof(h)); //初始化邻接链表的头节点们
	...
}

相关例题见:AcWing 846. 树的重心 (对于树的重心,可在DFS的搜索的过程中记录以遍历节点为根的树所含节点个数 s u m sum sum,同时可记录去掉该点后每个连通块中点的个数的最大值 r e s = m a x { d f s ( 1 ) , d f s ( 2 ) , . . . } res=max\{dfs(1),dfs(2),...\} res=max{dfs(1),dfs(2),...},再与 n − s u m n-sum nsum比较, r e s res res取两者中的最大值,最终结果取所有得到的 r e s res res当中的最小值即可)。


2.2 图的BFS遍历

BFS,即广度优先遍历,首先先和自己邻近的人打招呼,其次再和有一个中间人的人打招呼,以此类推,就像剥洋葱一样,只不过这次是从洋葱芯自内向外罢了。对于BFS常用的实现辅助数据结构为队列(queue),在遍历时,将头遍历节点从队列中取出并出队,再将头遍历节点的所有邻接点加入到队列当中去,如此进行循环,直到队列的队头和队尾的索引值相等/队列为空为止,即可实现图的BFS遍历。

BFS遍历的实现代码如下:

//这里不使用stl当中的queue容器实现,而是使用一个数组模拟队列的行为
#include <cstring>
#include <iostream>
#include <queue>
using namespace std;
const int N = 100010;
int h[N], e[N], ne[N], idx;
int q[N], hh, tt; //hh表示队列的头节点索引,tt表示队尾节点索引
bool st[N]; //记录节点是否遍历过

void add(int a, int b) {
	e[idx] = b; ne[idx] = h[a]; h[a] = idx++;
}

void bfs(int u) {
	st[u] = true;
	q[hh] = u;
	
	while(hh <= tt) { //hh<=tt判断队列不为空
		int t = q[hh++]; //将队头节点出队,并将队头索引值自增1
		
		//对队头节点的所有邻接点进行遍历
		for(int i = h[t]; i != -1; i = ne[i]) { 
			int j = e[i]; //取出队头邻接点的实际值,即节点j
			if(!st[j]) { //若是节点j尚未遍历过,则将其加入队列
				q[++tt] = j; //将节点j加入队列
				printf("%d ", j);
			}	
		}
	}		
}

int main() {
	memset(h, -1, sizeof(h));
	...
}

正是这个实现的过程,所以可以料想到,若是当图中的边的边权值都为相同的正值时,则根据BFS遍历的层次可以求得所遍历的点到出发点的最短距离。相关例题见:AcWing 844. 走迷宫(本人使用稠密图的表示方式实现的BFS遍历过程)。



3. 图的拓扑排序(Topology Sort)

若一个由图中所有点构成的序列 A A A 满足:对于图中的每条边 ( x , y ) (x,y) (x,y) x x x A A A 中都出现在 y y y 之前,则称 A A A 是该图的一个拓扑序列。可以理解为图中不含有环形链表。

对于图的拓扑排序,是有向无环图(DAG)的特有性质,即若一个图是有向无环图,则其必存在一拓扑排序,使得序列前面的点只有向后的出边而无来自后边点的入度。同时另一性质为,若图是有向无环图,则图中一定存在入度为0的节点。

例题见:848. 有向图的拓扑序列。求一个图的拓扑排序的方式为,在构造图的过程中,记录每个节点的入度。在求解时,首先遍历所有的图中节点,找出所有入度为 0 0 0的节点,将它们塞入队列当中,依次取出队头节点,将该节点的所有出边所指向的节点的入度减少(即删边操作),若是所指节点的入度变为 0 0 0,则将所指节点加入队列中。重复以上操作,直至队列为空,若是队列数组所用的索引数等于图中节点数,则说明有拓扑排序序列,且队列数组中所存储的序列即为拓扑序列之一,否则没有拓扑排序序列。使用的遍历方法为BFS遍历。

求解拓扑排序的代码实现如下:

 #include <iostream>
 #include <cstring>
 
 using namespace std;
 
 const int N = 100010, M = N * 2;
 int n, m;
 int h[N], e[M], ne[M], idx;
 int d[N];
 int q[N], tt, hh;
 
 void add(int a, int b) {
    e[idx] = b;
    ne[idx] = h[a];
    h[a] = idx++;
 }
 
 bool topologySort() {
 	//对图中的节点进行遍历,找出入度为0的点加入到队列当中
    for(int i = 1; i <= n; ++i) 
        if(!d[i]) q[tt++] = i; 
        
    while(hh <= tt) { //若队列不为空
        int t = q[hh++]; //取出队头元素,并将队头弹出
        for(int i = h[t]; i != -1; i = ne[i]) { //遍历队头节点的邻接点
            int j = e[i];
            d[j]--; //删减相应所指向点的入度
            if(!d[j]) q[tt++] = j; //若是入度为0,则加入到队列当中
        }
    }
    return tt == n; //若是曾存在于队列中的节点数等于图的节点数,则有拓扑排序序列
 }
 
 int main() {
    memset(h, -1, sizeof(h));
    cin >> n >> m;
    
    for(int i = 0; i < m; ++i) {
        int a, b;
        cin >> a >> b;
        d[b]++; //记录节点b的入度
        add(a, b); //添加a->b的有向边
    }
    
    if(topologySort())
        for(int i = 0; i < n; ++i) 
            cout << q[i] << " ";
    else puts("-1");
     
    return 0;
 }
import pandas as pd import networkx as nx import numpy as np import itertools # --- Constants and Assumptions --- # These should be clearly stated and can be modified. VOLTAGE_KV = 10.0 # Line voltage in kV ROOT_3 = np.sqrt(3) BASE_DG_CAPACITY_KW = 300.0 # Initial capacity for each DG N_DG = 8 # Failure rates from problem description FAILURE_RATE_DG_PERCENT = 0.5 / 100.0 # FAILURE_RATE_USER_PERCENT = 0.5 / 100.0 # Not directly used in this simplified line-fault model for widespread outages # FAILURE_RATE_SWITCH_PERCENT = 0.2 / 100.0 # Assuming switch failures manifest as line failures or inability to operate tie FAILURE_RATE_LINE_PER_KM = 0.002 # Per km per year (assuming rates are annual) # Costs (placeholders - these are critical for actual risk values) # Value of Lost Load (VoLL) in monetary units per kW per hour. # For risk = P * C, if P is annual probability, C should be impact of one event. # Let's define C_loss as total kW unserved * a severity factor. # Or, if we want an annual risk cost: P_annual_fault * kW_unserved * hours_outage * cost_per_kWh # For simplicity, using $/kW of unserved load for the consequence C. COST_VOLL_PER_KW = 10.0 # Example: $10 per kW of unserved load AVG_OUTAGE_DURATION_H = 4 # Example: average hours for an outage, if converting to energy # Cost of Overload (Consequence C_over) # This can be complex: accelerated aging, tripping, damage. # Simplified: A penalty if any line is overloaded in a given state. COST_PENALTY_FOR_ANY_OVERLOAD = 1000.0 # Example: $1000 penalty if system is in an overloaded state # Or, a cost per MWh of overloaded energy, or per overloaded line. # Line and Feeder Capacities # Main feeder rated current from problem: 220A. # P_rated_feeder_kW = ROOT_3 * VOLTAGE_KV * FEEDER_RATED_CURRENT_A * 1.0 (pf=1) # = 1.732 * 10 * 220 = 3810.4 kW (approx 3.8 MW, problem says 2.2MW for 220A, implies lower pf or different basis) # Let's use current as the primary limit. FEEDER_RATED_CURRENT_A = 220.0 # Assumption for individual line segments: For this model, we'll assume all lines # have a rated current equal to the main feeder. This is a strong simplification. # A more detailed model would assign ratings based on conductor types or downstream load. LINE_RATED_CURRENT_A = 100.0 # More conservative assumption for individual segments than 220A. Needs proper engineering values. # For lines directly from substation, perhaps 220A is more appropriate. # Let's use a dictionary for specific line ratings if known, else default. DEFAULT_LINE_RATED_CURRENT_A = 100.0 # Tie Line Capacity TIE_LINE_RATED_CURRENT_A = 150.0 # Assumption, should be based on tie switch/line capacity # DG Locations (Node IDs from 1 to 62) - Based on Figure 1 interpretation DG_LOCATIONS_KW = { 6: BASE_DG_CAPACITY_KW, 10: BASE_DG_CAPACITY_KW, 15: BASE_DG_CAPACITY_KW, 27: BASE_DG_CAPACITY_KW, 31: BASE_DG_CAPACITY_KW, 37: BASE_DG_CAPACITY_KW, 50: BASE_DG_CAPACITY_KW, 58: BASE_DG_CAPACITY_KW } # Tie Switches: (node1, node2, switch_id_text) - normally open # Interpretation based on careful review of Figure 1: # S13-1: (13, 22) - Intra-Feeder 1 (Connects two branches of Feeder 1) # S29-2: (29, 42) - Intra-Feeder 2 (Connects two branches of Feeder 2) # S62-3: (62, 19) - Inter-Feeder (Connects Feeder 3 (node 62) to Feeder 1 (node 19)) TIE_SWITCHES_INFO = [ {'nodes': (13, 22), 'id': 'S13-1', 'type': 'intra-F1', 'capacity_A': TIE_LINE_RATED_CURRENT_A}, {'nodes': (29, 42), 'id': 'S29-2', 'type': 'intra-F2', 'capacity_A': TIE_LINE_RATED_CURRENT_A}, {'nodes': (62, 19), 'id': 'S62-3', 'type': 'inter-F3_F1', 'capacity_A': TIE_LINE_RATED_CURRENT_A} ] # This interpretation means Feeder 2 cannot directly receive support from F1 or F3. # If problem implies all feeders can support each other, TIE_SWITCHES_INFO would need redefinition. # Substation connection points (source nodes for feeders) # CB1 -> Node 1, CB2 -> Node 23, CB3 -> Node 43 # Node 0 will represent the main grid / infinite source. SOURCE_NODE = 0 SUBSTATION_CONNECTIONS = { 'CB1': (SOURCE_NODE, 1), 'CB2': (SOURCE_NODE, 23), 'CB3': (SOURCE_NODE, 43) } # Capacity of connection from source to substation nodes (effectively feeder capacity) SUBSTATION_LINE_CAPACITY_A = FEEDER_RATED_CURRENT_A # --- Data Loading Functions --- def load_load_data(filename="C题附件:有源配电网62节点系统基本参数.xlsx - 表1 有源配电网62节点系统负荷参数.csv"): df = pd.read_csv(filename) df.columns = ['node_id', 'load_kw'] # Convert node_id to int if it's not already df['node_id'] = df['node_id'].astype(int) return df.set_index('node_id')['load_kw'].to_dict() def load_topology_data(filename="C题附件:有源配电网62节点系统基本参数.xlsx - 表2 有源配电网62节点系统拓扑参数.csv"): df = pd.read_csv(filename) # Rename columns for easier access (assuming standard Chinese headers) df.columns = ['line_num', 'from_node', 'to_node', 'length_km', 'resistance_ohm', 'reactance_ohm'] # Convert relevant columns to numeric for col in ['from_node', 'to_node', 'length_km', 'resistance_ohm', 'reactance_ohm']: df[col] = pd.to_numeric(df[col], errors='coerce') return df # --- Core Power Grid Model Class --- class PowerGridModel: def __init__(self, load_data, topology_data, dg_locations_kw, tie_switches_info, substation_connections): self.loads_kw = load_data self.topology_df = topology_data self.dg_kw = dg_locations_kw.copy() # Allow modification for different scenarios self.tie_switches_info = tie_switches_info self.substation_connections = substation_connections self.graph = self._build_graph() self.feeder_info = self._identify_feeders() def _build_graph(self): G = nx.Graph() # Use Graph for undirected, or DiGraph if flow direction is fixed by sources # Add nodes with load and DG info all_nodes = set(self.topology_df['from_node']) | set(self.topology_df['to_node']) for node_id in all_nodes: node_id = int(node_id) # Ensure int G.add_node(node_id, load_kw=self.loads_kw.get(node_id, 0), dg_kw=self.dg_kw.get(node_id, 0)) # Add lines from topology data for _, row in self.topology_df.iterrows(): u, v = int(row['from_node']), int(row['to_node']) G.add_edge(u, v, id=row['line_num'], length_km=row['length_km'], resistance_ohm=row['resistance_ohm'], # reactance_ohm=row['reactance_ohm'], # Ignoring reactance as per problem rated_current_a=DEFAULT_LINE_RATED_CURRENT_A, # Default, can be refined failed=False) # Add substation connections (virtual lines from a common source) # These represent the main feeder lines from CBs G.add_node(SOURCE_NODE, type='source') for cb_id, (src, dest_node) in self.substation_connections.items(): G.add_edge(src, dest_node, id=cb_id, length_km=0.01, resistance_ohm=0.001, # Minimal impedance rated_current_a=SUBSTATION_LINE_CAPACITY_A, type='substation_link', failed=False) return G def _get_subgraph_with_operational_lines(self, graph_to_copy, faulty_line_edge=None): """Creates a subgraph considering only non-failed lines and open tie switches.""" g_op = graph_to_copy.copy() # Remove failed lines lines_to_remove = [] if faulty_line_edge: # faulty_line_edge is (u,v) if g_op.has_edge(*faulty_line_edge): lines_to_remove.append(faulty_line_edge) for u, v, data in list(g_op.edges(data=True)): if data.get('failed', False): lines_to_remove.append((u,v)) g_op.remove_edges_from(lines_to_remove) # Normally, tie switches are open. For restoration, specific ones might be closed. # This base function assumes they are open unless explicitly handled by restoration logic. return g_op def _identify_feeders(self): """Identifies nodes belonging to each feeder under normal operation (tie switches open).""" g_normal = self._get_subgraph_with_operational_lines(self.graph) feeder_info = {} # {'CB1': {nodes}, 'CB2': {nodes}, ...} for cb_id, (src_node, start_node) in self.substation_connections.items(): if g_normal.has_node(start_node) and g_normal.has_node(src_node) and nx.has_path(g_normal, src_node, start_node): # Find all nodes reachable from start_node without passing through another substation's start_node # or the main source node again, after removing other substation links. temp_g = g_normal.copy() other_cb_links = [] for other_cb, (s,d) in self.substation_connections.items(): if other_cb != cb_id and temp_g.has_edge(s,d): other_cb_links.append((s,d)) temp_g.remove_edges_from(other_cb_links) if nx.has_path(temp_g, src_node, start_node): # All nodes in the component connected to start_node, excluding the source itself component_nodes = nx.node_connected_component(temp_g.subgraph( [n for n in temp_g.nodes if n != src_node or n == start_node] # Consider start_node part of feeder ), start_node) feeder_info[cb_id] = component_nodes else: # Should not happen if graph is built correctly feeder_info[cb_id] = {start_node} if start_node in g_normal else set() else: feeder_info[cb_id] = {start_node} if start_node in g_normal else set() return feeder_info def _calculate_line_current_kw(self, power_kw): """Calculates current (A) given power (kW) at VOLTAGE_KV (line-to-line).""" if VOLTAGE_KV <= 0: return float('inf') return abs(power_kw) / (ROOT_3 * VOLTAGE_KV * 1.0) # Assumed PF=1 for current calculation from P def _get_downstream_info(self, G, line_u, line_v, source_nodes_for_feeder): """ Calculates total load and DG power downstream of a directed line (u,v), assuming v is further from the source_node for this path. G: graph to operate on (can be a faulted graph) source_nodes_for_feeder: list of possible source nodes for the current connected component. """ # Temporarily make graph directed from source to loads to find downstream nodes # This is tricky if the graph is not purely radial or has loops after closing ties. # A simpler approach for radial sections: # Check connectivity from sources to line_v, if line_u is removed. # If line_v is disconnected from all sources when (u,v) is cut, then everything # in the component of line_v is downstream. # Create a copy of G to modify temp_g = G.copy() if not temp_g.has_edge(line_u, line_v): return [], 0, 0 # Line doesn't exist temp_g.remove_edge(line_u, line_v) downstream_nodes = set() # Check which side (u or v) is disconnected from the source(s) # Assume u is closer to source, v is further. # If v is still connected to a source, then (u,v) might be part of a loop or fed from elsewhere. # A robust way: find path from source to v. If (u,v) is on all paths, then v is downstream of u via this line. # For radial feeders (normal operation): # If we consider (u,v) where u is parent of v: component_of_v = set() q = [line_v] visited = {line_u, line_v} # Start by marking u as visited (as if coming from u) # If line_v is connected to any source_node without passing through line_u v_connected_to_source_alt_path = False for src in source_nodes_for_feeder: if nx.has_path(temp_g, src, line_v): v_connected_to_source_alt_path = True break if v_connected_to_source_alt_path: # (u,v) is part of a loop or v is fed from elsewhere # This simple downstream logic is insufficient for meshed networks. # For now, assume radial for this part of flow calculation. # A more complex flow calculation (Newton-Raphson) would be needed for meshed. # Given problem constraints, assume feeders are normally radial. pass # This line might not have a clear "downstream" if looped. # If v is disconnected from source when (u,v) is cut, then its component is downstream. # Check connectivity for v in temp_g (where (u,v) is removed) v_still_connected = any(nx.has_path(temp_g, src, line_v) for src in source_nodes_for_feeder if src in temp_g) if not v_still_connected: # v is now isolated from source, so its component is downstream of (u,v) component_of_v = nx.node_connected_component(temp_g, line_v) else: # v is still connected, means (u,v) might be redundant or complex. # Try to determine direction based on distance from source dist_u = float('inf') dist_v = float('inf') for src in source_nodes_for_feeder: if src not in G: continue if nx.has_path(G, src, line_u): dist_u = min(dist_u, nx.shortest_path_length(G, src, line_u)) if nx.has_path(G, src, line_v): dist_v = min(dist_v, nx.shortest_path_length(G, src, line_v)) if dist_v > dist_u : # v is downstream of u # Find component of v if (u,v) is removed and v is not connected to source g_temp_removed_edge = G.copy() g_temp_removed_edge.remove_edge(line_u,line_v) is_v_conn_to_src = False for src_node_feeder in source_nodes_for_feeder: if src_node_feeder in g_temp_removed_edge and nx.has_path(g_temp_removed_edge, src_node_feeder, line_v): is_v_conn_to_src = True break if not is_v_conn_to_src: component_of_v = nx.node_connected_component(g_temp_removed_edge, line_v) else: # v is still connected, (u,v) is likely a loop closing line. Flow is complex. # For simplicity, this function will return 0 flow for loop lines if direction is ambiguous. return [], 0, 0 elif dist_u > dist_v: # u is downstream of v (swap them) # similar logic for u g_temp_removed_edge = G.copy() g_temp_removed_edge.remove_edge(line_u,line_v) is_u_conn_to_src = False for src_node_feeder in source_nodes_for_feeder: if src_node_feeder in g_temp_removed_edge and nx.has_path(g_temp_removed_edge, src_node_feeder, line_u): is_u_conn_to_src = True break if not is_u_conn_to_src: component_of_v = nx.node_connected_component(g_temp_removed_edge, line_u) # component_of_v is actually comp of u else: return [], 0, 0 else: # Equidistant or complex, cannot determine simple downstream for this line return [], 0, 0 downstream_nodes = component_of_v total_downstream_load_kw = sum(G.nodes[n]['load_kw'] for n in downstream_nodes) total_downstream_dg_kw = sum(G.nodes[n]['dg_kw'] for n in downstream_nodes if G.nodes[n]['dg_kw'] > 0) return list(downstream_nodes), total_downstream_load_kw, total_downstream_dg_kw def calculate_power_flows_and_currents(self, current_graph_state, active_dgs_kw): """ Simplified power flow for radial networks or parts of networks. Returns dict of line flows and currents, and substation powers. Flows: {(u,v): power_kw} where power_kw > 0 means u to v. Currents: {(u,v): current_A} Substation_powers: {'CB1': power_kw_drawn_from_substation} """ line_flows_kw = {} line_currents_a = {} substation_powers_kw = {} # Update DG outputs in the graph state for node_id, dg_val in active_dgs_kw.items(): if node_id in current_graph_state: current_graph_state.nodes[node_id]['dg_kw'] = dg_val for node_id in current_graph_state.nodes(): # Reset others if not in active_dgs_kw if node_id not in active_dgs_kw and 'dg_kw' in current_graph_state.nodes[node_id]: if current_graph_state.nodes[node_id].get('type') != 'source': # Don't zero out if it was never a DG current_graph_state.nodes[node_id]['dg_kw'] = 0 # Determine connected components and their sources # This is a very simplified load flow. It assumes power flows from sources (substations) # down to loads. DG power reduces the load seen by upstream sections. # It does not handle loops well without iterative methods (e.g. Hardy Cross or Newton-Raphson). # For each feeder, calculate flows assuming radial structure # This is an approximation. A full AC or DC power flow is more accurate. # Initialize all line flows to 0 for u, v in current_graph_state.edges(): line_flows_kw[(u,v)] = 0 line_flows_kw[(v,u)] = 0 # For bi-directional calculation needs line_currents_a[(u,v)] = 0 # Iterate multiple times for flow distribution in case of ties or complex paths # This is a placeholder for a proper iterative flow solution. # For now, a topological sort based flow for radial parts. processed_nodes_for_flow_calc = set() for cb_id, (src_node, start_node) in self.substation_connections.items(): if not current_graph_state.has_node(start_node) or not nx.is_connected(current_graph_state.subgraph([n for n in current_graph_state.nodes() if n != SOURCE_NODE])): # This feeder might be entirely down if start_node is disconnected from actual nodes substation_powers_kw[cb_id] = 0 continue # Get nodes for this feeder (dynamic based on current_graph_state) # Nodes connected to start_node, excluding SOURCE_NODE, if start_node is connected to SOURCE_NODE feeder_nodes_component = set() if current_graph_state.has_edge(src_node, start_node): temp_g_for_feeder = current_graph_state.copy() # Remove other substation links to isolate this feeder's component other_links_to_remove = [] for other_cb, (s,d) in self.substation_connections.items(): if other_cb != cb_id and temp_g_for_feeder.has_edge(s,d): other_links_to_remove.append((s,d)) temp_g_for_feeder.remove_edges_from(other_links_to_remove) if nx.has_path(temp_g_for_feeder, src_node, start_node): try: # Consider only the part of the graph reachable from start_node, not crossing back to SOURCE_NODE # except via the designated start_node path. search_nodes = [n for n in temp_g_for_feeder.nodes if n != src_node] sub_graph_feeder = temp_g_for_feeder.subgraph(search_nodes) if start_node in sub_graph_feeder: feeder_nodes_component = nx.node_connected_component(sub_graph_feeder, start_node) except nx.NetworkXError: # if start_node not in subgraph feeder_nodes_component = set() if not feeder_nodes_component: substation_powers_kw[cb_id] = 0 continue # Order nodes from furthest to closest to substation for power accumulation # This is for radial feeders. If loops exist, this is not sufficient. # Using BFS layers from start_node # Net load at each node (Load - DG) node_net_power_kw = {} for node in feeder_nodes_component: node_net_power_kw[node] = current_graph_state.nodes[node]['load_kw'] - current_graph_state.nodes[node]['dg_kw'] # Accumulate power up towards the substation # This requires a tree traversal (e.g., DFS post-order traversal) from leaves to root. # For simplicity, if the feeder is a tree rooted at start_node: if nx.is_tree(current_graph_state.subgraph(feeder_nodes_component | {start_node})): # Check if it's a tree # Create a directed tree towards the source for easier traversal # This part is complex if graph is not a tree. # For now, sum all net loads on the feeder as the substation power (approximation) total_feeder_net_load = sum(node_net_power_kw[n] for n in feeder_nodes_component) substation_powers_kw[cb_id] = total_feeder_net_load # Distribute this flow down the lines (highly simplified) # A proper method: for each line, sum net_power of all nodes in subtree rooted by that line. # This simplified flow calculation is a major placeholder. # For overload risk, we need per-line flows. # Simplified: assume current_graph_state is a tree rooted at start_node for this feeder # Use BFS to assign flow from start_node downwards # This is not a full power flow, but an estimation for line loading. # Build a directed graph for this feeder based on BFS from start_node # This is just for flow assignment direction. # Actual flow needs to sum up demands from downstream. # For each edge (u,v) in the feeder: # Determine parent (closer to start_node) and child # Power on (parent, child) = sum of net loads in subtree rooted at child. # This is complex to implement robustly here without a full flow algorithm. # Fallback: Use the _get_downstream_info logic if possible, iterate edges # This needs to be called carefully to avoid double counting or misdirection in non-radial. # For now, this function will primarily return substation_powers_kw and # leave detailed line_flows_kw and line_currents_a for a more robust implementation # or accept its high level of approximation. # Let's try a slightly better approximation for line flows on a tree: # For each edge (u,v) in the feeder tree (rooted at start_node) # Assume u is parent of v. Power(u,v) = sum of net_power for all nodes in subtree of v. if start_node in feeder_nodes_component: # Should be try: dfs_edges = list(nx.dfs_edges(nx.bfs_tree(current_graph_state.subgraph(feeder_nodes_component), start_node), source=start_node)) # Calculate power for each node including its children's power node_total_subtree_power = node_net_power_kw.copy() for u, v in reversed(dfs_edges): # From leaves up to root if u in node_total_subtree_power and v in node_total_subtree_power: node_total_subtree_power[u] += node_total_subtree_power[v] for u,v in dfs_edges: # From root down if v in node_total_subtree_power: flow = node_total_subtree_power[v] line_flows_kw[(u,v)] = flow line_flows_kw[(v,u)] = -flow # Convention for direction line_currents_a[(u,v)] = self._calculate_line_current_kw(flow) except Exception as e: # print(f"Warning: Could not perform tree-based flow for {cb_id} due to {e}") pass # Keep substation power as sum, line flows might be inaccurate else: # Not a tree, flow calculation is more complex. # print(f"Warning: Feeder {cb_id} is not a tree. Simplified flow may be inaccurate.") total_feeder_net_load = sum(node_net_power_kw[n] for n in feeder_nodes_component if n in node_net_power_kw) substation_powers_kw[cb_id] = total_feeder_net_load # Line flows in meshed networks require iterative solvers. # For now, this part will be very approximate for meshed sections. # Handle reverse power flow and inter-feeder DG adjustment # "分布式能源不得向上级电网倒送功率" # "可以在相邻馈线间进行调节" for cb_id in list(substation_powers_kw.keys()): if substation_powers_kw[cb_id] < 0: # Reverse power flow excess_dg_on_feeder = -substation_powers_kw[cb_id] # Try to transfer to other feeders via inter-feeder tie lines # This logic is complex and needs careful state management. # For Q1, a simpler approach might be DG curtailment on that feeder. # Find inter-feeder tie switches connected to this feeder # Example: S62-3 connects Feeder 3 (node 62) to Feeder 1 (node 19) # If Feeder 1 has excess_dg, it might try to send to Feeder 3 via (19,62) # This is an advanced feature. For now, assume DG curtailment if倒送. # To implement curtailment: identify DGs on this feeder, reduce their output # proportionally until substation_powers_kw[cb_id] >= 0. # This would require re-calculating flows. # For now, just flag it. # print(f"Warning: Reverse power flow on {cb_id} of {substation_powers_kw[cb_id]} kW. DG curtailment or transfer needed.") # A simple curtailment: dG_on_feeder_nodes = [n for n in self.feeder_info.get(cb_id, set()) if n in active_dgs_kw and active_dgs_kw[n] > 0] total_dg_cap_on_feeder = sum(active_dgs_kw[n] for n in dG_on_feeder_nodes) if total_dg_cap_on_feeder > 0: curtail_ratio = min(1.0, excess_dg_on_feeder / total_dg_cap_on_feeder) if total_dg_cap_on_feeder >0 else 0 for dg_node in dG_on_feeder_nodes: active_dgs_kw[dg_node] *= (1 - curtail_ratio) # Flows need to be recalculated after curtailment. This suggests an iterative solution. # For this submission, we'll assume this check is done *before* final flow calc, # or simply note the violation. # To avoid recursion here, this function should ideally take DGs as fixed input. # The adjustment logic should be outside or iterative. pass return line_flows_kw, line_currents_a, substation_powers_kw def calculate_overload_risk(self): """ Calculates overload risk for the current DG setup. Assumes DGs are at their BASE_DG_CAPACITY_KW. """ # Get current operational graph (no faults, ties normally open) g_op = self._get_subgraph_with_operational_lines(self.graph) # Calculate power flows and currents # Need to handle DG outputs properly. current_dg_outputs = self.dg_kw.copy() # Use the model's current DG settings # Iterative step for DG curtailment if reverse power flow: # This is a simplified loop. A more robust solution uses optimization or better heuristics. for _iter in range(3): # Max 3 iterations for adjustment line_flows, line_currents, substation_powers = self.calculate_power_flows_and_currents(g_op, current_dg_outputs) reverse_power_detected = False for cb_id, power_kw in substation_powers.items(): if power_kw < -1e-3: # Small threshold for倒送 reverse_power_detected = True # print(f"Info: Reverse power on {cb_id} ({power_kw:.2f} kW). Attempting curtailment.") excess_dg_on_feeder = abs(power_kw) feeder_nodes_for_cb = self.feeder_info.get(cb_id, set()) dG_on_feeder_nodes = [n for n in feeder_nodes_for_cb if n in current_dg_outputs and current_dg_outputs[n] > 0] total_dg_cap_on_feeder = sum(current_dg_outputs[n] for n in dG_on_feeder_nodes) if total_dg_cap_on_feeder > 1e-3 : # Avoid division by zero curtail_amount_total = excess_dg_on_feeder for dg_node in dG_on_feeder_nodes: # Proportional curtailment proportion = current_dg_outputs[dg_node] / total_dg_cap_on_feeder curtail_this_dg = proportion * curtail_amount_total current_dg_outputs[dg_node] = max(0, current_dg_outputs[dg_node] - curtail_this_dg) else: # No DG to curtail, reverse power might be from other sources or model issue pass if not reverse_power_detected: break # Final flows after potential curtailment line_flows, line_currents, substation_powers = self.calculate_power_flows_and_currents(g_op, current_dg_outputs) overloaded_lines = [] for u, v, data in g_op.edges(data=True): if data.get('type') == 'substation_link': continue # Don't check substation links themselves for overload here current = line_currents.get((u,v), 0) # If flow is from v to u, current might be stored as current_uv = -current_vu # Take absolute value of flow for current calculation, or ensure current is always positive. # The _calculate_line_current_kw uses abs(power_kw) so current should be positive. rated_current = data.get('rated_current_a', DEFAULT_LINE_RATED_CURRENT_A) if current > 1.1 * rated_current: overloaded_lines.append({'edge': (u,v), 'current': current, 'rated': rated_current, 'over_by_%': (current/(1.1*rated_current)-1)*100 if rated_current>0 else float('inf')}) if overloaded_lines: # print(f"System Overload Detected. Overloaded lines: {overloaded_lines}") # P_over = 1 (for this deterministic scenario) # C_over = fixed penalty or sum of penalties risk_overload = 1.0 * COST_PENALTY_FOR_ANY_OVERLOAD # Or, sum of consequences for each overloaded line, if C_over is per line. # risk_overload = sum(some_cost_function(ol['over_by_%']) for ol in overloaded_lines) else: # print("System is NOT overloaded in the base case.") risk_overload = 0.0 return risk_overload, overloaded_lines, substation_powers, current_dg_outputs def calculate_load_loss_risk(self): """ Calculates total load loss risk by considering single line faults. R_loss = sum(P_fault_i * C_loss_i) P_fault_i = annual probability of fault i C_loss_i = consequence of fault i (e.g., unserved_load_kw * COST_VOLL_PER_KW) """ total_load_loss_risk = 0.0 detailed_fault_impacts = [] # Iterate through all operational lines (excluding substation virtual links for fault simulation) original_edges = [ (u,v,data) for u,v,data in self.graph.edges(data=True) if data.get('type') != 'substation_link' and not data.get('is_tie', False)] for u_fault, v_fault, line_data_faulted in original_edges: faulty_edge = (u_fault, v_fault) line_length_km = line_data_faulted.get('length_km', 0) # Probability of this specific line failing (annual) # Assuming failure rates are independent and this is the probability of this line being the one to fail. prob_line_fault = line_length_km * FAILURE_RATE_LINE_PER_KM if prob_line_fault == 0: continue # --- Simulate fault --- g_faulted = self.graph.copy() if not g_faulted.has_edge(*faulty_edge): continue g_faulted.edges[faulty_edge]['failed'] = True # Mark as failed g_after_fault_isolation = self._get_subgraph_with_operational_lines(g_faulted, faulty_edge=faulty_edge) # --- Identify initial load loss --- initial_unserved_load_kw = 0 disconnected_load_nodes = {} # {node: load_kw} # Check connectivity for all load nodes for node_id, load_kw in self.loads_kw.items(): if load_kw <= 0: continue is_connected_to_source = False for cb_id, (src_node, start_node) in self.substation_connections.items(): if nx.has_path(g_after_fault_isolation, src_node, node_id): is_connected_to_source = True break if not is_connected_to_source: initial_unserved_load_kw += load_kw disconnected_load_nodes[node_id] = load_kw if initial_unserved_load_kw == 0: # Fault does not cause load loss (e.g. redundant line) detailed_fault_impacts.append({'fault': faulty_edge, 'unserved_kw_initial': 0, 'unserved_kw_final':0, 'restored_kw':0, 'risk_contrib':0}) continue # --- Attempt restoration via tie lines --- # This is a complex part. Needs to: # 1. Identify disconnected areas and loads. # 2. Identify available tie switches that can connect these areas to healthy feeders. # 3. Check capacity of tie lines and the supporting feeder. # 4. Prioritize restoration (e.g., maximize load restored). # For this model, a simplified restoration: # Iterate over available tie switches. If closing one helps, simulate it. # This should be greedy or more optimized. restored_load_kw_total_for_this_fault = 0 # Create a graph state for restoration attempts g_for_restoration = g_after_fault_isolation.copy() # Sort disconnected loads by size (optional, for prioritization) sorted_disconnected_loads = sorted(disconnected_load_nodes.items(), key=lambda item: item[1], reverse=True) # Try closing tie switches one by one (if they connect a live part to a dead part) # This is a very simplified greedy approach. # A proper approach would evaluate all combinations or use optimization. # Identify current live sources/feeders live_feeder_sources = [] # (cb_id, start_node_of_live_feeder) for cb_id, (src,start) in self.substation_connections.items(): if nx.has_path(g_for_restoration, src, start): # Check if substation itself is connected live_feeder_sources.append(start) for tie in self.tie_switches_info: tie_n1, tie_n2 = tie['nodes'] tie_capacity_a = tie['capacity_A'] if not g_for_restoration.has_node(tie_n1) or not g_for_restoration.has_node(tie_n2): continue if g_for_restoration.has_edge(tie_n1, tie_n2): continue # Already closed or part of main graph (should not be for ties) # Check if one end is live and other is dead (or part of the disconnected component) tie_n1_is_live = any(nx.has_path(g_for_restoration, src, tie_n1) for src in live_feeder_sources) tie_n2_is_live = any(nx.has_path(g_for_restoration, src, tie_n2) for src in live_feeder_sources) if tie_n1_is_live == tie_n2_is_live: continue # Both live or both dead, closing doesn't bridge outage for now live_tie_node, dead_tie_node = (tie_n1, tie_n2) if tie_n1_is_live else (tie_n2, tie_n1) # Check if dead_tie_node is part of the current outage we are trying to fix # This requires knowing which component dead_tie_node belongs to. # For now, assume if it's not live, it's part of some outage. # Simulate closing this tie switch g_for_restoration.add_edge(live_tie_node, dead_tie_node, id=tie['id'], type='tie_closed', rated_current_a=tie_capacity_a, resistance_ohm=0.001, length_km=0.01) # Check how much load can be restored through this tie without overloading tie or new path # This requires a flow calculation on g_for_restoration. # Simplified: Check loads now connected. newly_restored_load_kw_this_tie = 0 temp_restored_nodes_this_tie = [] for node_id, load_val in disconnected_load_nodes.items(): if node_id not in g_for_restoration: continue # Should not happen # Check if this node is now connected to ANY source is_now_connected = any(nx.has_path(g_for_restoration, src, node_id) for src in live_feeder_sources) if is_now_connected and node_id not in temp_restored_nodes_this_tie: # And not already counted as restored by previous ties # More checks needed: # 1. Tie line capacity: Power through (live_tie_node, dead_tie_node) <= tie_capacity_a # 2. Path capacity on the live feeder. # This is where the simplified flow becomes a bottleneck. # For now, assume if connected, it can be restored up to a certain limit. # This is a MAJOR simplification. newly_restored_load_kw_this_tie += load_val temp_restored_nodes_this_tie.append(node_id) # Here, we'd need to check if adding newly_restored_load_kw_this_tie overloads the tie or feeder. # If current_through_tie > tie_capacity_a, then not all of this load can be restored. # This part needs a proper constrained flow allocation. # For now, let's assume a fraction can be restored if connected, or all if small. # This is a placeholder for a more robust restoration algorithm. # Let's assume, for now, if connected, it's restored. If this overloads things, # the overload risk model should capture it (but that's for normal state). # Here, the goal is to minimize unserved load. # If this tie leads to overload, we shouldn't use it or only partially. # For now, naively accept all newly connected load. if newly_restored_load_kw_this_tie > 0: restored_load_kw_total_for_this_fault += newly_restored_load_kw_this_tie # Update disconnected_load_nodes: for r_node in temp_restored_nodes_this_tie: if r_node in disconnected_load_nodes: del disconnected_load_nodes[r_node] # No longer disconnected else: # Closing this tie didn't help, revert if g_for_restoration.has_edge(live_tie_node, dead_tie_node): g_for_restoration.remove_edge(live_tie_node, dead_tie_node) # Final unserved load for this fault scenario final_unserved_load_kw = initial_unserved_load_kw - restored_load_kw_total_for_this_fault final_unserved_load_kw = max(0, final_unserved_load_kw) # Cannot be negative consequence_c_loss = final_unserved_load_kw * COST_VOLL_PER_KW risk_contribution = prob_line_fault * consequence_c_loss total_load_loss_risk += risk_contribution detailed_fault_impacts.append({ 'fault_type': 'line', 'component_id': faulty_edge, 'prob_fault': prob_line_fault, 'unserved_kw_initial': initial_unserved_load_kw, 'restored_kw': restored_load_kw_total_for_this_fault, 'unserved_kw_final': final_unserved_load_kw, 'consequence_c_loss': consequence_c_loss, 'risk_contribution': risk_contribution }) # TODO: Add DG faults, Switch faults, User faults if they cause wider outages. # For DG faults: prob_dg_fault = FAILURE_RATE_DG_PERCENT # A DG fault primarily impacts system's ability to meet load or avoid overload. # It doesn't directly cause load loss unless it's islanded and the DG is the only source. # The problem implies grid-connected DGs. return total_load_loss_risk, detailed_fault_impacts # --- Main Execution --- if __name__ == '__main__': print("--- 配电网风险评估模型 Q1 ---") # 1. Load Data print("\n1. 加载数据...") loads = load_load_data() topology = load_topology_data() # print(f"负荷数据: {len(loads)} 点") # print(f"拓扑数据: {len(topology)} 条线路") # 2. Initialize Power Grid Model print("\n2. 初始化电网模型...") grid = PowerGridModel(loads, topology, DG_LOCATIONS_KW, TIE_SWITCHES_INFO, SUBSTATION_CONNECTIONS) # print(f"电网: {grid.graph.number_of_nodes()} 个节点, {grid.graph.number_of_edges()} 条边") # print(f"馈线信息: {grid.feeder_info}") # --- 问题1: 失负荷风险和过负荷风险计算模型 --- print("\n--- 问题1: 风险计算 ---") # A. 过负荷风险模型 (R_over = P_over * C_over) # For Q1, DGs are at BASE_DG_CAPACITY_KW. This is a deterministic check for this state. # P_over = 1 if overload occurs, 0 otherwise. C_over is the penalty. print("\nA. 计算过负荷风险...") # Note: The calculate_power_flows_and_currents is highly simplified. # Results for overload depend heavily on its accuracy and line ratings. try: risk_overload, overloaded_lines_details, substation_p, final_dg_out = grid.calculate_overload_risk() print(f" 计算得到的过负荷风险 (R_over): ${risk_overload:.2f}") if overloaded_lines_details: print(f" 检测到过负荷线路 ({len(overloaded_lines_details)} 条):") # for ol in overloaded_lines_details[:3]: # Print first 3 # print(f" - 线路 {ol['edge']}, 电流: {ol['current']:.2f}A, 额定: {ol['rated']:.2f}A, 超出: {ol['over_by_%']:.2f}%") else: print(" 在当前DG配置下,未检测到线路过负荷。") # print(f" 变电站出口功率 (kW): {substation_p}") # print(f" 最终DG出力 (kW) (可能经过削减): {final_dg_out}") except Exception as e: print(f" 计算过负荷风险时发生错误: {e}") risk_overload = -1 # Indicate error # B. 失负荷风险模型 (R_loss = sum(P_fault_i * C_loss_i)) print("\nB. 计算失负荷风险...") # Note: Restoration logic is simplified. try: total_r_loss, fault_details = grid.calculate_load_loss_risk() print(f" 计算得到的总失负荷风险 (R_loss): ${total_r_loss:.2f} (基于所选成本)") # print("\n 部分故障场景详情:") # for fd in fault_details[:3]: # Print first 3 # print(f" - 故障线路: {fd.get('component_id')}, " # f"初始失负荷: {fd.get('unserved_kw_initial'):.2f} kW, " # f"最终失负荷: {fd.get('unserved_kw_final'):.2f} kW, " # f"风险贡献: ${fd.get('risk_contribution'):.2f}") except Exception as e: print(f" 计算失负荷风险时发生错误: {e}") total_r_loss = -1 # Indicate error print("\n--- 模型执行完毕 ---") print("注意: 此模型包含多项简化和假设 (如线路额定电流, 成本参数, 潮流计算简化, 恢复逻辑简化).") print("结果的准确性取决于这些假设的合理性和参数的精确性。") 对此代码进行分析
最新发布
05-12
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值